analysis of low-cost testing methods for led lumen

196
ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN MAINTENANCE OF OFF-GRID LIGHTING PRODUCTS HUMBOLDT STATE UNIVERSITY By Christopher Carlsen A Thesis Presented to The Faculty of Environmental Resources Engineering In Partial Fulfillment Of the Requirements for the Degree Master of Science Environmental Resources Engineering May, 2011

Upload: others

Post on 31-Jan-2022

0 views

Category:

Documents


0 download

TRANSCRIPT

Page 1: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

MAINTENANCE OF OFF-GRID LIGHTING PRODUCTS

HUMBOLDT STATE UNIVERSITY

By

Christopher Carlsen

A Thesis

Presented to

The Faculty of Environmental Resources Engineering

In Partial Fulfillment

Of the Requirements for the Degree

Master of Science

Environmental Resources Engineering

May, 2011

Page 2: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

MAINTENANCE OF OFF-GRID LIGHTING PRODUCTS

HUMBOLDT STATE UNIVERSITY

By

Christopher Carlsen

Approved by the Master‟s Thesis Committee:

Dr. Arne Jacobson, Major Professor Date

Dr. Charles Chamberlin, Committee Member Date

Dr. Eileen Cashman, Committee Member Date

Dr. Chris Dugaw, Graduate Coordinator Date

Dr. Jená Burges, Vice Provost Date

Page 3: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

iii

ABSTRACT

ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

MAINTENANCE OF OFF-GRID LIGHTING PRODUCTS

Christopher R. Carlsen

Two low-cost test methods for measuring the lumen maintenance of off-grid

lighting products, a box-photometer and a tube-photometer, were compared to a high-

accuracy integrating sphere system in order to evaluate the measurement error of the low-

cost testing devices. Results from this study indicate that the box- and tube-photometers

can be used to accurately measure the relative change in light output of typical off-grid

lighting products. The tube and box are simple and inexpensive to construct, and they are

quicker and easier to use than an integrating sphere system. The box-photometer and

tube-photometer are recommended for use in testing off-grid lighting products according

to the Long-Term Lumen Degradation Test specified in Lighting Africa‟s Quality Test

Method and Initial Screening Method.

The box‟s strengths lie in its versatility and ease-of-use. The box-photometer can

be used for testing product run-time and lumen maintenance, and a single measurement

can be conducted in a matter of seconds by a minimally trained technician. The tube-

Page 4: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

iv

photometer, while nearly as easy to use as the box, excels in terms of constructability,

cost, and physical size. A tube-photometer can be built of basic materials that are widely

available at very low cost. No specialized tools or skills are necessary to build the

device. The tube-photometer is small, light and portable, allowing for easy storage and

minimal space for operation.

Results from this study indicate that a small degree of measurement error can

result from improper use of the low-cost devices. Box-photometer measurements are

susceptible to error caused by changes in the light source location and orientation,

especially for highly directional and adjustable products. The tube-photometer is prone

to small errors due to false identification of the maximum illuminance measurement.

These errors, however, can be easily minimized by repetition of measurements and

training test operators to use the equipment properly.

Page 5: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

v

ACKNOWLEDGEMENTS

This research would not have been possible without the support of several

individuals and organizations over the past year. I would like to thank Dr. Arne Jacobson

for his guidance and for the opportunity to become involved in the lighting lab at Schatz

Energy Research Center (SERC). Peter Alstone at SERC was also a valuable resource, as

well as the creator of the first tube-photometer design.

A bulk of this study was conducted at the National Lighting Test Center (NLTC)

in Beijing, China, who so graciously allowed me to take over their Testing Method

Research Lab for a couple months. Thanks to Dr. Klaus Mehl at NLTC for arranging my

cooperation with NLTC. Mr. Xin Hong Zheng at NLTC was a critical player in this

research, as he took me under his wing and patiently walked me through the applicable

photometric principles and measurement methods. My work at NLTC was funded by the

National Science Foundation‟s (NSF) International Research Experience for Engineers

(IREE) program.

I would like to thank James Wafula at the University of Nairobi (UoN) in Kenya

for his participation in qualitative analysis of the test methods. Leo Blyth at Lighting

Africa was also of assistance in Nairobi, as we worked together to fabricate the box-

photometer for UoN and improve the device construction plans. Our cooperation in

Kenya and the U.S. was funded by the International Finance Corporation‟s Lighting

Africa Program.

Page 6: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

vi

Thanks to Kevin Gauna for his steady stream of photometric testing advice and

his practical perspective on evaluating off-grid lighting products. I‟d like to acknowledge

Dr. Robert Van Kirk and Dr. Charles Chamberlin at Humboldt State University for their

assistance in the statistical analysis of the test results. My master‟s degree studies at HSU

for the 2010-2011 term were funded by the NSF through the Professional Environmental

Resources Engineering Science Master‟s Program.

Page 7: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

vii

TABLE OF CONTENTS

ABSTRACT ....................................................................................................................... iii

ACKNOWLEDGEMENTS .................................................................................................v

TABLE OF CONTENTS .................................................................................................. vii

LIST OF TABLES ............................................................................................................. xi

LIST OF FIGURES ......................................................................................................... xiii

LIST OF APPENDICIES ............................................................................................... xvii

INTRODUCTION ...............................................................................................................1

Objectives .................................................................................................................... 3

Structure of Thesis ....................................................................................................... 4

1 BACKGROUND .............................................................................................................6

1.1 Off-grid Lighting in Sub-Saharan Africa ......................................................... 7

1.2 Fuel-based Lighting .......................................................................................... 8

1.3 Off-Grid Lighting Products (OLPs) ............................................................... 10

1.3.1 Currently Available OLPs ................................................................... 12

1.3.2 Market Spoilage ................................................................................... 14

1.3.3 OLP Market ......................................................................................... 16

1.4 Lighting Africa – Catalyzing Markets for Modern Lighting .......................... 20

1.5 OLP Quality Assurance .................................................................................. 21

1.5.1 Lighting Africa Quality Test Method (QTM) ..................................... 22

1.5.2 Lighting Africa Initial Screening Method (ISM) ................................ 25

1.6 OLP Systems .................................................................................................. 26

Page 8: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

viii

1.6.1 Energy Conversion: Photovoltaic Modules ......................................... 28

1.6.2 Batteries ............................................................................................... 29

1.6.3 Control Circuitry.................................................................................. 30

1.6.4 Light Emitting Diodes (LEDs) ............................................................ 31

1.6.5 Lumen Maintenance of LEDs .............................................................. 35

1.6.6 Lumen Maintenance Testing ............................................................... 37

1.7 Photometry...................................................................................................... 41

1.7.1 Luminous Flux ..................................................................................... 43

1.7.2 Luminous Intensity .............................................................................. 44

1.7.3 Illuminance .......................................................................................... 44

1.7.4 Determining Luminous Flux of a Light Source................................... 45

1.8 Light Measuring Devices ................................................................................ 46

1.8.1 Spectroradiometer ................................................................................ 46

1.8.2 Illuminance meter ................................................................................ 47

1.8.3 Goniophotometer ................................................................................. 47

1.8.4 Integrating Sphere ................................................................................ 48

1.8.5 Box-Photometer ................................................................................... 54

1.8.6 Tube-Photometer ................................................................................. 58

2 MATERIALS AND METHODS ..................................................................................60

2.1 Materials ......................................................................................................... 60

2.1.1 Devices under Test (DuTs) .................................................................. 60

2.1.2 LED Driver .......................................................................................... 62

2.1.3 Box-Photometer ................................................................................... 63

Page 9: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

ix

2.1.4 Tube-Photometer ................................................................................. 65

2.1.5 Integrating Sphere – Spectroradiometer System ................................. 66

2.2 Integrating Sphere Calibration ........................................................................ 68

2.2.1 Measuring non-uniform reflectance of interior sphere surface ........... 69

2.2.2 Correcting for Light Distribution Mismatch ....................................... 70

2.2.3 Correcting for Self Absorption ............................................................ 73

2.2.4 Color Mismatch ................................................................................... 74

2.3 Simulated Lumen Maintenance Test .............................................................. 76

2.3.1 Experimental Design ........................................................................... 78

2.3.2 Driving the DuT................................................................................... 80

2.3.3 Determining the LED Nominal Drive Current .................................... 80

2.3.4 Adjusting the LED Drive Current ....................................................... 81

2.3.5 DuT Orientation ................................................................................... 82

2.3.6 DuT Testing Order............................................................................... 82

2.3.7 Ambient Temperature Regulation ....................................................... 83

2.3.8 Integrating Sphere Procedure .............................................................. 83

2.3.9 Box-Photometer Procedure.................................................................. 84

2.3.10 Tube-Photometer Procedure ................................................................ 85

3 RESULTS ......................................................................................................................87

3.1 Calibration Plots ............................................................................................. 87

3.2 Relative Change in Initial Light Output ......................................................... 90

3.3 Box-Photometer Calibration Plots .................................................................. 92

3.4 Tube-Photometer Calibration Plots ................................................................ 96

Page 10: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

x

3.5 Relative Change in Initial Light Output ......................................................... 99

3.6 Statistical Analysis of Test Data ................................................................... 101

3.6.1 Standard Error of Regression for Calibration Plots ........................... 101

3.6.2 95% Confidence Intervals of the Calibration Plot ............................. 103

3.6.3 Error in Relative Light Output Calculations...................................... 107

3.6.4 Analysis of Variance (ANOVA) for R2 of Calibration Plots ............ 109

4 DISCUSSION OF RESULTS .....................................................................................120

4.1 Worst Case: Error Analysis of Firefly in the Box-Photometer ................... 121

4.2 Sources of Error ............................................................................................ 123

5 CONCLUSIONS AND RECOMMENDATIONS ......................................................125

5.1 Impact of Error on Lumen Maintenance Test Results .................................. 125

5.2 Benefits and Disadvantages of the Box-Photometer .................................... 126

5.3 Benefits and Disadvantages of the Tube-Photometer ................................... 128

5.4 Multiple Test Operators ................................................................................ 129

5.5 Potential Improvements to Lumen Maintenance Testing Devices ............... 130

REFERENCES ................................................................................................................133

Page 11: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

xi

LIST OF TABLES

Table 1: Specifications of Devices under Test (DuTs) used in the Simulated Lumen

Maintenance Test. .............................................................................................62

Table 2: Summary of the light distribution mismatch correction for the DuTs. ...............73

Table 3: Auxiliary lamp luminous flux measurements for all of the light sources

according to body color. ....................................................................................76

Table 4: Structure of the Simulated Lumen Maintenance Test. .......................................79

Table 5: Calibration plots whose linear models do not pass through the origin within the

95% confidence interval.. ................................................................................107

Table 6: Summary of null hypotheses tested in ANOVA and interpretation of the

statistical test results. .......................................................................................109

Table 7: Hierarchy of experimental design for ANOVA................................................113

Table 8: Summary of the categorical predictor variables used in the Simulated Lumen

Maintenance test statistical analysis. ...............................................................113

Table 9: Statistical results from the general linear model for the transformed R2 values of

the calibration plots. ........................................................................................117

Table 10: Illuminance map describing irregularity in the interior integrating sphere

surface used in testing. ....................................................................................140

Table 11: Raw data from comparison testing of the Firefly product in the integrating

sphere, box-photometer and tube-photometer. ................................................149

Table 12: Raw data from comparison testing of the Kiran product in the integrating

sphere, box-photometer and tube-photometer. ................................................150

Table 13: Raw data from comparison testing of the Solux product in the integrating

sphere, box-photometer and tube-photometer. ................................................151

Table 14: Raw data from comparison testing of the Aishwarya product in the integrating

sphere, box-photometer and tube-photometer. ................................................152

Page 12: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

xii

Table 15: Maximum error in the calculation of relative decrease in the initial light output

according to light source, round of testing, testing apparatus, and orientation

within the box-photometer. .............................................................................159

Table 16: R-squared values, standard error of the linear regression, and the standard error

as a percentage of the average luminous flux estimated by the linear regression

model for each round of Simulated Lumen Maintenance testing ...................166

Page 13: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

xiii

LIST OF FIGURES

Figure 1: Total cost of illumination services for select off-grid lighting products. ..........12

Figure 2: Solar Portable Light (SPL) Quality Matrix that describes the range of solar

charged portable lighting products according to the autonomous run time .....14

Figure 3: Solar Portable Light (SPL) market growth scenarios ........................................19

Figure 4: Conceptual system description of a typical off-grid lighting product. ..............27

Figure 5: Off-grid lighting product solar panel prices are expected to continue rapid cost

decline ..............................................................................................................29

Figure 6: Electrical model of a light emitting diode (LED). .............................................32

Figure 7: Components of a typical through-hole type LED package.. .............................32

Figure 8: Trends and projections for luminous efficacy of commercially available cool

and warm white LEDs .....................................................................................34

Figure 9: Electromagnetic spectrum, including the range of visible light wavelengths

detectable by the human eye. ...........................................................................41

Figure 10: V-lambda curve describing the sensitivity of the „standard observer‟ (i.e.

typical human vision) to different wavelengths in the visible spectrum. ........43

Figure 11: Lamp measurement sphere using a detector mounted directly at the view port

..........................................................................................................................50

Figure 12: Box-photometer. ..............................................................................................55

Figure 13: Optimal measuring configuration for a box-photometer. ................................57

Figure 14: Line drawing of a tube-photometer, indicating the basic device components.

..........................................................................................................................58

Figure 15: Current shunt circuit used to measure current through the LEDs. ..................63

Figure 16: Tube-photometer similar to that used in the study. .........................................65

Figure 17: Example of the software output for a measurement of the Firefly luminous

flux.. .................................................................................................................67

Page 14: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

xiv

Figure 18: Device for measuring the reflectance characteristics of the integrating

sphere‟s interior surface. ..................................................................................70

Figure 19: Computer rendering of the simplified three dimensional light distribution

shapes for the standard lamp, Solux, Firefly, Kiran, and Aishwarya ..............72

Figure 20: Photos of self absorption color dependency measurements using the

integrating sphere .............................................................................................76

Figure 21: Tube calibration plot for one round of measurements with the Firefly...........88

Figure 22: Relative change in initial light output for the Firefly, as measured by the

integrating sphere and the tube-photometer. ....................................................91

Figure 23: Box-photometer calibration plot for the Firefly ..............................................94

Figure 24: Box-photometer calibration plot for the Kiran ................................................94

Figure 25: Box-photometer calibration plot for the Solux................................................95

Figure 26: Box-photometer calibration for the Aishwarya ...............................................95

Figure 27: Tube-photometer calibration plot for the Firefly ............................................97

Figure 28: Tube-photometer calibration plot for the Kiran ..............................................97

Figure 29: Tube-photometer calibration plot for the Solux ..............................................98

Figure 30: Tube-photometer calibration plot for the Aishwarya ......................................98

Figure 31: Plots of measured relative luminous flux decrease as a percentage of the

original light output for Firefly over three rounds of testing. ........................100

Figure 32: Tube calibration plot for the second round of testing with the Firefly,

including the best fit line and 95% confidence interval bounds ....................104

Figure 33: Enlarged section of the tube calibration plot for the Firefly. ........................105

Figure 34: Box calibration plot for round 3 of testing the Firefly with the light directed at

the lid of the box, including the linear model and the 95% confidence interval

bounds ............................................................................................................105

Figure 35: Lumen depreciation plot for the third round of testing with the Aishwarya..

........................................................................................................................108

Figure 36: Histogram of R-squared values for the experimental calibration plots .........112

Page 15: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

xv

Figure 37: Residual plots of the R-squared values for the experimental calibration plots.

........................................................................................................................115

Figure 38: Histogram of R2 data after applying the transformation shows normal

distribution. ....................................................................................................116

Figure 39: Interaction plot showing how the mean R2 value varies across the box- and

tube-photometer testing methods and four light sources. ..............................119

Figure 40: Light measurement apparatus used at the Lighting Research Center for lumen

maintenance and run time testing of off-grid lighting products ....................131

Figure 41: Off-grid lighting products used in this study.................................................138

Figure 42: Radial light distribution of the Firefly product .............................................142

Figure 43: Radial light distribution of the Aishwarya product .......................................142

Figure 44: Radial light distribution of the Solux product ...............................................143

Figure 45: Radial light distribution of the Kiran product ...............................................143

Figure 46: Radial light distribution of the standard lamp ...............................................144

Figure 47: Plots of measured relative luminous flux decrease for Firefly for three rounds

of testing.........................................................................................................154

Figure 48: Plots of measured relative luminous flux decrease for Kiran for three rounds

of testing.........................................................................................................155

Figure 49: Plots of measured relative luminous flux decrease for Solux for three rounds

of testing.........................................................................................................156

Figure 50: Plots of measured relative luminous flux decrease for Aishwarya for three

rounds of testing. ............................................................................................157

Figure 51: Comparison of integrating sphere lumen measurement to tube- and box-

photometer lux measurements for each round of Firefly Simulated Lumen

Maintenance testing. ......................................................................................161

Figure 52: Comparison of integrating sphere lumen measurement to tube- and box-

photometer lux measurements for each round of Kiran Simulated Lumen

Maintenance testing. ......................................................................................162

Page 16: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

xvi

Figure 53: Comparison of integrating sphere lumen measurement to tube- and box-

photometer lux measurements for each round of Solux Simulated Lumen

Maintenance testing. ......................................................................................163

Figure 54: Comparison of integrating sphere lumen measurement to tube- and box-

photometer lux measurements for each round of Aishwarya Simulated Lumen

Maintenance testing. ......................................................................................164

Page 17: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

xvii

LIST OF APPENDICIES

APPENDIX A: LIGHT SOURCES USED IN TESTING .............................................137

APPENDIX B: INTEGRATING SPHERE INTERNAL REFLECTANCE MAP ........139

APPENDIX C: DUT LIGHT DISTRIBUTION .............................................................141

APPENDIX D: INTEGRATING SPHERE PROCEDURE ...........................................145

APPENDIX E: RAW TEST DATA ...............................................................................148

APPENDIX F: RELATIVE LUMEN DECREASE PLOTS ..........................................153

APPENDIX G: MAXIMUM ERROR IN RELATIVE LUMEN DECREASE .............158

APPENDIX H: CALIBRATION PLOTS ......................................................................160

APPENDIX I: STANDARD ERROR OF TEST RESULTS .........................................165

APPENDIX J: BOX-PHOTOMETER CONSTRUCTION PLANS AND

INSTRUCTIONS ..................................................................................168

APPENDIX K: TUBE-PHOTOMETER CONSTRUCTION PLANS ..........................176

APPENDIX L: LIST OF EQUIPMENT USED IN THE STUDY .................................178

Page 18: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

1

INTRODUCTION

Kerosene-fueled lamps and candles are the main sources of light for

approximately 600 million people (69.5% of the population) in Sub-Saharan Africa who

lack access to electricity (International Energy Agency, 2010). Every year, African

households and small businesses spend upwards of $17 billion on lighting (Lighting

Africa, 2010d). Many low-income African households spend approximately 2% of their

income on fuel for lighting (Mills, 2005; World Bank, 2009).1 Not only does this

population spend a significant portion of their household income on fuel, but they are

consequently subjected to poor light quality and health hazards due to emission of air

pollutants (Apple, 2010; Mills, 2005). A transition away from dirty, inefficient kerosene

combustion will drastically reduce the environmental, health and economic problems

resulting from current off-grid lighting practices.

Recent technological advancement and decreasing cost of Light Emitting Diodes

(LEDs), photovoltaic (PV) modules, and rechargeable batteries offer a potentially

sustainable means of addressing the multi-faceted problem of off-grid lighting in Sub-

Saharan Africa. Clean, efficient and reliable; off-grid lighting products (OLPs) are

beginning to gain a foothold in a market dominated by fuel-based lighting. New OLPs

are entering the market with affordable prices that may position them to replace fuel-

1 Percent of household income dedicated to the purchase of kerosene for fuel-based lighting calculated

based on 3.2 L/month-person kerosene consumption, $0.50/Liter of kerosene, and $1,125 gross national

income per capita.

Page 19: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

2

based lighting in Sub-Saharan Africa and throughout the developing world.

Unfortunately, the off-grid lighting market is becoming flooded with poor quality

products. Inferior goods are contributing to “market spoilage,” whereby low-income

end-users sustain significant financial losses due to rapid failure of the lighting products.

Negative experiences undermine consumer confidence, ultimately creating a barrier to

the widespread adoption of LED products (Lighting Africa 2010c; Tracy, 2010).

In an effort to reduce market spoilage, the author has been working with the

Schatz Energy Research Center (SERC) and the National Lighting Test Center (NLTC)

to develop a standardized method for testing the performance of off-grid lighting

products. SERC is working with Lighting Africa, a project of the World Bank Group, to

establish an internationally accepted testing protocol and performance standards that will

be used for certification of off-grid lighting products. This testing method, currently

titled Lighting Africa Quality Test Method (QTM), is the culmination of contributions

made by Fraunhofer Institute for Solar Energy Systems (FISE), SERC, NLTC and

others.2

This thesis specifically addresses the QTM lumen maintenance test, referred to as

the Long-Term Lumen Degradation Test, which is conducted to evaluate the irreversible

decrease in light output from a light source over time. Tests performed by Lighting

Africa have indicated that some LED lighting products on the market will lose a large

percentage of their light output in the first few days or months of operation (Lighting

2 The most recently published version of the QTM is available for download online at

http://lightingafrica.org/resources/technical-research.html

Page 20: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

3

Africa, 2010b). The lumen maintenance test is designed to distinguish between products

that experience rapid lumen depreciation and those that are able to provide thousands of

hours of lighting service. Verifying that an OLP can maintain appropriate levels of light

output instills confidence in the consumer that over its lifetime, the product will cost less

than fuel-based alternatives.

The photometric industry standard method for determining lumen maintenance of

a lighting product established by the Illuminating Engineering Society (IES) requires

expensive, specialized equipment and training that is generally not accessible to Lighting

Africa‟s affiliates in developing countries (IES, 2008). The Long-Term Lumen

Degradation Test that is being developed by Lighting Africa utilizes relatively simple

and inexpensive devices that are more appropriate for testing centers and manufacturers

with small budgets.

Objectives

This research compares the measurement accuracy of two low-cost light

measuring devices to a spectroradiometer-integrating sphere system, which is the

photometric industry standard testing apparatus. One low-cost apparatus, a box-

photometer, was custom built according to specifications described in the QTM (Lighting

Africa, 2010c). The other low-cost device, a tube-photometer, was custom made

according to a prototype developed at SERC. The three testing devices were used to

conduct parallel measurements of the light output from four different OLPs. Illuminance

Page 21: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

4

measurements made with the box- and tube-photometers are compared to the luminous

flux determined with the integrating sphere over a range of light output levels.

„Calibration plots‟ were generated for each round of testing in order to evaluate how

closely the low-cost devices replicate the integrating sphere measurements. Statistical

analysis of the test results is used to describe the behavior of the systems under different

conditions, including varying light levels, spatial light distributions, and physical size.

The box- and tube-photometers are also evaluated qualitatively in terms of ease of use,

repeatability and general appropriateness for use in the QTM. The ultimate objective of

this research is to provide Lighting Africa and the off-grid lighting industry with well-

supported recommendations for improving the Long-Term Lumen Degradation Test.

Structure of Thesis

To put this research in context, Sections 1.1 and 1.2 begin with a general

overview of off-grid lighting in Sub-Saharan Africa. Section 1.3 I addresses the typical

forms of fuel-based lighting as well as OLPs that have been recently entering the market.

A broad overview of the current OLP market and forecasts for market development over

the coming decade are presented. Lighting Africa and the quality assurance strategy to

which this research directly applies is then introduced in Sections 1.4 and 1.5. Beginning

in Section 1.6, the thesis delves into a technical, component-level evaluation of OLP

systems, with an emphasis on LEDs and the phenomenon of lumen depreciation. Next,

Sections 1.7 and 1.8 describe the basic photometric concepts and devices that are used in

Page 22: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

5

this research. After laying out the setting and general scientific principles associated with

this study, the specific materials and methods used to compare the accuracy of the lumen

maintenance testing apparatuses are outlined in Section 2. The experimental data from

this research are presented in Section 3, including statistical analysis of the results.

Finally, in Sections 4 and 5 the results are interpreted and conclusions are drawn about

the use and appropriateness of low-cost lumen maintenance test methods for OLP quality

assurance testing.

Page 23: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

6

1 BACKGROUND

The underlying premise of this study is that lighting is a human necessity. Since

the discovery of fire, human development has been inextricably linked to artificial light.

Without light sources that range from a simple fire, to modern incandescent, fluorescent,

and LED luminaires, we would not have achieved the levels of general well-being and

productivity that most of us take for granted. In his 2005 article about off-grid lighting,

Evan Mills states that “illumination is one of the core end-use services sought by society”

(Mills, 2005).

The narrow focus of this research, low-cost testing of OLP lumen maintenance, is

a small but important issue in the big picture of delivering affordable, high quality

lighting service to low income people in Sub-Saharan Africa and throughout the

developing world. In fact, lumen maintenance testing of off-grid lighting products is

only one of a suite of tests included in Lighting Africa‟s Quality Test Method (Lighting

Africa, 2010c). In order to understand the importance of quality assurance testing of off-

grid lighting products, and specifically lumen maintenance testing, one must first

understand the context in which the testing occurs.

Page 24: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

7

1.1 Off-grid Lighting in Sub-Saharan Africa

Joseph Conrad‟s 1902 novel, Heart of Darkness, alludes in part to the dark

African nights before the dawn of modern illumination. While much has changed since

the late 19th

century, much of Sub-Saharan Africa is still literally in the dark. Currently,

585.2 million people in Sub-Saharan Africa lack access to electricity (International

Energy Agency, 2010). Expansion of the electric grid to rural areas in developing

countries has proven to be very slow due to low load densities, coupled with high capital

costs and low efficiencies associated with thermal power generation (Mills, 2005). The

IEA forecasts that by 2030 Africa‟s un-electrified population will grow to 700 million

(International Energy Agency, 2010).

Some have turned to solar home systems (SHS) as a solution to the energy access

dilemma. The market for SHS has experienced growth in Sub-Saharan Africa (Lighting

Africa, 2010d). The cost of system components is dropping and quality is improving, yet

SHS still remain financially out of reach for many of Africa‟s rural poor and this is not

expected to change in the next decade (GTZ, 2010). Households that are able to afford

SHS, however, may only allocate a small proportion of the energy to lighting. A 2004

study on rural electrification in Kenya asserts “that in many households, especially those

with small systems, intra-household dynamics constrain key social uses (e.g. children‟s

studying), as the energy is allocated to other uses” (Jacobson, 2007).

Page 25: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

8

1.2 Fuel-based Lighting

For the nearly 70% of Sub-Saharan Africans that do not have grid electricity, fuel-

based light sources like wick lamps and candles are a common means of household and

commercial illumination. Fuel-based lighting has dominated the market due to the

availability of fuels and lamps, as well as the relatively low initial and recurring costs of

operation. Even very rural villages have access to fuel-based lighting, and fuel can be

purchased in small, affordable quantities. Simple, low-cost and robust lamps are

available in rural markets. Candles are also a common commodity. Inexpensive wick

lamps can be fashioned from recycled metal cans with basic hand tools. From the

perspective of a low-income rural household living without electricity, fuel-based

lighting makes sense. The consumers may not fully consider, however, a laundry list of

negative impacts associated with burning biomass and fossil fuels for light.

From an economic standpoint, the average low-income African family spends

approximately 2% of their monthly budget on recurring fuel expenses for lighting (Mills,

2005; World Bank, 2009). With an enormous population off the grid, these fuel expenses

add up. In 2005, Evan Mills estimated that fuel-based lighting is responsible for annual

energy consumption of 77 billion liters of fuel worldwide (or 2800 petajoules, PJ), at a

cost of $38 billion/year or $77 per household (Mills, 2005). Price volatility due to

instability of subsidy regimes and ongoing increases in kerosene prices, estimated at 4%

annually over the next few years, will make fuel-based lighting less and less attractive

(Lighting Africa, 2010d).

Page 26: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

9

When evaluated in terms of the illumination provided, fuel-based lighting proves

to be inadequate and expensive. The amount of light emitted from candles and wick

lanterns is too low for many basic tasks. A simple wick lantern provides about 1 lux

(lumens/m2) at 1 meter from the source, compared with levels on the order of 500 lux

routinely provided in industrialized countries (Mills, 2005). Mills also estimates that the

cost per unit of useful lighting energy service delivered ($/lux-hour of light, including

capital and operating costs) for fuel-based lighting is up to ~150 times that for premium

efficiency fluorescent lighting (Mills, 2005).

Fuel-based lighting also raises health concerns. Combustion of biomass and fossil

fuels emits particulate matter into the air that increases the risk of respiratory illnesses. A

recent study led by Dustin Poppendieck found that vendors who use a single simple wick

lamp in high-air-exchange market kiosks will likely be exposed to levels of PM2.5

(particulate matter with a diameter less than or equal to 2.5 m) that are an order of

magnitude greater than ambient health guidelines (Apple et al., 2010) . When placed in

homes with lower air exchange rates, fuel-based lights can lead to even higher

concentrations of respirable particulate matter in the air.

It is also well known that burning fossil fuels emits greenhouse gases into the

atmosphere. A single kerosene lamp that is used four hours per day emits over 100 kg of

CO2 into the atmosphere each year (Mills, 2005). Evan Mills found that the combustion

of fuel for lighting results in 190 million metric tonnes per year of carbon dioxide

emissions, equivalent to one-third the total emissions from the U.K. (Mills, 2005).

Page 27: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

10

1.3 Off-Grid Lighting Products (OLPs)

Increasing fuel prices as well as health and environmental concerns are shifting

consumers‟ attention towards alternatives to fuel-based lighting. With SHS prices still

too high and grid electrification too far away, some low-income households in Sub-

Saharan Africa are purchasing off-grid lighting products (OLPs). A working definition

of OLPs for this research is based on a description of “solar portable lights” presented by

Lighting Africa in their recent publication, titled Solar Lighting for the Base of the

Pyramid. OLPs can be differentiated from SHS and other small scale lighting devices

according to function, technology, and quality.

Function - Lighting systems range from the task specific (torches/flashlights) to the

general ambient lighting functions. These products can include added functions such

as mobile phone charging, but light has to be the primary design driver. The

functionality also has to allow easy portability and therefore is distinct from the solar

home system market.

Technology - The light – typically LED-based, though many products still feature CFL

bulbs – has to be rechargeable.

Quality - Recognizing the emerging issue of market spoilage from poor quality products,

[this] analysis excludes ultra-cheap (typically battery-powered, nonsolar) LED

torches/flashlights ($1-10), which have experienced substantial sales over the past few

years in Africa (Lighting Africa, 2010d).

Although relatively new to the market, high quality OLPs have already proven to

offer substantial benefits to off-grid households in Sub-Saharan Africa. Evan Mills

succinctly touts the advantages of solar powered LED lights: “WLED [White Light

Emitting Diode] technologies provide more and better illumination (with easier optical

Page 28: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

11

control) than do fuels, dramatically reducing operating costs and greenhouse gas

emissions, while increasing the quality and quantity of lighting services” (Mills, 2005).

The light output by LEDs in currently available OLPs ranges between about 20

and 70 lumens, compared to 12 lumens emitted from a typical candle (Lighting Africa,

2010a). Well designed OLPs also deliver higher quality light. Commonly available

LEDs and inexpensive optics are capable of providing a highly uniform light distribution

with color rendering and color temperatures that are superior to most fuel-based lights.

Over the lifetime of an OLP, the cost of lighting service ($/lux-hour) is much less

than fuel-based lights. Mills‟ research indicates that OLPs can be the most cost effective

solution for off-grid applications. Figure 1 shows that the cost of lighting service for a

typical OLP is not only drastically less than fuel-based lights, but is also slightly less than

fluorescent and incandescent lamps in grid-connected homes.

Page 29: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

12

Figure 1: Total cost of illumination services for select off-grid lighting products. Costs include

equipment purchase price amortized over three years, fuel, electricity, wicks, mantles, replacement

lamps and batteries. Assumptions are four hours/day operation over a one year period in each case,

$0.1/kWh electricity price, $0.5/liter fuel price (reproduced from Mills, 2005).

1.3.1 Currently Available OLPs

As the OLP market matures, manufacturers have begun to sell products that cater

to the specific demands of consumers in Sub-Saharan Africa. Improvements in

technology, decreasing component costs and continued product development will surely

yield OLPs of diverse form and function. Nonetheless, familiarity with the current OLPs

is needed in order to understand the context in which this thesis research was conducted.

Page 30: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

13

Lighting Africa‟s experience with products that are currently available in Sub-

Saharan Africa has shown that OLPs fall into the following general categories:

Flashlights/Torches - portable handheld devices offering directional lighting at low

lumen output.

Task lamps/work lights – portable or stationary handheld devices, including desk

lamps, in a range of light output levels utilized for specific tasks (i.e. reading,

weaving etc.).

Ambient lamps /“lanterns” – portable or stationary devices that resemble the kerosene

hurricane lamp form factor. They typically offer multi-directional light along with a

wide variety of size and functionality depending on technology (e.g., from heavy,

powerful CFL lanterns to smaller LED-based systems).

Multi-functional devices – portable or stationary devices that can provide directional

and multi directional light, a variety of value-added features (i.e. mobile phone

recharge), and can be utilized for either task based or ambient lighting needs.

Micro-SHS – semi-portable lighting devices associated with a small portable solar

panel that powers or charges 1-3 small lights, mobile phones, and other low-power

accessories (e.g., radio, mini-fan) (Lighting Africa, 2010d).

A quality matrix is a useful way to view the variety of OLP options on the market.

Data from the Renewable Energy and Energy Efficiency Partnership (REEEP) and

Lawrence Berkley National Laboratory has been compiled to create a quality matrix of

battery life vs. lumen output for 12 different “Solar Portable Lights” (SPLs). Shown in

Figure 2; the SPL Quality Matrix demonstrates the range of performance and

corresponding prices that have been witnessed in the market. While this thesis addresses

the broader category of off-grid lighting (both solar-powered and non-solar-powered

products), the SPL Quality Matrix is indicative of the general state of the off-grid lighting

market as a whole. The plot indicates that even in its infancy, the OLP market has begun

to develop segments according to price and performance.

Page 31: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

14

Figure 2: Solar Portable Light (SPL) Quality Matrix that describes the range of solar charged

portable lighting products according to the autonomous run time (i.e. the amount of time that the

OLP provides useable light on a fully charged battery) and the luminous flux (i.e. the total power of

light emitted by the OLP when fully charged)(reproduced from Lighting Africa, 2010c).

1.3.2 Market Spoilage

Not shown in Figure 2 are the products in the $1-$10 range that are of extremely

poor quality. Most of these cheap products exhibit endemic failures that arise from low

quality components, poor design, and poor craftsmanship. OLPs that quickly break or fail

to function properly are causing market spoilage, wherein “consumers have increasingly

become cautious and have at times chosen to continue using kerosene lamps, the

economic, health and social disadvantages notwithstanding” (Lighting Africa, 2010d).

Page 32: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

15

Market spoilage is a serious concern for the growth of the high quality OLP

market. Component costs are decreasing and performance is improving to the point

where high quality OLPs are becoming physically and economically accessible to low

income households in Sub-Saharan Africa. Yet, the wide availability of cheap, low

quality lights threatens to bias the consumers against OLPs, regardless of quality. When

a modern lighting product rapidly fails, the total cost of illumination service can be much

higher than fuel-based alternatives.

With large numbers of poor quality lighting products available, African

consumers have already developed some bias against OLPs. A study conducted in 2007

indicated that some degree of market spoilage was probably already occurring at that

time (Mills & Jacobson, 2008). The good news is that all sectors of the off-grid lighting

market are not completely spoiled. Many new high quality products with affordable

retail prices are entering the market. OLPs are being manufactured by a range of

companies, from small social entrepreneurs to large, multinationals. Lighting Africa

expects that some extent of consumer education will occur naturally as higher quality

OLPs gain a larger market share. Field studies offer evidence that the willingness to pay

for quality OLPs increases as much as fivefold with experience (Lighting Africa, 2010d).

Page 33: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

16

1.3.3 OLP Market

Currently, the market penetration of OLPs in Africa is relatively low. Recent

studies by the World Bank and Dalberg Global Development Advisors estimates that

market penetration of solar lighting products is currently around 1%, with less than a

0.5% share for solar portable lights (Lighting Africa, 2010d). The OLP market, however,

is still young and rapid growth is expected over the coming years. In a recent interview

with the New York Times, Stewart Craine, co-founder of Barefoot Power, which has sold

solar desk lamps and other clean lighting products to 120,000 households in Africa and

elsewhere, likened the current OLP market to the African mobile phone market. Craine

said that the OLP industry, while worth less than $1 billion now, is about the same size of

the African mobile phone industry in the 1990s. Africa is now the fastest-growing mobile

phone market in the world. This comparison is encouraging for both the rural poor and

companies in the OLP industry. Craine quoted, "We would expect precisely the same

behavior from the microenergy market in the next five or 10 years, and that's what's

going to reach a lot of people, even if we haven't reached a whole lot just yet" (Friedman,

2010).

Recent studies by Lighting Africa and the German Company for International

Cooperation (GTZ) suggest that the OLP market will experience rapid growth over the

coming years. In their 2010 publication titled “What difference can a PicoPV system

make?” GTZ lists several reasons why PicoPV systems (small-scale, solar powered

OLPs) are expected to rapidly replace fuel-based lights in Sub-Saharan Africa:

Page 34: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

17

Pico PV prices are coming down fast.

Pico PV systems are over-the-counter consumer products and don‟t need specific

know-how for installation or O&M. Therefore, distribution has lower transaction

costs than for all other grid or off-grid alternatives.

The welfare gain from electrification at household level is arguably largest after

stepping from flame-based lighting to efficient electric lights.

Consumers do not fear that Pico PV lamps will bar them from future grid roll-out,

as they often do in the case of SHSs (GTZ, 2010).

Lighting Africa‟s analysis in their 2010 SPL market report suggests that the

African market for off-grid renewable lighting will experience exceptional growth.

Based on current growth trends, the market will easily experience 40-50% annual sales

growth, and 5-6 million African households will own OLPs by 2015 (Lighting Africa,

2010d). Lighting Africa projects that “by 2015, SPLs that are of the same cost as

currently available products will be more robust, lighter weight, longer lasting,

environmentally cleaner, and two to three times brighter than today‟s SPLs” (Lighting

Africa, 2010d). More specifically, projections indicate that the manufactured cost of

OLPs will decrease by 40% in the next five years and the consumer payback period will

be in the range of two to eight months (Lighting Africa, 2010d). Additional growth of

the OLP market beyond the conservative projection is expected to come from

technological advancements, entrepreneurial innovation, improved distribution networks

and financing mechanisms.

The OLP market may be further supported by clean investment capital. In 2010

the United Nations Framework Convention on Climate Change (UNFCCC) has included

Page 35: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

18

OLPs in a list of Clean Development Mechanism (CDM) methods that developed

countries can use to earn certified emission reduction (CER) credits (UNFCCC, 2010) .

These CERs can be traded and sold, and used by industrialized countries to meet a part of

their emission reduction targets under the Kyoto Protocol. “The mechanism stimulates

sustainable development and emission reductions, while giving industrialized countries

some flexibility in how they meet their emission reduction limitation targets” (UNFCCC,

2007). Inclusion of OLPs in the list of accepted CDM methods may result in growth of

the OLP market beyond what would occur without any additional incentives.

In Figure 3, Lighting Africa presents three projected scenarios for the growth of

the SPL market over the next five years. Even the most conservative estimate forecasts a

45% growth in SPL sales, which suggests that the off-grid lighting market, as a whole,

will also experience similar expansion.

Page 36: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

19

Figure 3: Solar Portable Light (SPL) market growth scenarios (reproduced from Lighting Africa,

2010c).

Short term exponential growth is expected for the OLP market, but there still exist

several barriers to wide scale use and market penetration. From top to bottom, the OLP

market is hindered by inadequate financial structures. Lighting Africa found that many

manufacturers lack the capital to procure components and produce finished goods before

receiving payment. In the middle, many distributors are stretched thin when they

simultaneously purchase wholesale products and extend credit to dealers. Further down

the line, the consumers also experience financial challenges. Most low-income African

consumers are unable to make lump-sum payments and have limited access to credit. As

a result, good quality OLPs are still out of reach for much of the target population.

Page 37: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

20

Current trade and economic policies are also hindering the success of the OLP market.

Many African countries are collecting tariffs and taxes on OLPs that result in higher retail

prices for the consumers. Although some countries are moving to reduce or remove

taxation of OLPs, others continue to assess customs duties and value added taxes that can

add 10% - 30% to the product price (Lighting Africa, 2010d).

1.4 Lighting Africa – Catalyzing Markets for Modern Lighting

“Lighting Africa, a joint IFC and World Bank program, is helping develop

commercial off-grid lighting markets in Sub-Saharan Africa as part of the World

Bank Group‟s wider efforts to improve access to energy. Lighting Africa is

mobilizing the private sector to build sustainable markets to provide safe, affordable,

and modern off-grid lighting to 2.5 million people in Africa by 2012 and to 250

million people by 2030” (Lighting Africa, 2011).

Lighting Africa‟s approach to supporting development of the off-grid lighting

market is divided into five areas: quality assurance, market intelligence, consumer

education, business support, and policy research.

“Lighting Africa lowers market entry barriers of the off-grid lighting market at

every step, from the design of lighting products, to their commercial production and

distribution. The program works with manufacturers of lighting products, distributors,

consumers, financial institutions and governments to build a lasting market for

reliable, practical and affordable lighting products” (Lighting Africa, 2011).

By addressing off-grid lighting at all levels, from individual components and

system design, to regulation, distribution and consumer awareness, Lighting Africa is

working to build a self-sustaining market that can ultimately improve the lives of the

world‟s population that lack access to suitable lighting service.

Page 38: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

21

1.5 OLP Quality Assurance

This study is intended to be directly applicable to Lighting Africa‟s quality

assurance strategy, which “supports market development, provides technical advisory

services to quality oriented companies, and protects the interests of low-income

consumers” (Lighting Africa, 2011). As seen in other related industries such as PV

panels and compact fluorescent light bulbs, product testing and minimum performance

standards are needed in order to maintain the integrity of the market.

As previously stated, minimal product quality standards in the emerging OLP

market are leading to market spoilage that is detrimental to consumers and to the outlook

for small-scale off-grid lighting products as a whole. Problems associated with poor

quality, mislabeling, counterfeiting and lack of consumer awareness “can be addressed

through the growth of quality testing and certification programs at the national level…

Well funded and heavily promoted region-wide product quality testing solutions will be

necessary to reduce information asymmetries for consumers and improve the quality of

existing products by providing vital feedback to manufacturers” (Lighting Africa, 2010d).

Mills and Jacobson stress the urgency of establishing a quality assurance framework for

OLPs:

“Given the rising popularity of the LED lighting concept for developing countries,

and the impending launch of major deployment programs, there is a specific urgency

to formalize a product quality and performance testing process, and ensure that the

results reach key audiences. The failure to do so will invite market-spoiling problems

that will ultimately inhibit the penetration of good products and the achievement of

significant energy, economic, and environmental benefits. Indeed, this process may

already have begun” (Mills & Jacobson, 2008).

Page 39: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

22

Currently, Lighting Africa is working with OLP market stakeholders to develop a

quality assurance strategy. Whatever the final outcome, performance testing of OLPs

will be a critical piece in any quality assurance program. There are, however, no

internationally accepted test methods or performance standards for OLPs. Standard test

methods exists for some of the individual components (e.g. LEDs, batteries, PV

modules), yet no system level tests or standards have been widely accepted for OLPs

since they are an emerging application of relatively new technologies.

1.5.1 Lighting Africa Quality Test Method (QTM)

Lighting Africa has developed a set of standardized test methods to evaluate the

performance of off-grid lighting products sold in Africa. The primary test method for

product performance verification is the Lighting Africa Quality Test Method (QTM).

“The QTM is designed to be faster and less expensive (in terms of personnel time and test

instrument requirements) than many existing test methods that can be applied to solar and

lighting products” (Lighting Africa, 2011). The QTM is freely available for use by

product manufacturers, government agencies, multi-lateral institutions, bulk-purchasing

agents, non government organizations (NGOs), importers, and others who need to

identify good-quality products or verify compliance with minimum performance levels

(Lighting Africa, 2011).

Lighting Africa is currently supporting OLP manufacturers whose products meet

minimum performance criteria. The QTM is being used to measure key performance

Page 40: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

23

metrics and to verify claims made by manufacturers on specification sheets. OLPs that

have proven to be of high quality gain qualification for business, marketing and product

development benefits provided by Lighting Africa. The QTM also serves as the

foundation for the UNFCC CDM methodology for evaluating the impact of substituting

fuel-based lighting with LED lighting systems. Some African governments have also

shown interest in the QTM as a means of enforcing quality control standards on a

country-wide level.

The QTM, therefore, is being crafted with the African OLP market and

infrastructure in mind. Internal regulation of products that enter a country will require

testing centers that verify compliance to standards. Testing bodies in developing

countries are often poorly funded and operate on a shoestring. As such, quality assurance

test methods that are appropriate for use throughout Africa and the developing world

should not require expensive equipment. Nor should the tests require operators with

highly specialized education that may not be available locally. The QTM is also intended

for OLP manufacturers who would like to conduct in-house performance testing for

quality assurance and research and development. The ideal test method is quick and

affordable to conduct, while delivering test results that are useful for analysis of

component and system level performance.

The QTM originates from a report prepared by the Fraunhofer Institute for Solar

Energy Systems (FISE) titled Stand-Alone LED Lighting Systems Quality Screening,

which was developed to evaluate the performance and quality of LED-based off-grid

Page 41: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

24

luminaires. Existing standards and test methods for OLP components such as LEDs, PV

modules, batteries and charge controllers serve as a reference for the QTM. These

include specifications from the Global Approval Program for Photovoltaics (PVGAP),

the International Electrotechnical Commission (IEC), and the International Commission

on Illumination (CIE). The QTM is undergoing continuous modifications to correct

shortcomings in the procedure and improve the appropriateness in terms of the Lighting

Africa mission. The current version of the QTM requires a total of 15 product samples,

costs about $6,000 per product, and requires approximately four months for completion.

The QTM is comprised of nine tests that are conducted on six different product samples

(n=6). The QTM consists of the following test procedures:

1. Visual screening of reported performance and general workmanship

2. PV module I-V characterization

3. Battery capacity determination

4. Charge controller characterization

a. Deep discharge protection

b. Overcharge protection

5. Autonomous run time determination

6. Lighting service

a. Luminous flux

b. Light distribution

c. Color characterization

7. Charging behavior characterization

a. Solar charging

Page 42: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

25

b. Grid charging

8. Mechanical durability

9. Long-term lumen degradation test (2,000 operational hours)

1.5.2 Lighting Africa Initial Screening Method (ISM)

The Lighting Africa Initial Screening Method (ISM) is a pared down, rapid

version of the QTM intended as a preliminary quality check for OLPs. The ISM is

designed to provide feedback on critical performance criteria in approximately six weeks.

The ISM requires that a single product sample be used for each test (n = 1) and only three

samples are required to test an OLP according to the ISM. The cost for testing a product

according to the ISM is considerably less than the complete QTM. The ISM is useful for

importers, bulk purchasers and government regulators who seek a low-cost test to verify

that a product meets some minimum performance standards. Off-grid lighting

manufacturers that do not have the capacity to conduct advanced research and

development testing in-house can use the ISM for internal performance testing and

quality control. For manufacturers that seek business and technical services from

Lighting Africa, the ISM serves as a low-cost „gateway‟ test to evaluate a product‟s

potential to pass the more rigorous and expensive QTM. The ISM is comprised of the

following tests:

1. Visual screening of reported performance and general workmanship

2. PV module I-V characterization

Page 43: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

26

3. Battery capacity determination

4. Autonomous run time determination

5. Lighting service

a. Light distribution

6. Charging behavior characterization

a. Solar charging

b. Grid charging

7. Mechanical durability

8. Long-term lumen degradation test (500 operational hours)

1.6 OLP Systems

To understand the specific testing methodology addressed in this research, one

must first be familiar with off-grid lighting systems and components. In this section,

OLP systems are broken down into four sub-systems: light source (particularly LEDs),

control circuitry (battery charge/discharge and LED driver), energy conversion (PV

modules), and energy storage (rechargeable batteries). The individual system

components are not complex, but integration into a complete system that is inexpensive,

useful, efficient and durable can be complicated and requires well-educated design

choices. “Ideally, the lighting design process results in a solution that balances the user‟s

needs, the economics and environment” (Freyssinier et al., 2009). A conceptual diagram

of a typical OLP system and sub-systems is shown in Figure 4.

Page 44: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

27

Figure 4: Conceptual system description of a typical off-grid lighting product (reproduced from

Mills & Jacobson, 2008).

Page 45: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

28

1.6.1 Energy Conversion: Photovoltaic Modules

Energy for powering the OLP is often provided by a photovoltaic (PV) module.

Due to the low power draw and high luminous efficacy of LEDs, the daily energy

required to operate most OLPs can be provided by PV modules with peak output less

than 10 watts. Typical OLPs include PV modules that are about 2.5 watts and the size of

a small book (Lighting Africa, 2010d). PV modules made of mono- and polycrystalline

silicon as well as amorphous silicon are commonly used in OLP applications. PV

modules are integrated into the body of the OLP or connected remotely, according to the

form and function of the product.

The largest costs in today‟s SPLs are concentrated in the solar panel, which often

accounts for well over 30% of a typical solar lantern or torch component costs (Lighting

Africa, 2010d). The good news for the future of off-grid power systems is that the cost of

PV modules (on a dollar per peak watt basis) has been rapidly declining. The recent

trend and short term forecast for crystalline and amorphous silicon PV panel prices are

shown in Figure 5. Continued improvement in PV panel efficiency and further cost

reductions are expected in the coming years. As a result, the price of high quality OLPs

will become more affordable for low-income consumers.

Page 46: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

29

Figure 5: Off-grid lighting product solar panel prices are expected to continue rapid cost decline

driven by crystalline PV price declines along with a shift to amorphous thin-film technology

(reproduced from Lighting Africa, 2010c).

1.6.2 Batteries

OLPs rely on rechargeable batteries to store energy from the PV module and to

power lights and other product functions. The batteries commonly used in OLP products

are of four chemistry types: sealed lead acid (SLA), nickel cadmium (NiCd), nickel

metal hydride (NiMH), and lithium ion (Li-ion). Most batteries used in OLPs are widely

available commercially. Each battery chemistry offers a unique combination of

attributes, including cost, physical size and weight, storage capacity, lifecycle, and

toxicity, to name a few. Product designers, therefore, have access to a variety of energy

storage options that can be selected according to the specific system demands. The

availability of these common battery types also allows for replacement by the consumer.

Page 47: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

30

Currently, NiMH batteries account for over half of the batteries found in OLPs, but

Lighting Africa research suggests that Li-ion batteries will gain an 80% market share by

2020 (Lighting Africa, 2010d).

1.6.3 Control Circuitry

Electronic control circuitry is essentially the „brains‟ of the OLP that connects all

of the individual components, forming a functional system. Properly designed control

circuitry efficiently regulates the battery charging and discharging, drives the light source

at appropriate (and often adjustable) levels of current and voltage, allows for other system

functions like mobile phone charging, and protects the OLP components from electrical

damage. Additionally, portable lighting products are subjected to a range of environments

and operating conditions that require the integrated circuitry to be physically robust. In

order to create an OLP that is appropriate for the African off-grid market, the control

circuitry must accomplish all of the aforementioned functions while maintaining low

hardware and manufacturing costs.

The OLP control circuitry is of particular interest to the research conducted in this

thesis since it is directly tied to the lumen maintenance and overall lifetime of a lighting

product. Experience with testing a broad range of high and low quality products has

shown that the circuit design is often the root of poor performance and device failure.

Especially in the lowest cost OLPs, simply designed drive circuitry has been witnessed to

push too much current through the LEDs and ineffectively regulate the battery state of

Page 48: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

31

charge. If the control circuit, itself, does not first experience catastrophic failure, the

battery and LEDs will soon fail to operate properly. Poorly controlled batteries are prone

to drastically reduced storage capacity and over-driven LEDs will rapidly become so dim

that the OLP is essentially unusable.

1.6.4 Light Emitting Diodes (LEDs)

The Illuminating Engineering Society of North America (IESNA) has issued TM-

16: Technical Memorandum on Light Emitting Diode (LED) Sources and Systems, which

is one of the foremost technical references on LEDs. The report‟s general description of

LEDs is useful here as a working definition:

“LEDs are solid-state semiconductor devices that convert electrical energy into

visible light. When certain elements are combined in specific configurations and

electrical current is passed through them, photons (light) and heat are produced. The

heart of LEDs, often called a „die‟ or „chip,‟ is composed of two semiconductor layers

– an n-type layer that provides electrons and a p-type layer that provides holes for the

electrons to fall into. The actual junction of the layers (called the p-n junction) is

where electrons and holes are injected into an active region. When the electrons and

holes recombine, photons (light) are created. The photons are emitted in a narrow

spectrum around the energy band gap of the semiconductor material, corresponding to

visible and near-UV wavelengths” (IESNA, 2005).

The symbol for an LED used in circuit diagrams is shown in Figure 6. When

sufficient current flows across the p-n junction of an LED, visible light and heat is

produced. Figure 7 shows the basic parts of a through-hole type LED.

Page 49: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

32

Figure 6: Electrical model of a light emitting diode (LED) (reproduced from IESNA, 2005).

Figure 7: Components of a typical through-hole type LED package. The epoxy encapsulant, wire

bond, reflective cavity, semiconductor die and leadframe are common to all types of LED packages

(reproduced from Wikipedia, 2009).

The first practical LED, invented in 1962, emitted light in the red portion of the

visible spectrum. Over the next two decades, the technology had developed such that

LEDs could emit other colors of light. The subsequent invention of two semiconductor

Page 50: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

33

materials used in LEDs, Aluminum gallium indium phosphide (AlGaInP) and Indium

gallium nitride (InGaN), finally “enabled LEDs to become a readily available commercial

product” (IESNA, 2005). LEDs made of AlGaInP and InGaN have a much higher light

output than the earlier LEDs. “In addition, these materials allowed, for the first time,

LEDs with peak wavelengths at any part of the visible spectrum to be made” (Bullough,

2003). AlGaInP and InGan LEDs are coated with phosphors that convert the emitted

light into white light, much like the phosphors that are used in fluorescent tubes.

Commonly available LEDs are now capable of generating white light that is both „warm‟,

indicating a yellowish appearance, and „cool‟, which appears bluish in color.

LEDs are also experiencing rapid improvement in luminous efficacy, which is

defined as “the luminous flux (lumens) produced by the system divided by the system

power input (Watts) and is expressed lm/W” (IESNA, 2005). The U.S. Department of

Energy (DOE) has been tracking the luminous efficacy of LEDs, and the trend suggests

that warm and cool LEDs will reach 160 lm/W and 220 lm/W by the year 2020,

respectively (Welsh, 2009). The DOE‟s forecast of LED luminous efficacy is shown in

Figure 8. In terms of luminous efficacy, the outlook for LEDs is quite promising.

Conventional light sources like fluorescent and high intensity discharge (HID) lamps

currently have luminous efficacies slightly above 100 lm/W, with little expected

improvement over the coming decade. Compact fluorescent lamps, which are perhaps a

more direct competitor with LEDs in the short term, generally have luminous efficacies

on the order of 40 – 60 lm/W. With energy efficiency becoming a key design criterion, it

is likely that LEDs will replace conventional light sources in many applications.

Page 51: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

34

Figure 8: Trends and projections for luminous efficacy of commercially available cool and warm

white LEDs. The line labeled as “2008 MYPP Comm Warm White” is the U.S. Department of

Energy‟s multi-year program plan projection for the increase in luminous efficacy of commercially

available, warm white LEDs (reproduced from Welsh, 2009).

From an economic perspective, LEDs still lag behind conventional light sources.

In 2009, the DOE found that on a normalized light output basis, LEDs are more than 430

times the cost of incandescent light bulbs and more than 50 times the cost of a CFL. Yet,

cost and performance trends suggest that over the next several years, LED light sources

are projected to become competitive on a first-cost basis (U.S. Dept. of Energy, 2010).

While LEDs may still be insufficient for several illumination applications, the

current cost, performance and unique characteristics of LEDs have already proven to be

appropriate for use in OLPs. LEDs are solid-state devices, which means that there are no

Page 52: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

35

filaments or moving parts that are prone to mechanical failure. The size and versatility of

LEDs have also proven to be extremely useful in OLP applications. LEDs come in a

range of sizes from tenths of millimeters to packages more than 1mm2. The reflective

cavity and epoxy encapsulant in LED packages can be used to create different light

distribution patterns. As such, LEDs can be used individually or combined in arrays to

provide light in a wide variety of light output, color and spatial distributions. The fact

that LEDs are direct current, low voltage devices allows them to be easily and efficiently

integrated into photovoltaic or other DC systems (Freyssinier et al., 2009).

1.6.5 Lumen Maintenance of LEDs

One of the important benefits of using LEDs in OLPs and other lighting devices is

the potential for service life greater than 50,000 hours, more than nearly any other light

source. Unlike conventional light sources, LEDs do not tend to fail catastrophically.

Instead, they experience an irreversible decrease in light output over time, called lumen

depreciation. The ability of an LED to emit a constant level of light over its operational

life is referred to as lumen maintenance, which is the inverse of lumen depreciation. In

their approved method for measuring lumen maintenance of LED light sources, denoted

as LM-80, the Illuminating Engineering Society (IES) defines lumen maintenance as “the

luminous flux output remaining (typically expressed as a percentage of the maximum

output) at any selected elapsed operating time” (IES, 2008). The Alliance for Solid-State

Illumination Systems and Technologies (ASSIST) recommends that LED life for general

Page 53: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

36

illumination be defined by the time it takes for the light output to reach 70% of its initial

light level, denoted by L70 (Narendran et al., 2007). A 30% decrease in the luminous

flux emitted by an LED has been determined to be close to the threshold at which the

human eye can detect a reduction in light output. An LED product that has reached the

L70 light level can be considered a failed unit even though it still produces light. The

L70 level has become an accepted lifetime level for LED systems and is used by Lighting

Africa as the standard LED lifetime in OLPs (Lighting Africa, 2010b).

The non-reversible decrease of luminous flux during extended use is primarily

caused by heat generated at the LED junction. “LEDs are notoriously more sensitive to

temperature effects than any other light source” (Freyssinier et al., 2009). This is due to

the fact that LEDs do not emit heat as infrared radiation like other light sources. As a

result, the heat must be removed from the device by conduction or convection. Without

adequate heat sinking or ventilation, the LED junction temperature will rise, resulting in

lower light output. While the effects of short-term exposure to high temperatures can be

reversed, continuous high temperature operation will cause permanent reduction in light

output (U.S. Dept. of Energy, 2006).

In addition to insufficient heat sinking, excessive electrical current through the

LED also results in elevated junction temperatures. With increasing power there is

increased thermal load and more heat to dissipate. This phenomenon is referred to as

thermal runaway (Cooper, 2007). Other causes of LED lumen depreciation include poor

Page 54: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

37

LED chip quality, encapsulant degradation, and LED bond wire electric resistance

(Lighting Africa, 2010b).

Testing low quality off-grid lighting products at SERC has shown that many of

the products available to consumers in Africa exhibit rapid lumen depreciation (Mink et

al., 2010). Ineffective heat sinking and overdriving the LEDs are common design flaws

in these devices. Rapid lumen depreciation of off-grid LED lights is a major concern

when attempting to protect the interests of consumers and the integrity of the OLP

market. Products that provide high forward current through the LEDs and poor heat

dissipation can greatly mislead the purchasers. When a product is just off the shelf, the

consumer is impressed by the brightness achieved by overdriven LEDs. Yet, the users

are disappointed shortly thereafter when the light is drastically dimmer, perhaps even

unusable. Ultimately, the consumer loses money and OLP market spoilage is propagated.

1.6.6 Lumen Maintenance Testing

An important criterion for assessing the quality of an off-grid lighting system is

the product lifetime. A well-designed product provides useful lighting service for a

length of time that reduces the operating cost to a level lower than fuel-based alternatives.

Modern off-grid lighting technologies must be competitive on the basis of cost-per-time

of lighting service if they are to be adopted by low-income consumers. Unfortunately,

many products exhibit such rapid lumen depreciation that the potential economic savings

are never achieved.

Page 55: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

38

With this in mind, lumen maintenance testing of OLPs is an important part of the

Lighting Africa quality assurance strategy. The QTM includes an OLP-specific method

for measuring lumen maintenance that is derived from existing standards and

internationally accepted methods. The test method established in LM-80-08 by the IES

serves as the primary methodological reference for determining the lumen maintenance

of LED light sources.

LM-80 specifies that luminous flux is to be measured “in conformance with the

appropriate laboratory method for the LED light source under test.” To comply with

LM-80, the LED must be operated “for at least 6,000 hours with data collection at a

minimum of 1,000 hours”, resulting in a minimum of six measurements over the course

of the test. These measurements are typically conducted with an integrating sphere-

photometer, as described by the CIE technical report on the measurement of luminous

flux (CIE 84-1989). LM-80 states that the LED is to be driven at a constant current that

is specified by the LED manufacturer, and must be regulated to within +/-3% of the rated

value over the testing period. The ambient air temperature shall be maintained at 25oC

+/-2oC and airflow must be minimized in order to reduce the effects of convective heat

transfer away from the LED package. A lumen maintenance test report must include all

of the operating conditions and the percent change in initial luminous flux at each

measurement interval.

LM-80 has proven to be an accurate method for measuring the change in LED

luminous flux over time, but it can be inappropriate for widespread use in OLP quality

Page 56: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

39

assurance testing. Slight alterations to LM-80 have been integrated into the Long-Term

Lumen Degradation Test in Lighting Africa‟s QTM and ISM in order to reduce the time,

expense, and complexity of OLP lifetime testing.

The most significant deviation from LM-80 is the means by which OLP light

output is measured. The QTM and ISM specify that, in addition to an integrating sphere-

photometer system, illuminance measuring devices can be used to measure the relative

change in light output over time (Lighting Africa, 2010c). Since illuminance (lux, or

lm/m2) is derived from luminous flux (lm), relative change in illuminance can serve as a

proxy for the change in luminous flux.

“A measurement of illuminance in a fixed geometry (such as a dark room or

isolated box) is always directly proportional in a linear fashion to the luminous flux of

a lamp. Therefore, fixed-geometry measurements of illuminance can be used in place

of luminous flux measurements for this test, which relies on relative light output to

indicate the change in useful power emitted by a light source over time” (Lighting

Africa, 2010c).

This directly proportional relationship between lux and lumens allows lumen

maintenance measurements to be made with much simpler and inexpensive equipment.

Detailed descriptions of the integrating sphere and the alternative illuminance measuring

devices are in Section 1.8.

Lighting Africa‟s Long-Term Lumen Degradation Test also includes a few other

deviations from the LM-80 method. Firstly, the QTM and ISM state that lumen

maintenance measurements are to be conducted on OLPs over a period of 2,000 and 500

hours, respectively. The abbreviated lumen maintenance testing period is intended to

shorten the overall product testing time, thereby reducing the cost of testing. The

Page 57: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

40

rationale for a 2,000 hour test in the QTM is based on the expected lifetime and daily use

of typical OLPs. The 500 hour test in the ISM is intended to identify products that

experience extremely fast lumen depreciation before proceeding to the QTM.

Another divergence of the QTM from the LM-80 procedure is how the OLP is

powered during the lumen maintenance test. Since OLPs are battery powered, the light

source in a product is driven at different levels according to the battery state of charge

and the particular behavior of the control circuitry. The constant current specified in LM-

80, therefore, does not apply to OLPs. Instead, the QTM and ISM require that the OLP is

driven at the product‟s nominal battery voltage.

Finally, the Lighting Africa Long-Term Lumen Degradation Test requires more

frequent measurement intervals than LM-80. With a relatively short 2,000 hour operating

period, the light output measurements in the QTM must be made at least every week in

order to effectively track the change in luminous flux. The 500 hour lumen maintenance

test in the ISM also requires frequent measurements to record the trend in OLP light

output.

Now that the OLP components have been described and lumen maintenance

testing of LEDs has been outlined, it is important to understand the theory behind light

measuring practices. The following sections delve into the science of photometry and the

methods for measuring light.

Page 58: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

41

1.7 Photometry

Visible light describes the part of the electromagnetic spectrum that can be sensed

by the human eye. From a physiological perspective, light waves are those having a

periodicity of such value that when they are received by the retina of the eye they

produce the sensation of vision in the brain (Barrows, 1912). Humans are able to see

light with wavelengths between 380 nm and 770 nm. The location of the visible

wavelengths within the electromagnetic spectrum is shown in Figure 9. Note that the

human eye perceives light with smaller wavelengths as violet in color and larger

wavelengths are perceived as red.

Figure 9: Electromagnetic spectrum, including the range of visible light wavelengths detectable by

the human eye (reproduced from Irvine, 2010).

Page 59: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

42

Photometry is the science of the measurement of light in terms of its perceived

brightness to the human eye (Bass, 1995). In fact, photometry is the only system of

physical measurement that is based entirely on human perception. The spectral response

of the human eye, however, varies from person to person as well as between well-lit and

low light conditions. Our vision in well-lit conditions, or photopic vision, is most

sensitive to a wavelength of 555 nm (green). Under low light conditions, our perception

of light is called scotopic vision, and our eyes are most sensitive to wavelengths around

498 nm (green-blue). In 1924, the Commission Internationale de l‟Eclairage (CIE)

recorded the spectral response of 52 experienced observers. The data resulted in a

standard luminosity curve, or V-lambda (V) function, commonly called the photopic

response of the standard observer (Labsphere, 2008a). The typical photopic and scotopic

response of the human eye are shown in Figure 10.

Page 60: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

43

Figure 10: V-lambda curve describing the sensitivity of the „standard observer‟ (i.e. typical human

vision) to different wavelengths in the visible spectrum, according to scotopic (low light) and photopic

(well-lit) conditions (reproduced from Electro Optical Industries, Inc., 2010).

1.7.1 Luminous Flux

Luminous flux is a measure of the power of visible light and is expressed in terms

of lumens (lm). Luminous flux () is derived from radiant flux (e), which is the energy

per unit time (watts) that is radiated from a source with wavelengths from 0.01m to

1,000 m. Luminous flux is determined by evaluating radiant flux according to its action

upon the CIE standard photometric observer. For photopic vision, luminous flux is

calculated according to Equation 1.

Page 61: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

44

Equation 1: Luminous flux is calculated by integrating the spectral distribution of a light source‟s

radiant flux over the spectral luminosity function specified by the CIE curve (CIE, 1996).

where:

Luminous flux (lumens).

= Lumens per watt conversion factor for the CIE function (683 lm/W).

= Spectral distribution of the radiant flux (lm/nm).

= Spectral luminous efficiency of the standard observer.

1.7.2 Luminous Intensity

Luminous intensity (I) of a light source is a measure of visible power emitted per

solid angle, expressed in candela, which is equal to lumens per steradian (lm/sr). A

steradian is defined as the solid angle, which, having its vertex at the center of a sphere,

cuts off a spherical surface area equal to the square of the radius of the sphere (Ryer,

1997). A sphere contains 4 steradians.

1.7.3 Illuminance

Illuminance is a measure of the luminous flux incident on a surface per unit area.

Illuminance at a point of a surface (EV) is the quotient of the luminous flux incident

on an element of the surface containing the point, by the area dA of that element (CIE,

Page 62: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

45

1996). Illuminance is measured in units of lux (lx), which is equal to lm/m2

, and is

calculated by Equation 2.

Equation 2: Illuminance is a measure of the total amount of visible light on a surface (CIE, 1996).

where:

Illuminance at a point of a surface (lux).

= Luminous flux incident on an element of the surface containing the point (lumen).

= Area of the element of the surface (m2).

1.7.4 Determining Luminous Flux of a Light Source

According to CIE 84-1989, the internationally recognized standard for measuring

luminous flux, the luminous flux of a light source can be calculated or measured by the

following methods:

Calculation from the luminous intensity distribution

Calculation from the illuminance distribution

Measurement with a sphere photometer by photometric or spectral measurements

Measurement with a box photometer

Relative measurements via illuminance, luminous intensity or luminance (CIE,

1996)

Page 63: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

46

This study follows the sphere photometer (integrating sphere) method to measure

the luminous flux of light sources. The illuminance distribution and box photometer

methods were also referenced in this research to evaluate light distribution patterns of the

tested products and to measure the relative change in product light output, respectively.

1.8 Light Measuring Devices

The research described in this thesis is based on the previously outlined

photometric principles and their application in a laboratory setting. Several light

measuring devices, ranging from simple to complex, were used to quantify the light

output of the OLPs and evaluate the accuracy of the low-cost test methods in question.

The NLTC made available devices such as a spectroradiometer, a goniophotometer, and

an integrating sphere for use in this research. Less complex devices like an illuminance

meter as well as the tube- and box-photometers were also required. The following

sections describe the light measuring devices used in this research.

1.8.1 Spectroradiometer

A spectroradiometer is a device that measures light power as a function of

wavelength. The device consists of a detector coupled to a spectral separation device

such as a diffraction grating monochromator or a Fourier transform interferometer

(Labsphere, 2008b). In simplified terms, a spectroradiometer disperses the light

Page 64: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

47

spectrum, then incrementally scans each wavelength, and measures the intensity of each

narrow wavelength band. The device then weighs and integrates the light power

distribution according to the spectral response of the human eye (V) in order calculate

the luminous flux of a light source.

1.8.2 Illuminance meter

An illuminance meter, often referred to as a lux meter, uses a photocell to convert

incident light energy into electrical current. The amount of current output by the

photocell is directly dependent on the amount of light that strikes the photocell. The

meter then converts the electrical signal to units of lux according to the photopic V()

spectral sensitivity curve established by CIE.

1.8.3 Goniophotometer

A goniophotometer is a device that measures the directional light distribution

characteristics of light sources, luminaires, media and surfaces (CIE, 1996). A

goniophotometer consists of a goniometer and an illuminance meter. A goniometer is an

instrument that rotates an object about a fixed axis in precise angular increments.

In practice, a light source is placed in the center of the device and the goniometer

arm with an attached mirror rotates 360 degrees around the light source in a single plane.

An illuminance meter located at a fixed distance perpendicular from the plane of

Page 65: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

48

goniometer rotation records the illuminance at each angular step of the arm. The

goniophotometer repeats the illuminance measurements on multiple planes. By

combining the illuminance measurements for each plane, a spherical „map‟ of the

source‟s light distribution can be created. The goniophotometer can be calibrated using a

light source of known luminous flux to calculate the luminous flux of the tested lamp.

1.8.4 Integrating Sphere

An integrating sphere, also known as an Ulbricht sphere, is a hollow sphere

whose internal surface is a diffuse reflector that spatially integrates radiant flux

(Labsphere, 2008a). The integrating sphere is a photometric device used for several

common photometric measurements, including radiant and luminous flux, laser power,

reflectance and transmittance of materials. The sphere can also be used as a large area

diffuse light source. This study focuses on using an integrating sphere in combination

with a spectroradiometer to measure the luminous flux of a light source.

An integrating sphere is so named due to the behavior of light within the hollow

cavity. The highly diffuse reflectivity of the sphere‟s interior white coating results in

uniform scattering of light. An ideal interior surface is Lambertian, meaning that it

exhibits 100% reflectance and completely uniform angular spreading of the light energy

on the first bounce. When an ideal Lambertian surface is combined with a spherical

enclosure, the geometry of the sphere ensures that every point within a sphere receives

the same intensity of light as every other part of the sphere at the first bounce

Page 66: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

49

(SphereOptics, 2007). Typical integrating spheres are coated with barium sulfate-based

optical paints with reflectance near 0.8, as recommended by CIE-84.

When a light source is placed in the center of an ideal integrating sphere, the

illuminance on a unit area of the interior surface is equal to the illuminance on all other

unit areas within the sphere. The luminous flux of the light source, therefore, can be

calculated by integrating over the entire sphere surface. By definition, the luminous flux

can be derived from the distribution of illuminance over a closed surface around the light

source using the relation in Equation 3.

Equation 3: The luminous flux of a light source in an ideal integrating sphere is calculated by

integrating the illuminance measured on a discrete surface over the entire surface area of the sphere

(CIE, 1996).

where:

Luminous flux (lm).

A = Area of closed surface surrounding the light source.

E = Illuminance at a point on the closed surface.

In reality, however, an integrating sphere exhibits non-ideal behavior since the

interior coating is not completely Lambertian, the sphere surface contains some irregular

geometry, and the light source itself interacts with the reflection of light off the sphere

walls. As discussed below, the accurate measurement of luminous flux also requires that

Page 67: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

50

two baffles and a second „auxiliary‟ lamp be placed in the sphere. A diagram of the basic

integrating sphere components is shown in Figure 11.

Figure 11: Lamp measurement sphere using a detector mounted directly at the view port

(reproduced from Labsphere, 2008a)

The spectroradiometer labeled as “Detector” in the Figure 11 receives light

through a small, light diffusing viewing port that is set flush with the interior surface of

the sphere. Notice that a baffle is placed between the light source and the viewing port.

This is to prevent the detector from being directly illuminated by the light source and to

prevent the detector from directly viewing a part of the sphere wall that is directly

illuminated. Either of these situations introduces a false response since the luminous flux

of the source is directly proportional to the illumination of the sphere wall (Labsphere,

2008b). The baffle is coated with the same paint as the sphere and it has a size and

placement that only allow light to enter the viewing port after two reflections from the

sphere wall.

Page 68: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

51

With additional surface irregularity introduced by the viewing port and an

additional obstacle created by the baffle, the sphere can no longer determine the luminous

flux of a light source according to Equation 3. The luminous flux of a light source can,

however, be measured in an integrating sphere by a comparison with a luminous flux

standard lamp (CIE, 1996). A standard lamp is a reference light source that produces a

known luminous flux when driven at a specified voltage and current. Standard lamps are

calibrated at national standardizing laboratories according to National Institute of

Standards and Technology (NIST) total spectral radiant flux measurements. These „NIST

traceable‟ standard lamps are necessary for accurate measurement of a test lamp‟s

luminous flux in an integrating sphere. Incandescent lamps are generally used for lamp

standards because of their inherently good stability and convenience in handling. The

total uncertainty in the assigned values of total luminous flux for an incandescent lamp

calibrated at NIST ranges from 1.4% to 1.8% (Labsphere, 2008a).

The luminous flux of a light source is determined by placing the test lamp and the

standard lamp successively at the same location in the integrating sphere (CIE, 1996).

The luminous flux of the test lamp is calculated from a simple ratio of the test lamp

illuminance measured in the sphere to the standard lamp illuminance measured in the

sphere, multiplied by the NIST traceable lumen output of the standard lamp. This

relationship is calculated according to Equation 4.

Page 69: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

52

Equation 4: Since the luminous flux of the standard lamp is known, the luminous flux of the test

lamp can be calculated according to a ratio of measured illuminance multiplied by the lumen output

of the standard lamp (Labsphere, 2008a).

where:

Luminous flux of the test lamp (lm).

N = Luminous flux of standard lamp (lm).

Eind = Indirect illuminance of the luminous flux from the test lamp (lx).

Eind, N = Indirect illuminance of the luminous flux N from the standard lamp (lx).

The calculation of luminous flux using Equation 4 is only valid when the test

lamp has the same size, shape and color as the standard lamp. Physical dissimilarities

between the test and standard lamp result in different „self absorption‟ responses within

the integrating sphere, which causes measurement error. An auxiliary lamp and an

associated baffle, shown in Figure 11, are used to correct for the self absorption of the

standard and test lamps. “The auxiliary lamp remains inside the integrating sphere at all

times. It is usually mounted diametrically opposite the detector port and baffled from

direct view and direct illumination of lamps mounted at the sphere center” (Labsphere,

2008a).

Correction for self absorption is accomplished by placing each unilluminated

lamp in the integrating sphere and operating the auxiliary lamp. The illuminance

measured by the detector is recorded with each lamp in the sphere and the self

absorption-corrected test lamp luminous flux is given by Equation 5.

Page 70: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

53

Equation 5: Self absorption correction for the luminous flux of a light source is the ratio of

measured lux when the auxiliary lamp is illuminating the standard lamp to the lux measured when

the test lamp is in the sphere (Labsphere, 2008b).

where:

Eaux, s = Indirect illuminance measured by the detector for the auxiliary lamp with the

unilluminated standard lamp in the sphere (lx).

Eaux, t = Indirect illuminance measured by the detector for the auxiliary lamp with the

unilluminated test lamp in the sphere (lx).

Finally, if the standard and test lamps have different spatial luminous flux

distributions, the test lamp luminous flux may need to be corrected to account for

dissimilar responses in the integrating sphere. This correction is necessary when the

interior sphere surface exhibits high levels of non-uniformity and when a highly

directional test lamp is compared to a uniformly distributed standard lamp.

Light distribution mismatch is corrected through characterization of the spatial

light distribution of both lamps and measurement of the average sphere reflectance over

the area illuminated by each light source. The luminous flux of the test lamp calculated

in Equation 5 is corrected for light distribution mismatch by multiplying by the following

ratio in Equation 6:

Equation 6: Light distribution mismatch correction is used to adjust the measured luminous flux of

the test lamp when the standard lamp has a different spatial light distribution pattern.

Page 71: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

54

where:

Average reflectance of the sphere surface over area that is encompassed by the

full width half maximum (FWHM) angle of the standard lamp.

Average reflectance of the sphere surface over area that is encompassed by the

full width half maximum (FWHM) angle of the test lamp.

Refer to the Sections 2.2.1 and 2.2.2 for a detailed explanation of the methods used in this

study to determine the terms in Equation 6.

1.8.5 Box-Photometer

Luminous flux is usually measured with an integrating sphere in accordance with

CIE 84. However, if an integrating sphere is not available, a self-made device, referred

to as a box-photometer, can be used for measuring luminous flux (Lighting Africa,

2010c). A box-photometer is an integrating photometer employing an arbitrarily shaped,

hollow box or cavity that can be used to compare the luminous flux of light sources of the

same type (CIE, 1996). An interior view and a three dimensional rendering of a box-

photometer are shown in Figure 12.

Page 72: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

55

Figure 12: Box-photometer (left: Interior view, right: 3-D view) (reproduced from Lighting Africa,

2009).

When comparing a box-photometer to an integrating sphere, one must understand

the difference in behavior of light within the cavity. Unlike an integrating sphere, the

light striking a discrete area on the interior box surface is not necessarily equal to other

areas of the interior surface. This non-uniformity is due to irregularities from imprecise

construction, angular intersections of the box walls and inconsistency in the surface

coating. As a result, a box-photometer is much more sensitive than an integrating sphere

to changes in the location and orientation of the light source within the cavity, as well as

mismatches between the size and light distribution of the standard lamp and device under

test (DuT). A luminous flux measurement with a box-photometer only presents a direct

relationship between the luminous flux of the light source and indirect illuminance at an

arbitrary point at the inside surface of the box, provided that the reference light source

and light source to be measured have the same spatial luminous intensity distribution, the

same spectral distribution, and same dimensions (CIE, 1996). The use of box-

photometers to determine the luminous flux of light sources has been previously

Page 73: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

56

investigated, but “results with respect to accuracy and theoretical correctness seem to

favor the globe photometer [integrating sphere]. Its only advantage appears to be in the

mechanical simplicity of its construction” (Barrows, 1912).

The use of a box-photometer for measuring the luminous flux of a light source

was investigated in depth by Dr. Gerhard Krenzke in 1965. Krenzke determined the

optimal geometry and dimensions of a box-photometer for measurement accuracy in the

device. The optimal measuring configuration for a box-photometer is shown in Figure

13, where:

L = Light source

F = Measuring window

S = Measuring window shade

H = Auxiliary lamp

SH = Auxiliary lamp shade

Page 74: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

57

Figure 13: Optimal measuring configuration for a box-photometer (reproduced from Krenzke,

1965).

Although previous studies have shown that a properly constructed and calibrated

box-photometer can be used to accurately measure the luminous flux of a light source,

box-photometers are not commonly employed in modern photometric laboratories. In

practice, box-photometers can be much more difficult and time consuming to calibrate

than integrating spheres. Modern lighting laboratories generally rely on integrating

spheres to make lumen measurements on account of the ease of use, measurement

accuracy, and minimal impact of spatial irregularities. The box-photometer, however, is

useful and convenient for measuring the relative change in light output of a light source.

Relative light output measurements with the box-photometer do not require DuT-specific

Page 75: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

58

calibration to deliver quick and accurate results. The box-photometer used in this study

was used solely for determining the relative change in light output. The method for

measuring the change in light output with the box-photometer is described in greater

depth in Section 2.3.9.

1.8.6 Tube-Photometer

A tube-photometer is a device used to measure the relative change in illuminance

of a light source over time. The apparatus consists of tube with a cap at one end. The

end cap fits snugly on the tube such that it holds the illuminance meter sensor in a fixed

position and restricts stray light from entering the tube. Light emitted by a DuT is

directed into the open end of the tube and the illuminance incident on the sensor head is

measured by the lux meter. A diagram of the tube-photometer is shown in Figure 14.

Figure 14: Line drawing of a tube-photometer, indicating the basic device components.

Unlike the integrating sphere and box-photometer, the luminous flux of a light

source cannot be determined by the tube-photometer. The tube-photometer is used solely

Page 76: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

59

for evaluating the lumen depreciation characteristics of a light source. By placing the

DuT flush with the open end of the tube, the orientation and distance of the light source

from the sensor head can be replicated for each illuminance reading taken throughout the

lumen maintenance test. As long as the same orientation between the light source and

sensor is maintained for each reading, an accurate measurement of relative change in

luminous flux can be achieved.

Page 77: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

60

2 MATERIALS AND METHODS

An integrating sphere, a box-photometer, and a tube-photometer were used to

conduct parallel measurements over a range of light output levels from four different

OLPs. The test procedure followed in this study, herein referred to as „Simulated Lumen

Maintenance Test,‟ was conducted in order to compare the measurement accuracy of the

tube- and box-photometers for varying magnitudes of light output, spatial light

distribution patterns, and OLP physical characteristics.

2.1 Materials

Equipment for the Simulated Lumen Maintenance Test comprises of four OLPs

that are driven by a benchtop power supply and four light measurement apparatuses: a

goniophotometer, an integrating sphere, a tube-photometer, and a box-photometer. A

complete list describing specifications of the testing devices as well as additional

equipment associated with each device is included in Appendix L. The following

sections discuss the selection and characteristics of materials used in this research.

2.1.1 Devices under Test (DuTs)

Four different OLPs were used as DuTs for testing the box- and tube-

photometers. The lights were selected from several LED products that had been procured

by random market sampling. The lighting devices used in this research were selected

Page 78: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

61

according to specific performance characteristics that represent the diversity of OLPs

currently on the market. The first criterion for choosing which products to use for the

research was long lumen maintenance. Stability of the light source is necessary for

comparison of the test methods. Secondly, the products were selected according to light

distribution characteristics. One product has the form of a lantern and emits relatively

uniform light in 360o. The other three products exhibit varying degrees of directional

light distributions that are less than 180o. Finally, the chosen products represent a range

of form factors and luminous flux. Table 1 gives relevant specifications of the four DuTs

used in the research. Photos of the lighting products are in Appendix A and additional

descriptions of the light distribution patterns are in Section 2.2.2.

Page 79: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

62

Table 1: Specifications of Devices under Test (DuTs) used in the Simulated Lumen Maintenance

Test.

SAMPLE NAME

FORM FACTOR &

LED TYPE

NOMINAL BATTERY VOLTAGE

(V)

APPROX. NOMINAL LED DRIVE CURRENT

(mA)

APPROX. NOMINAL

LUMINOUS FLUX (lm)

Aishwarya Ambient lantern Through-hole array

6 121 62

Firefly Desktop tasklight Through-hole array

3.6 104 25

Kiran Ambient lantern Single Surface mount

3.6 70 15

Solux Flashlight / Ambient Single Surface mount

3.6 288 71

2.1.2 LED Driver

A simple LED driver circuit was created to provide an adjustable and precise

power source for the DuTs. The LED driver is a critical element in the test procedure

since a lamp‟s light output has a strong dependence on the lamp drive parameters;

accurate and stable electronics are vital to a good photometry system (Labsphere, 2008a).

A precision bench top power supply was used to drive the LEDs at constant current. The

positive and negative leads of the power supply were connected to the DuTs by paired

conductor wire for ease of transfer between the light measuring devices. A simple

current shunt circuit, shown schematically in Figure 15, was used to measure the current

Page 80: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

63

flowing through the DuT LEDs. A high precision voltmeter measured the voltage at the

resistor. The current through the resistor is calculated using Ohm‟s law (I = V/R). Since

the DuT and the resistor are connected in series, the current through the resistor is equal

to the current through the DuT LEDs. Specifications of the LED driver circuit elements

and voltmeters are available in Appendix L.

Figure 15: Current shunt circuit used to measure current through the Device under Test (DuT)

LEDs.

2.1.3 Box-Photometer

Much of the following description of the box-photometer used in this study

originates from Section 3.2 of the most recent revision of the QTM available for

download on the Lighting Africa website (Lighting Africa, 2011). Results from this

thesis research as well as the author‟s experience constructing, using and training

technicians to use box-photometer have been integrated into the “Photometer Box”

section of the QTM.

Page 81: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

64

The box-photometer is designed to be simple and economical to produce. These

attributes are essential for light testing centers in locations that do not have access to

expensive photometric equipment or advanced manufacturing facilities. The box-

photometer used in this study is constructed of basic materials that are generally available

in most countries. Due to their availability, relatively low cost and ease of manufacture,

the materials specified in the plans are considered appropriate for building a photometer

box. Functionally similar materials may be used in place of the recommended materials,

according to local price and availability.

The box is built from plywood, but other strong wood or timber materials can also

be used. A thickness of at least 2.5 cm is recommended to ensure clean processing. On

grounds of cost and procurement difficulty, a highly reflective special coating (which is

normally used for integrating spheres) was not used. Instead, the inner surfaces were

painted with white emulsion paint. Multiple coats of paint (5 or more) improve the

uniformity and reflection of the walls. The finished interior surface is intended to reflect

light as diffusely and homogeneously as possible. Care was taken to apply the paint such

that the surface texture is smooth. Since the box-photometer was not to be used for

luminous flux measurements, the auxiliary light and associated baffle were omitted from

the device used in this study. The baffle located in front of the photometer sensor port is

built of 0.5cm-thick plywood and threaded metal rods with a 5 mm diameter. The rods

and baffles were also painted with white emulsion paint. A thermocouple was mounted

to one of the baffle rods to monitor the temperature inside the box-photometer during

testing. An Extech 401036 datalogging photometer sensor measured light in the box

Page 82: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

65

through a port made of PVC pipe. The test stand in the center of the box was custom

made of a machined aluminum base and a threaded metal rod welded to a square piece of

sheet metal with approximate dimensions of 10cm x 10cm. Refer to Appendix L for a

list of the box-photometer components, and Appendix J for detailed plans and

construction directions.

2.1.4 Tube-Photometer

Much of the following description of the tube-photometer used in this study

comes from section 3.3 of the most recent revision of the QTM available for download on

the Lighting Africa website (Lighting Africa, 2011). The author wrote the “Photometer

Tube” section of the QTM according to his experience constructing, using and

researching the tube-photometer.

The tube-photometer is a simple, hand-made device for taking measurements of

relative luminous flux. Like the box-photometer, the tube- photometer also consists of

low-cost materials that are readily available in most countries. A photo of the tube-

photometer used in this study is shown in Figure 16.

Figure 16: Tube-photometer used in the study.

Page 83: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

66

The tube of the photometer used in this research is made of cardboard, and was

procured free of cost from a fabric seller in Beijing, China. PVC pipe is also a relatively

inexpensive material that is appropriate for use as a tube-photometer. The tube used in

this study measures 0.5 m in length, with a 6 cm inside diameter. The end cap fit to one

end of the tube was fashioned from cardboard and packaging tape. Although not

necessary, the end cap can be machined from a more rigid material like wood or plastic.

An Extech 401036 datalogging illuminance meter was used to measure light incident on

the end of the tube. No coating was applied to the internal surface of the photometer

tube. During measurements, dark, opaque fabric was placed over the DuT and the open

end of the tube to ensure that no stray light was measured by the light meter. Due to the

availability, extremely low cost and ease of manufacture, the specified materials are

appropriate for a tube-photometer. Functionally similar materials may be used in place of

the recommended materials, according to local price and availability.

2.1.5 Integrating Sphere – Spectroradiometer System

An integrating sphere with a 1.3-meter diameter was used in this study to measure

the luminous flux of the light sources. The integrating sphere diameter is large enough to

provide sufficient distance between the largest DuT and the sphere wall to permit

adequate multiple reflections of the light within the sphere without undue interference

from the source itself. A high precision spectroradiometer, manufactured by the Beijing

Optical Instrument Factory, was used to measure the light incident on the integrating

Page 84: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

67

sphere port. The spectroradiometer has an optical bandwidth of wavelengths from 350 –

780 nanometers, with light intensity measured at 5 nm intervals. The National Lighting

Test Center (NLTC) of China created custom computer software, aptly named Electrical

Light Source Spectral Test and Color Analysis System, which performs sphere

calibrations, compiles the light meter data, applies self-absorption corrections, displays

the measured light intensity over the visible spectrum, and calculates the luminous flux of

the light source. An example of the software output for a single luminous flux

measurement of the Firefly product is shown in Figure 17.

Figure 17: Example of the software output for a measurement of the Firefly luminous flux. Of

importance to this study is the lumen value after being corrected for self-absorption. In this

example, the integrating sphere – spectroradiometer system measured a luminous flux of 24.778 lm.

Page 85: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

68

As previously described in Section 1.8.4, the system includes a standard and

auxiliary lamp, as well as baffles for the measurement port and the auxiliary lamp. The

standard lamp was a 15.4 watt, NIST traceable incandescent bulb from OSRAM and the

auxiliary lamp was an OSRAM 10 watt halogen bulb. A thermocouple was fixed inside

of the sphere to monitor the ambient air temperature during testing. The procedure for

measurement of luminous flux specified in CIE 84 was strictly followed for all lumen

measurements conducted in this research.

2.2 Integrating Sphere Calibration

Each integrating sphere has a specific and unique throughput. The throughput of

the sphere is affected by the detector that is placed at a port in the sphere. Because each

sphere and detector combination is unique, the sphere and detector are calibrated as a unit

(Labsphere, 2008a). The integrating sphere – spectroradiometer system used in this

research was calibrated according to CIE 84, which has been previously described in

detail in the Section 1.8.4. Following are the methods used to calibrate the integrating

sphere system to account for non-uniformity of the sphere surface, light distribution and

body color mismatch between the standard and test lamps, and light source self

absorption.

Page 86: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

69

2.2.1 Measuring Non-uniform Reflectance of Interior Sphere Surface

The interior surface of integrating spheres, especially older models or those that

have not been recently painted, can exhibit non-uniform reflectance. This can result in

increased measurement error if the standard lamp has a different spatial light distribution

than the DuT. Error due to non-uniform reflectance and mismatched light distribution

can be corrected by a technique that requires characterization of the sphere‟s interior

surface reflectivity and the light distribution of the standard lamp and DuT.

A „map‟ of the sphere‟s interior surface reflectivity is composed of 666 points that

are spaced at 10-degree intervals in the horizontal and vertical planes. A directional light

source is located in the center of the sphere and directed at each of the 666 points. The

illuminance measured by the photometer is recorded at each light orientation.

Accurate rotation of the light source is achieved by a device that is mounted in the

top port of the sphere. The device includes protractors that allow the operator to rotate

the light source at known vertical and horizontal angles. Figure 18 shows the device that

was used to characterize the interior surface reflectivity of the integrating sphere used in

this study. A „reflectance map‟ of the integrating sphere is shown in Table 10 of

Appendix B. Measurement of the sphere‟s interior surface reflectance shows deviations

from the average of up to 7%, which may be large enough to affect the accuracy of

luminous flux measurements made in the sphere.

Page 87: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

70

Figure 18: Device for measuring the reflectance characteristics of the integrating sphere‟s interior

surface.

2.2.2 Correcting for Light Distribution Mismatch

The light distribution spatial pattern of the standard lamp is not the same as that of

the lighting products used in this study. A difference between standard lamp and DuT

light distribution patterns can be a source of error when the interior surface of the

integrating sphere exhibits non-uniform reflectivity. The integrating sphere used in this

study was expected to have a slightly irregular surface reflectance that could distort the

luminous flux measurements due to the dissimilar and directional light distributions. As

a result, the measurements made in the integrating sphere must be corrected for light

distribution mismatch.

Page 88: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

71

The first step in making this correction is to characterize the light distribution of

the standard lamp and all of the DuTs. A goniophotometer was used to measure the

illuminance on a plane that is parallel to the direction of light emitted by the DuT. Radial

plots of light distribution for all of the lights are shown in Figure 42 through Figure 46 in

Appendix B. All of the lamps have extremely symmetrical light distribution patterns

about the vertical (0o – 180

o ) axis. The assumption, therefore, is that the light

distribution on the single plane can be rotated around the vertical axis to create a three

dimensional shape that characterizes the complete spatial light distribution of the source.

Next, the full width half maximum (FWHM) angle is determined for each light

source. The FWHM angle is a way of expressing how wide of a beam is produced by the

light source. FWHM is the angle in which the light output is at least 50% of the

maximum measured on a plane. For example, the Firefly‟s maximum illuminance

measured by the goniophotometer was 1,296 lux. Half of this value (648 lux) occurs at

about 39o and 318

o, which results in a FWHM angle of 81

o. The FWHM angles for

Solux, Kiran, and the standard lamp are, 81o, and 141

o, and 309

o, respectively. As a

lantern, the Aishwarya‟s light distribution differs from the other, more directional

sources. Aishwarya emits light 360o around the lantern, with a FWHM angle of 99

o.

Computer renderings of the nearly omnidirectional light distribution of the standard lamp,

the simplified cone shaped light distribution from the directional light sources and the

inverse hourglass shape of the Aishwarya lantern are shown in Figure 19.

Page 89: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

72

Figure 19: Computer rendering of the simplified three dimensional light distribution shapes for the

standard lamp, top left; directional light sources (Solux, Firefly, and Kiran), top right; Aishwarya

lantern, bottom.

Now that the light distribution for each source and the internal reflectance of the

integrating sphere is determined, the interaction between light and sphere can be

characterized. Here, one must find the average measure of sphere reflectance that occurs

within the FWHM angle of the light source. For ease of calculation, the light sources

(excluding Aishwarya) are placed in the center of the sphere with the light directed

towards the bottom of the sphere (sphere coordinates of 0o horizontal, 0

o vertical). Using

the Firefly as an example, the FWHM angle (approx. 80o) encompasses points on the

sphere‟s interior surface that are between the vertical angles of 0o and 40

o and horizontal

angles from 0o to 360

o. The Aishwarya was also located in the center of the sphere, yet

the FWHM angle spans vertically from 40o to 140

o, and horizontally from 0

o to 360

o.

The reflectance values encompassed by each FWHM angle over this region are averaged.

Page 90: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

73

The average reflectance values over the FWHM angle of each product are then

compared to the average reflectance over the FWHM angle of the standard lamp. A ratio

of the standard lamp FWHM reflectance to the DuT FWHM angle reflectance is

calculated according to Equation 6, resulting in a correction factor for the light

distribution mismatch of each DuT. The luminous flux measured by the integrating

sphere-spectroradiometer for each sample is multiplied by the correction factor indicated

in Table 2 in order to correct for the light distribution mismatch.

Table 2: Summary of the light distribution mismatch correction for the DuTs, calculated according

to the full width half maximum (FWHM) angle.

SAMPLE NAME FWHM ANGLE

(DEGREES)

FWHM AVERAGE ILLUMINANCE (lx)

CORRECTION FACTOR

Aishwarya 99 2644.1 0.989

Firefly 84 2528.4 1.035

Kiran 141 2550.1 1.026

Solux 81 2528.4 1.035

Standard Lamp 312 2616.9 -

2.2.3 Correcting for Self Absorption

Since the light sources used in this research are of different size and shape than

the standard lamp, luminous flux measurements made in the integrating sphere require a

self absorption correction specific to each product. Self absorption of the standard lamp

and DuT in the integrating sphere is corrected according to the process described in

Section 1.8.2 and calculated according to Equation 5. A new self absorption correction

was calculated and applied to each round of Simulated Lumen Maintenance testing. The

Page 91: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

74

custom computer software that accompanies the integrating sphere-spectroradiometer

system at NLTC includes a function that performs the self absorption correction based on

measurements made while illuminating the auxiliary lamp.

2.2.4 Color Mismatch

Luminous flux measurements made using the spectroradiometer are sensitive to

differences in color between the physical body of a test lamp and the standard lamp. It is

possible that the various colors of the DuTs can introduce error in the spectral intensity

measurements that are used to calculate luminous flux. Prior to conducting luminous flux

measurements with the integrating sphere that were used in this study, it was necessary to

understand the extent to which color mismatch affects the measurements for each

product.

An object appears to be of a certain color because its surface reflects visible light

over a range of wavelengths that compose that particular color. All other visible

wavelengths incident on the surface are absorbed by the object. For example, a green

object appears as such since it reflects light with an approximate wavelength of 510 nm,

and absorbs the other visible wavelengths. A body placed in the integrating sphere that is

of a color other than white will absorb some wavelengths and reflect others. When a

spectroradiometer is used to measure light in the sphere, the selective reflection and

absorption of a body within the sphere can result in measurement error.

Page 92: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

75

The auxiliary lamp was used to determine the affect of body color on the self

absorption correction for each light source. First, the unilluminated DuT was placed in

the center of the sphere and multiple measurements of the luminous flux from the

auxiliary lamp were made. The DuT was then covered with white paper and the

luminous flux measurements were repeated. Photos of the covered and uncovered Solux

light in the integrating sphere are included in Figure 20. The difference in the average

lumen values between the covered and uncovered product were compared to ascertain

whether excessive error was caused by body color. Results shown in Table 3 indicate

that the DuT body colors resulted in small changes in the luminous flux measured from

the auxiliary lamp. A difference of less than 1% for all of the products is assumed to

have an insignificant effect on the self absorption correction factor that is applied to the

luminous flux measurements. The error associated with color mismatch, therefore, has

been considered negligible in this study.

Page 93: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

76

Figure 20: Measurements of self absorption color dependency were conducted to verify that the

various body colors of the light sources did not introduce undue error in the determination of

luminous flux. Self absorption of the Devices under Test (DuTs) was measured with the original

colored surface (left) and with a completely white surface (right).

Table 3: Auxiliary lamp luminous flux measurements for all of the light sources indicate that the

error introduced by body color are small enough to be assumed negligible in the study.

SAMPLE NAME AVERAGE LUMINOUS FLUX (lm) DIFFERENCE

UNCOVERED COVERED

Aishwarya 62.05 62.31 0.42%

Firefly 25.84 25.89 0.21%

Kiran 13.90 13.81 0.65%

Solux 71.53 71.78 0.35%

2.3 Simulated Lumen Maintenance Test

Determining the lumen maintenance of a light source according to LM-80 and

Lighting Africa‟s QTM requires that the DuT operate for 6,000 hours (250 days) and

2,000 hours (83 days), respectively. The time required to follow these procedures is

Page 94: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

77

beyond the scope of this study. More importantly, the intent of this research is to

evaluate the accuracy of the testing devices and methods rather than the lumen

maintenance of the light sources. With these factors in mind, a „Simulated Lumen

Maintenance Test‟ was conducted in order to determine if the tube- and box-photometers

are acceptable substitutes for an integrating sphere for use in actual lumen maintenance

testing.

Like normal lumen maintenance testing, the Simulated Lumen Maintenance Test

uses a power supply to drive the LEDs at a constant current. The simulated test differs

from actual lumen maintenance testing in that the decrease in light output is not caused

by the modes of irreversible degradation previously described in the Section 1.6.2.

Instead, decreases in the DuT light output are achieved by reducing the current through

the LEDs at discrete steps. Each level of current through the LEDs is held constant over

the time required to make light output measurements with each of the three testing

devices.

Ideally, the proportional change in measured light output between each level of

LED drive current should be the same across the integrating sphere, box- and tube-

photometers. Assuming that the luminous flux measured with the integrating sphere is

highly accurate, the illuminance measurements made with the alternative devices can be

compared to the „actual‟ values. By comparing the relative change in light output that is

measured in the low-cost devices to the corresponding measurements made in the

integrating sphere, one can evaluate the accuracy of the alternative low-cost methods. If

Page 95: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

78

measurements made in an alternative light measuring device exhibit a highly linear

correlation to those made in the integrating sphere, the alternative device is considered an

acceptable substitute for use in lumen maintenance testing.

2.3.1 Experimental Design

Simply stated, the Simulated Lumen Maintenance Test is a series light output

measurements made at various levels of light intensity. For a single round of testing, one

light source is driven at ten different levels of electrical current, resulting in ten different

light output levels. At each level, the light output is measured once in the sphere, once

with the tube, and twice in the box. The two box-photometer readings are made for two

different DuT orientations within the box. For example, in the first round of Firefly

testing, the product was oriented such that the light was directed towards the back wall of

the box for one measurement and directed towards the lid of the box for the second

measurement. Data for different DuT orientations in the box, especially for highly

directional light sources, are important for investigating the potential for measurement

error caused by rectilinear geometry and irregularities of the interior surface. Three

rounds of Simulated Lumen Maintenance testing are conducted for each of the four

OLPs. A breakdown of the experimental design is shown in Table 4.

Page 96: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

79

Table 4: The experiment consists of three rounds of testing per product. Each round includes

measurements at ten different light levels made once in the sphere, twice in the box at two different

orientations, and once with the tube.

Each round of testing follows the same procedure that consists of the following steps:

1. Drive the LEDs at a constant current until stability has been reached.

2. Verify that ambient temperature is within acceptable bounds.

3. Specify the test lamp orientations in the box-photometer.

4. Determine the apparatus testing order.

5. Measure the light output using each apparatus.

6. Decrease the LED drive current.

7. Repeat.

Detailed descriptions of each step are given below in Sections 2.3.2 through 2.3.10.

Page 97: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

80

2.3.2 Driving the DuT

In order to improve the stability of light output during testing, the batteries and

internal drive circuitry in the DuT were bypassed. The LED array was isolated from the

drive circuitry and an external power supply was connected directly to the LED(s) in each

sample and set to deliver constant current. An adjustment to the LED drive current is

followed by a stabilization period. According to CIE-84, stability is reached when the

variation (maximum – minimum) of at least 3 readings of the light output and electrical

power over a period of 30 min, taken 15 minutes apart, is less than 0.5% (CIE, 1996). A

voltmeter with a precision of 0.001 mV was used to measure the electrical power

delivered to the DuT and light output measurements were conducted with the integrating

sphere system.

2.3.3 Determining the LED Nominal Drive Current

At the start of each round of testing, the LEDs are driven at approximately the

same current that is achieved when the DuT is powered by its battery at nominal voltage.

To determine the current through the LEDs, the battery is removed from the circuit and

replaced by an external power supply. The power supply is set to provide the battery‟s

nominal voltage (refer to Table 1). The DuT is driven at the nominal voltage until a

steady state condition is reached according to CIE-84, as described above. A digital

multimeter and a current shunt circuit, as shown in Figure 15, were used to measure the

Page 98: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

81

voltage and current through the LED array. This current measurement is the approximate

initial condition for each round of the Simulated Lumen Maintenance Test.

2.3.4 Adjusting the LED Drive Current

A single round of testing for one DuT consists of 40 light output measurements.

The light output is measured at ten different levels using each of the three testing devices.

One measurement is made at each light level with the integrating sphere and tube-

photometer. Two measurements at each light level are made with the box-photometer for

two different orientations of the DuT within the box. The light output at each level is

adjusted by changing the current delivered to the LEDs from the power supply. After

each current adjustment, the DuT is allowed to reach a steady state, which is determined

according to CIE-84. Upon reaching equilibrium, the light output is measured in the box,

tube and sphere. After performing measurements in each of the three testing apparatuses,

the LED drive current is reduced by 5% to 15% and the cycle of stabilization and

measurement is repeated.

Preliminary testing showed that the four products used in this research exhibit

constant light output when the specified stability criteria were met. In fact, for the small

changes in drive current used in this study, the LED light output stabilized more quickly

than the 30 minutes required by CIE-84. Yet, since the stability of light output at each

level is critical for comparison of results across the three testing devices, the stability

criteria established in CIE-84 were strictly followed.

Page 99: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

82

2.3.5 DuT Orientation

The placement and orientation of the light sources within the box and sphere and,

in some cases, the position of the tube were documented and photographed to improve

repeatability. For products with broad light distribution patterns, the operator must select

a side or region of the light source that contacts the end of the tube. Products with

directional light sources are shined directly into the tube.

2.3.6 DuT Testing Order

Ideally, the tube, box, and sphere measurements would be made at exactly the

same time to remove the effect of any temporal changes in the DuT light output. Since

the DuT cannot be in all three devices at the same time, the sampling order had to be

randomly assigned. A random number generator was used to determine the order of

testing apparatuses used at each light level. This randomization method effectively

removed bias that may result from the order of testing methods. Even so, a detectable

change in light output over the time to test the DuT in all three devices is assumed

negligible since the stability criteria are met for each light level. To further assure that

the light output did not change during the four measurements, an effort was made to

minimize the time to transfer the DuTs between testing devices.

Page 100: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

83

2.3.7 Ambient Temperature Regulation

As emphasized in Section 1.6.1, LED light output is highly dependent on

temperature. A constant temperature in and around the testing devices is necessary for

maintaining steady light output from the LEDs. Throughout all of the light output

measurements, the room temperature and temperatures within each testing apparatus

were monitored and held within 2oC of 25

oC, as specified by CIE-84.

2.3.8 Integrating Sphere Procedure

The integrating sphere-spectroradiometer system is the most complex of the three

testing apparatuses. The procedure for operating, calibrating, and measuring luminous

flux of the light sources, therefore, is much more involved than the box- and tube-

photometers. The basic steps for determining the luminous flux of the DuTs are:

1. Calibrate the integrating sphere system using the standard lamp.

2. Determine the self absorption correction using the auxiliary lamp.

3. Measure the test lamp luminous flux.

4. Adjust the measured lumen value according to the self absorption correction and

light distribution mismatch correction factors.

A more detailed, step-by-step description of how the integrating sphere system was

operated in this research is included in Appendix D.

Page 101: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

84

2.3.9 Box-Photometer Procedure

One of the obvious benefits of the box-photometer is its simplicity of use.

Compared to the integrating sphere, the procedure for measuring light output in the box is

quick and uncomplicated. This is primarily due to the fact that absolute measurements

of luminous flux are not conducted with the box. Since no calibration or correction is

required, the DuT is simply placed in the center of the box and the illuminance measured

by the lux meter is recorded. Below is a detailed description of the procedure followed

for all measurements made with the box-photometer in this study.

1. Connect the test lamp to a power supply and current shunt circuit, using paired,

white insulated wire.

2. Adjust the power supply to the specified current for DuT.

3. Allow the lamp to reach steady state according to CIE-84.

4. Place the DuT on the test stand in the box-photometer. Care is taken to place the

DuT such that its location and orientation can be replicated at each light output level

in a particular round of testing.

5. Route the paired wire through a small notch in the lid-wall interface of the box.

Care is taken to keep the wire at a consistent length and orientation within the box

for all measurements in a particular round of testing.

6. Record the illuminance value measured by the box-photometer‟s lux meter. Adjust

the range of the meter to achieve maximum precision.

Page 102: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

85

2.3.10 Tube-Photometer Procedure

The tube-photometer is also advantageous for use in lumen maintenance testing

due to its inherent simplicity. The method for measuring lumen maintenance with the

tube is much faster and easier than the integrating sphere, yet slightly more complicated

than the box-photometer method.

As put forth in Section 1.6.3, photometers like the box and tube can be used in

place of an integrating sphere to measure lumen maintenance since they maintain a fixed

geometry between the light source and the photometer sensor. The tube element of the

tube-photometer effectively keeps a constant distance between the DuT and the sensor.

The design of the tube-photometer used in this study, however, does not fix the

orientation of the light source.

To account for this, the “Max. Hold” function that is built into the Extech light

meter is used. When the “Max” button on the light meter is depressed, the meter displays

only the maximum illuminance reading from the sensor. The DuT is placed flush against

the open end of the tube and the light is directed into the tube. Slight adjustments to the

angle between the light source and the tube are made, searching for the maximum

illuminance measured by the light meter. Searching continues until the maximum

illuminance reading displayed on the light meter no longer increases for any orientation

of the DuT. For highly directional light sources like the small task light and flashlight

used in this study, all of the light can be shined into the open end of the tube. More

broadly distributed light sources like the Kiran and Aishwarya products used in this study

require that the operator select a specific region of the light source in which to search for

Page 103: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

86

the maximum illuminance. Below is a detailed description of the measurement procedure

followed for all measurements made with the tube-photometer in this study.

1. Connect the test lamp to a power supply and current shunt circuit. Use paired,

insulated wire.

2. Adjust the power supply to the specified current for test lamp.

3. Allow the lamp to reach steady state according to CIE-84.

4. Hold the DuT flush to the open end of the tube with the light directed into the tube.

5. Cover the light source and tube opening with an opaque cloth to inhibit stray light

from entering the tube.

6. Search for the maximum illuminance using the light meter‟s “Max Hold” function.

7. Record the maximum illuminance value measured by the tube-photometer‟s lux

meter. Adjust the range of the meter to achieve maximum precision.

Page 104: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

87

3 RESULTS

Illuminance data acquired from the box- and tube-photometers for all rounds of

testing all four light sources showed exceptional correlation to the luminous flux data

from the integrating sphere-spectroradiometer system. Graphical and statistical analysis

of the test results is used to evaluate the deviation of the alternative measurements from

the ideal values and to draw conclusions about the use of the box- and tube-photometers

for determining the lumen maintenance of OLPs. Relationships between the low-cost

device measurements and the integrating sphere measurements for each light source are

represented in two different formats: „calibration plots‟ and in terms of relative change in

light output.

3.1 Calibration Plots

Calibration plots show the alternative (box and tube) measurements on the x axis

and the corresponding sphere measurements on the y axis. This comparison is useful for

assessing the proportionality between alternative and sphere measurements. Calibration

plots also provide information for statistical analysis of the test results across the different

light sources. An example of a calibration plot is shown in Figure 21 for a single round

of measurements of the Firefly using the tube-photometer.

Page 105: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

88

Figure 21: Tube calibration plot for one round of measurements with the Firefly. The linear

equation that best describes the data has a very good fit, with an R2 value of 0.9995.

For the ideal case where no variability exists in the measurement methods, the

illuminance measured by the box- and tube-photometers for a particular DuT can be

multiplied by an experimentally derived factor, with the product equal to the integrating

sphere measurement. The ideal calibration plot is perfectly linear, with a slope that

represents a „calibration factor‟. This simple, linear relationship between ideal measuring

devices is shown in Equation 7. Note that the equation has a y-intercept of zero, which is

due to both ideal methods being „zeroed out‟ when measuring a DuT that is not emitting

light.

y = 0.8572x - 0.0662R² = 0.9995

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

0 5 10 15 20 25 30 35

Inte

grat

ing

Sph

ere

me

asu

rem

en

t (L

um

en

)

Tube-photometer measurement (Lux)

Firefly - Tube Calibration Plot (Round 2)

Page 106: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

89

Equation 7: Calibration equation describing the ideal relationship between illuminance

measurements made with the alternative methods and luminous flux measurements made with the

integrating sphere.

where:

= Luminous flux of a light source measured using the integrating sphere-

photoradiometer (lm).

= Illuminance measured by an alternative photometer, such as the tube- or box

photometers used in this study (lx).

C = Calibration factor for a specific light source. When the light source is enclosed by a

cavity like the box-photometer, the calibration factor is applicable to a unique, fixed

orientation of the DuT within the cavity.

The illuminance values measured in the tube- and box- photometers form very

linear calibration plots and the y-intercepts are very near zero. In fact, most of the

calibration plot linear models have y-intercepts that are statistically indistinguishable

from zero. Further discussion of the linear regression confidence intervals is presented in

Section 3.6.2. All of the measurements do contain some degree of error that deviates

from the ideal behavior described in Equation 7. For example, the series of

measurements plotted in Figure 21 have a slightly non-ideal relationship to the

integrating sphere measurements, indicated by an R2 value of 0.995. The y-intercept of

the best fit line is non-zero, but relatively small, at -0.0662. In this particular case, the

origin is encompassed by the 95% confidence interval, indicating that the y-intercept of

the calibration plot follows Equation 7. Calibration plots for all rounds of Simulated

Lumen Maintenance testing are discussed in greater detail in Section 3.3.

Page 107: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

90

3.2 Relative Change in Initial Light Output

The second format for displaying data from each round of the Simulated Lumen

Maintenance Test is a plot of the relative change in light output. This means of showing

the data is similar to that specified by LM-80 and Lighting Africa‟s QTM. Each

measurement made using the alternative method and the integrating sphere is plotted as a

percentage of the initial light output measurement. The x axis shows the „measurement

number‟, which is the sequence that the measurements of decreasing light output were

conducted. The y axis values are expressed as percentage of the initial light

measurement, calculated by:

Equation 8: Percent of the initial luminous flux for the light output of a product measured in the

integrating sphere.

Equation 9: Percent of the initial illuminance for the light output of a product measured in the box-

and tube-photometers.

where:

= Percent of initial light output for measurement number x (%).

= Initial illuminance for a DuT measured in the integrating sphere (lm).

= Luminous flux of a DuT measured in the integrating sphere for measurement

number x (lm).

= Initial illuminance for a DuT measured by the box- or tube-photometer (lx).

= Illuminance of a DuT measured in the integrating sphere for measurement number x

(lx).

Page 108: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

91

Ideally, the percent of initial light output measured at each level in the box- and

tube-photometers would be exactly the same as the integrating sphere. An example of a

plot of relative change in light output for the Firefly using the tube-photometer and

integrating sphere is shown in Figure 22. Notice that in this example, the relative change

in light output measured with the tube very closely follows the sphere.

Figure 22: Relative change in initial light output for the Firefly, as measured by the integrating

sphere and the tube-photometer.

A major consideration for plotting the test data in these relative terms is that all

data points are dependent on the initial light output measurement. This is of concern

because any error in the initial measurement is propagated through all subsequent relative

light output calculations. As a result, the accuracy of the alternative measurements may

appear to be worse than they actually are. The importance of the initial light output

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Firefly - Round 2

Sphere

Tube

Page 109: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

92

measurement in actual lumen maintenance testing and the potential for propagation of

error throughout the entire lifetime test are addressed in greater depth in Sections 3.6.3

and 5.1.

3.3 Box-Photometer Calibration Plots

Previous box-photometer studies indicate that measurement accuracy can be

affected by the orientation and directionality of light sources within the box. Each round

of this experiment, therefore, includes two different orientations of the DuT within the

box. The two highly directional light sources used in testing, the Firefly and Solux, were

placed in the box such that their light was directed towards all four walls and the box lid.

The Kiran and Aishwarya are lantern-like in form and were placed in the box at different

rotational angles in the horizontal plane. The rotational orientations of the Kiran and

Aishwarya for each round of testing are specified by the direction that the integrating PV

module and on/off switch are facing, respectively.

Calibration plots for the box-photometer are shown in Figures 23 through 26,

according to the light source and the specific orientation within the box. Full page

calibration plots are also included in Appendix H. At first glance, one can see that the

data points for the Kiran and Aishwarya are much more tightly grouped than the Firefly

and Solux plots. This is attributed to the directionality of the lights. For DuTs that are

highly directional, changes in testing orientation result in changes to the calibration

factor. Measurements of light sources like the Aishwary and Kiran that have more

Page 110: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

93

omnidirectional light distribution are less sensitive to changes in DuT orientation, which

is indicated by calibration factors that are nearly the same for all orientations.

The calibration plots also show that the relationship between the box

measurements and the sphere measurements for all of the DuTs and all of the orientations

are highly linear. Some error is apparent in the box-photometer measurements, indicated

by R2 values that are less than 1 and y-intercepts that are non-zero. None of the testing

rounds have a completely ideal relationship between box and sphere measurements, but

all of the data sets exhibit R2 values that are very near 1 and y-intercepts that are near the

origin. Detailed statistical analysis of the y-intercepts, linearity, and error in the box-

photometer measurements is conducted in Section 3.6.

Page 111: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

94

Figure 23: Box-photometer calibration plot for the Firefly

Figure 24: Box-photometer calibration plot for the Kiran

y = 0.653x - 0.3847R² = 0.9994

y = 0.5732x + 0.2106R² = 0.9836

y = 0.6331x - 0.0007R² = 0.9993

y = 0.5322x - 0.0575R² = 0.9994

y = 0.5788x - 2.0243R² = 0.9805

y = 0.5277x - 0.1712R² = 0.9836

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

0 5 10 15 20 25 30 35 40 45 50 55

Sph

ere

me

asu

rem

en

t (L

um

en

)

Box measurement (Lux)

Firefly - Box Calibration Plot

Back Wall

Lid

Right Wall

Baffle

Lid

Front Wall

y = 0.6562x + 0.4786R² = 0.9993

y = 0.6583x + 0.4211R² = 0.9993

y = 0.6518x + 0.0585R² = 0.9994

y = 0.6485x + 0.0693R² = 0.9994

y = 0.6417x + 0.3793R² = 0.9983

y = 0.6428x + 0.416R² = 0.9987

0

2

4

6

8

10

12

14

16

0 2 4 6 8 10 12 14 16 18 20 22 24

Sph

ere

me

asu

rem

en

t (L

um

en

)

Box measurement (Lux)

Kiran - Box Calibration Plot

PV Forward

PV Back

PV Right

PV Left

PV Front

PV Back

Page 112: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

95

Figure 25: Box-photometer calibration plot for the Solux

Figure 26: Box-photometer calibration for the Aishwarya

y = 0.5193x - 0.4622R² = 0.9999

y = 0.5752x - 0.5666R² = 0.9999

y = 0.5492x + 2.2295R² = 0.9952

y = 0.4732x + 0.4222R² = 0.9998

y = 0.5161x + 0.2624R² = 0.9999

y = 0.5404x - 1.2725R² = 0.9956

0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160

Sph

ere

me

asu

rem

en

t (L

um

en

)

Alternative method measurement (Lux)

Solux - Box Calibration Plot

Lid

Back Wall

Right Wall

Baffle

Lid

Front Wall

y = 0.578x - 0.6286R² = 0.9999

y = 0.5793x - 0.7289R² = 1

y = 0.5714x + 0.0282R² = 0.9998

y = 0.5748x + 0.0503R² = 0.9998

y = 0.5914x - 0.7922R² = 0.9999

y = 0.5901x - 0.7166R² = 0.99980

10

20

30

40

50

60

70

0 10 20 30 40 50 60 70 80 90 100 110 120

Sph

ere

me

asu

rem

en

t (L

um

en

)

Alternative method measurement (Lux)

Aishwarya - Box Calibration Plot

Switch Back

Switch Front

Switch Right

Switch Left

Switch Front

Switch Back

Page 113: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

96

3.4 Tube-Photometer Calibration Plots

Calibration plots of the tube-photometer measurements for all four light sources

also show near-ideal behavior. Data from each round of testing have a very linear

relationship between the tube and sphere measurements. Like the box-photometer

calibration plots, y-intercepts of the tube calibration plots are relatively near the origin.

Yet, unlike the box calibration plots, there is not an obvious distinction between the tube-

photometer data for directional and omnidirectional light distributions. Tube-photometer

calibration plots for the Firefly, Kiran, Solux and Aishwarya are shown below in Figures

27 through 30.

Page 114: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

97

Figure 27: Tube-photometer calibration plot for the Firefly

Figure 28: Tube-photometer calibration plot for the Kiran

y = 0.8681x - 0.3155R² = 0.9989

y = 0.8572x - 0.0662R² = 0.9995

y = 0.7883x - 0.0709R² = 0.9981

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

0 5 10 15 20 25 30 35

Sph

ere

me

asu

rem

en

t (L

um

en

)

Tube-photometer measurement (Lux)

Firefly - Tube Calibration Plot

Tube-1

Tube-2

Tube-3

y = 2.2426x + 0.4398R² = 0.9981

y = 2.2787x + 0.0199R² = 0.9993

y = 2.2826x + 0.0685R² = 0.9985

0

2

4

6

8

10

12

14

16

0 1 2 3 4 5 6 7

Sph

ere

me

asu

rem

en

t (L

um

en

)

Tube-photometer measurement (Lux)

Kiran - Tube Calibration PlotTube-1

Tube-2

Tube-3

Page 115: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

98

Figure 29: Tube-photometer calibration plot for the Solux

Figure 30: Tube-photometer calibration plot for the Aishwarya

y = 0.6809x + 1.7985R² = 0.9997

y = 0.6779x + 0.907R² = 0.9989

y = 0.7006x + 1.7014R² = 0.9978

0

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120

Sph

ere

me

asu

rem

en

t (L

um

en

)

Tube-photometer measurement (Lux)

Solux - Tube Calibration Plot

Tube-1

Tube-2

Tube-3

y = 3.85x + 0.2008R² = 0.9957

y = 3.7935x + 0.5366R² = 0.9993

y = 4.0358x - 1.3748R² = 0.9966

0

10

20

30

40

50

60

70

0 5 10 15 20

Sph

ere

me

asu

rem

en

t (L

um

en

)

Tube-photometer measurement (Lux)

Aishwarya - Tube Calibration Plot

Tube-1

Tube-2

Tube-3

Page 116: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

99

3.5 Relative Change in Initial Light Output

When the light output measurements for the sphere, box and tube in a single

round are plotted as a percentage of the original light output, one can view the data in the

format of an actual lumen maintenance test. More importantly, deviations of the

alternative measurements from the sphere measurements become apparent in these plots.

Data from all three rounds of testing for the Firefly are shown in Figure 31 as a reference

for discussion. Plots of the relative change in initial light output for all products and all

rounds of testing are included in Appendix F. Qualitatively, the graphs in Figure 31

indicate that the box and tube relative measurements correspond closely with the relative

sphere measurements. The plots also show that when the Firefly‟s light is directed

towards the lid of the box-photometer in Rounds 1 and 3, the relative light measurements

noticeably deviate from the sphere measurements. Detailed evaluation of the maximum

percent error for each set of box and sphere measurements is included in Section 3.6.3.

Page 117: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

100

Figure 31: Plots of measured relative luminous flux decrease as a percentage of the original light

output for Firefly over three rounds of testing.

40%

50%

60%

70%

80%

90%

100%

1 2 3 4 5 6 7 8 9 10 11 12

Pe

rce

nt

of

init

ial l

igh

t o

utp

ut

Measurement Number

Firefly - Round 1

Sphere

Box - Back

Box - Lid

Tube

40%

50%

60%

70%

80%

90%

100%

1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Firefly - Round 2

Sphere

Box - Right Wall

Box - Baffle

Tube

65%

70%

75%

80%

85%

90%

95%

1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Firefly - Round 3

Sphere

Box - Lid

Box - Front Wall

Tube

Page 118: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

101

3.6 Statistical Analysis of Test Data

Data from the Simulated Lumen Maintenance Tests were evaluated using statistical

methods in order to compare the measurement accuracy of the tube- and box-

photometers. The Simulated Lumen Maintenance calibration plots and relative change in

light output data were used as the basis for analysis. The calibration plots were examined

according to the linear regression R2 value, standard error of the regression, and 95%

confidence interval. The R2 value and standard error of the regression for each round of

testing are tabulated in Appendix I. The tube and box relative light output data were

compared to the integrating sphere measurements on a percent error basis, which is

addressed in Section 3.6.3.

3.6.1 Standard Error of Regression for Calibration Plots

The standard error of the regression is “an overall indication of the accuracy with

which the fitted regression function depicts the dependence of the Y on X” (Zar, 1999).

The standard error of the regression was determined for points on the calibration plot

linear models in order to calculate the 95% confidence intervals. The formula used to

determine the standard error of a point on the regression line is given by Equation 10.

Equation 10: Standard error of a point on the regression line (Zar, 1999).

Page 119: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

102

where:

= Standard error of a point on the linear regression model.

= Standard error of the regression.

= Number of samples in each round of measurements.

= Measured illuminance from box- or tube-photometer.

= Mean measured illuminance from a round of testing with the box- or tube

photometer.

and

where:

= Sum of the measured illuminance values for a round of testing with the

box- or tube-photometer.

= Sum of the squared measured illuminance values for a round of testing with the

box- or tube-photometer.

A summary of the standard error for each round of testing is in Appendix I. The

tables include the standard error for each round of testing as a percentage of the average

luminous flux estimated by the linear regression model (SE/ ). These values, ranging

from 0.2% to 3.0%, with an average of 0.9%, indicate that the regression lines accurately

model the calibration plots. Furthermore, comparison of the percentages across all of the

tests highlights specific rounds that deviate from ideal linear behavior more than the

others, which is addressed in Section 4.

Page 120: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

103

3.6.2 95% Confidence Intervals of the Calibration Plot

Looking at the 95% confidence interval of the data is also useful for evaluating

how closely the tube-and box-photometers follow ideal measurement accuracy. The

upper and lower bounds for the confidence intervals were calculated according to

Equation 11.

Equation 11: 95% confidence interval for a y-value estimated by the linear regression model

where:

= Sphere luminous flux measurement estimated by the linear regression model for a

particular alternative measurement, Xi.

= t-value of the Student‟s t-distribution as a function of 95% probability and the

degrees of freedom of the data set.

The 95% confidence intervals for all rounds of testing draw tight bounds around

the linear regression models. Examples of two calibration plots that include the 95%

confidence interval bounds are shown in Figures 32 through 34. The linear model is

shown as a solid line and the dashed lines draw the 95% confidence interval bounds. The

tube calibration plot for the Firefly in Figure 32 is representative of the typical results

across all of the methods and products. Qualitatively, the plot shows that the 95%

confidence interval tightly hugs the linear model and that the regression line passes very

near to the origin. Figure 34 shows a box calibration plot for the Firefly shining on the

lid, which is the most non-ideal round of testing. Although the 95% confidence interval

Page 121: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

104

is relatively broader in this particular series of measurements, bounding the calibration

plot by +-0.4 lumens (2%) from the predicted values, the measurements are still within a

reasonable tolerance for lumen maintenance testing of OLPs.

Figure 32: Tube calibration plot for the second round of testing with the Firefly, including the best

fit line (solid) and 95% confidence interval bounds (dashed). The inset of the origin illustrates non-

ideality, as the linear model does not pass through the origin.

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30 32

Inte

grat

ing

Sph

ere

me

asu

rem

en

t (L

um

en

)

Tube-photometer measurement (Lux)

Firefly - Tube Calibration Plot (Round 2)

See Enlarged

Page 122: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

105

Figure 33: Enlarged section of the tube calibration plot for the Firefly shown in Figure 32.

Figure 34: Box calibration plot for round 3 of testing the Firefly with the light directed at the lid of

the box, including the linear model (bold line) and the 95% confidence interval bounds (thin lines).

This series of measurements showed some of the greatest divergence from ideal behavior, yet still has

an acceptable range of expected error (2%) for use in lumen maintenance testing.

16

17

18

19

20

21

22

23

24

25

26

20 21 22 23 24 25 26 27 28 29 30 31 32

Inte

grat

ing

Sph

ere

me

asu

rem

en

t (L

um

en

)

Tube-photometer measurement (Lux)

Firefly - Tube Calibration (Round 2)with 95% Confidence Interval

15

16

17

18

19

20

21

22

23

24

25

30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47

Inte

grat

ing

Sph

ere

me

asu

rem

en

t (L

um

en

)

Box-photometer measurment (Lux)

Firefly - Box Calibration Plot (Lid, Round 3)

Page 123: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

106

The 95% confidence interval can also be used to evaluate whether the linear

calibration models are expected to pass through the origin. Regression statistics show

that eight of the 36 calibration plot models do not pass through the origin within the 95%

confidence interval range. Although the linear calibration models for these non-ideal

series of measurements do pass near the origin, the expected alternative measurements do

not „zero out.‟ In practice, zeroing out the photometer is not an issue since measurements

at 0 lux are never made in the lumen maintenance testing methods specified by LM-80

and Lighting Africa‟s QTM. More importantly, the fact that the origin is not within the

95% confidence interval of a calibration plot is indicative of some degree of non-ideality

in the measurements. Little emphasis should be placed on these instances of relatively

minor non-ideality, however, since the regression lines project far beyond the range of

measured data. A summary of the calibration plots that are not expected to pass through

the origin is given in Table 5. Notice that the y-intercepts of the regression models are all

relatively near the origin. The difference between the 95% confidence interval bounds

and the origin (ORIGIN – 95% C.I. BOUND) further illustrates that these non-ideal

models narrowly miss the origin.

Page 124: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

107

Table 5: Calibration plots whose linear models do not pass through the origin within the 95%

confidence interval. The difference between the 95% confidence interval bound and the origin for

each round of measurements is shown in the right column.

LIGHT SOURCE NAME

LIGHT SOURCE ORIENTATION Y-

INTERCEPT (lm)

ORIGIN - 95% C.I. BOUND

(lm)

Box

Solux Shining on lid - Round 1 -0.46 -0.06

Shining on back wall - Round 1 -0.57 -0.13

Aishwarya On/off switch facing front wall - Round 1 -0.63 -0.48

On/off switch facing back wall - Round 1 -0.73 -0.35

On/off switch facing front wall - Round 3 -0.80 -0.39

On/off switch facing back wall - Round 3 -0.72 -0.23

Tube

Solux Round 1 1.80 1.01

Kiran Round 1 0.44 0.08

3.6.3 Error in Relative Light Output Calculations

The Relative Lumen Decrease plots in Appendix F give additional information

about the applicability of the tube- and box-photometers to lumen maintenance testing.

Calculating the percent decrease in original light output according to Equation 9 results in

propagation of the error from both illuminance measurements. If the initial measurement

is inaccurate, that error is passed on to the subsequent measurements at different light

levels. This propagation of error is visible in Figure 35, where the relative values from

the tube-photometer are all shifted slightly higher than the other methods.

Page 125: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

108

Figure 35: Lumen depreciation plot for the third round of testing with the Aishwarya. The relative

measurements made with the tube-photometer are shifted above the other methods due to a low

initial measurement.

Evaluating the data in relative terms adds the error associated with the initial

measurement, which replicates the propagation of error in actual lumen maintenance

testing. Looking at the data from this perspective, therefore, is a more realistic

representation of the error in actual lumen maintenance testing.

The largest discrepancy in a single relative measurement has an error of 7.5%.

This outlier occurs in Round 1 of testing the Firefly in the box-photometer, with the light

directed towards the box lid. The average percent error across all of the relative light

measurements in the box and tube are much lower, at 1.4% and 1.5%, respectively. In

terms of the accuracy required for lumen maintenance measurement in the QTM and

50%

55%

60%

65%

70%

75%

80%

85%

90%

95%

100%

1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Aishwarya - Round 3

Sphere

Box - Switch Front

Box - Switch Back

Tube

Page 126: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

109

ISM, this degree of error is acceptable. A summary of the maximum percent error for all

sets of relative light output measurements is included in Appendix G.

3.6.4 Analysis of Variance (ANOVA) for R2 of Calibration Plots

An Analysis of Variance (ANOVA) General Linear Model (GLM) method of

statistical analysis is used to determine if there is a statistically significant difference

between the accuracy of measurements made in the box- and tube-photometers. In

addition, the GLM compares the accuracy of box and tube measurements according to the

specific light source that is being measured. A summary of the three null hypotheses and

the GLM test results is shown in Table 6.

Table 6: Summary of null hypotheses tested in ANOVA and interpretation of the statistical test

results.

Null Hypothesis (HO) ANOVA Result

(Reject / Fail to Reject)

Interpretation of Result

R2 of the calibration plot

linear model does not

depend on the light source.

Fail to Reject

No statistically significant

difference in measurement

accuracy is attributed to the

different light sources.

R2 of the calibration plot

linear model does not

depend on the method used

to measure the light output.

Fail to Reject

No statistically significant

difference in measurement

accuracy between the tube-

and box-photometers.

Response of R2 to the light

measurement method does

not depend on the light

source.

Reject

Statistically significant

differences exist in the

measurement accuracy for

specific light source/testing

apparatus combinations.

Page 127: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

110

One way of comparing the accuracy of measurements between the different

products, testing devices and rounds of testing is the coefficient of determination. “The

coefficient of determination, R2, is the proportion of variability in a data set that is

accounted for by the statistical model” (Steel & Torrie, 1960). In this analysis, R2

indicates the goodness of fit of a linear model for the relationship between illuminance

measurements made with low-cost methods and the corresponding luminous flux

measurements made with the integrating sphere. The R2 value serves as a measure of the

ideality of the box and tube measurements, where the perfect round of measurements can

be described by a linear equation with R2 = 1. By using R

2 values from the calibration

plots as the continuous response variable in a GLM, one can generate statistical data that

rejects or fails to reject specific hypotheses about the accuracy of the testing methods.

The null hypotheses for analysis of data from this study are:

HO, 1: R2 of the calibration plot linear model does not depend on the light source.

HO, 2: R2 of the calibration plot linear model does not depend on the method used to

measure the light output.

HO, 3: Response of R2 to the light measurement method does not depend on the light

source.

The R-squared values of the calibration plot linear models are useful for

comparing the accuracy of each series of measurements. Yet, one complication

associated with the R2 values used in this particular study is that each round of testing

includes two series of box measurements, each with its own linear regression model. For

use in this statistical analysis, the two R2 values for each round of box-photometer

Page 128: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

111

measurements are averaged. This yields 24 data points that are the GLM response

variables (four DuTs, each with three rounds of measurements, using two measurement

apparatuses).

Another complication encountered when generating the response variables for the

statistical analysis is that the R2 values of the calibration plots are very similar in

magnitude. The R2 values range from 0.982 to 0.999, with an average value of 0.998.

The histogram in Figure 36 shows how the data are right-skewed, with a majority of R2

values lying between 0.997 and 1. The data grouped near R2=1 are so similar that many

are equal up to two and even three significant digits (i.e. 0.99 and 0.999). In order to

differentiate between such similar magnitudes, the R2 values had to be calculated out to

as many digits as possible. This was accomplished by using the Microsoft Excel “Data

Analysis - Regression” tool, which reports the R2 values out to 15 digits.

Page 129: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

112

1.00000.99750.99500.99250.99000.98750.98500.9825

12

10

8

6

4

2

0

R-Squared

Fre

qu

en

cy

Histogram of R-Squared

Figure 36: Histogram of R-squared values for the experimental calibration plots shows that the data

are right-skewed and two outliers are less than 0.99.

Minitab statistical analysis software was used to create a GLM with the high

precision R-squared values as the response variable. The statistical model consists of

three different categorical predictor variables: Name, Round, and Method. Name

identifies which OLP was used (i.e. Firefly, Kiran, Solux, or Aishwarya). Round

identifies the series of parallel measurements conducted with the box- and tube-

photometers (three rounds of measurements were conducted for each Name). Method

identifies whether the tube- or box-photometer was used to make the light output

measurements. The structure of the experiment for use in the GLM is shown

schematically in Table 7 and a summary of the categorical predictor variables is given in

Table 8.

Page 130: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

113

Table 7: Hierarchy of experimental design for ANOVA.

Table 8: Summary of the categorical predictor variables used in the Simulated Lumen Maintenance

statistical analysis.

General Linear Model: R2 versus Name, Round, Method

Factor Type Levels Values

Name Fixed 4 Firefly, Kiran, Solux, Aishwarya

Round Random 12 1, 2, 3 for each Name

Method Fixed 2 Box, Tube

When creating the GLM, the relationship between the factors in the experimental

design must be specified. First, Round is nested within Name. This means that each OLP

is subjected to three unique rounds of testing that cannot be repeated exactly with the

other light sources. Secondly, Method is crossed with Name. This means that all of the

DuTs are measured with the same box- and tube-photometers in all rounds of the

experiment.

The types of factors used in the GLM must also be specified according to the

experimental design. Both Name and Method are fixed factors, indicating that they are

controlled in the experiment. Round is specified as a random factor in the GLM. This is

due to the fact that the exact electrical current and sequence at which LEDs are being

driven can never be exactly replicated.

Page 131: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

114

The experiment is described by the following equation for the GLM:

Equation 12: General Linear Model used for statistical analysis of the Simulated Lumen

Maintenance test data.

Y = NAME + ROUND(NAME) + METHOD + METHOD*NAME + ERROR

where:

ROUND(NAME) = Testing round is nested in the light source

METHOD + METHOD*NAME = Testing device is crossed with the light source

ERROR = Error not accounted for by the model

Using Minitab software, the 24 R-squared values were input into the GLM

described in Equation 12. The normal probability plot of R2 and the „Versus Fits‟ plot,

shown in Figure 37, indicate that the data are right-skewed. Most of the R-squared

values lie between 0.997 and 1.0, with only a few less than 0.997. The residuals are not

normally distributed, which shows that the R2 values do not have equal variance. This

violates the normality assumption for a general linear model. Also, note that some fitted

values in the „Versus Fits‟ plot are greater than 1. This model is incorrect since R2 is, by

definition, never greater than one.

Page 132: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

115

3210-1-2-3

99

95

90

80

70

60

50

40

30

20

10

5

1

Standardized Residual

Pe

rce

nt

Normal Probability Plot(response is R SQUARE)

1.0051.0000.9950.9900.985

3

2

1

0

-1

-2

-3

Fitted Value

Sta

nd

ard

ize

d R

esid

ua

l

Versus Fits(response is R SQUARE)

Figure 37: Residual plots of the R-squared values for the experimental calibration plots. The

Normal Probability Plot and Versus Fits plot indicate that the data are right-skewed and not

normally distributed.

A transformation of the response variable, therefore, is necessary. Equal variance

of the R2 values is achieved by applying the following transformation:

Equation 13: Transformation applied to the R2 data in order to normalize a skewed distribution.

The natural logarithm (ln) of the quantity (1 – R2) adjusts the data such that they

exhibit normal distribution and equal variance. The natural log is negative in order to

keep the transformed values positive. As shown in Figure 38, the transformed data have

normally distributed residuals and equal variances, thereby meeting the assumptions for a

general linear model.

Page 133: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

116

9.68.88.07.26.45.64.84.0

7

6

5

4

3

2

1

0

-ln(1-R^2)

Fre

qu

en

cy

Histogram of -ln(1-R^2)

3210-1-2-3

99

95

90

80706050403020

10

5

1

Standardized Residual

Pe

rce

nt

Normal Probability Plot(response is -ln(1-R2))

10987654

2

1

0

-1

-2

Fitted Value

Sta

nd

ard

ize

d R

esid

ua

l

Versus Fits(response is -ln(1-R^2))

Figure 38: The Histogram of R2 data after applying the transformation shows normal distribution.

The Normal Probability Plot has equal variance and the Versus Fits plot also shows a distribution

that satisfies the assumptions of the general linear model.

Page 134: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

117

A GLM of the same form as Equation 12 was generated using Minitab for the

transformed R2 values. A summary of the statistical results is shown in Table 9.

Table 9: Statistical results from the general linear model for the transformed R2 values (R

2TRANS) of

the calibration plots.

ANOVA for R2TRANS

Source DF Seq. SS Adj. SS Adj. MS F P

NAME 3 7.9800 7.9800 2.6600 1.32 0.335

ROUND(NAME) 8 16.1607 16.1607 2.0201 2.92 0.076

METHOD 1 1.4723 1.4723 1.4723 2.13 0.183

NAME*METHOD 3 15.6348 15.6348 5.2116 7.53 0.010

Error 8 5.5395 5.5395 0.6924

Total 23 46.7874

S = 0.832129 R-Sq = 88.16% R -Sq (adj) = 65.96%

The general linear model has a moderately high adjusted R2 value of 65.96%,

indicating that the GLM accounts for nearly 66% of the variability of the data. An

adjusted R2 value of this magnitude is a fair approximation of the data points and is

acceptable for use in evaluating the null hypotheses.

A p-value of 0.335 for the Name term indicates that the difference in R2 values of

the calibration curves for all four DuTs have a 33.5% chance of being ascribed to chance

alone. In other words, it is statistically plausible that the measurement accuracy for each

of the four light sources used in testing are the same. Furthermore, an F statistic of 1.32

means that the variance in the transformed R2 due to brand is 1.32 times greater than that

Page 135: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

118

due to error. The first null hypothesis that „R2 of the calibration plot linear model does

not depend on the light source,‟ therefore, cannot be rejected.

A p-value of 0.183 and F-statistic of 2.13 for the Method term indicates that the

second null hypothesis, too, cannot be rejected. The difference in R2 values of the linear

models for the calibration plots cannot be attributed the method alone.

The third null hypothesis, that the „response of R2 to the light measurement

method does not depend on the light source,‟ is rejected. The Name*Method term has a

p-value of 0.010 and an F-statistic of 7.53, which indicates that the interaction between

Name and Method causes statistically significant differences in R2 values. The tube- and

box-photometers have statistically different measurement accuracy for some light sources

used in the experiment. The interaction plot in Figure 39 demonstrates how the mean R2

value of the calibration plots for the Firefly measured in the box-photometer is

significantly less than all other Name*Method combinations. This outlier supports the

results of the ANOVA, indicating that rejection of HO, 3 is justified. The error in the

Firefly‟s box-photometer measurements are discussed in greater detail below.

Page 136: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

119

TubeBox

1.000

0.998

0.996

0.994

0.992

0.990

METHOD

Me

an

Aishwarya

Firefly

Kiran

Solux

NAME

Interaction Plot for R-SquareFitted Means

Figure 39: Interaction plot showing how the mean R2 value varies across the box- and tube-

photometer testing methods and four light sources. Calibration plots of the Firefly measured in the

box-photometer exhibit significantly smaller R2 values than the other light source-measuring device

combinations.

Page 137: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

120

4 DISCUSSION OF RESULTS

The Simulated Lumen Maintenance Test data indicate that the tube- and box-

photometers can be used to accurately measure the relative decrease in light output of

various OLPs. Qualitatively, the near-ideal behavior of the box and tube is easily

identifiable in the highly linear calibration plots and tightly grouped relative light output

plots. Quantitative analysis of the box and tube measurements not only indicates a close

correlation to the sphere measurements, but also allows for further comparative

conclusions to be drawn.

The Simulated Lumen Maintenance Tests were evaluated from several different

statistical perspectives in order to compare measurement accuracy and identify strengths

and weaknesses of the tube- and box-photometers. The R2 values of the calibration plots,

ranging from 0.982 to 0.999, are a clear indication of the highly linear, near-ideal

relationship between the tube, box and integrating sphere measurements. The 95%

confidence intervals for the calibration plots indicate that light measurements made with

the box- and tube-photometer have little deviation from the linear model. The relative

light output plots for tube and box measurements also exhibit a very close correlation to

the integrating sphere measurements in most cases. The average percent error across all

of the relative light measurements in the box and tube are 1.4% and 1.5%, respectively.

Simple comparisons of these error metrics across all rounds of testing are useful for

identifying apparatus-DuT combinations that had relatively high levels of error.

Page 138: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

121

Analysis of variance of the calibration plot R2 values shows that the tube and box are

equally accurate in measuring the light output of the light sources used in this study. The

ANOVA also confirms that the outlying rounds of Firefly measurements did, in fact,

have statistically significant deviation from the error in other rounds of testing.

4.1 Worst Case: Error Analysis of Firefly in the Box-Photometer

A closer look at the rounds of testing that have the greatest amount of variability

is useful for understanding some possible ways in which the low-cost test methods are

prone to measurement inaccuracy. Evaluating how the measurements deviate from ideal

behavior allows for conclusions to be drawn about potential improvements to the testing

devices and informs recommendations for their use in quality assurance testing of OLPs.

When measured in the box-photometer, the Firefly exhibited the greatest

deviation from ideal behavior. Error in the measurements was primarily due to two

rounds of testing with the light directed towards the box-photometer lid. The calibration

plots for these two outliers have R2 values of 0.984 and 0.981, indicating that the linear

models for these two „worst case‟ rounds of testing account for about 98% of the

variability in the data sets. The standard error of the regression as a percentage of the

average predicted values (SE/ ), listed in Appendix I also indicate that these two

rounds of Firefly measurements have relatively elevated levels of error (3.0% and 1.8%).

Although statistical analysis does show that more error is present when the Firefly light

Page 139: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

122

output is measured in the box, one must keep in mind that the degree of variability in

even this worst case scenario is quite small.

In the case of the Firefly oriented towards the box lid, the slight non-linearity of

the data is most likely due to two main factors. One source of measurement error may be

the particular orientation of the light source. When the Firefly‟s directional light (FWHM

angle of 81o) shines towards the lid of the box, the light may strike the photometer sensor

after a single reflection. As described earlier, hollow cavity photometers like the box-

photometer require that the light reflect off the interior surface at least twice before

striking the sensor. Directing the Firefly towards the box‟s lid may allow some of the

light to strike the photometer sensor after the first bounce, resulting in a detectable

increase in measurement error. The measurement error may also be attributed to the

Firefly‟s flexible neck. A relatively high degree of adjustability introduces more

variability in the orientation of the light source. Repeating the exact same orientation of

the light source for each measurement in the box-photometer is much more difficult for

the Firefly. This results in increased measurement error, especially since the Firefly is a

directional light source. Statistically significant increase in the measurement error of the

Firefly in the box-photometer highlights the importance of maintaining a constant

location and orientation of a DuT in the box. Care should be taken to set up an easily

repeatable placement of the DuT. Gooseneck-style lamps like the Firefly and other

highly adjustable products should be fixed in a configuration that restricts movement of

the light source.

Page 140: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

123

4.2 Sources of Error

Deviations of light output measured with the box-photometer from the integrating

sphere measurements can be primarily attributed to random error that is due to

unpredictable changes in the test. The most likely source of random error in box-

photometer measurements is inconsistency in the way that the operator conducts the tests.

Inaccuracy in the box-photometer measurements occur when the operator fails to repeat

the same placement and orientation of the DuT throughout the lumen maintenance test.

Even seemingly small changes in the orientation of the DuT within the box can yield

inaccurate test results. Fortunately, random errors can be reduced by making repeated

measurements.

The dominant cause of error in tube-photometer measurements is also likely due

to operator error. The method for measuring light output with the tube-photometer

requires the test operator to search for and identify the maximum illuminance. Failure to

find the maximum illuminance will result in a measured value that is too low. Like the

box-photometer, error due to misidentification of the maximum illuminance with the

tube-photometer can be reduced by making multiple measurements. The maximum

illuminance value from the repeated tube-photometer measurements can be assumed to

be the most accurate light output value.

Systematic errors, on the other hand, cannot be reduced by repeating

measurements. Systematic errors in the box and tube measurements, while extremely

small, are most likely due to the apparatuses, themselves. Some possible sources of

Page 141: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

124

systematic error may include irregularities in the tube and box cavities, light intrusion in

the cavities, and the illuminance meters. The low-cost illuminance meters used in the

box- and tube-photometers may have introduced a relatively small degree of error. The

precision of the low-cost meters is 0.01 lux for a range of measured values up to 20 lux,

and 0.1 lux for measured values up to 200 lux. A majority of the box- and tube-

photometer measurements made in this study lie in the 20 – 200 lux range, which is

typical for OLPs currently on the market. While higher quality, more precise illuminance

meters may be used in place of the Extech meters used in this study, it is unlikely that any

significant measurement accuracy will be achieved. Furthermore, the Extech meter is

relatively inexpensive, which is an important criterion for selecting components used in

the low-cost testing apparatuses.

Page 142: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

125

5 CONCLUSIONS AND RECOMMENDATIONS

The results of this study clearly indicate that the box- and tube-photometer can be

used to accurately measure the relative decrease in light output of OLPs. The box- and

tube-photometers, therefore, are recommended for use in the QTM and ISM Long-Term

Lumen Degradation Test. While the statistical analyses do point out some divergences

from perfect correlation to the integrating sphere measurements, the box- and tube-

photometers have proven to be highly accurate substitutes for an integrating sphere in

lumen maintenance testing. Nonetheless, the type and magnitude of measurement errors

identified in this research point to potential flaws in the low-cost testing devices.

5.1 Impact of Error on Lumen Maintenance Test Results

The current versions of Lighting Africa‟s QTM and ISM indicate that the box-

and tube-photometers are acceptable substitutes for an integrating sphere for use in the

Long-Term Lumen Degradation Test. Approval of the alternative light measuring

devices is based on the assumption that additional error associated with box and tube

measurements is negligible. Results from this study indicate that while the box- and

tube-photometers can accurately measure the relative change in light output, there are

instances where the alternative measurements deviate from the sphere by up to 7.5%. If a

high-error measurement occurs at the beginning or end of the lumen maintenance test,

significant miscalculation of the product‟s life may result. Consider for example, a

Page 143: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

126

product that reaches its L70 lifetime just before 2,000 hours and, as a result, does not

meet the Lighting Africa minimum performance target. If the final relative light output

measurement has an error on the order of 7.5%, the maximum seen in this study, there is

a possibility that the product may have actually met the performance target. As

mentioned above, the best way to address this type of error is to make repeated

measurements. Each light output measurement, especially the initial reading, should

therefore be repeated at least twice in order to identify incorrect readings.

5.2 Benefits and Disadvantages of the Box-Photometer

The box-photometer has been used for quality assurance testing of OLPs at

SERC, FISE and the University of Nairobi in Kenya. Experience has shown that boxes

can be affordably and easily constructed according to the specifications shown in

Appendix J. The box-photometer also has a good track record of use for illuminance and

luminous flux measurements. The device has been verified as a legitimate light

measuring apparatus by authoritative sources and, more recently, it has been used

successfully at the aforementioned institutions.

A major strength of the box-photometer is its relative ease of use. Light

measurements made in the box can be conducted very quickly and simply. The box-

photometer is also versatile. The box is currently a critical piece of equipment in light

testing labs for conducting the QTM Autonomous Run Time and Solar Run Time tests, in

addition to the Long-Term Lumen Degradation test.

Page 144: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

127

The box-photometer does, however, exhibit some weaknesses. This study has

shown that the box is susceptible to small errors when measuring the light output of

highly directional and adjustable light sources. These errors, which can generally be

avoided by repeated measurements and diligent repetition of DuT orientation within the

box, may actually be less important than the cost and size of the device. Box-

photometers constructed for the University of Nairobi lighting lab cost approximately

100 USD each, not including the cost of the light meters. This is a relatively minor

expense when compared to the cost of other equipment required for a fully functional

lighting test center. Yet, from the perspective of a low-budget laboratory in a developing

country, or a lighting manufacturer seeking to conduct low-cost in-house testing, this may

be a considerable expense.

The physical size of the box-photometer has already proven to be an issue at the

University of Nairobi. The small room that houses their lighting lab had to be

strategically organized to fit all of the necessary equipment. As testing operations

expand, space is at a premium, and any additional equipment must be compact in size.

An additional box-photometer requires nearly 1m2 of floor space. Other lighting test

centers with similar budgetary and space constraints are also likely to experience similar

problems.

Page 145: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

128

5.3 Benefits and Disadvantages of the Tube-Photometer

The tube-photometer excels in its physical simplicity and exceptionally low cost.

The device consists of only three parts: a tube, an end cap and an illuminance meter.

Both the tube and end cap can be constructed of common, inexpensive (often free)

materials. No specialized tools are needed for assembly of the tube-photometer. The

device used in this study was constructed using a utility knife and packaging tape. In

addition, the tube-photometer does not present the size issues encountered by the box-

photometer. The device used in this study was 0.5 m long, with a 6 cm diameter. Unlike

the box-photometer, the tube does not require any floor space in the lighting laboratory.

The tube is lightweight and easily portable. The tube-photometer can be stored in any

convenient location, as long as the light meter sensor remains fixed to the tube in the

same orientation and free of dust.

While the tube-photometer clearly offers financial, constructability, and spatial

advantages over the box-photometer, the apparatus does have its own particular

weaknesses. Firstly, the tube-photometer, as configured in this study, cannot be used for

the QTM Autonomous Run Time and Solar Run Time tests. The tube is intended solely

for the measure of relative change in light output for lumen maintenance testing.

Secondly, measurements made with the tube-photometer are not as simple or as quick as

the box-photometer. For each reading, the operator must „search‟ for the maximum

illuminance using the light meter‟s “Max. Hold” function. This not only requires more

time and skill using the device, but introduces the possibility for increased operator error.

Page 146: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

129

When measuring DuT light output, the light source must be held flush to the open

edge of the tube as angular adjustments are made to the orientation of the light. For

lantern-shaped lights like the Aishwarya, the operator has to „search‟ for a maximum

illuminance over the surface of the product. The operator must continue searching for the

maximum lux value until the light meter‟s “Max. Hold” function no longer increases.

This presents the problem of „giving up the search‟ before the actual maximum

illuminance has been recorded by the meter. The tube-photometer may be especially

prone to this type of measurement error when more than one operator is conducting the

lumen maintenance test for a particular DuT.

5.4 Multiple Test Operators

An important caveat of this study is that all of the measurements made in the

Simulated Lumen Maintenance Tests were conducted by the same experienced operator.

The tests were performed by the author who understands the theories and concepts

behind the measurements and is very familiar with all of the equipment and methods. In

addition, the author had a substantial stake in carefully executing the measurements.

Many technicians employed to conduct these tests by developing country lighting

laboratories or off-grid lighting manufacturers may not have the same level of

comprehension, familiarity, or meticulous execution of the test methods. The

measurement accuracy for the tube- and box-photometers described herein, therefore,

may be greater than the accuracy achieved by similar tests conducted by less experienced

Page 147: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

130

operators. While user experience and skill level may affect measurement error, one must

keep in mind that accurate use of the box- and tube-photometers is much simpler and

easier than integrating sphere and goniophotometer systems.

Additionally, lumen maintenance testing in lighting laboratories is likely to be

conducted by more than one operator. This too, may lead to decreased measurement

accuracy due to inconsistencies between test operators. In the case of the box-

photometer, different people may place the DuT in the box in slightly different

orientations. The tube-photometer may be even more susceptible to operator-based

variability if one operator does not thoroughly „search‟ for the maximum illuminance

value. With these additional sources of variability, one must take caution when

translating the results and recommendations from this study to actual implementation in

lighting laboratories. When training lighting lab technicians and selecting the most

appropriate means of measuring lumen maintenance, one must take into account the

operators‟ skills, levels of expertise, and attention to detail. Since the variability

introduced by multiple, less skilled technicians was not addressed in this study, further

research that explores the significance of error introduced by these factors may be

warranted.

5.5 Potential Improvements to Lumen Maintenance Testing Devices

The results of this study point to variability in the light source orientation as a

major contributor to the error in tube- and box-photometer measurements. In the context

Page 148: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

131

of quality assurance testing of OLPs in developing countries, these errors are quite

acceptable when considering the economic and ease-of-use benefits achieved by the low-

cost testing devices. The next logical step in developing the ideal lumen maintenance

testing device is to explore how to minimize measurement error while maintaining or

even improving the benefits exhibited by the box- and tube- photometers.

The Lighting Research Center (LRC) at Rensselaer Polytechnic Institute is

currently contracted with Lighting Africa to conduct quality assurance testing of select

OLPs available in the African market. LRC has taken a slightly different approach to

lumen maintenance testing that may have drastically reduced the measurement error

while maintaining the ease-of-use and low-cost characteristics of the box- and tube-

photometers. They have constructed a simple apparatus, shown in Figure 40, that

maintains the DuT in a fixed position over the course of the lumen maintenance test by

affixing the product to a block of wood.

Figure 40: Light measurement apparatus used at the Lighting Research Center for lumen

maintenance and run time testing of off-grid lighting products. Photo courtesy of Erik Page.

Page 149: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

132

The most novel aspect of LRC‟s light measurement device, however, is the use of

low-cost photodiodes to measure the light output of the DuTs. Each DuT is fixed in a

cavity and shines light on a dedicated photodiode. The photodiodes, manufactured by

OSRAM, cost about 1 USD each and have an accurate V() spectral response. The

photodiodes are connected to a computer that uses LabView software to convert the

voltage signals to an illuminance reading and automatically samples and records the

illuminance measurements at specified time intervals. This method for measuring the

relative change in light output eliminates the human error associated with the box- and

tube-photometers, but it does require more space and knowledge of a complex software

package.

As demand for low-cost performance testing of OLPs continues to grow, new

iterations of test methods and improvements to testing devices, like the apparatus created

by LRC, are likely to emerge. Nonetheless, this study confirms that the tube- and box-

photometers are both effective tools for measuring the lumen maintenance of OLPs. Yet,

when selecting the most appropriate device for lumen maintenance testing, lighting

laboratories must use discretion to choose apparatuses according to their particular time,

training, monetary, spatial, and accuracy constraints.

Page 150: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

133

REFERENCES

Apple, Vincente, Yarberry, Lohse, Mills, Jacobson, Poppendieck. (2010). Characterization of

particulate matter size distributions and indoor concentrations from kerosene and diesel

lamps. Indoor Air , 399-411.

Barrows, W. (1912). Light, photometry and illumination. New York: McGraw-Hill Book

Company.

Bass, M. (1995). Handbook of Optics Volume II - Devices, Measurements and Properties, 2nd

Ed. New York: McGraw-Hill.

Bullough, J. D. (2003, May). Lighting Answers: Light Emitting Diode Lighting Systems.

Retrieved January 2011, from Lighting Research Center web site:

http://www.lrc.rpi.edu/nlpip/publicationDetails.asp?id=885&type=2

CIE. (1996). Technical Report: The Measurement of Luminous Flux (CIE 84-1989). International

Commission on Illumination.

Cooper, J. (2007, August). Driving LED lamps-some simple design guidelines. LEDs Magazine .

Durell, C. (2009, March 26). Practical Photometric Testing of SSL Products. Retrieved January

2011, from LEDs Magazine web site:

http://img.ledsmagazine.com/objects/features/6/2/15/SphereOptics-260309.pdf

Electro Optical Industries, Inc. (2010). What is Visible Radiation. Retrieved March 2011, from

Electro Optical Industries, Inc.: http://www.electro-optical.com/eoi_page.asp?h=What Is

Visible Radiation?

Freyssinier, Taylor, Frering, Rizzo. (2009). Considerations for Successful LED Applications.

Shenzhen: China Solid State Lighting Alliance.

Friedman, L. (2010, October 20). Bringing Clean Light to Poor Nations and Moving Beyond

Charity. The New York Times .

Goeble, D. G. (1967). Generalized Integrating-Sphere Theory. Applied Optics , 125-128.

GTZ. (2010, May). What difference can a PicoPV system make? Retrieved January 2011, from

GTZ web site: http://www2.gtz.de/dokumente/bib/gtz2010-0210en-picopv-system.pdf

IES. (2008). Approved Method: Measuring Lumen Maintenance of LED Light Sources (LM-80-

08). New York: Illuminating Engineering Society (IES).

Page 151: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

134

IESNA. (2005). Technical Memorandum on Light Emitting Diode (LED) Sources and Systems,

TM16-16-05. New York, New York, USA: Illuminating Engineering Society of North

America.

International Energy Agency. (2010). Electricity Access Database. Retrieved January 2011, from

World Energy Outlook 2010:

http://www.iea.org/weo/database_electricity10/electricity_database_web_2010.htm

Irvine, J. (2011). What are the uses and hazards of waves that form the electromagnetic

spectrum? Retrieved March 2011, from Antonine Education Physics GCSE:

http://www.antonine-

education.co.uk/physics_gcse/Unit_1/Topic_5/topic_5_what_are_the_uses_and_ha.htm

Jacobson, A. (2007). Connective Power: Solar Electrification and Social Change in Kenya. World

Development , 35 (1), 144 - 162.

Krenzke, G. (1965). Die Optimierung der Meßanordnung in runden und eckigen Hohlräumen zur

Lichtstrombestimmung ausgedehnter Lichtquellen.

Labsphere. (2008a). A Guide to Integrating Sphere Radiometry and Photometry. Retrieved

January 2011, from Labsphere Web site: http://labsphere.org/tecdocs.aspx

Labsphere. (2008b). A Guide to Integrating Sphere Theory and Applications. Retrieved January

2011, from Labsphere Web site: http://www.labsphere.com/uploads/technical-guides/a-

guide-to-integrating-sphere-theory-and-applications.pdf

Lighting Africa. (2011). Retrieved 1 2011, from Lighting Africa Web site:

http://www.lightingafrica.org

Lighting Africa. (2010a). Briefing Notes, Issue 1: Light Emitting Diode (LED) Lighting Basics.

Retrieved January 2011, from Lighting Africa web site:

http://www.lightingafrica.org/files/Lighting_Africa_Briefing_Notes_21.Dec_.2009.pdf

Lighting Africa. (2010b, March). Briefing Notes, Issue 2: LED Lumen Depreciation and Lifetime.

Retrieved January 2011, from Lighting Africa web site:

http://www.lightingafrica.org/files/Briefing_Notes_Lifetime_April_2010_Issue2.pdf

Lighting Africa. (2010c). Lighting Africa Quality Test Method (QTM). Retrieved January 2011,

from Lighting Africa web site: http://lightingafrica.org/resources/technical-research.html

Lighting Africa. (2010d). Solar Lighting for the Base of the Pyramid - Overview of an Emerging

Market.

Page 152: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

135

Lighting Research Center. (2003, May 28). NPLIP: LED Lighting Systems. Retrieved October

2010, from Lighting Research Center Web Site:

http://www.lrc.rpi.edu/publicationDetails.asp?id=885

Mills, E. (2007, May 3). Assessing the Performance of 5mm White LED Light Sources for

Developing-Country Applications. Retrieved January 2011, from The Lumina Project

web site: http://light.lbl.gov/pubs/rn/lumina-rn1.pdf

Mills, E. (2005, May 27). The Specter of Fuel-Based Lighting. Science , pp. 1263-1264.

Mills, E., & Jacobson, A. (2008). The need for independent quality and performance testing of

emerging off-grid white-LED illumination systems for developing countries. Light &

Engineering , 16 (2), 5-24.

Mink, Alstone, Tracy, Jacobson. (2010, May 1). LED Flashlights in the Kenyan Market: Quality

Problems Comfirmed by Laboratory Testing. Retrieved April 2011, from Lighting Africa

Web Site: http://lightingafrica.org/resources/technical-research.html

Narendran, Gu, Jayasinghe, Freyssinier, Zhu. (2007, November). Long-term performance of

white LEDs and systems. Retrieved January 2011, from Lighting Resource Center web

site: http://www.lrc.rpi.edu/programs/solidstate/pdf/Narendran-

WhiteLEDsTokyo2007.pdf

Philips Lumileds. (2010, May 3). Technology White Paper: Understanding power LED lifetime

analysis. Retrieved January 2011, from Philips Lumileds web site:

http://www.philipslumileds.com/support/documentation/white-papers

Practical Action. (2010). Poor People's Energy Outlook 2010. Retrieved January 2011, from

Practical Action web site: http://www.practicalaction.org/energy-advocacy/ppeo-report-

poor-peoples-energy-outlook

Ruttan, V. (2001). Technology, Growth and Development. New York: Oxford University Press.

Ryer, A. (1997). Light Measurement Handbook. Newburyport, MA: International Light.

SphereOptics. (2007). Integrating Sphere Design and Applications. Retrieved January 2011, from

SphereOptics web site: http://www.sphereoptics.com/assets/sphere-optic-pdf/sphere-

technical-guide.pdf

Steel, R., & Torrie, J. (1960). Principles and Procedures of Statistics. New York: McGraw-Hill.

Tracy, J. (2010). Flashlights in Kenya: Revealing the Social, Economic, Health and

Environmental Implications in the Absence of Quality Assurance. Humboldt State

University - Master's Thesis.

Page 153: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

136

U.S. Dept. of Energy. (2006, August). Lifetime of White LEDs. Retrieved January 2011, from

U.S. Dept. of Energy: Building Technologies Program:

http://apps1.eere.energy.gov/buildings/publications/pdfs/ssl/lifetime_white_leds_aug16_r

1.pdf

U.S. Dept. of Energy. (2010, March). Solid-State Lighting Research and Development: Multi-

Year Program Plan. Retrieved 2 2011, from Energy Efficiency & Renewable Energy

Web Site:

http://apps1.eere.energy.gov/buildings/publications/pdfs/ssl/ssl_mypp2010_web.pdf

UNFCCC. (2010, November 26). EB 58, Annex 19: Indicative simplified baseline and

monitoring methodologies for selected small-scale CDM project activity categories.

UNFCCC/CCNUC: CDM - Executive Board.

UNFCCC. (2007, March 11). The CDM. Retrieved March 2011, from United Nations Framework

Convention on Climate Change (UNFCCC): http://cdm.unfccc.int/

Welsh, F. (2009). U.S. Dept. of Energy. Retrieved 2 2011, from Energy Efficiency & Renewable

Energy: http://apps1.eere.energy.gov/buildings/publications/pdfs/ssl/welsh_rd_ssl09.pdf

Wikipedia. (2009, April 6). Light-emitting diode. Retrieved 3 2011, from Wikipedia:

http://en.wikipedia.org/wiki/Light-emitting_diode

World Bank. (2009). Gross national income per capita, 2009. Retrieved April 2011, from The

World Bank Website:

http://siteresources.worldbank.org/DATASTATISTICS/Resources/GNIPC.pdf

Zar, J. H. (1999). Biostatistical Analysis. Upper Saddle River, New Jersy: Prentice-Hall, Inc.

Page 154: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

137

APPENDIX A: LIGHT SOURCES USED IN TESTING

Page 155: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

138

Figure 41: Off-grid lighting products used in this study. Clockwise from top left: Aishwarya Wow,

D light Nova, Solux LED-50, Barefoot Firefly.3

3 Product images downloaded from the following websites:

Aishwarya: www.akashdeep.org; Kiran: www.metaefficient.com; Firefly: www.enejipwop.com;

Solux: www.solux.org

Page 156: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

139

APPENDIX B: INTEGRATING SPHERE INTERNAL REFLECTANCE MAP

Page 157: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

140

Table 10: Illuminance map describing irregularity in the interior integrating sphere surface used in testing. The horizontal and vertical

angles are measured in reference to the bottom of the integrating sphere. The illuminance values are measured by the photospectrometer

when the light source was directed to the indicated horizontal and vertical angles.

Page 158: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

141

APPENDIX C: DEVICE UNDER TEST (DuT) LIGHT DISTRIBUTION PLOTS

Page 159: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

142

Figure 42: Radial light distribution of Firefly product, measured in lux on a vertical plane.

Figure 43: Radial light distribution of Aishwarya product, measured in lux on a horizontal plane.

06 12 1824

3036

4248

5460

667278849096102

108114

120126

132138

144150

156162168174180

186192198204210

216222

228234

240246

252258264270276282288294

300306

312318

324330

336342348354

06 121824

3036

4248

5460

667278849096102

108114

120126

132138

144150

156162168174180

186192198204210

216222

228234

240246

252258264270276282288294

300306

312318

324330

336342348354

Page 160: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

143

Figure 44: Radial light distribution of Solux product, measured in lux on a horizontal plane.

Figure 45: Radial light distribution of Kiran product, measured in lux on a vertical plane.

06 12 18 24

3036

4248

5460

667278849096102

108114

120126

132138

144150

156162168174180

186192198204210

216222

228234

240246

252258264270276282288294

300306

312318

324330

336342348354

06 12 18 24

3036

4248

5460

667278

84

90

96

102108

114120

126132

138144

150156162168174

180186192198204

210216

222228

234240

246252

258

264

270

276

282288

294300

306312

318324

330336342348354

Page 161: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

144

Figure 46: Radial light distribution of the standard lamp, measured in lux on a vertical plane.

06 12 18 24

3036

4248

5460

667278849096102

108114

120126

132138

144150

156162168174180

186192198204210

216222

228234

240246

252258264270276282288294

300306

312318

324330

336342348354

Page 162: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

145

APPENDIX D: INTEGRATING SPHERE PROCEDURE

Page 163: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

146

Each day that the integrating sphere was used to measure luminous flux, it was calibrated

according to the known light output of the NIST traceable standard lamp. The following

steps were followed to perform the sphere calibration.

1. Turn on the spectroradiometer and the standard lamp power supply.

2. Insert the standard lamp in the socket connected to the power supply.

a. Use cotton gloves so no oils are left on the optics.

3. Use the bulb‟s element as a reference for location in the center of the sphere and for

maintaining the same rotational orientation (for improved repeatability).

4. Slowly increase the standard lamp power supply to 220V AC (to reduce thermal

shock).

5. Allow at least 20 min. for the lamp to reach steady state temperature and light

output.

6. Close the spectroradiometer aperture.

7. “Zero” the spectroradiometer using the accompanying computer software.

8. Open the aperture to the largest setting.

9. Using the software, calibrate the sphere system according to the standard lamp‟s

rated lumen output.

10. Save the standard lamp measurement file to the computer for subsequent

calculations of test lamp luminous flux.

11. Slowly decrease the standard lamp power supply to 0 V to minimize thermal shock.

Each round of testing requires that the test lamp luminous flux measurement be corrected

for self absorption. The following steps were followed at the beginning of each round of

testing to determine the self absorption correction factor that was applied to each

luminous flux measurement.

1. Connect the auxiliary lamp to a power supply and current shunt circuit.

2. Slowly increase the power supply voltage until 0.84715 A is reached to reduce

thermal shock.

3. Allow at least 20 min. for the auxiliary lamp to reach steady state.

4. Using the computer software, measure the luminous flux of the auxiliary lamp with

the unilluminated standard lamp in the sphere.

Page 164: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

147

5. Save the measurement file on the computer for use in the self absorption correction.

6. Remove the standard lamp from the sphere.

7. Mount the unilluminated test lamp in the center of the sphere.

a. Hang using thin white cord.

b. Orient so that the light source is even with the middle of the baffle.

c. For lamps with focused light distribution, the beam is directed to the

bottom of the sphere.

8. Using the software, measure the luminous flux of the auxiliary lamp.

9. Save the measurement file on the computer for use in the self absorption correction.

10. Slowly decrease the auxiliary lamp‟s power supply voltage to zero to reduce thermal

shock.

Luminous flux of the test lamps at each light output level was measured and corrected for

self absorption and light distribution mismatch according to the following steps.

8. Connect the test lamp to a power supply and current shunt circuit

a. Use paired, white insulated wire

9. Adjust the power supply to the specified current for test lamp.

10. Allow the lamp to reach steady state according to CIE-84.

11. Using the software, measure the luminous flux of the test lamp.

12. Save the measurement file to the computer.

13. Using the software, select the appropriate files for self-absorption correction and the

luminous flux calculation.

a. Test lamp auxiliary lamp measurement file

b. Standard lamp auxiliary lamp measurement file

c. Test lamp measurement file

14. Save the light intensity vs. wavelength plot

15. Apply the light distribution mismatch correction to determine the luminous flux of

the test lamp.

Page 165: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

148

APPENDIX E: RAW TEST DATA

Page 166: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

149

Table 11: Raw data from comparison testing of the Firefly product in the integrating sphere, box-

photometer and tube-photometer.

Page 167: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

150

Table 12: Raw data from comparison testing of the Kiran product in the integrating sphere, box-

photometer and tube-photometer.

Page 168: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

151

Table 13: Raw data from comparison testing of the Solux product in the integrating sphere, box-

photometer and tube-photometer.

Page 169: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

152

Table 14: Raw data from comparison testing of the Aishwarya product in the integrating sphere,

box-photometer and tube-photometer.

Page 170: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

153

APPENDIX F: RELATIVE LUMEN DECREASE PLOTS

Page 171: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

154

Figure 47: Plots of measured relative luminous flux decrease for Firefly for three rounds of testing.

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10 11 12Pe

rce

nt

of

init

ial l

igh

t o

utp

ut

Measurement Number

Firefly - Round 1

Sphere

Box - Back

Box - Lid

Tube

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Firefly - Round 2

Sphere

Box - Right Wall

Box - Baffle

Tube

65%

70%

75%

80%

85%

90%

95%

100%

0 1 2 3 4 5 6 7 8 9 10Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Firefly - Round 3

Sphere

Box - Lid

Box - Front Wall

Tube

Page 172: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

155

Figure 48: Plots of measured relative luminous flux decrease for Kiran for three rounds of testing.

30%

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Kiran - Round 1

Sphere

Box - PV Front

Box - PV Back

Tube

30%

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Kiran - Round 2

Sphere

Box - PV Right

Box - PV Left

Tube

60%

65%

70%

75%

80%

85%

90%

95%

100%

0 1 2 3 4 5 6 7 8 9 10Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Kiran - Round 3

Sphere

PV - Front

PV - Back

Tube

Page 173: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

156

Figure 49: Plots of measured relative luminous flux decrease for Solux for three rounds of testing.

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Solux- Round 1

Sphere

Box - Lid

Box - Back

Tube

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Solux- Round 2

Sphere

Box - Right

Box - Baffle

Tube

30%

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Solux- Round 3

Sphere

Box - Lid

Box - Front

Tube

Page 174: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

157

Figure 50: Plots of measured relative luminous flux decrease for Aishwarya for three rounds of

testing.

40%

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Aishwarya - Round 1

Sphere

Box - Switch Back

Box - Switch Front

Tube

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10

Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Aishwarya - Round 2

Sphere

Box - Switch Right

Box - Switch Left

Tube

50%

60%

70%

80%

90%

100%

0 1 2 3 4 5 6 7 8 9 10Pe

rce

nt

of

Init

ial L

igh

t O

utp

ut

Measurement Number

Aishwarya - Round 3

Sphere

Box - Switch Front

Box - Switch Back

Tube

Page 175: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

158

APPENDIX G: MAXIMUM ERROR IN RELATIVE LUMEN DECREASE

Page 176: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

159

Table 15: Maximum error in the calculation of relative decrease in the initial light output according to light source, round of testing, testing

apparatus, and orientation within the box-photometer (where applicable).

Page 177: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

160

APPENDIX H: CALIBRATION PLOTS

Page 178: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

161

Figure 51: Comparison of integrating sphere lumen measurement to alternative method (tube- and box-photometer) lux measurements for

each round of Firefly Simulated Lumen Maintenance testing.

y = 0.653x - 0.3847R² = 0.9994

y = 0.5732x + 0.2106R² = 0.9836

y = 0.8681x - 0.3155R² = 0.9989

y = 0.6331x - 0.0007R² = 0.9993

y = 0.5322x - 0.0575R² = 0.9994

y = 0.8572x - 0.0662R² = 0.9995

y = 0.5788x - 2.0243R² = 0.9805

y = 0.5277x - 0.1712R² = 0.9836

y = 0.7883x - 0.0709R² = 0.9981

0

2

4

6

8

10

12

14

16

18

20

22

24

26

28

30

0 5 10 15 20 25 30 35 40 45 50 55

Sph

ere

me

asu

rem

en

t (L

um

en

)

Alternative method measurement (Lux)

Firefly Calibration Plots Box-1

Box-2

Tube-1

Box-3

Box-4

Tube-2

Box-5

Box-6

Tube-3

Page 179: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

162

Figure 52: Comparison of integrating sphere lumen measurement to alternative method (tube- and box-photometer) lux measurements for

each round of Kiran Simulated Lumen Maintenance testing.

y = 0.6562x + 0.4786R² = 0.9993

y = 0.6583x + 0.4211R² = 0.9993

y = 2.2426x + 0.4398R² = 0.9981

y = 0.6518x + 0.0585R² = 0.9994

y = 0.6485x + 0.0693R² = 0.9994

y = 2.2787x + 0.0199R² = 0.9993

y = 0.6417x + 0.3793R² = 0.9983

y = 0.6428x + 0.416R² = 0.9987

y = 2.2826x + 0.0685R² = 0.9985

0

2

4

6

8

10

12

14

16

0 2 4 6 8 10 12 14 16 18 20 22 24

Sph

ere

me

asu

rem

en

t (L

um

en

)

Alternative method measurement (Lux)

Kiran Calibration Plots Box-1

Box-2

Tube-1

Box-3

Box-4

Tube-2

Box-5

Box-6

Tube-3

Page 180: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

163

Figure 53: Comparison of integrating sphere lumen measurement to alternative method (tube- and box-photometer) lux measurements for

each round of Solux Simulated Lumen Maintenance testing.

y = 0.5193x - 0.4622R² = 0.9999

y = 0.5752x - 0.5666R² = 0.9999

y = 0.6809x + 1.7985R² = 0.9997

y = 0.5492x + 2.2295R² = 0.9952

y = 0.4732x + 0.4222R² = 0.9998

y = 0.6779x + 0.907R² = 0.9989

y = 0.5161x + 0.2624R² = 0.9999

y = 0.5404x - 1.2725R² = 0.9956

y = 0.7006x + 1.7014R² = 0.99780

10

20

30

40

50

60

70

80

0 20 40 60 80 100 120 140 160

Sph

ere

me

asu

rem

en

t (L

um

en

)

Alternative method measurement (Lux)

Solux Calibration Plots Box-1

Box-2

Tube-1

Box-3

Box-4

Tube-2

Box-5

Box-6

Tube-3

Page 181: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

164

Figure 54: Comparison of integrating sphere lumen measurement to alternative method (tube- and box-photometer) lux measurements for

each round of Aishwarya Simulated Lumen Maintenance testing.

y = 0.578x - 0.6286R² = 0.9999

y = 0.5793x - 0.7289R² = 1

y = 3.85x + 0.2008R² = 0.9957

y = 0.5714x + 0.0282R² = 0.9998

y = 0.5748x + 0.0503R² = 0.9998

y = 3.7935x + 0.5366R² = 0.9993

y = 0.5914x - 0.7922R² = 0.9999

y = 0.5901x - 0.7166R² = 0.9998

y = 4.0358x - 1.3748R² = 0.99660

10

20

30

40

50

60

70

0 10 20 30 40 50 60 70 80 90 100 110 120

Sph

ere

me

asu

rem

en

t (L

um

en

)

Alternative method measurement (Lux)

Aishwarya Calibration Plots Box-1

Box-2

Tube-1

Box-3

Box-4

Tube-2

Box-5

Box-6

Tube-3

Page 182: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

165

APPENDIX I: R2 VALUES AND STANDARD ERROR OF TEST RESULTS

Page 183: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

166

Table 16: R-squared values and standard error of the linear regression for each round of Simulated Lumen

Maintenance testing, including a calculation of the standard error as a percentage of the average luminous

flux estimated by the linear regression model (SE/ ).

Page 184: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

167

Page 185: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

168

APPENDIX J: BOX-PHOTOMETER CONSTRUCTION PLANS AND

INSTRUCTIONS

Page 186: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

169

Page 187: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

170

Page 188: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

171

Page 189: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

172

Page 190: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

173

Page 191: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

174

ASSEMBLY INSTRUCTIONS FOR BOX-PHOTOMETER

1. Cut the “MAIN PIECES FOR ASSEMBLY” to the dimensions shown on

PHOTOMETER BOX PLANS – Use a table saw if available

2. Pre-drill pilot holes on 65cm x 60cm wall pieces

3. Apply glue along 2,5cm x 60cm area of wall pieces with pilot holes

4. Use four clamps (one at top, one at bottom for each side)to hold the four walls

together as shown on in the top view of PHOTOMETER BOX PLANS

5. Drive finish nails into pre-drilled pilot holes

6. Allow 12 hours for glue to cure

7. Remove clamps

8. Check butt joints for structural integrity

9. Apply glue to bottom edges of walls

10. Align and clamp bottom piece to walls

11. Allow 12 hours for glue to cure

12. Remove clamps

13. Check for structural integrity

14. Cut 4cm x 4cm feet (as shown in front view of PHOTOMETER BOX PLANS)

and glue to four exterior corners of bottom piece. The weight of the photometer

box will hold the feet in place while the glue cures

15. Apply glue to 2,5cm width of “TOP RIM” pieces along top outside perimeter of

walls

16. Place “TOP RIM” pieces on outside walls as shown in PHOTOMETER BOX

PLANS and clamp in place

17. Allow 12 hours for glue to cure

18. Remove clamps

19. Check for structural integrity

20. Drill hole in corner of photometer box for placement of PVC section – Use a hole

saw if available

Page 192: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

175

21. Cut PVC pipe to appropriate length and miter cut to tightly fit against hole in

photometer box

22. Affix PVC section to photometer box with caulk sealant, making sure to seal

against all possible light intrusion at joint

23. Assemble test stand and attach with wood screws to center bottom of photometer

box as shown in top view

24. Cut plastic screens and drill holes for insertion of dowels

25. Drill holes at appropriate angles and locations for screen dowels, refer to top view

(interior)

26. Insert screen dowels into holes in photometer box walls. No glue should be

required

27. Place top lid piece onto photometer box

28. Align hinge as shown in top view (lid)

29. Pre-drill pilot holes and attach hinge with wood screws

30. Pre-drill hole for handle (as shown in top view) and attach to lid

31. Paint ALL interior surfaces of photometer box. Several coats are recommended

32. Refer to the Lighting Africa Quality Test Method, Section 3.2 for additional

details (Lighting Africa, 2010c). Available for download at:

http://lightingafrica.org/resources/technical-research.html

Page 193: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

176

APPENDIX K: TUBE-PHOTOMETER CONSTRUCTION PLANS

Page 194: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

177

ASSEMBLY INSTRUCTIONS FOR TUBE-PHOTOMETER

1. Cut the tube to a length of 0.5 m

2. Cut a piece of cardboard into a square with sides that are at least twice as long as

the tube diameter

3. Create the end cap by cutting a circle out of the center of the cardboard square

that is of the same diameter as the photometer sensor.

a. It is important that the sensor fit snugly into the hole so that no light

intrusion occurs and the sensor remains in a fixed position.

b. If more advanced tools and materials are available, a more robust and rigid

end cap can be constructed of wood or other equivalent materials.

4. Fit the end cap to one end of the tube.

a. The cardboard square is placed such that the hole is located at the center of

the tube opening. The cap is formed by bending the cardboard over the

end of the tube.

5. Using packaging tape, duct tape, or an equivalent method, affix the end cap to the

end of the tube.

a. The end cap must be solidly mated to the tube such that it does not shift

and no light can pass between the cap and the tube.

6. Insert the photometer sensor into the end cap and affix with packaging tape, duct

tape, or the equivalent.

a. The sensor must be solidly mated to the end cap such that it does not shift

and no light can pass between the sensor and the cap.

b. Check that no unwanted light is intruding into the tube by blocking the

open end of the tube and reading the photometer measurement. The

photometer must read 0 lx.

Page 195: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

178

APPENDIX L: LIST OF EQUIPMENT USED IN THE STUDY

Page 196: ANALYSIS OF LOW-COST TESTING METHODS FOR LED LUMEN

179

Test Apparatus

Equipment Name

Manufacturer / Model Specifications

Box-photometer

Thermocouple Reader

Omega HH806AU 0.1oC precision

Thermocouple Omega K-type, insulated

Lux Meter Extech 401036

Datalogging, precision: 0.01 - 0.1lx (depending on range)

Box Custom fabricated* Refer to Appendix J

Tube-photometer

Thermocouple Reader

Omega HH806AU 0.1oC precision

Thermocouple Omega K-type, insulated

Lux Meter Extech 401036

Datalogging, precision: 0.01 - 0.1lx (depending on range)

Tube w/ end cap Custom fabricated** 0.5m long, 6cm diameter

Integrating Sphere

Thermocouple Reader

Omega HH806AU 0.1oC precision

Thermocouple Omega K-type, insulated

Spectrophoto-meter

Beijing Optical Instrument Factory, WDM1-1

Optical bandwidth: 350 - 780 nm; resolution: 5nm

Software for integrating sphere luminous flux calculation

NLTC, proprietary: Electrical Light Source Spectral Test and Color Analysis System

Calculates luminous flux of DuT based on spectrophotometer reading, standard lamp and auxiliary lamp lux measurements

DC Power Supply DH1719-1 1 mA precision

Standard Lamp OSRAM

15.4W incandescent, NIST traceable

Auxiliary Lamp OSRAM 10W halogen

Sphere 1.3m diameter -

LED Driver

Power Supply DH1719-1 1 mA precision

Digital Multimeter

Fluke 179 Precision: 0.01 mA, 0.1 mV

Digital Multimeter

Solartron Schlumberger 7150

0.001 mV precision

Resistor 0.01 High precision

Goniophotometer NLTC, custom -

*Refer to APPENDIX J for a detailed description of the Box-photometer assembly.

**Refer to APPENDIX K for a detailed description of the Tube-photometer assembly.