Technical University – Sofia Bulgaria

Texas Instruments European Analog Design Contest 2013 Project Report

Atmospheric electrostatic

field acquisition system with

charge sensing Team Leader: Tsvetan Marinov ts_marinov@tu-sofia.bg

Team Members: Team Petya Popova p_popova@tu-sofia.bg

Advising Professor: Mityo Mitev mitev@ecad.tu-sofia.bg

University: Technical University of Sofia, Bulgaria

Date: 31.07.2013

Qty. TI Part Number & URL Qty. TI Part Number & URL

1 IVC102 1 DRV8834

1 OPA369 1 REG101-5

1 ADS8519 1 TPS7A4901

1 MSP430FR5739 1 TPS7A3001

1 LP38690

Project abstract: In this project а complete system for atmospheric electrostatic field monitoring and logging was developed. “Field Mill” type of sensor was chosen in combination with charge sensing amplifier. The system was developed as self sustained and autonomous, capable of intelligent power management and continues data acquisition. Also live data is transmitted via wireless connection, with capability of complete device control. Working prototype was developed, proving the overall concept and practicality, also recording real life data from thunderstorms.

Introduction

The motivation for this project came from the fascination with nature. To be able to predict the weather gives us a great advantage. Not only it aids our everyday lives, it helps us plan and execute different activities and even a rough estimation could mean saving a lot of lives in different exploratory missions and conducting things that we take for granted as bridge maintenance and tall building maintenance for example. The device not only detects nearby lightning, but can detect the atmospheric conditions which precede lightning. This means that ELFIS (Electrostatic field acquisition system) spots trends in weather which could help the prediction of approaching storms. Devices of this type are often exclusively used in weather stations, mainly because of their price, the complexity of their software and their big dimensions. The device presented in this project is compact and can be easily installed in any household. This makes it valuable for agricultural uses and secluded households.

Design Overview

Block diagram of the electrostatic field acquisition system is shown on Fig.1 The building blocks of the device are “field mill” sensor, charge sensing, power supply, motor control, data acquisition + RF link. The following chapters will discuss design decisions for each block.

Field Mill Sensor

The first step of the development was the “field mill” sensor [Fig. 2]. It consists of a

base plate – “ground”, sensing electrodes, rotating shield, acting as a shutter, and a

position sensor. The shutter periodically exposes and shields the electrodes from the

applied electric field. This field “chopping” technique allows compensating leakage,

charge injection from switches and other unwanted effects of real life amplifiers. For

the position sensor, a Hall Effect magnetic sensor with combination of magnets was

Fig. 1 – Block diagram of the system

chosen, glued to the shutter. This way the microcontroller knows if the electrodes are

being shielded or exposed to the field. The physical dimensions, construction and

shape of the “field mill” sensor, were chosen in favor of easy manufacturing and

small case dimensions. For the case for the system (also base plate for the “field mill”

sensor) was chosen an of-the-shelf water tight (IP65) die cast aluminum case.

In order to calculate the total amount of charge produced by the sensor, assumption

that the electric field is homogenous in every part of the construction was made. The

sensitivity of the sensor can be expressed by the following equation:

( )

( )

[

⁄]

Where: 6 is the number of electrodes; S is the surface area of one electrode; ε0 is the

vacuum permittivity; d is the distance between the electrodes and the base; E is

applied electric field (V/m). This calculated coefficient will be used later for amplifier

gain calculation.

Charge sensing

The charge sensing is based on TI’s IVC102 transimpedance amplifier. The IVC102

was chosen because it provides more precise, highly integrated alternative to

conventional transimpedance op amp circuits and its output voltage can be held for

accurate measurement using the internal analog switches. The charge sensing is

performed in 2 stages, each stage consists of 2 measurements and each

measurement is based on 4 ADC samples. The first stage is with the electrodes

exposed to the field, the second stage is with the electrodes shielded from the field.

First measurement is the offset of the amplifier; second measurement is the end of

integration voltage. Four samples per measurement were made in order to reduce

the error caused by noise. The result from each stage is the resulted voltage from

the end of integration cycle minus the offset voltage. Electric field strength is the

result from stage one minus stage two. This way leakage is being compensated, as

a. b. c.

Fig. 2: a – fully assembled sensor; b – 1-base plate, 2-sensor electrodes, 3-hall

effect sensor; c – 1-rotating shield (shutter), 2-magnets

2

1

3

4 1 2

well as charge injection and offset of the operational amplifier. In order to

compensate errors from mechanical tolerances, the field strength is calculated using

the average of all shutter positions in one complete turn of the shield. The high input

impedance of the charge sensing amplifier brings 50Hz interference from the mains.

In order to reject these interferences integration period of 40ms was chosen which is

equal to 2 periods of the mains frequency. The described procedure was simulated

using TINA-TI [Fig. 3]. Fig.4 shows result from transient analysis. It is clear that the

output voltage is shifted because of leakage. The expected amount of charge for this

two integration cycles is the same, but with opposite signs. The leakage on the other

hand is always in the same direction. Therefore subtracting the first measurement

from the second and dividing it by 2 will return the actual value, canceling the

leakage.

Fig. 3: schematic for simulation of IVC102 with TINA-TI

V+

V-

V-

V+

Vout

+ S1

C1

10

0p

+

VS2 -12

+

VS1 12

+ S2

E_FIELDHUM

Agnd DgndV+

V-

Out

S1

S2

Iin

C1

C2

C3

-In

U1 IVC102M

Field Mill sensor is represented by 2 current generators:

HUM - interference from the mains

E_FIELD - chopped electrostatic field

Fig. 4: results from transient analysis of IVC10 with TINA-TI

-3.87V

1.25mV

4.49V

1.21mV

Time (s)

0.00 50.00m 100.00m 150.00m 200.00m

E_FIELD

-20.00n

20.00n

HUM

-1.00n

1.00n

S1

0.00

4.00

S2

0.00

4.00

Vout

-5.00

5.00

1.21mV

4.49V

1.25mV

-3.87V

The ADC is configured for +/-5V input range and a +/-30kV/m range is desirable. Because of that the integration capacitor for IVC102 (setting the gain) must be:

Initially the internal capacitor was configured for 100pF, but the real transfer coefficient of the sensor was lower and configuration with 40pF integration capacitor was used. While designing the PCB [Fig. 5] the following considerations were taken into account: ensuring as low as possible leakage from other traces, ensuring no ground loop and the absence of all other possible ground problems. In order to lower the leakage a guard ring was made, around the signal trace, in every part of the PCB. The guard ring is connected to the non-inverting input of IVC102, which is Analog Ground. In order to avoid ground problems, single point grounding was chosen for Analog Ground. Digital Ground and Analog Ground are connected together near the Analog to Digital Converter. The ADC itself is ADS8519. The ADS8519 was chosen because of its bipolar input ranges, precision internal resistors, high resolution, low error and high sample rate. The design is also provided with header for external reference, but the internal reference was used.

Power supply

The power supply should output 3.3V for all digital ICs, 5V for the ADC and +/-12V for the IVC102, from 2 li-ion cells. The first prototype of the system was built on universal prototyping boards and an of-the-shelf +/-12V switching convertor was being used. This DC-DC convertor was emitting a lot of switching noise and filtering proved to be difficult. In order to eliminate this problem in the final design a decision was made to design a custom low noise DC-DC convertor. The decision was to implement resonant Royer convertor. The pros of this convertor are high efficiency, simple hardware design, low part count and very narrow spectrum. The convertor works with sinusoidal wave form, ideally only one harmonic is present. Full schematic of the converter with transient analysis is shown on Fig. 6 and Fig. 7.

a. b.

Fig. 5: Charge sensing PCB, a – front side, b – back side

The resonance frequency is tuned to ~37kHz. In order to achieve desired inductance and turn ratio, the transformer was wounded by the team. On the secondary side, after the rectifier, the frequency is double ~74kHz. This relatively high frequency, in combination with low current consumption (around 5mA), is suitable for filtering using RC low pass filter. A triple RC filter was designed, tuned with cut frequency one decade lower than the frequency after the rectifier. This gives 60db suppression in total, with only 70mV losses. As this convertor doesn’t offer any regulation, TI’s TPS7A3001 and TPS7A4901 adjustable positive and negative linear regulators were added. These linear regulators were chosen for their high power supply rejection ratio, low quiescent current and especially for their wide bandwidth, which is not typically the case with linear regulators. The 5V rail is directly powered from the battery using TI’s REG101 low-dropout regulator. This regulator was chosen for its low noise, low quiescent current and high input voltage. The 3.3V rail is also powered from the battery and regulated using TI’s LP38690 low dropout regulator, with 1A output current and TO-252 package this regulator is more than capable to supply all digital circuits and radio peak power consumption with plenty of thermal and current reserve. All regulators have very low quiescent currents, which improves overall efficiency, especially at light loads. The PCB for the PSU [Fig. 8]

Fig. 6: Royer resonant DC-DC convertor

R_

LO

AD

_2

5k

R_

LO

AD

_1

5k

SE

C_

FIL

T

SE

C

PRIM_1 PRIM_2

C7

4,7

u

R11 4,7

C6

4,7

u

R10 4,7

C5

4,7

u

R9 4,7

C4

4,7

u

R8 4,7

C3

4,7

u

R7 4,7

C2

4,7

u

R6 4,7

L5 800u

SD

6 M

BR

11

00

SD

5 M

BR

11

00

SD

4 M

BR

11

00

SD

3 M

BR

11

00+

VS1 8,4

R5 8,2

R4

10

0k

R3

1k

R2

1k

R1

10

0k

SD2 1N5819

T2 BC337-40T1 BC337-40

SD1 1N5819

C1 20n

L4 470u

L3 800u

L2 220u L1 221u

K

K1 1

Controlling inductors L1,L2,L3,L5

Fig. 7: Royer transient analysis.

Time (s)

38.30m 38.40m 38.50m 38.60m

PRIM

-30.00

30.00

SEC

22.00

23.00

SEC_FILT

22.00

23.00

was designed to accommodate all regulators and the Royer convertor. Also battery voltage sense was added using 0.1% 1:4 resistor divider. The divided voltage was buffered using TI’s OPA369 and fed to the microcontroller internal ADC. An important role for choosing OPA369 was its very low supply current and its rail to rail inputs and output, which are very important for battery powered devices. Magnetic shield was soldered on top of the PCB in order to shield the rest of the device from the dissipated magnetic field from the transformer (not present on the photos).

Motor control

For propulsion of the shutter a stepper motor was chosen. The motor is used as synchronous AC motor rather than to hold stationary position, which is what stepper motors are typically used for. The stepper motor driver is used as inverter with variable frequency (rotation speed). For the stepper motor driver TI’s DRV8834 was used, because of its integrated power transistors, 1/32 step microstepping and integrated current control. The rotation speed and the motor drive current are controlled from the microcontroller. As this is switch mode IC, it is a source of interference. In order to reduce the amount of noise from the motor driver, heavy decoupling was provided [Fig. 9].

a. b.

Fig. 8: Power supply PCB, a – front side, b – back side

a. b.

Fig. 9: Motor driver PCB, a – front side, b – back side

The DRV8834 is configured for 32 micro stepping, this way the motor works with pseudo sinusoidal waveform, ensuring smooth rotation.

Data acquisition and RF communication

MSP430FR5739 was chosen. Its low power, high speed and high integrated peripheral count make it ideal for controlling the whole system and handling data logging and communications. The PCB [Fig. 10] accommodates the microcontroller itself, header for the Hall effect position sensor, watch crystal for real time clock, CR2032 batter backup for the clock function, header for radio module, programming and serial communication, EEPROM for data logging and microSD card slot also for data logging. The EEPROM chip is Cypress FM25V10, 1Mb with SPI interface, capable of holding up to 17 hours of log data. The radio module is based on Nordic Semiconductor nRF24L01+ transceiver chip. The IC uses 2.4GHz ISM band and has internal packet handling. The radio module also has additional PA and LNA amplifiers, greatly improving communication range.

Software

Embedded and Windows based software were developed for this project. The

embedded part was developed using IAR Embedded Workbench for MSP430 and

written in C programming language. The current version of the embedded software is

performing the following tasks: power management of peripheral ICs and PSU,

speed control for the synchronous motor, external ADC control and sample

acquisition, shutter position detection, control for the analog switches in the

transimpedance amplifier, data processing and data logging (on external EEPROM),

serial port handling and command processing, radio packet handling, automatic

radio channel selection, command processing from radio. microSDcard and FAT file

system will be supported in later versions of the embedded software. The Windows

based software was developed using Matlab, its purpose is visualization of live data

stream from the device via the radio [Fig. 11].

a. b.

Fig. 10: Microcontroller PCB

a – front side and nRF24L01+ radio module; b – back side

Experimental data

Due to the nature of these types of measurements, gathering data is difficult and

depends heavily on weather conditions. Stormy weather conditions were available

during the development of the first prototype. Recorded data [Fig. 12] is from

10/06/2013. Record started at 16:45 at location +42° 39’ 48.90”, +23° 22’ 55.17’’. On

the chart X axis is time offset in seconds and on the Y axis is Electrostatic field

strength in V/m. Clouds are usually positively charged on the top and negatively

charged at the bottom. As the cloud approaches the sensor location the field is with

positive sign, this field can be represented with force lines coming from the top of the

cloud and closing to the earth. After 600 seconds the field became negative, when

the cloud approached the sensor location. The field strength is also higher. Thunder

strikes can be observed on the record as sudden drop in the field strength, followed

by slow gaining of field strength, multiple drops are multiple strikes near each other.

Fig. 11: Matlab based GUI for live data visualization.

Fig. 12: Record of thunderstorm in Sofia

-14000

-12000

-10000

-8000

-6000

-4000

-2000

0

2000

4000

0 100 200 300 400 500 600 700 800 900 1000 1100

Ele

ctri

c Fi

eld

[V

/m]

Seconds [s]

10/06/2013 Thunderstorm - Sofia

Future plans

For future development we will implement microSD card support with FAT file

system. We also wish to add additional meteorological sensors for temperature,

humidity, wind speed and direction, amount of rain fall etc. Another idea is to

implement solar power. The idea is to develop another power supply module

incorporating solar panel power management and charging circuit for li-on battery.

Conclusion

A complete system for measurement and acquisition of atmospheric electrostatic

field was developed in this project. The design was implemented on a completely

modular basis, which allows easy maintenance and upgradability [Fig. 13, 14]. Newly

developed modules can be installed without redesigning the whole system. All major

goals are achieved. The charge sensing method for measuring electric field, proved

to be appropriate and overall the performance of the system was good. The recorded

real life data proved the practicality of the system. During the development important

analog design experience was gained by the team.

a. b.

Fig. 13: Partially assembled device; a – only interconnection board and electric

motor; b – fully populated board, all modules installed

Fig. 14: Partially disassembled device, case

is opened and battery is disconnected