introduction to mimo ota environment simulation

Introduction to MIMO OTA Environment

Simulation, Calibration, Validation, and

Measurement Results

Dr. Michael D. Foegelle

Director of Technology Development

Garth D’Abreu

Director of RF Engineering

Outline

The Meaning of MIMO

MIMO and the RF Environment

The Channel Emulator

Understanding the Channel Model

Implications of OTA Testing

Spatial Environment Simulation

ETS-Lindgren AMS-8700

2

Outline

Calibrating the System

System Validation

Wi-Fi and LTE Throughput Measurement Results

with the ETS-Lindgren AMS-8700 OTA

Environment Simulator

Metrics

Reference MIMO Antennas

3

MIMO stands for Multiple Input, Multiple Output and refers to the characteristics of the communication channel(s) between two devices.

In communication theory, a channel is the path by which the data gets from an input (transmitter) to an output (receiver).

For Ethernet or USB, the channel is the cable used.

For wireless, the channel includes the RF frequency bandwidth, the space between antennas, and anything that reflects RF energy from one point to the other.

Often includes antennas and cables too.

The Meaning of MIMO

4

The term MIMO is often used to represent a range of bandwidth/performance enhancing technologies that rely on multiple antennas in a wireless device.

These can be classified into several categories: “True” Spatial Multiplexing MIMO.

SIMO (Single Input, Multiple Output) technologies like beam forming and receive diversity.

While this discussion will concentrate on downlink MIMO, uplink MIMO/MISO concepts are similar.

The Meaning of MIMO

5

“True” MIMO uses multiple transmit and receive

antennas to increase the total information

bandwidth through time-space coding.

Multiple channels of communication (streams) share the

same frequency bandwidth allocation simultaneously.

The Meaning of MIMO

6

MIMOTransmitter

MIMOReceiver

1101

0010 1010 1001 0000 1100 1110 0110 0

01 101 1001 0 1 00 111 01

1

11101 111 0

1 00

SIMO technologies use the multiple (receive) antennas to improve single channel performance under edge-of-link (EOL) conditions.

Beam forming allows creating a stronger gain pattern in the direction of the desired signal while simultaneously rejecting undesired signals from other directions.

The Meaning of MIMO

7

Null (Low Gain)oriented in direction of interfering signal

DesiredSignal PathUnwanted

Interferer

Main Lobe(Highest Gain)

oriented in desiredcommunication direction

Receive diversity uses multiple antennas to overcome channel fades by using additional antenna(s) to capture additional information that may be missing from the first channel.

Includes simple switching diversity or more complicated techniques like maximal ratio combining or other combinatorial diversity techniques.

The Meaning of MIMO

8 1905 19101906 1907 1908 1909

-140

-80

-130

-120

-110

-100

-90

1905 19101906 1907 1908 1909

-140

-80

-130

-120

-110

-100

-90

1905 19101906 1907 1908 1909

-140

-80

-130

-120

-110

-100

-90

All of these multiple antenna technologies share one thing in common – their performance is a function of the environment in which they’re used.

The device adapts to its environment through embedded algorithms that change its (effective) radiation pattern.


9

Radiated Performance

Po

we

r (

dB

m)

-85

-55

-80

-75

-70

-65

-60

Y

Z

X

Azimuth = 104.6

Elevation = -25.1

Roll = -51.5


Po

we

r (

dB

m)

-80

-55

-75

-70

-65

-60

Y

Z

X

Azimuth = 104.6

Elevation = -25.1

Roll = -51.5


Po

we

r (

dB

m)

-85

-55

-80

-75

-70

-65

-60

Y

Z

X

Azimuth = 104.6

Elevation = -25.1

Roll = -51.5

Traditional TRP and TIS metrics are properties

of the mobile device only. These represent

the average performance of the device to

signals from any direction.


10

Y

Z

X

TRP

Metrics like Near Horizon Partial Radiated

Power/Sensitivity terms or Mean Effective

Gain apply simple environmental models to

fixed pattern data, but the basic behavior of

the device does not change.


11

Y

Z

X

TRP

26.3 dBm

Y

Z

XY

Z

X

NHPRP +/-45° (Pi/4)

25.2 dBm

NHPRP +/-30° Pi/ 6)

23.7 dBm

(

For MIMO technologies, performance is a

function of the system and cannot be restricted to

the mobile device.

Individual device performance can only be

evaluated or compared in a given environment.

This implies the need for environment simulation.


12

Environment #1 Environment #2

A channel emulator is typically used for

conducted testing of MIMO radios.

The channel emulator simulates the wireless

channel between transmit and receive radios

using a channel model.

Channel models simulate not only a given

environment, but also properties of the base

station and mobile device including antenna

patterns, antenna separation, and angles of

departure/arrival (AOD/AOA).


13


A typical RF channel emulator consists of a

number of VSA receivers and VSG transmitters

connected to a DSP modeling core that

introduces delay spreads, fading, etc. at

baseband.

14

Receiver(VSA)

Transmitter(VSG)

DSPModeling

Core

Receiver(VSA)

Transmitter(VSG)

Receiver(VSA)

Transmitter(VSG)


The ideal channel emulator routes multiple inputs

to multiple outputs after applying appropriate

modeled delay spreads, fading, etc.

15

Transmitter 1

Receiver1

Transmitter2

Receiver2

ChannelEmulation


In the real world, various objects in the

environment cause reflections of the transmitted

signal that are seen at the receiver.

16

Reflecting Objectsin Environment

Propagation Ray Paths

Transmitter

Receiver


Signal paths are often classified as Line-of-Sight

(LOS) and Non-Line-of-Sight (NLOS).

17

Transmitter

Receiver

NLOS

LOS

NLOS


Each path has a different length (propagation

delay).

18

Transmitter

Receiver

B

A

C

D

A

B

C

D


Plotting the signal strength vs. time gives a

power delay profile (PDP) for the model.

19

A

B

C

D

Time

Sig

nal

Str

en

gth

Power Delay Profile

A

B

C

D


The individual times of arrival are called taps,

drawing from the concept of a tapped delay line.

20

Time

Sig

na

l S

tre

ng

th

Power Delay Profile

A

B

C

D

Taps


Reflecting objects typically don’t just cause one

reflected signal.

Instead, scatterers produce a cluster of

reflections with slightly different delays and

varied magnitude and phase.

21

Transmitter

Receiver

“Cluster” of Reflections


Each cluster produces its own unique statistical

PDP.

22

Transmitter

Receiver

“Cluster” of Reflections

Time

Sig

na

l S

tre

ng

th

Cluster PDP

Combining the cluster concept with the tap

concept produces realistic time domain profiles.


23

Time

Sig

na

l S

tre

ng

th

Power Delay Profile

A

B

C

D

Taps

Now the modeled data looks a lot like real

measured time domain data acquired using a

vector network analyzer.


24


Motion of the transmitter, receiver, or other

objects within the environment causes Doppler

shift of the frequency.

25

Transmitter

Receiver


Moving towards a wave increases its frequency,

while moving away decreases the frequency.

A moving radio results in Doppler

spread since it moves towards

some reflections and away from

others.

26

=+

=+

Spatial channel models include geometric

information about the location of scatterers, and

determine channel behavior based on angles of

departure and arrival (AOD/AOA) and the angular

spread for each cluster.


27

Spatial channel models for conducted testing

also apply assumed antenna patterns for the

source and receiver.


28

2x2 Channel Emulation

Two Transmitter

BSE

Two Receiver

Radio

Simulated TransmitAntenna Patterns

Simulated ReflectionClusters

Simulated ReceiveAntenna Patterns

A primary goal of OTA testing is to determine radio performance of the DUT with the actual antenna patterns, orientation, and spacing.

If this was all that was required, a combination of antenna pattern measurement and conducted channel modeling would suffice.

However, traditional OTA measurements of TRP/TIS perform simultaneous evaluation of the entire RF signal chain for a variety of reasons:


29

Platform Desensitization – interference from

platform components enters radio through

attached antennas.


30

GPS

WCDMA

Baseband

ProcessorWi-Fi

Backlight

Bluetooth

Antenna Another

Antenna

And Yet

Another

Antenna

One More

Antenna

Near Field Influences – including platform

structure, head, hands, body, table top, etc.


31

Mismatch and other Interaction Factors – performance of a radio into a matched 50 Ohm load may not be the same as that into a mismatched or detuned antenna, resulting in non-linear behavior.

Antenna-Antenna Interactions – mutual coupling of antennas may not be accounted for properly in pattern tests.

Cable Effects – currents on feed cables can alter the radiation pattern, especially for small DUTs.


32

MIMO relies on a complex multipath environment to provide the information necessary to reconstruct multiple source signals that have been combined into multiple receive signals.


33



MIMOTransmitter MIMO

Receiver


The goal of the OTA Environment Simulator is to

place the DUT in a controlled, isolated near field

environment and then simulate everything

outside that region.



MIMOTransmitter MIMO

Receiver

34






35






36


From inside the bubble, everything looks the

same, even though everything outside the bubble

is simulated.

37


Practical limitations may result in a low resolution

picture of the environment.

Using active spatial channel emulation provides

motion simulation, etc.

38


We may also only care about a portion of the

environment.

E.g. Most reflections cluster near the horizon.

39


For comparison, using a reverberation chamber

averages out the spatial picture. The same

(statistical) signal comes from all directions.

40


The alternate two antenna method proposed is a

less accurate representation of the real

environment.

41


The two stage method uses antenna pattern data

applied to a conducted channel emulation model.

42


Example: Typical Multi-Path Power Delay Profile

from a Real World Environment

43


Using a fully anechoic chamber to isolate the DUT, a

matrix of antennas arrayed around the DUT can be

used to produce different angles of arrival (AOA).

DUT

Path 1 (LOS)AOA = 0

Path 2 AOA ~135°

Path 3 AOA ~225°

Path 4 AOA ~45°

44

A spatial channel emulator (a channel emulator

with modified channel models) simulates the

desired external environment between BSE and

DUT.


45

DUT

MIMOTester

SpatialChannelEmulator

Evaluation of SIMO functions like beam forming

and receive diversity likely require only

rudimentary environment simulation.

Sufficient to simulate only basic directional effects and

spatial fading.

While there are a variety of simplistic ways to

create an external environment containing delay

spread, fading, and even repeatable reflection

“taps”, they may be insufficient for proper

evaluation of MIMO performance.


46

Spatial channel models include clusters of

scatterers with each tap having an angular spread

as well as a delay spread.


47

The angular spread of a given cluster is

simulated by feeding multiple antennas with an

appropriate statistical distribution of the source

signal.

Designing the OTA Environment Simulator

DUT

48

Converting a conducted channel model to an

OTA channel model:

Conducted model simulates TX and RX antennas.



Two Transmitter

BSE

Two Receiver

Radio

Simulated TransmitAntenna Patterns

Simulated ReflectionClusters

Simulated ReceiveAntenna Patterns

49

Conducted channel model:

Ray paths from reflections in simulated environment are

collected at each simulated receive antenna.



Two Transmitter

BSE

Two Receiver

Radio

Simulated Ray PathsBetween TX and RX Antennas

50


OTA channel model:

Clusters produced different angles of arrival (AOA)



Two Transmitter

BSE

Two Receiver

Radio

Directions of Received Signals(Angles of Arrival)

51


OTA channel model:

Grouping AOAs, we can remove virtual RX antennas.



Two Transmitter

BSE

Two Receiver

Radio

Region around Simulated DUT

52

OTA channel model:

2xN channel emulator used to feed N antennas for AOA

simulation around DUT with real antennas.


DUT withIntegrated DualReceivers and

Antennas

2xN Environment Simulation

Two Transmitter

BSE

53


Ideally, the sphere around the DUT would define a

perfect boundary condition that exactly reproduces

the desired field distribution inside the test region.

Practicality and physical limitations impose

restrictions that create a less than ideal environment

simulation.

The chosen number of antenna positions limits the

available range of “Real” propagation directions.

Splitting clusters across discrete antennas does not

produce true plane wave behavior in test region.

Results in an interference pattern with wave-like distribution in

center of test region.

54


Discretization results in an interference pattern with

wave-like distribution in center of test region.

Quality depends on angular spacing and number of antennas

used to create interference pattern.

Plane Wave Source at 15°

Y

(m)

X (m)

-0.002

0.002

-0.001

0

0.001

-0.125 0.125-0.05 0 0.05 0.1

-0.125

0.125

-0.1

-0.05

0

0.05

0.1

Simulated 15° Source using Plane Waves at 0° and 45°

Y

(m)

X (m)

-0.002

0.002

-0.001

0

0.001

-0.125 0.125-0.05 0 0.05 0.1

-0.125

0.125

-0.1

-0.05

0

0.05

0.1

55


Effect of angular resolution on simulated AOA.

Plane Wave Illumination with 10 cm Wavelength

Ele

ctr

ic F

ield

Y

(m)

X (m)

-0.002

0.002

-0.001

0

0.001

-0.125 0.125-0.1 -0.05 0 0.05 0.1

-0.125

0.125

-0.1

-0.075

-0.05

-0.025

0

0.025

0.05

0.075

0.1

56



Coherence Region for Antennas at 10° Spacing with 10 cm Wavelength

Ele

ctr

ic F

ield

Y

(m)

X (m)

-0.002

0.002

-0.001

0

0.001

-0.125 0.125-0.1 -0.05 0 0.05 0.1

-0.125

0.125

-0.1

-0.075

-0.05

-0.025

0

0.025

0.05

0.075

0.1

57




Ele

ctr

ic F

ield

Y

(m)

X (m)

-0.002

0.002

-0.001

0

0.001

-0.125 0.125-0.1 -0.05 0 0.05 0.1

-0.125

0.125

-0.1

-0.075

-0.05

-0.025

0

0.025

0.05

0.075

0.1

58




Ele

ctr

ic F

ield

Y

(m)

X (m)

-0.002

0.002

-0.001

0

0.001

-0.125 0.125-0.1 -0.05 0 0.05 0.1

-0.125

0.125

-0.1

-0.075

-0.05

-0.025

0

0.025

0.05

0.075

0.1

59




Ele

ctr

ic F

ield

Y

(m)

X (m)

-0.002

0.002

-0.001

0

0.001

-0.125 0.125-0.1 -0.05 0 0.05 0.1

-0.125

0.125

-0.1

-0.075

-0.05

-0.025

0

0.025

0.05

0.075

0.1

60




Ele

ctr

ic F

ield

Y

(m)

X (m)

-0.002

0.002

-0.001

0

0.001

-0.125 0.125-0.1 -0.05 0 0.05 0.1

-0.125

0.125

-0.1

-0.075

-0.05

-0.025

0

0.025

0.05

0.075

0.1

61




Ele

ctr

ic F

ield

Y

(m)

X (m)

-0.002

0.002

-0.001

0

0.001

-0.125 0.125-0.1 -0.05 0 0.05 0.1

-0.125

0.125

-0.1

-0.075

-0.05

-0.025

0

0.025

0.05

0.075

0.1

62


A perfect spherical boundary condition produces

perfect spherical symmetry.

63 E

lec

tric

Fie

ld

(V/m

)

Y

(m)

X (m)

0

0.00011

1e-05

2e-05

3e-05

4e-05

5e-05

6e-05

7e-05

8e-05

9e-05

0.0001

-0.5 0.5-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

-0.5

0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

Using only eight antennas at 45° spacing produces a

uniform test volume that’s only about a wavelength.

Ele

ctr

ic F

ield

(V

/m)

Y

(m)

X (m)

0

3.5e-05

2e-06

4e-06

6e-06

8e-06

1e-05

1.2e-05

1.4e-05

1.6e-05

1.8e-05

2e-05

2.2e-05

2.4e-05

2.6e-05

2.8e-05

3e-05

3.2e-05

-0.5 0.5-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

-0.5

0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4


64

Using 24 antennas at 15° spacing produces a

much larger uniform test volume.

Ele

ctr

ic F

ield

(V

/m)

Y

(m)

X (m)

0

6e-05

1e-05

2e-05

3e-05

4e-05

5e-05

-0.5 0.5-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

-0.5

0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4


65


Radial fall-off from traditional antennas in close proximity to DUT does not behave like reflections from distant objects (i.e. non-plane-wave behavior).

Relative Signal Strength for Point Source at 1.0 m Distance

Re

lati

ve

Po

we

r (

dB

)

Y

(m)

X (m)

-2.5

2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

-0.25 0.25-0.2 -0.1 0 0.1 0.2

-0.25

0.25

-0.2

-0.1

0

0.1

0.2

Relative Signal Strength for Point Source at 3.0 m Distance

Re

lati

ve

Po

we

r (

dB

)

Y

(m)

X (m)

-2.5

2.5

-2

-1.5

-1

-0.5

0

0.5

1

1.5

2

-0.25 0.25-0.2 -0.1 0 0.1 0.2

-0.25

0.25

-0.2

-0.1

0

0.1

0.2

66

Standardization of MIMO OTA Testing

CTIA, 3GPP, COST all looking at MIMO OTA.

WiMAX Forum is also interested.

CTIA MIMO Anechoic Chamber subgroup

(MACSG) has been merged with Reverb

Chamber subgroup (RCSG) to create the new

MIMO OTA subgroup (MOSG)

This has slowed the pace of CTIA development

similar to 3GPP as more approaches are

proposed to MOSG.

67

Standardization of MIMO OTA Testing

3GPP finished its second round robin to

evaluate methodologies.

Round robin originally planned to prove

reproducibility of MIMO anechoic method.

Was expanded to add comparison between all

methodologies.

Usefulness of the results is limited due to some

limitations in the process.

68

ETS-Lindgren Intellectual Property

We published first papers on this topic in 2006, after applying for multiple patents.

20080056340 – “Systems and methods for over the air performance testing of wireless devices with multiple antennas”

Anechoic array MIMO OTA concept using channel emulator, delay lines, etc.

20080305754 – “Systems and methods for over-the-air testing of wireless systems”

Covers anechoic, reverb, lossy reverb, delay lines, etc. as well as combinations thereof.

69

ETS-Lindgren Intellectual Property

Patent #7,965,986 “Systems and methods for

over-the-air testing of wireless systems” has

been granted based on 20080305754.

70

AMS-8700 Environment Simulator

The MIMO OTA Environment Simulation System

is referred to as the

AMS-8700 series.

The baseline AMS-8700

has eight dual polarized

antennas and one 8-ch

channel emulator.

The mounting ring allows

different spacing to support

single cluster & distributed configurations.

71


The baseline provides only eight active

elements, switchable between vertical & horiz.

72


A base station with throughput

testing options is used for MIMO.

A VNA with multiple channels can

be used for evaluating correlation

of embedded MIMO antennas.

A MAPS is provided for 3-D tests.

An additional SISO-only antenna

can be added for TRP/TIS.

Other antenna configuration/test

options are available.

73


EB’s was the first to market with an OTA channel

emulator solution.

They include an OTA modeling tool for creating

OTA spatial channel models.

74


Additional antennas can be added (up to the

available space on the ring) by adding the

required number of

channel emulators.

Minor incremental

cost to the chamber

but multiplies the

instrumentation cost.

75


An AMS-8900 can be combined with an AMS-

8700 to allow high speed APM and TRP/TIS.

Chamber cost is a fraction of overall system

cost.

User must weigh

value of doing

APM/TRP/TIS

vs. having

inactive channel

emulator(s).

76

EMQuest EMQ-108 MIMO OTA Testing

An optional EMQ-108 MIMO OTA expansion

module has been added to EMQuest.

Option adds support for channel emulators as

variable gain devices.

Includes calibration/validation tests for spatial

channel emulation.

Includes special vector APM post processing for

calculating antenna envelope correlation.

Will be sold as a separate option as well.

77

EMQuest EMQ-108 MIMO OTA Testing

Requirements of MIMO OTA introduce a higher

level of complexity to system and test

automation.

Interactions between BSE/VSA, channel emulator, and

amplifiers.

Uplink port isolation.

Validation tests produce huge data sets.

An adequate level of software control is required

to automate the calibration and test routines.

78


A typical RF channel emulator consists of a

number of VSA receivers and VSG transmitters

connected to a DSP modeling core that

introduces delay spreads, fading, etc. at

baseband.

79

Receiver(VSA)

Transmitter(VSG)

DSPModeling

Core

Receiver(VSA)

Transmitter(VSG)

Receiver(VSA)

Transmitter(VSG)


The ideal channel emulator routes multiple inputs

to multiple outputs after applying appropriate

modeled delay spreads, fading, etc.

80

Transmitter 1

Receiver1

Transmitter2

Receiver2

ChannelEmulation


In a real system, there are external path losses

that must be accounted for.

For an OTA system, this includes cable losses,

antenna gains, and range path losses.

External amplifiers are also typically required

81

ChannelEmulator

Receiver1

Receiver2

Transmitter 1

Transmitter2

Total Emulated Channel


For input calibration, the requirement is that

G1 1 + G1 2 = G2 1 + G2 2 = GX 1 + GX 2

so that PA = PB = PX, when PTX is applied to each

input cable.

82

Channel Emulator

Receiver1

Receiver2

Transmitter 1

Transmitter2


Net Input Path Loss Net Output Path Loss

A

B

C

D

G1 1 G1 2

G2 1 G2 2

G1 3 G1 4

G2 3 G2 4

GA-C

GB-D

GB-C

GA-D

Channel Gain

IN1

IN2

OUT1

OUT2


Similarly, for output calibration, requirement is

that

G1 3 + G1 4 = G2 3 + G2 4 = GX 3 + GX 4

so that net losses to test volume are equal.

83

Channel Emulator

Receiver1

Receiver2

Transmitter 1

Transmitter2



A

B

C

D

G1 1 G1 2

G2 1 G2 2

G1 3 G1 4

G2 3 G2 4

GA-C

GB-D

GB-C

GA-D

Channel Gain

IN1

IN2

OUT1

OUT2


Simple channel model with four equivalent inputs

(red) and eight resulting outputs (blue).

84

-35

-30

-25

-20

-15

-10

-5

0

1 2 3 4 5 6 7 8

Po

we

r (d

Bm

)

Port Number

Simple Channel Model

Input

Output


Result of input and output path losses applied to

channel model.

85

-60

-50

-40

-30

-20

-10

0

1 2 3 4 5 6 7 8

Po

we

r (d

Bm

)

Port Number

Effect of Input and Output Cables on Channel Model

Input

Output

Net Output


Correcting for relative input losses (purple)

flattens inputs to channel model. Net output is

still wrong.

86

-60

-50

-40

-30

-20

-10

0

10

1 2 3 4 5 6 7 8

Po

we

r (d

Bm

)

Port Number

Applying Input Gain to Offset Input Cable Differences

Input

Input Gain (dB)

Channel Input

Output

Net Output


Correcting the relative output levels reproduces

the desired impact of clusters within the

environment.

87

-70

-60

-50

-40

-30

-20

-10

0

1 2 3 4 5 6 7 8

Po

we

r (d

Bm

)

Port Number

Applying Output Gain to Offset Output Cable Differences

Input

Channel Input

Channel Output

Output Gain (dB)

Output

Net Output


Finally, to predict the average power level in the

center of the test volume, the average path loss

must be calculated, including the loss (gain) of

the channel model.

This requires detailed knowledge of the channel

model gains, as well as assumptions about the

relative input levels of each input path for MIMO.

Application of a SIMO output calibration to a

MIMO model requires understanding/adjustment

of source power definition.

88


where GChannel is the gain of a single channel path.

89

Channel Emulator

Receiver1

Receiver2

Transmitter 1

Transmitter2



A

B

C

D

G1 1 G1 2

G2 1 G2 2

G1 3 G1 4

G2 3 G2 4

GA-C

GB-D

GB-C

GA-D

Channel Gain

IN1

IN2

OUT1

OUT2

N

j

GGG

iiijjjiChannelGGg

1

214310log10


However, when properly calibrated

and likewise

so that

which can be simplified to:

90

N

j

gG

inputioutputjiChannelgg

1

10log10

413143 GGGGg jjoutput

21 iiinput GGg

N

j

G

outputinputijiChannelggg

1

10log10


This simplification shows that both the input and

output gain of the system can be easily altered to

address output levels and modulation headroom

of the source, and to vary total path loss in the

system.

Such changes must be accounted for in determining the

power in the test volume.

When changing channel models, the total

channel model gain must be recalculated.

Changing the number of active inputs and

outputs also alters the gain.

91


When evaluating the gain of a MIMO system

where the same power, PTX, is applied to each

input cable, the total gain to the test volume can

be given by:

Note that this sum includes the array gain of the

multiple transmitters. For average power gain,

92

M

i

g

totalig

1

10log10

M

i

g

averagei

Mg

1

101

log10

System Validation

A wide range of tests are possible to evaluate:

Chamber/RF system quality

Channel emulation quality

Calibration quality

Combined system performance

Many tests are more interesting for research

purposes rather than system validation

It’s important to separate out tests that provide

useful system information vs. component level

performance.

E.g. correlation or field distribution vs. Doppler spread

93

System Validation

A re-configurable ETS-Lindgren AMS-8700 MIMO

OTA system with 16 dual polarized antennas and

two 8-output channel emulators was evaluated.

94

A linear positioner and turntable were used to

map a 1 m diameter disc in the center of the test

volume at 1 cm by 1 degree (0.87 cm at edge)

resolution.

System Validation

95

System Validation

Spatial Field Mapping is used to compare the

measured environment to a theoretical model.

96

Calculated Far-field Field Structure, 45 Degree Spacing

Ele

ctr

ic F

ield

(V

/m)

Y

(m)

X (m)

0

3.2e-05

1.1e-06

2.2e-06

3.3e-06

4.4e-06

5.5e-06

6.6e-06

7.7e-06

8.8e-06

9.9e-06

1.1e-05

1.21e-05

1.32e-05

1.43e-05

1.54e-05

1.65e-05

1.76e-05

1.87e-05

1.98e-05

2.09e-05

2.2e-05

2.31e-05

2.42e-05

2.53e-05

2.64e-05

2.75e-05

2.86e-05

2.97e-05

3.08e-05

-0.5 0.5-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

-0.5

0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

Calculated Far-Field Field Structure with High Resolution Boundary Condition

Ele

ctr

ic F

ield

(V

/m)

Y

(m)

X (m)

0

7.83e-05

2.7e-06

5.4e-06

8.1e-06

1.08e-05

1.35e-05

1.62e-05

1.89e-05

2.16e-05

2.43e-05

2.7e-05

2.97e-05

3.24e-05

3.51e-05

3.78e-05

4.05e-05

4.32e-05

4.59e-05

4.86e-05

5.13e-05

5.4e-05

5.67e-05

5.94e-05

6.21e-05

6.48e-05

6.75e-05

7.02e-05

7.29e-05

7.56e-05

-0.5 0.5-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

-0.5

0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

Ideal Free-Space or Continuous Boundary Condition Interference Pattern from Eight Evenly Spaced Plane Waves

Calculated Field Structure at 2.0 m Radius, 45 Degree Spacing

Ele

ctr

ic F

ield

(V

/m)

Y

(m)

X (m)

0

0.016

0.0006

0.0012

0.0018

0.0024

0.003

0.0036

0.0042

0.0048

0.0054

0.006

0.0066

0.0072

0.0078

0.0084

0.009

0.0096

0.0102

0.0108

0.0114

0.012

0.0126

0.0132

0.0138

0.0144

0.015

-0.5 0.5-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

-0.5

0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

Measured Field Structure, 45 Degree Spacing, Vertically Polarized

Ma

gn

itu

de

X

(cm

)

Angle (°)

0

0.0192

0.0007

0.0014

0.0021

0.0028

0.0035

0.0042

0.0049

0.0056

0.0063

0.007

0.0077

0.0084

0.0091

0.0098

0.0105

0.0112

0.0119

0.0126

0.0133

0.014

0.0147

0.0154

0.0161

0.0168

0.0175

0.0182

Scale: 5/div

Min: 0

Max: 50

0

180

30

210

60

240

90270

120

300

150

330

System Validation

Modeling a 2 m range length instead of a plane

wave shows excellent correlation.

97

Measured Interference Pattern from Eight Antennas, r = 2 m Calculated Interference Pattern from Eight Antennas , r = 2 m

Comparing a single cut through the test volume.

System Validation

98

Comparison of Measured Field Structure to Theory for 8 Antenna Array (45° Spacing)

Re

lati

ve

Fie

ld L

ev

el

X (cm)

-50 50-40 -30 -20 -10 0 10 20 30 40

0

1

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Measured Field Theoretical Field Ideal Free-Space Field

Measured Field Structure, 22.5 Degree Spacing

Ma

gn

itu

de

X

(cm

)

Angle (°)

0

7.04

0.25

0.5

0.75

1

1.25

1.5

1.75

2

2.25

2.5

2.75

3

3.25

3.5

3.75

4

4.25

4.5

4.75

5

5.25

5.5

5.75

6

6.25

6.5

6.75

Scale: 5/div

Min: 0

Max: 50

0

180

30

210

60

240

90270

120

300

150

330

Increasing the resolution of the boundary condition

from 8 to 16 antennas increases usable test volume.

Calculated Field Structure at 2.0 m Radius, 22.5 Degree Spacing

Ele

ctr

ic F

ield

(V

/m)

Y

(m)

X (m)

0

0.0226

0.0008

0.0016

0.0024

0.0032

0.004

0.0048

0.0056

0.0064

0.0072

0.008

0.0088

0.0096

0.0104

0.0112

0.012

0.0128

0.0136

0.0144

0.0152

0.016

0.0168

0.0176

0.0184

0.0192

0.02

0.0208

0.0216

-0.5 0.5-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3 0.4

-0.5

0.5

-0.4

-0.3

-0.2

-0.1

0

0.1

0.2

0.3

0.4

System Validation

99

Measured Interference Pattern from 16 Antennas, r = 2 m Calculated Interference Pattern from 16 Antennas , r = 2 m

System Validation

Spatial Correlation evaluates field structure and

channel model behavior.

Move one dipole through test volume and evaluate

correlation vs. separation.

Requires replay of channel model at each position.

Single cluster behavior most straightforward to evaluate.

100

SingleCluster

1 m slice through test volume on 1 cm steps

Spatial Correlation evaluates RF system +

emulation.

System Validation

101

Spatial Correlation for 8 Antenna (45° Spacing) Configuration

Co

rre

lati

on

X (cm)

-50 50-40 -30 -20 -10 0 10 20 30 40

0.2

1

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Both tests show similar system performance

results. Comparison of Spatial Correlation and Field Structure for 22.5° Resolution Configuration

X (cm)

-50 50-40 -30 -20 -10 0 10 20 30 40

0

1

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

Spatial Correlation Field Structure Free-Space Field Structure

System Validation

102

System Validation

Channel Model Pattern – Using a narrow beam

antenna, the generated angular spread profile can

be mapped.

Works well as a quick verification for single cluster

Not as agile for more complicated models

Antenna-by-Antenna Mapping – Measure channel

frequency response of each antenna across

statistically large set of IR steps and TD

transform.

Evaluates PDP and AS of channel model.

Can be numerically compared to summation of all antennas

active (requires valid phase calibration).

103

Throughput Measurement Results

Unlike traditional TRP/TIS tests, which provide edge

of link performance metrics, MIMO performance is all

about high bandwidth with large SNRs.

The corresponding metric for measuring bandwidth

is throughput, and the equivalent evaluation would be

to determine when the throughput begins to fall off.

Initial tests were performed with 802.11n devices

supporting 2x2 MIMO, to prove the capabilities of the

system and methodology.

Now that LTE communication testers are available, it

is possible to show the first LTE MIMO OTA results.

104

Wi-Fi Throughput Measurement Results

A re-configurable MIMO OTA system was

installed in ETS-Lindgren’s Cedar Park facility for

research and development of test requirements.

Eight dual polarized antenna elements were

mounted on adjustable fixtures and arranged

around a DUT positioning turntable.

The Elektrobit Propsim F8 channel emulator was

used to provide the spatial channel emulation

required for the OTA environment simulation.

Eight 30 dB gain power amplifiers drive eight

vertical antenna elements.

105


An 802.11n 2x2 MIMO Wireless Router with

removable, adjustable external antennas was

chosen as the DUT.

A matching NIC was used as the downlink

source.

Directly cabled conducted tests

were used to verify MIMO

operation with appropriately

higher throughput compared to

SIMO/SISO cabled

configurations.

106


Conducted tests of throughput vs. attenuation were performed with Propsim F8 using circulators/isolators to provide a single return uplink.

Direct single tap models were used to replicate cabled results.

Several 2x2 MIMO models suitable for OTA testing were evaluated to determine typical MIMO performance.

Modified SCME Urban Micro w/ 3 km/h fading & zero delay spread.

Modified TGn-C w/ AOD/AOA based on SCME

Modified TGn-C w/ low TX correlation (10 wavelength sep.)

107


Using standard 20 MHz 802.11 channels,

conducted tests show maximum SIMO

throughput around 25 MBPS, with MIMO

performance around 40-45 MBPS with typical

channel models.

Initial OTA tests with stock antennas using low

correlation TGn-C OTA model produces similar

results but shows angular dependence of MIMO

performance while SIMO (diversity) performance

remains uniform.

108


Throughput vs. Total Path Loss

Th

rou

gh

pu

t (

Mb

ps

)

Attenuation (dB)

30 8035 40 45 50 55 60 65 70 75

10

50

15

20

25

30

35

40

45

0° 30° 60° 90° 120° 150° 180° 210° 240°

270° 300° 330° 360° SIMO TX1 SIMO TX2

109



Th

rou

gh

pu

t (

Mb

ps

)

Attenuation (dB)

30 8035 40 45 50 55 60 65 70 75

10

50

15

20

25

30

35

40

45

0° 30° 60° 90° 120° 150° 180° 210° 240°

270° 300° 330° 360° SIMO TX1 SIMO TX2

MIMO

Operating

Region

110



Th

rou

gh

pu

t (

Mb

ps

)

Attenuation (dB)

30 8035 40 45 50 55 60 65 70 75

10

50

15

20

25

30

35

40

45

0° 30° 60° 90° 120° 150° 180° 210° 240°

270° 300° 330° 360° SIMO TX1 SIMO TX2

SIMO

Operation

111



Th

rou

gh

pu

t (

Mb

ps

)

Attenuation (dB)

30 8035 40 45 50 55 60 65 70 75

10

50

15

20

25

30

35

40

45

0° 30° 60° 90° 120° 150° 180° 210° 240°

270° 300° 330° 360° SIMO TX1 SIMO TX2

TX

Beam- Forming

Region

112

LTE Throughput Measurement Results

LTE USB modem on test pedestal in middle of chamber

113


114

Throughput vs. Power vs. Orientation, SCME Urban Micro, 16 QAM LTE DUT

Th

rou

gh

pu

t (

Mb

ps

)

Power (dBm)

-79 -60-78 -76 -74 -72 -70 -68 -66 -64 -62

8

24

10

12

14

16

18

20

22

0° 45° 90° 135° 180° 225° 270° 315°

20 Mbps Throughput Sensitivity Pattern, 16 QAM LTE DUT

Po

we

r (

-dB

m)

Angle (°)

Scale: 1/div

Min: 71

Max: 79

0

180

30

210

60

240

90 270

120

300

150

330

6.2.1 - SCME Urban Micro, Avg = -74.0 dBm 6.2.2 - Modified SCME Urban Micro, Avg = -74.4 dBm

6.2.3 - SCME Urban Macro, Avg = -74.5 dBm 6.2.4 - Modified WINNER2, Avg = -73.1 dBm


115

40 Mbps Throughput Sensitivity Pattern, 64 QAM LTE DUT

Po

we

r (

-dB

m)

Angle (°)

Scale: 1/div

Min: 56

Max: 67

0

180

30

210

60

240

90 270

120

300

150

330

6.2.1 - SCME Urban Micro, Avg = -61.1 dBm 6.2.2 - Modified SCME Urban Micro, Avg = -61.0 dBm

6.2.3 - SCME Urban Macro, Avg = -58.8 dBm 6.2.4 - Modified WINNER2, Avg = -60.6 dBm


116

Metrics

The data acquired thus far can be evaluated in a

number of ways to define different metrics for

MIMO performance.

Removing the position axis produces average

throughput vs. power (attenuation) curves.

This could be done as a post processing step,

but if position (pattern) information is not

needed, average throughput performance can be

determined by moving DUT continuously

through simulated environment.

117

Metrics

118

Average Azimuthal Throughput vs. Total Path Loss

Th

rou

gh

pu

t (M

bp

s)

Attenuation (dB)

30 7535 40 45 50 55 60 65 70

5

45

10

15

20

25

30

35

40

TGn-C Low Correlation TGn-C Normal Correlation

Metrics

This test can be further reduced by choosing to

determine average throughput performance at a

given field level (no power level search).

E.g. At an attenuation value of 50 dB, this DUT has an

average throughput of 36 Mbps for the low correlation

TGn-C model and 30 Mbps for the normal correlation

TGn-C model.

This is similar to many conformance tests with a

simple pass/fail result, and assumes a minimum

expected network capability.

119

Metrics

By retaining angular information, or by

measuring throughput over short dwell times as

the DUT moves, peak throughput performance

can be determined.

This metric may have limited usefulness, but

does illustrate a slightly different reaction to the

two models.

120

Metrics

121

Peak Azimuthal Throughput vs. Total Path Loss

Th

rou

gh

pu

t (M

bp

s)

Attenuation (dB)

30 7535 40 45 50 55 60 65 70

5

50

10

15

20

25

30

35

40

45


Metrics

By retaining throughput vs. attenuation or using

a throughput vs. attenuation search mode, one

can define a “MIMO Sensitivity” where

throughput falls below a certain target.

This can be defined in two ways, with varying

test time requirements.

Average power required to produce the target throughput

at each angle (integrated TIS pattern)

Power required to produce desired average throughput as

device is rotated through all angles

122

Metrics

123

(Linear) Average Attenuation vs. Throughput

Av

era

ge

Att

en

ua

tio

n (

dB

)

Throughput (Mbps)

10 3515 20 25 30

40

80

45

50

55

60

65

70

75


Metrics

While the statistics of these two metrics are

slightly different and provide slightly different

results, both provide considerably more

information on the DUT, offering an “edge of

MIMO link” performance indicator.

Such information can be used to rank products

and influence improvements, while the previous

pass/fail options only offer basic acceptability

criteria.

124

“Good” device antenna patterns.

Reference MIMO Antenna

125 Images courtesy of Motorola Mobility

“Nominal” device antenna patterns.



“Bad” device antenna patterns.



Tests were performed in an AMS-8700 MIMO OTA

system using eight vertically polarized elements

evenly spaced every 45 degrees.


128

DUT

MIMOTester

SpatialChannelEmulator


129

SCME Urban Micro (TR 37.976 Section 6.2.1), 30 km/h

Av

era

ge

Th

rou

gh

pu

t (

Mb

ps

)

EPRE (dBm)

-115 -90-110 -105 -100 -95

15

24

16

17

18

19

20

21

22

23

Conducted 1:1 Constant Tap Conducted Fading Model "Good" MIMO Antenna "Nominal" MIMO Antenna

"Bad" MIMO Antenna


130

" Optimum Drop" Model, 30 km/h

Av

era

ge

Th

rou

gh

pu

t (

Mb

ps

)

EPRE (dBm)

-115 -90-110 -105 -100 -95

15

24

16

17

18

19

20

21

22

23

Conducted 1:1 Constant Tap Conducted Fading Model "Good" MIMO Antenna "Nominal" MIMO Antenna

"Bad" MIMO Antenna


With a measured <10 dB delta between a “good”

and “bad” antenna, this method is good at

differentiating between devices within the

expected measurement uncertainty.

Orientation of the device and polarization are

expected to have an effect. This device was

optimally oriented in the test system.

Plans to repeat test with a dual polarized setup

and possibly different orientations will follow.

131

Conclusion

Extensive efforts are underway to standardize on

a next generation platform for wireless testing.

The ability to perform realistic RF environment

simulation and evaluate end user metrics in real-

world scenarios is an invaluable resource to

wireless technology developers.

Detailed calibration and validation methods are

required to ensure the validity of measured data.

While a throughput related metric is the logical

choice, the industry must still choose the desired

target metric (e.g. throughput sensitivity).

132

Thank You!

133

QUESTIONS? Dr. Michael D. Foegelle

Director of Technology

Development

[email protected]

+1-512-531-6444

Garth D’Abreu

Director of RF Engineering

Garth.dabreuets-lindgren.com

+1-512-531-6438

mailto:[email protected]






introduction to mimo ota environment simulation

Documents