Low-Distortion

Signal Generation

for ADC Testing Fumitaka Abe, Yutaro Kobayashi

Kenji Sawada, Keisuke Kato

Osamu Kobayashi, Haruo Kobayashi

Gunma University

STARC

2014 IEEE International Test Conference

2/41

Research Objectives

• Use an Arbitrary Waveform Generator (AWG)

as low-cost low-distortion signal source

for ADC testing

• Validate our 3rd-harmonic-cancelling

phase-switching signal generation technique

• Perform analysis, simulation, and experiments

using AWGs for ADC testing

3/41

Expected Advantage of Our Solution

𝐟𝒊𝒏

100kHz 𝟑𝐟𝒊𝒏

𝟑𝟎𝟎𝐤𝐇𝐳

Freq.

Filter A

𝐟𝒊𝒏

300kHz 𝟑𝐟𝒊𝒏

𝟗𝟎𝟎𝐤𝐇𝐳

Freq.

Filter B

𝐟𝒊𝒏

100kHz

𝟏

𝟐𝐟𝐬 − 𝐟𝒊𝒏

𝟏𝟒𝟎𝟎𝐤𝐇𝐳

Freq.

Filter C

𝐟𝒊𝒏

300kHz

𝟏

𝟐𝐟𝐬 − 𝐟𝒊𝒏

𝟏𝟐𝟎𝟎𝐤𝐇𝐳

Freq.

Filter C

Conventional Method Tough LPF requirements

Our Method Relaxed LPF requirements

𝟏

𝟐𝐟𝒔

1500kHz

𝟏

𝟐𝐟𝒔

1500kHz

𝟒𝟎𝐝𝐁 𝟒𝟎𝐝𝐁

𝟏𝟎𝐝𝐁

4/41

Outline

• Background to this research

• Proposed solution

• Problems with proposed solution

• Remedy 1

• Remedy 2

• Experimental results

• Conclusions

5/41

Outline

• Background to this research

• Proposed solution

• Problems with proposed solution

• Remedy 1

• Remedy 2

• Experimental results

• Conclusions

6/41

Background to this Research

• ADCs are important part of mixed-signal SOCs.

• Need a low-cost, low-distortion signal source

for ADC linearity testing.

• #1: Use existing AWGs in testers low cost

• # 2: Develop a technique that doesn’t require

identification of AWG nonlinearity,

just a DSP program change.

7/41

Freq.

ADC HD3

Freq.

“No Go”

“Go”

Freq.

Reference level

Allowable

fundamental

Freq.

Test result

ADC HD3 measurement

ADC linearity test

ADC Testing with Sine Signal

fundamental

AWG ADC

ADC test signal

8/41

Problem with Conventional Method

fundamental

Freq.

Freq.

AWG HD3+ADC HD3

fundamental HD3

Test result

AWG HD3

ADC HD3 cannot be

measured accurately.

Low quality test

AWG ADC

Test signal

Generation

ADC test signal

9/41

Research Objective

AWG ADC

Test signal

Generation

fundamental

Freq.

ADC test signal

Freq.

Only ADC HD3

fundamental

HD3

ADC test result

AWG HD3 reduction

High quality test

only

program

change

10/41

Outline

• Background to this research

• Proposed solution

• Problems with proposed solution

• Remedy 1

• Remedy 2

• Experimental results

• Conclusion

11/41

AWG

DAC

CLK

DSP

Conventional Signal Generation with AWG

X X X X

Din

Din

A

0 0.1 0.2 0.3 0.4 0.5

-300

-200

-100

0P

ow

er

[dB

]

Normalized frequency f/fs

AWG sampling frequency: fs(AWG) = 1/Ts

DAC has 3rd order nonlinearity

fin

3fin

12/41

AWG

DAC

CLK

DSP

Proposed Signal Generation with AWG

X0 X1 X0 X1

Din

Din

𝐗𝟎 = 1.15Acos(2πfinnTs − π/6)

𝐗𝟏 = 1.15Acos(2πfinnTs + 𝜋/6)

Θ = π/3

0 0.1 0.2 0.3 0.4 0.5

-300

-200

-100

0

Pow

er

[dB

]

Normalized frequency f/fs

HD3 is cancelled

fs/2-fin

fs/2-3fin fin

13/41

Co

nve

nti

on

al

Ph

as

e S

wit

ch

ing

Θ = π/3 3Θ = π

Principle of 3rd Harmonics Cancellation

3rd

ord

er

non

-lin

ear

syste

m

Phase r

ota

tion b

y x

3

Two waves with phase difference π

are cancelled

fundamental: fin 3rd harmonics: 3 fin

14/41

Conventional and Phase Switching Signals

Time [sec]

Vo

lta

ge

[V

]

Frequency [kHz]

3rd harmonic Po

wer

[dB

m]

fundamental：9.0dBm

HD3：-68.0dBm

Frequency [MHz]

Waveform Measured Power Spectrum

3rd harmonic

1V

fs/2-fin ：-0.91dBm

Frequency [kHz]

3rd harmonic Po

wer

[dB

m]

fundamental：9.0dBm

HD3：-78.2dBm

Frequency [MHz]

Time [sec]

Vo

ltag

e [

V]

Waveform

3rd harmonic

π/3

1.15V

Conventional

Phase Switching

15/41

Outline

• Research background

• Proposed solution

• Problems of proposed solution

• Remedy 1

• Remedy 2

• Experimental results

• Conclusion

16/41

0 0.1 0.2 0.3 0.4 0.5-400

-300

-200

-100

0

Normalized Frequency f/fs

Pow

er

[dB

]

0 0.1 0.2 0.3 0.4 0.5-400

-300

-200

-100

0

Normalized Frequency f/fs

Pow

er

[dB

]

0 0.1 0.2 0.3 0.4 0.5-400

-300

-200

-100

0

Normalized Frequency f/fs

Pow

er

[dB

]

0 0.1 0.2 0.3 0.4 0.5-400

-300

-200

-100

0

Normalized Frequency f/fs

Pow

er

[dB

]

0 0.1 0.2 0.3 0.4 0.5-400

-300

-200

-100

0

Normalized Frequency f/fs

Pow

er

[dB

]

0 0.1 0.2 0.3 0.4 0.5-400

-300

-200

-100

0

Normalized Frequency f/fs

Pow

er

[dB

]

3rd

nonlinearity 3次

非線形

Phase

switching

Conventional

Input signal

cancellation cancellation cancellation

generation generation generation

Conditions of HD3 Cancellation in Cascaded System

3rd

nonlinearity

3rd

nonlinearity

17/41

Problem of Phase Switching Signal Generation

ADC

fundamental

Freq.

fundamental

Freq.

HD3

HD3 due to

ADC nonlinearity

Nonlinearity Model

ADC

Reduction of HD3

due to ADC nonlinearity

fundamental

Freq.

fundamental

Freq.

HD3

Reduction of HD3

due to AWG nonlinearity

Spurious

due to phase switching

Big problem !

Ideal

Phase

switching

18/41

Model for Theoretical Analysis

n: odd

n: even

AWG Input with Phase Switching

AWG Nonlinearity Model

ADC Nonlinearity Model

AWG ADC w/o LPF

For simplicity

fs(AWG)=fs(ADC)

19/41

AWG Output

signal

spurious

small

large

20/41

Direct Application of Phase Switching Signal

: :

ADC output Z(n) components

By calculation

CANNOT measure ADC HD3.

signal

HD3

21/41

Outline

• Research background

• Proposed solution

• Problems of proposed solution

• Remedy 1

• Remedy 2

• Experimental results

• Conclusion

22/41

AWG Output and Low Pass Filtering

LPF AWG ADC w/ LPF

23/41

Application of Phase Switching and LPF

: :

ADC output Z(n) components

By calculation

ADC HD3 can be measured.

signal

HD3

24/41

Spurious Attenuation Effect

q

AWG output and low pass filtering

25/41

ADC

Freq. Freq.

LPF reduction

0 20 40 60 80 10010

-10

10-8

10-6

10-4

10-2

100

102

fs/2-finスプリアスの減衰量 [dB]

図2.1

0(B

)の3次

高調

波の

本来

値か

らの

ずれ

[％]

Spurious reduction at fs/2-fin [dB]

MATLAB Simulation Results

Spurious reduction

at fs/2-fin HD3 measurement

Error

10dB

20dB

1%

0.1%

AD

C H

D3 M

ea

su

rem

ent E

rro

r [%

]

26/41

Outline

• Research background

• Proposed solution

• Problems of proposed solution

• Remedy 1

• Remedy 2

• Experimental results

• Conclusion

27/41

ADC

Freq. Freq.

Inside ADC

π/3 π HD3 cancellation

Reproduction of phase switching signal

Digital output

Vo

ltag

e

Time

Analog input

0 5 10 15 20 25 30 35 40

0

Sample Number

Codes

ADC HD3 can NOT be measured accurately

Similar

28/41

Inside ADC

π/3 π HD3 cancellation

Reproduction of phase switching signal

Digital output Analog input

ADC

Freq. Freq.

folded

0 10 20 30

1000

2000

3000

Sample Number

Codes

ADC HD3 can be measured accurately

Different

Vo

ltag

e

Time

29/41

ADC

Freq. Freq.

folded

LPF reduction

Change sampling frequency by folding

Accurate measurement of ADC output HD3 with phase switching signal

①

②

Conditions for Accurate ADC HD3 Measurement

30/41

Outline

• Research background

• Proposed solution

• Problems of proposed solution

• Remedy 1

• Remedy 2

• Experimental results

• Conclusion

31/41

AWG LPF ADC

PC Oscilloscope,

Spectrum Analyzer FFT

AD7356

output ..01011..

Agilent

33220A

ADI

AD7356

12bit SAR ADC

EVAL

CED1Z 3.478261

MHz

CLK Evaluation

board

digital

output

ADC 3rd Harmonic Measurement Diagram

Satisfying Remedy 2

ADC test signal generation

・Conventional

・Phase switching

14bit DAC

Satisfying Remedy 1

32/41

AWG 33220A

for test signal

generation

ADC AD7356 (12bit)

6 samples

Device Under Test

Low Pass Filters

PC1 for test signal program Spectrum Analyzer

for test signal analysis Oscilloscope for test signal analysis

EVAL-CED1Z for generating

sampling clock , etc.

Common mode

noise suppression

by a choke coil PC2 for

ADC output

analysis

Experimental Environment for ADC Testing fs(AWG)=10MHz, fin=200kHz

fs(ADC)=3.476261MHz

33/41

AWG LPF ADC

PC Oscilloscope,

Spectrum Analyzer

DUT

FFT

AD7356

output ..01011..

Agilent

33220A

ADI

AD7356

EVAL

CED1Z 3.478261

MHz

CLK

Design

Implementation

Evaluation

Evaluation

board

digital

output

Analog LPF Design for ADC Testing

Freq.

HD3＠600kHz

removal

fundamental

@ 200kHz

fs/2-fin

fundamental

@ 200kHz

Freq. fs/2-fin

Phase switching

spurious @4.8MHz

attenuation

fc=250kHz

5th order LC Butterworth LPF

4th order LC Butterworth LPF

fc= 1MHz, 2MHz, 2.7MHz, 3.7MHz

34/41

AWG LPF ADC

PC Oscilloscope,

Spectrum Analyzer

DUT

FFT

AD7356

output ..01011..

Agilent

33220A

ADI

AD7356

EVAL

CED1Z 3.478261

MHz

CLK

Design

Implementation

Evaluation

Evaluation

board

digital

output

LPF Implementation

fc=2MHz fc=3.7MHzfc=2.7MHzfc=1MHzfc=1MHz fc=2MHz fc=2.7MHz fc=3.7MHz fc=250kHz

for HD3 reduction for fs/2-fin spurious reduction

35/41

AWG LPF ADC

PC Oscilloscope,

Spectrum Analyzer

DUT

FFT

AD7356

output ..01011..

Agilent

33220A

ADI

AD7356

EVAL

CED1Z 3.478261

MHz

CLK

Design

Implementation

Evaluation

fc=250kHz

fc=1MHz

fc=2MHz

fc=2.7MHz

fc=3.7MHz

- 17 dB@4.8MHz

- 30 dB@4.8MHz

- 54 dB@4.8MHz

- 8.0 dB@4.8MHz

- 40dB @ 600kHz

104

105

106

107

-100

-80

-60

-40

-20

0

Frequency[Hz]

Gain

[dB

]

HD3

Spurious

@~fs/2

Evaluation

board

digital

output

LPF Evaluation with Frequency Response Analysis

36/41

Measured Waveforms of ADC Input and Output

0 10 20 30

1000

2000

3000

Sample Number

Codes

0 10 20 30

1000

2000

3000

Sample Number

Codes

0 10 20 30

1000

2000

3000

Sample Number

Codes

0 10 20 30

1000

2000

3000

Sample Number

Codes

0 10 20 30

1000

2000

3000

Sample Number

Codes

Analog Input

Conventional

-17dB

fs/2-fin

Spurious

-30dB

-54dB

A/D

Convers

ion

Spectru

m A

naly

sis

Vo

ltag

e [

V]

2

-2

0

1

-1

Time

Phase Switching Signal

Fundamental

@ 200kHz

fs/2-fin

Phase switching

spurious@4.8MHz

attenuation

LPF

Vo

ltag

e [

V] 2

-2

0

1

-1

Time

Vo

ltag

e [

V] 2

-2

0

1

-1

Time

Vo

ltag

e [

V] 2

-2

0

1

-1

Time

Vo

ltag

e [

V] 2

-2

0

1

-1

Time

Vo

ltag

e [

V] 2

-2

0

1

-1

Time

-8dB

Digital output

HD

3 M

ea

su

rem

en

t

37/41

Frequency [MHz]

po

we

r [

dB

Fs

]

3rd

①+ ①-

Conventional w/o LPF

Frequency [MHz]

Po

wer

[d

BF

s]

Phase switching w/ LPF (fc=1MHz)

Comparison of Conventional and Proposed Methods

Frequency [MHz]

Po

wer

[d

BF

s]

3rd

Conventional w/ LPF (fc=250kHz)

True ADC HD3

AWG由来3次高調波をフィルタによりカットした手法

②- ②+

LPF

- 54dB

3rd 3rd

True ADC HD3: -94.6dBFs

Conventional method

Measured HD3: -88.1dBFs

Error 6.8%

Proposed method

Measured HD3: -92.6dBFs

Error 2.1%

38/41

0 20 40 60

0

3

6

9

12

15

18

Supurious@fs/2-fin attenuation [dB]

AD

C 3

rd h

arm

onic

s d

ete

cti

on e

rror

[%]

0 20 40 60

0

3

6

9

12

15

18

Supurious@fs/2-fin attenuation [dB]

AD

C 3

rd h

arm

onic

s d

ete

cti

on e

rror

[%]

0 20 40 60

0

3

6

9

12

15

18

Supurious@fs/2-fin attenuation [dB]

AD

C 3

rd h

arm

onic

s d

ete

cti

on e

rror

[%]

ADC HD3 Measurement Results

0 20 40 60

0

3

6

9

12

15

18

Supurious@fs/2-fin attenuation [dB]

AD

C 3

rd h

arm

onic

s d

ete

cti

on e

rror

[%]

0 20 40 60

0

3

6

9

12

15

18

Supurious@fs/2-fin attenuation [dB]

AD

C 3

rd h

arm

onic

s d

ete

cti

on e

rror

[%]

0 20 40 60

0

3

6

9

12

15

18

Supurious@fs/2-fin attenuation [dB]

AD

C 3

rd h

arm

onic

s d

ete

cti

on e

rror

[%]

Sample 1 Sample 2 Sample 3

Sample 4 Sample 5 Sample 6

conventional

phase switching ADC (AD7356) HD3 measurement error reduction is verified

Spurious @ fs/2-fin

attenuation [dB]

Spurious @ fs/2-fin

attenuation [dB]

Spurious @ fs/2-fin

attenuation [dB]

Spurious @ fs/2-fin

attenuation [dB] Spurious @ fs/2-fin

attenuation [dB]

Spurious @ fs/2-fin

attenuation [dB]

39/41

Outline

• Research background

• Proposed solution

• Problems of proposed solution

• Remedy 1

• Remedy 2

• Experimental results

• Conclusion

40/41

Low distortion sine signal generation

Without AWG hardware modification

Just AWG program change

and a simple analog LPF

Verified with

AWG (Agilent 33220)

6 samples of 12bit SAR ADC (AD7356)

Greatly improved quality of ADC Linearity testing

at virtually no extra cost.

Conclusions

41/41

Future Work

Generalization to other types of

low distortion sinusoidal signal generation

- HD2, HD2&HD3 cancellation for 1-tone

- IMD3 cancellation for 2-tone

We have partially verified these.

Detailed theoretical analysis, simulations,

and experiments are underway.