Using Distortion Shaping Technique to Equalize ADC THD Performance Between ATEs
Download as PPTX, PDF1 like263 views
This document discusses using distortion shaping techniques to equalize the THD performance of ADCs between different automatic test equipment (ATEs). It presents a method of generating test signals with an arbitrary waveform generator (AWG) that shifts harmonic distortion components closer to the Nyquist frequency to improve the AWG's harmonic performance and allow more accurate testing of ADC THD specifications. Test results show the proposed method significantly reduces harmonic distortion in the AWG output and allows different testers to achieve identical ADC test results, resolving issues caused by performance differences between ATEs. The technique is especially effective for higher resolution ADCs where noise floors are lower.
![Confidential © ams AG
Page 7
Conventional Signal Generation with AWG
AWG
DAC
CLK
DSP
X XXX
Din
Din
AWG sampling frequency: fs(AWG) = 1/Ts
A
𝐗= Acos(2πf_in nT_s)
0 0.1 0.2 0.3 0.4 0.5
-300
-200
-100
0
Power[dB]
Normalized frequency f/fs
𝐟𝐢𝐧
𝐇𝐃𝟑](https://image.slidesharecdn.com/imstw16sarson-160719062118/85/Using-Distortion-Shaping-Technique-to-Equalize-ADC-THD-Performance-Between-ATEs-7-320.jpg)
![Confidential © ams AG
Page 8
Proposed Signal Generation with AWG
AWG
DAC
CLK
DSP
X0 X1X0X1
Din
Din
AWG sampling frequency: fs(AWG) = 1/Ts(AWG)
𝐗 𝟎= 1.15Acos(2πf_in nT_s−π/6)
𝐗 𝟏=1.15Acos(2πfin nTs+𝜋/6)
0 0.1 0.2 0.3 0.4 0.5
-300
-200
-100
0
𝐟𝐢𝐧
𝐟𝒔/2−𝟑𝐟in
𝐟𝒔/2−𝐟in
Power[dB]
Normalized frequency f/fs
Θ = π/3𝐓𝐬(𝐀𝐖𝐆)
X0
𝐗 𝟏](https://image.slidesharecdn.com/imstw16sarson-160719062118/85/Using-Distortion-Shaping-Technique-to-Equalize-ADC-THD-Performance-Between-ATEs-8-320.jpg)
Using Distortion Shaping Technique to Equalize ADC THD Performance Between ATEs
- 1. Confidential © ams AG Using Distortion Shaping Technique to Equalize ADC THD Performance Between ATEs IMSTW 16 Peter Sarson CMgr MCMI SMIEEE Full Service Foundry 4th July 2016 Prof Hauro Kobayashi Gunma University
- 2. Confidential © ams AG Page 2 Introduction • During development of any test program test engineers sometimes face the issue that the tester isn’t as good as the chip • This causes issues when device specs need testing and the limit cannot be reached • Kobayashi will show some mathematics: harmonic distortion contents can be shifted up close to Nyquist frequency. • Peter will show how the harmonic performance of an ATE’s AWG can be improved using this approach • Peter will show how this approach improved the result and reproducibility of a 12-bit ADC testing • Peter will show how this approach improved the result of a 16-bit ADC testing that was limited by the tester performance
- 3. Confidential © ams AG Page 3 Research Objective Low distortion signal generation with AWG for low cost and high quality testing of ADC Investigation of phase switching signal generation method with experiment at ATE environment • No need for AWG nonlinearity identification • Only DSP program change
- 4. Confidential © ams AG Page 4 ADC test signal Test result ADC Testing with Sine Signal Freq. ADC HD3 Freq. “No Go” “Go” Freq. Reference level Allowable fundamental Freq. Test result fundamental AWG ADC 𝑍 = 𝑏1 𝑌 + 𝑏3 𝑌3 fin 3fin fin 3fin fin 3fin fin
- 5. Confidential © ams AG Page 5 ADC test signal Test result Problem of Conventional Method AWG HD3+ADC HD3. Freq. fundamental Freq. Test result AWG ADC 𝑍 = 𝑏1 𝑌 + 𝑏3 𝑌3 fundamental fin 3fin fin Test signal generation Y=a1Din + a3Din 3 3fin AWG HD3 HD3 ADC HD3 cannot be measured accurately Low quality test
- 6. Confidential © ams AG Page 6 ADC test signal Test result Research Goal Only ADC HD3 Freq. fundamental Freq. Test result AWG ADC 𝑍 = 𝑏1 𝑌 + 𝑏3 𝑌3 fundamental fin 3fin fin Test signal generation Y=a1Din + a3Din 3 3fin AWG HD3 reduction HD3 High quality testOnly program change
- 7. Confidential © ams AG Page 7 Conventional Signal Generation with AWG AWG DAC CLK DSP X XXX Din Din AWG sampling frequency: fs(AWG) = 1/Ts A 𝐗= Acos(2πf_in nT_s) 0 0.1 0.2 0.3 0.4 0.5 -300 -200 -100 0 Power[dB] Normalized frequency f/fs 𝐟𝐢𝐧 𝐇𝐃𝟑
- 8. Confidential © ams AG Page 8 Proposed Signal Generation with AWG AWG DAC CLK DSP X0 X1X0X1 Din Din AWG sampling frequency: fs(AWG) = 1/Ts(AWG) 𝐗 𝟎= 1.15Acos(2πf_in nT_s−π/6) 𝐗 𝟏=1.15Acos(2πfin nTs+𝜋/6) 0 0.1 0.2 0.3 0.4 0.5 -300 -200 -100 0 𝐟𝐢𝐧 𝐟𝒔/2−𝟑𝐟in 𝐟𝒔/2−𝐟in Power[dB] Normalized frequency f/fs Θ = π/3𝐓𝐬(𝐀𝐖𝐆) X0 𝐗 𝟏
- 9. Confidential © ams AG Page 9 Principle of 3rd Harmonics Cancellation Two waves with phase difference π are cancelled ConventionalPhaseSwitching Θ = π/3 3Θ = π 3rdordernon-linearsystem Phaserotationbyx3 fundamental: fin 3rd harmonics: 3 fin
- 10. Confidential © ams AG Page 10 X0=Asin(2πfint+π/4) X1=Asin(2πfint-π/4) HD2 Cancellation 2-way interleave cancels HD2
- 11. Confidential © ams AG Page 11 X0=Asin(2πfint-π/4-π/6) X2=Asin(2πfint+π/4-π/6) X1=Asin(2πfint-π/4+π/6) X3=Asin(2πfint+π/4+π/6) HD2, HD3 Cancellation 4-way interleave cancels HD2 & HD3
- 12. Confidential © ams AG Page 12 Be Careful If you use this technique without filtering the generated high-frequency spurious Harmonic contents of the ADC under test will be under-estimated. These spurious components need filtering out to get back normal sinewave without harmonics. More detailed later.
- 13. Confidential © ams AG Page 13 How to Do in Software 2-signal interleaved Element 1 Element 2 Element 3 Element 4 Element 5 Element 6 Element 7 Element 8 Array 1 1 2 3 4 5 6 7 8 Array 2 9 10 11 12 13 14 15 16 Supressed 1 10 3 12 5 14 7 16
- 14. Confidential © ams AG Page 14 How to Do in Software 4-signal interleaved Element 1 Element 2 Element 3 Element 4 Element 5 Element 6 Element 7 Element 8 Array 1 1 2 3 4 5 6 7 8 Array 2 9 10 11 12 13 14 15 16 Array 3 17 18 19 20 21 22 23 24 Array 4 25 26 27 28 29 30 31 32 Supressed 1 10 19 28 5 14 23 32
- 15. Confidential © ams AG Page 15 Test Setup for Measuring Tester Performance AWG MEMORY AWG FILTER 50MHz DIGITIZER FFT Tester Result Loopback
- 16. Confidential © ams AG Page 16 Tester 1 3rd Harmonic Reduction Measured spectrum level improvement of reference vs suppressed 3rd harmonic for AWG.
- 17. Confidential © ams AG Page 17 Tester 1: 3rd and 5th Harmonic Reduction Measured spectrum levelimprovement of reference vs suppressed 3rd & 5th harmonicsforAWG.
- 18. Confidential © ams AG Page 18 Harmonic Reduction Measured Harmonic Performance Improvement for AWG Using the Proposed Method Harmonic Improvement 2nd 6 dB 3rd 16 dB 4th 18 dB 5th 23 dB
- 19. Confidential © ams AG Page 19 ADC Performance Test Setup AWG MEMORY AWG FILTER 50MHz DUT ADC DIGITIZER FFT Tester Result
- 20. Confidential © ams AG Page 20 12-bit ADC Results Tester 1 signal Thd std units ref 89,7 2,5 dB 2nd_red 89,8 2,1 dB 3rd_red 89,3 2 dB 4th_red 89,5 2,1 dB 5th_red 89,2 2,6 dB 2nd,4th_red 91 2,4 dB 3rd,5th_red 89,7 2,5 dB
- 21. Confidential © ams AG Page 21 12-bit ADC Results Tester 1 signal Thd std units ref 89,7 2,5 dB 2nd_red 89,8 2,1 dB 3rd_red 89,3 2 dB 4th_red 89,5 2,1 dB 5th_red 89,2 2,6 dB 2nd,4th_red 91 2,4 dB 3rd,5th_red 89,7 2,5 dB
- 22. Confidential © ams AG Page 22 12-bit ADC Results Tester 1 signal Thd std units ref 89,7 2,5 dB 2nd_red 89,8 2,1 dB 3rd_red 89,3 2 dB 4th_red 89,5 2,1 dB 5th_red 89,2 2,6 dB 2nd,4th_red 91 2,4 dB 3rd,5th_red 89,7 2,5 dB
- 23. Confidential © ams AG Page 23 12-bit ADC Harmonic Suppressed Result 2nd and 4th suppressed delta and overall THD Improvement THd ref THd supp Tester 2 88,7dB 91dB Tester 1 89,6dB 91dB Without suppression: • Different measured values of THD performance With Suppression: • Exactly the same result between Testers 1 & 2
- 24. Confidential © ams AG Page 24 Site to Site Comparison No difference between sites with the proposed method! 12-bit ADC testing case Tester 1 Site 1 Site 2 delta Red 91dB 91,1dB 0,1dB ref 89,6dB 89,6dB 0dB
- 25. Confidential © ams AG Page 25 3rd, 5th and 7th Suppressed
- 26. Confidential © ams AG Page 26 How 8-signal interleaved looks like 2^3 = 8
- 27. Confidential © ams AG Page 27 3-Harmonics (3rd , 5th & 7th Harmonics) Suppression Using 3-harmonics suppression algorithm, effect of the sidebands would be SB= fs/8 – 7*fin Maximum sampling freq. (fs) of AWGHSB : 250MHz Signal freq. (fin): 1MHz Filter with 24MHz cutoff is needed Not available on AWGHSB Need a load-board filter.
- 28. Confidential © ams AG Page 28 OVERSAMPLE = 1 FS: 259.2000 kHz THD = -89.83dB Another Example – Tester 1 (Conventional) Obelic 16-bit ADC – no suppression, direct conversion
- 29. Confidential © ams AG Page 29 OVERSAMPLE = 1 FS: 259.2000 kHz THD = -89.83dB Another Example – MX01 (Conventional) 16-bit ADC – no suppression, direct conversion
- 30. Confidential © ams AG Page 30 OVERSAMPLE = 600 FS: 155.5200 MHz THD = -115.09dB MUCH BETTER Obelic 16-bit ADC – 2nd Harmonic Suppressed (Proposed) Oversample 600 times – Tester 1
- 31. Confidential © ams AG Page 31 OVERSAMPLE = 600 FS: 155.5200 MHz THD = -113.87dB Same as Tester 1 16bit ADC – 2nd Harmonic Suppressed (Proposed) Oversample 600 times – Tester 2
- 32. Confidential © ams AG Page 32 Performance summary Using a 12-bit ADC, testing performance can be improved but you are limited to the noise performance of the ADC itself • If spec limits are an issue, the technique can help squeeze out some margin Using a 16-bit ADC, performance can be significantly improved (say, by 25 dB) as the noise floor is much lower due to the number of bits. Effectiveness of the distortion-shaping technique was verified with ATE environment.
- 33. Confidential © ams AG Thank you Please visit our website www.ams.com
- 34. Confidential © ams AG Page 34 Appendix Proposed signal generation method • Distortion components close to signal band are reduced • Distortion components far from signal band may appear. Similar to but different from noise-shaping Distortion-shaping Normalized frequency f/fs