1 / 22

Kostas Georgopoulos, Martin Burbidge, Andreas Lechner and Andrew Richardson

Behavioural Error Injection, Spectral Analysis and Error Detection for a 4 th order Single-loop Sigma-delta Converter Using Walsh transforms. Kostas Georgopoulos, Martin Burbidge, Andreas Lechner and Andrew Richardson. Presentation Overview. The Sigma-Delta A/D Converter

olin
Download Presentation

Kostas Georgopoulos, Martin Burbidge, Andreas Lechner and Andrew Richardson

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Behavioural Error Injection, Spectral Analysis and Error Detection for a 4th order Single-loop Sigma-delta Converter Using Walsh transforms Kostas Georgopoulos, Martin Burbidge, Andreas Lechner and Andrew Richardson

  2. Presentation Overview • The Sigma-Delta A/D Converter • The Walsh functions and Walsh series • Motivation for work • The FFT error simulation analysis • The Walsh error simulation analysis • Conclusions • Future Work

  3. The Sigma-Delta A/D Converter (I) • A device comprised by three stages • Anti-aliasing filter • Sigma-Delta modulator • Decimation phase

  4. The Sigma-Delta A/D Converter (II) • Sampling at a frequency much higher then the Nyquist , where fs is sampling frequency and fiis the input frequency

  5. The Sigma-Delta A/D Converter (III) • Transfer function: , where L is the order of modulator • Noise floor is moved out of bandwidth of interest by noise shaping Bandwidth of interest, 0-24 kHz

  6. The Walsh Functions (Theory) • Walsh functions form an ordered set of rectangular orthogonal waveforms • Only two amplitude values, +1 and –1 • Fast Walsh transforms exist • Any given signal can be represented through the combination of two or more Walsh functions

  7. The Walsh Series (I) • The Walsh series is similar to the Fourier Series expansion where parameter α determines the amplitude or weighting of each Walsh function and

  8. The Walsh Series (II) • Walsh functions can also be expressed in terms of even and odd waveform symmetry , where Walsh functions SAL and CAL can be visualised as the respective sine and cosine basis functions in Fourier Series

  9. The Walsh Series (III) • Employing SAL and CAL functions a Walsh Series similar to the sine-cosine series is given where f(t) is the sum of a series of square-wave shaped functions

  10. Signal Reconstruction using Walsh • A simple case of a sine-wave signal approximated with 3 Walsh functions

  11. Motivation & Methodology • FFT converges rapidly to sine wave hence use for classic dynamic performance testing • Walsh converges rapidly to square wave: • Idea to use square wave for input to modulator • Walsh transform of bit-stream should give single spectral peak • All other peaks in spectrum are due to noise and non-idealities • Higher potential for on-chip transform of fewer samples Methodology: • Determine modulator behavior and model parameters that lead to performance failure in FFT domain • Analyse effect of these failure modes on Walsh results

  12. FFT Analysis Setup • Use of initial C-based model provided by Dolphin • Ideal model FFT S/(N + THD) results: • Input 2.5 Vpk sine @ 1 kHz • BW approx 24 kHz (150 Hz to 24 kHz) • S/(N+THD) approx 100dB • Next step: Analyse how Walsh transforms could compare to FFT

  13. Fault Set For FFT Faulty capacitor • Input Offsets • Integrator Gain Variations • Corrupted feedback paths • Presence of noise on the modulator input • Integrating capacitor mismatch Gain mismatch

  14. Effect of Integrator Offset • 40 mV offset on the modulator input, S(N/THD) 99,6 dB • 50 mV offset on the modulator input, 50 dB drop in S(N/THD) Bandwidth of interest, 46 – 24 kHz

  15. Analyses of Walsh Testing • Setup: • Square wave test stimulus, @ 1.5kHz, 1.9 to 2.3 V amplitude • Bit-stream frequency of 3.072 MHz, analysis on the bit stream with 16384 and 65536 (1-bit) samples. • Analyses: • Investigation into test stimulus accuracy requirements • Assessment of Walsh-based modulator performance tests • Analysis of Walsh test coverage against modulator failure modes

  16. Test Stimulus Accuracy (I) • Finite rise/fall time: No significant effect • Overshoots: 4% of maximum amplitude, i.e. 0.08 V for 2 V input 16384 samples

  17. Test Stimulus Accuracy (II) • SNR with respect to different input amplitudes Analysis for both 16384 and 65536 samples

  18. Ideal Walsh Sequency Power Spectrum Walsh transform for 2Vpk square wave @ 1.5 kHz 0 dB -12 dB SNR = 95 dB SNR = 107 dB 16384 samples -120 dB -130 dB 65536 samples BW = 24 kHz (0 kHz to 24 kHz)

  19. Gain Deviation in 2nd Integrator • Gain deviation of 7.4% 0 dB Deviated output -72 dB -120 dB Ideal BW = 24 kHz (0 kHz to 24 kHz)

  20. Walsh Noise Test (I) Noise modelled at input can cause catastrophic failure (SNR ~30dB) 0 dB Catastrophic failure -130 dB For both cases N = 65536 Ideal BW = 24 kHz (0 kHz to 24 kHz)

  21. Walsh Noise Test (II) • FFT – Smoother transition to performance failure • Walsh – Sudden transition to catastrophic failure

  22. Summary and Future Work Summary • Failure Insertion in C-based models => FFT results • Usage of Walsh Transforms with square wave inputs for spectral analysis • Initial Potential for Walsh SNR test assessed • Test Stimulus Requirements • Challenges and Limitations Identified Future Work • Expansion of existing fault simulation data for Walsh applicability • Investigation into hardware implementation and test stimulus generation • Investigation into hybrid test solution

More Related