The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl Simulators >> RF Simulators >> Noise analysis of asynchronous SAR https://designers-guide.org/forum/YaBB.pl?num=1478278025 Message started by deba on Nov 4th, 2016, 9:47am |
Title: Noise analysis of asynchronous SAR Post by deba on Nov 4th, 2016, 9:47am Hi All, For asynchronous SAR ADC noise analysis, I am using transient noise analysis to verify the circuit performance. But it doesn't list out the noise contributors as other small signal noise analysis prints out. Is there a way by which one can get noise contributors using transient noise? Regards Debasish |
Title: Re: Noise analysis of asynchronous SAR Post by Ken Kundert on Nov 5th, 2016, 10:30pm Perhaps you should try PNoise analysis rather than using transient noise. -Ken |
Title: Re: Noise analysis of asynchronous SAR Post by deba on Nov 7th, 2016, 1:54am Hi Ken, Pnoise will not work for the asynchronous SAR. The comparator is not clocked synchronously. There is a noiseon parameter, I guess I need to run multiple .trannoise sims, and keep turning off noise sources one by one to find the dominant one. |
Title: Re: Noise analysis of asynchronous SAR Post by Ken Kundert on Nov 7th, 2016, 4:57pm Quote:
Why does that matter? -Ken |
Title: Re: Noise analysis of asynchronous SAR Post by deba on Nov 8th, 2016, 7:49am Hi Ken, I am doing noise analysis on the complete SAR. Since the comparator is clocked asynchronously, it has no relation to the sampling clock. It's non-periodic, thus PSS will not converge. |
Title: Re: Noise analysis of asynchronous SAR Post by Ken Kundert on Nov 9th, 2016, 6:58pm I am assuming that if the input is the same on each conversion, and the fundamental period equals the conversion interval, then the response on each conversion will be the same and so the waveforms will be periodic. But perhaps there is some aspect about the architecture that I do not understand. -Ken |
Title: Re: Noise analysis of asynchronous SAR Post by deba on Nov 9th, 2016, 8:32pm Hi Ken The SAR uses monotonic switching down sequence. The input common mode of the comparator keeps going down as the SAR conversion progresses. Thus, the comparator conversion time varies from cycle to cycle, resulting in non-periodic behaviour. Thanks for your inputs. |
Title: Re: Noise analysis of asynchronous SAR Post by Ken Kundert on Nov 10th, 2016, 5:19pm Wouldn't the whole ADC behave periodically in N cycles, where N is the number of cycles needed to do a conversion? -Ken |
Title: Re: Noise analysis of asynchronous SAR Post by deba on Nov 11th, 2016, 12:02am Hi Ken, As the input common mode for the comparator keeps dropping down, the conversion time is different for each conversion cycle. Thus, a non-periodic behaviour. A 10-bit 50-MS/s SAR ADC With a Monotonic Capacitor Switching Procedure. IEEE Journal of Solid-State Circuits ( Volume: 45, Issue: 4, April 2010 ) |
Title: Re: Noise analysis of asynchronous SAR Post by Ken Kundert on Nov 11th, 2016, 7:40pm According to the flow chart in figure 3, a conversion lasts N cycles. If the input is constant, isn't the converter behavior periodic in N cycles? -Ken |
Title: Re: Noise analysis of asynchronous SAR Post by deba on Nov 12th, 2016, 7:04am Hi Ken, Fig. 4(b) of the paper shows the behaviour of comparator inputs as the conversion progresses. As the common mode keep changing, the input gm of the transistor keeps changing from cycle to cycle. Thus, even with a constant input, the comparator clock is different for each conversion. |
Title: Re: Noise analysis of asynchronous SAR Post by Ken Kundert on Nov 12th, 2016, 9:28pm Don't the waveforms in Fig. 4 repeat after N phases, where N is resolution of the converter in bits? Perhaps we are getting tripped up in our terminology. When I say conversion time, I mean the time it takes for one input to be fully converted to digital. Thus, a conversion requires N phases, where each phase lasts the length of a clock cycle. So for a 10-bit converter running with a 100MHz clock, the conversion time would be 100ns, or 10 times the 10ns clock period). Given this example, wouldn't the waveforms in Fig. 4 be periodic with a period of 100ns? -Ken |
Title: Re: Noise analysis of asynchronous SAR Post by deba on Nov 14th, 2016, 5:25am Hi Ken, I got your point. Yes the comparator waveforms will be periodic. For your example, period will be 100 ns. I also checked the PSS does converge for the complete SAR. This is great. Thanks. Now I need to understand how to translate the noise seen at the comparator output, to the ADC noise spectrum. This has opened up a new perspective. |
Title: Re: Noise analysis of asynchronous SAR Post by deba on Nov 19th, 2016, 7:56pm Hi Ken and all, After establishing that the SAR has a PSS for constant DC input(in my sim it is 0). I did a strobed pnoise analysis for the comparator's output. The strobing is such that the comparator waveforms are in the exponential settling regime. Now I get 13-pnoise plots (for a 13-bit converter) for every conversion cycle. I am still trying to figure out how to relate this to the ADC spectrum plot? Do you have any suggestions on how to do this? On a separate note, it is not related to this thread. But I will ask anyways. For other ADC architectures, like pipeline, flash, algorithmic ADC etc. one can also do a PSS/PNOISE analysis. But in most of the appnotes, I have seen only transient noise is the recommended method for measuring noise performance. There seems to be very few papers which talk about applying SpectreRF analysis to ADC's. Mostly they are limited only to simple cases like noise analysis of a dynamic comparator. Do you have any comments on this? Thanks |
Title: Re: Noise analysis of asynchronous SAR Post by zeropond on Apr 9th, 2017, 8:31am Hello ;D I am zeropond Iam a new member of this forum site. but I have nothing any answer to your question about noise analysis of asynchronous SAR. so please ask to your friends. Thank you :) |
Title: Re: Noise analysis of asynchronous SAR Post by BackerShu on May 12th, 2017, 12:52am deba wrote on Nov 19th, 2016, 7:56pm:
This is an interesting point. My feeling is that although it is not common in papers/application notes, SpectreRF is widely used for clocked circuits. I think it will be a good idea that you open a new post for this discussion. |
The Designer's Guide Community Forum » Powered by YaBB 2.2.2! YaBB © 2000-2008. All Rights Reserved. |