The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Input Referred Noise with "timedomain" PNOISE
https://designers-guide.org/forum/YaBB.pl?num=1428049161

Message started by Rinoz on Apr 3^rd, 2015, 1:19am

Title: Input Referred Noise with "timedomain" PNOISE
Post by Rinoz on Apr 3^rd, 2015, 1:19am

Hi all,

I would ask you a help regarding a simple SC circuit. Attached you can find the circuit.

Target of my simulation is to find the Input Referred Noise in [0:Fs/2] at the first input Sv_in_TOT of the KTC and OPAMP contributions, using the PSS and the PNOISE (bottom picture).

I'm able to compute the OUTPUT referred noise at the output of the OPAMP with the timedomain option, but the problem is that with this option spectre doesn't compute the INPUT Referred Noise.

Why with timedomain option is not possible to see the input referred noise but only the output referred noise density?

I've read the paper of Ken Kundert about to compute the noise in SC filters and he advices to substitute the sampling process done by:

- timedomain option
- number of points = 1
- add specific point = 0

with the ideal S/H and using the psnoise standard (no timedomain).

But if I use this method, how can I remove the sinc(x) shaping done by the S/H to the spectral noise density?

Do you have any other solutions?

Thanks in advance, if you can find a minute for helping me,

I'm looking forward to hearing you,
Have a nice day at all,

Rino

Title: Re: Input Referred Noise with "timedomain" PNOISE
Post by Ken Kundert on Apr 3^rd, 2015, 11:39am

Did you specify the input source to the noise analysis?

-Ken

Title: Re: Input Referred Noise with "timedomain" PNOISE
Post by Rinoz on Apr 7^th, 2015, 9:26am

Hi Ken,

thanks for your answer.

Yes, I've specified the input noise source as "Input Voltage Source",

Ciao,
Rino

Title: Re: Input Referred Noise with "timedomain" PNOISE
Post by Ken Kundert on Apr 7^th, 2015, 12:17pm

I'm afraid that is all I have. You might ask Cadence why they are not generating the input referred noise in this case. I suspect it is because of over-generalization on Cadence's part. What most designer's want is 'sampled noise' or noise at a particular time. What Cadence provides is an over generalization of this: noise at many time points. Usually this is just a minor annoyance, forcing you to work though their awkward UI to specify that you only want one point. However, I suspect their over generalization causes them to disable input referred noise because it would be awkward to generated and display input referred noise at each output point.

So if you ask Cadence about this, you might want to suggest that they just support sampled noise rather than trying to compute input referred noise for every time-domain-noise sample point. You might also suggest that 'time-domain noise' is a misnomer. It is really 'noise versus sample time'.

-Ken

Title: Re: Input Referred Noise with "timedomain" PNOISE
Post by Rinoz on Apr 8^th, 2015, 12:33am

Hi Ken,

But can you suggest me another way to do this?
Do you think that a PAC or PXF could be used to refer the output noise to the input through a computed gain?

Or, Could I use your ideal sampler to sample the noise to the output and use the common pnoise (no timedomain)? If yes, how can I remove the sinc shape of the S/H?

I'm sorry for my question but I don't see a viable way to resolve my problem :(

Have a nice day,
Rino

Title: Re: Input Referred Noise with "timedomain" PNOISE
Post by Frank Wiedmann on Apr 8^th, 2015, 3:51am

I guess you could use sampled PXF to calculate the input-referred noise from your sampled output noise. I usually prefer to work with output noise and gain directly. With input-referred noise, you might think that you have a noise problem when you really have a gain problem.

Title: Re: Input Referred Noise with "timedomain" PNOISE
Post by Ken Kundert on Apr 8^th, 2015, 10:24am

Rino,
Either approach should work. Most of my experience with this kind of thing comes from before the time SpectreRF provided the modulation add-ons to the RF analyses (AM, PM, sampled, jitter, etc.) so I had to make due with ideal samplers and limiters and such using Verilog-A. Doing so generally worked out pretty well for me. But it would probably be easier to follow Frank's suggestion of using sampled PXF.

-Ken

Title: Re: Input Referred Noise with "timedomain" PNOISE
Post by Rinoz on Apr 15^th, 2015, 5:14am

Many thanks at all,

I've reached the target using PSS + sampledPNOISE + sampledPXF

A note probably useful for who wants to simulate the KTC noise using Pnoise is:
- do not use "port" as input source because spectre will compute the noise figure instead the input referred noise (noisy source). Use "Voltage source" instead;

- if you want to use the ideal "switch" of cadence AnalogLib you must insert a series resistance because I saw that the resistance inserted in the "switch" block probably doesn't generate the 4KTR noise

many thanks at all,
Rino