The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> RF Simulators >> Differences in sources or timedomain PNoise
https://designers-guide.org/forum/YaBB.pl?num=1220976977

Message started by Tlaloc on Sep 9^th, 2008, 9:16am

Title: Differences in sources or timedomain PNoise
Post by Tlaloc on Sep 9^th, 2008, 9:16am

I have been trying to verify that I am running some pnoise simulations correctly but am baffled by something. I have a simple integrator using an ideal amp with finite gain but infinite bandwidth. The switches are non-overlapping, and the INIT signal is only to quick-start the sim. I have two ideal track and holds using the code from this website to make the integrator closed-loop. The first has timing at the end of P2 and the second has timing at the end of P1. The period of the clocks (tper) is 122.88kHz with a 50% non-overlapping duty cycle.

My assumption is that a timedomain pnoise sim from 0 to 1/(2*tper) at the output of the amp (OUT) should result in the same number as a sources sim from 0 to inf after the first SH (SOUT). Everything about the noise simulations are otherwise identical, i.e. 200 maxsidebands and maxacfreq at 500/tper. My thought was that timedomain essentially performed the sample and hold that I am explicitly doing with the veriloga module. However, the simulations give me different results (integrated root noise).

sources: 11.38uV (output between SOUT and vss)
timedomain: 9.52uV (output between OUT and vss at time zero, i.e. end of P2)

None of the time-constants have changed, so what am I missing?

Title: Re: Differences in sources or timedomain PNoise
Post by pancho_hideboo on Sep 9^th, 2008, 11:06pm

Tlaloc wrote on Sep 9^th, 2008, 9:16am:

My assumption is that
a timedomain pnoise sim from 0 to 1/(2*tper) at the output of the amp (OUT)
should result in the same number as
a sources sim from 0 to inf after the first SH (SOUT).

I don't know your situation correctly. But I can't understand why you think so.

Tlaloc wrote on Sep 9^th, 2008, 9:16am:

sources: 11.38uV (output between SOUT and vss)
timedomain: 9.52uV (output between OUT and vss at time zero, i.e. end of P2)

"timedomain: 9.52uV" is ensemble averaged noise of amplifier only at time zero.
On the other hand, "sources: 11.38uV" is time averaged noise over all time during one period even though you use ideal S/H circuit.

Also in this case, ergotic process assumption can't be applied for OUT and SOUT although I don't know your situation correctly.

Title: Re: Differences in sources or timedomain PNoise
Post by Tlaloc on Sep 10^th, 2008, 8:02pm

I guess my confusion came from a line in http://www.designers-guide.org/Analysis/sc-filters.pdf. Quoting from page 13:

Quote:

Conceptually you can imagine SpectreRF sampling the output waveform every Tc seconds starting at t = 0 and then reporting the power spectral density of this sequence, Ss, which is shown in Figure 5.

From this I assumed that a sources pnoise taken after an ideal S/H that samples at the same time that a timedomain pnoise is sampled should result in the same answer.

As I was reviewing the document above, I did notice this time that the sources pnoise returns S_c while the timedomain returns S_s. If I understand, then, what Kundert is saying is that the timedomain noise will only report the sampled noise power while the sources noise will also include the direct filtered noise at the output. Is this correct? If so, when is a timedomain noise sim worthwhile if it is only a subset of the total noise?

By the way, which assumption assumed an ergodic process? Finally, what are some other references on this subject that could help clarify my understanding? Thanks.

Adam

Title: Re: Differences in sources or timedomain PNoise
Post by pancho_hideboo on Sep 10^th, 2008, 8:45pm

Tlaloc wrote on Sep 10^th, 2008, 8:02pm:

If I understand, then, what Kundert is saying is that the timedomain noise will only report the sampled noise power while the sources noise will also include the direct filtered noise at the output. Is this correct ?

Correct. But a sampling of tdnoise in Cadence Pnoise is impulse sampling.

Tlaloc wrote on Sep 10^th, 2008, 8:02pm:

If so, when is a timedomain noise sim worthwhile if it is only a subset of the total noise?

You can see noise at specific time.
For example, you can observe noise of Gilbert Cell Mixer over local signal period. Changing duty of local or changing local signal shape, e.g. sinusoidal, rectangular pulse, you can observe noise behaivior over one local period.

But here this is not same as real spectrum analyzer where short time interval FFT is used.
Many people think of tdnoise same as real spectrum analyzer. This is misunderstanding.

Tlaloc wrote on Sep 10^th, 2008, 8:02pm:

By the way, which assumption assumed an ergodic process?

I can't understand a meaning of your question.
Noise at inverter output under large signal drive is not ergodic.
On the other hand, noise at inverter output under no signal drive is ergodic.

See the following.
http://www.designers-guide.org/Forum/YaBB.pl?num=1207830622
http://www.designers-guide.org/Forum/YaBB.pl?num=1059089369
http://www.designers-guide.org/Forum/YaBB.pl?num=1218622880