pnoise timedomain

Hi Svilen,

Pick "Sampled Phase" as follows:

Having set it up with the output node and the points to add etc, hit the Add button. You can add more than one sampled noise analysis this way.

Andrew

Thanks, Andrew, for the prompt reply. Here is an attachment of my setup - hopefully it is correct. One additional question. There used to be an option of also specifying the input voltage source or port to which I could refer the noise. Now I see only an output node. Can I still refer the noise to the input?

Dear Svilen64,

First, I am not Andrew and will not even pretend to provide his insight to your latest question! However, I will add my comments if you bear with my response!

Svilen64 said:

There used to be an option of also specifying the input voltage source or port to which I could refer the noise. Now I see only an output node. Can I still refer the noise to the input?

I believe the sampled jitter capabilities inherently can refer the edge jitter to the input edge - which I think is what you mean. In other words, is it really meaningful to compare the non-sampled noise of the input to the sampled jitter of the output? Hence, in the GUI for noise, set the trigger for the output voltage to a transition of the input. Does this make sense Svilen64 - or did I misunderstand your question?

A full description of the sampled jitter pnoise analysis may be accessed from the Cadence On-line support portal at URL: 

Spectre Circuit Simulator and Accelerated Parallel Simulator RF Analysis in ADE Explorer User Guide -- Single Input Large and Small-signal Analyses - Noise Type=sampled(jitter)

Shawn

Hi Shawn,

Thanks for your input. Yes, I know we are discussing the sampled jitter noise type but actually my initial question was about noise in sampled or switched capacitos setting - for example a sample and hold or SC integrator. So, I don't really need jitter type of a noise but rather just regular noise with all the folding effects because of switching. In this regard if we find the noise at the output of the circuit, we should be also able to refer it to the input. At least this was possible in the old GUI for pnoise 3-4 years back.

Regards
Svilen

Whilst the old form did allow you to specify an input source (this is from an IC616 ISR dated from about 6 years ago):

The pnoise analysis (when in time domain mode) ignored it - there was nothing on the "tdnoise" Direct Plot form, nor is there any input-referred noise in the results browser.

I did some searching and as far as I know, this has never been possible. It's questionable whether it is going to be useful either - the noise you are seeing at the sample point may originate from earlier times in the PSS period (it may have been sampled and held for example, or as a result of some delay in the circuit propagated to the output); input referred noise is normally computed by dividing the output noise by the gain - and if we computed the input referred sampled noise it would be the gain at that instant which may be completely irrelevant (there may be no gain at that instant in time). So which gain would it be expected to use?

Of course, if you know the gain you want, you can simply divide the sampled output noise by that gain (if there is some specific meaning in your case).

If you are computing the normal time-averaged noise (a.k.a "sources" in the past) then you can of course specify an input source. Rather unhelpfully in current versions you have to enable the "Noise Figure" checkbox to enable the specification of the input field (even though you're probably not measuring noise figure in a switch cap circuit, probably).

Overall I don't understand how you managed to do this in the past - are you sure you really did? As I said, as far as I can see from my checks this wasn't possible (I even found a request for it and a discussion similar to the above from R&D when I just looked). 

Regards,

Andrew

Hi Andrew,

I have not done referring noise to the input in the past but I remembered there was a filed there for the input source. But reading your explanations it probably doesn't make sense to refer to input in this scenario.
However, what you wrote makes me think of something else. Look at noise in two switched capacitor cases - a S/H and an integrator.
The S/H circuit basically  clears its memory from sample to sample (same for SC amplifier). And it makes sense to use pnoise in sampled mode because we care about noise samples.
Integrators accumulate samples over time forming the output. So, does it mean that for switched capacitor integrators it is better to use timeaverage noise type? Or even in this case sampled type is better?

Thanks
Svilen

Whether you use sampled noise or timeaverage typically would depend on how the output of the block is used. If the output is sampled (which is often the case with a sample and hold; it's going to be sampled by an ADC) then sampled makes sense as the noise in the non-sampled parts of the period (at the output) is not relevant to the subsequent stage. If going into a more continuous time circuit (such as how you might measure the noise on a spectrum analyser), then timeaverage makes more sense. So it's more about what follows the circuit than the circuit itself.

Put another way, the sampler in the sampled noise analysis is emulating the sampling of the subsequent stage.

Andrew

Understood. Thanks

Understood. Thanks