Does PNOISE sim of a sample & hold circuit include clock jitter effect?

Yes.

You can think of the time domain noise (tdnoise) mode of pnoise as inserting an ideal sampler at the particular time instant at the output of the circuit - and then doing a pnoise as normal. Consequently noise generated earlier in the period (before the sampling event) can still end up being propagated to the output (e.g. stored on device capacitance or real capacitors and then transferred to the output). Because there's an ideal sampler present though, make sure you do not sweep beyond half the PSS fundamental, otherwise you will get multiple counting of the noise due to aliasing (this is only an issue when in tdnoise or PMjitter modes, not in normal "sources" pnoise mode).

Regards,

Andrew

Hi Andrew,

Although this thread is 9 years ago, I find it is helpful to understand pnoise noise sim for switched-cap circuit.

As you said that the total output noise of a track and hold circuit includes the clock jitter from the clock buffer, I have a question on interpreting the noise summary report.

In the noise summary report, noise contribution is listed in a unit "V^2/Hz" or "V^2". What does it mean for the device in a clock buffer?

Does the simulator convert the jitter (in unit of s) contributed by clock buffer to output voltage noise (in unit of V^/Hz)? 

Thanks and regards,

Yutao

Dear YutaoLiu,

First and foremost, I am not Andrew and am not attempting to respond with his insights and knowledge! Nevertheless, I thought I would try to respond to your questions to be best of my ability, albeit far less than Andrew's!

YutaoLiu said:

In the noise summary report, noise contribution is listed in a unit "V^2/Hz" or "V^2". What does it mean for the device in a clock buffer?

The noise summary setup GUI allows for both "spot noise" and "integrated noise". The former is the noise, measured on a per Hz basis, for a particular offsets frequency whereas the latter is an integrated noise over a range of frequencies. Hence, the units of the result will be either in V^2/Hz or V^2 depending if the result is a spot noise or integrated noise respectively.

The options for the noise summary function are defined on page 262 of the Product Version ICADVM20.1, March 2021 Ocean reference manual at the Cadence  On-line support URL:

https://support1.cadence.com/tech-pubs/Docs/oceanref/oceanrefICADVM20.1/oceanref.pdf

and discusses the available summaries and their associated options. This may be helpful to your understanding YutaoLin.

YutaoLiu said:

Does the simulator convert the jitter (in unit of s) contributed by clock buffer to output voltage noise (in unit of V^/Hz)? 

I believe it does as the simulator output noise computation is based on the steady-state output waveform which will be dependent on the clock buffer as it controls the manner in which the track and hold switch introduces noise.

Is there a reason you think it will not be included?

Shawn

Dear YutaoLiu,

First and foremost, I am not Andrew and am not attempting to respond with his insights and knowledge! Nevertheless, I thought I would try to respond to your questions to be best of my ability, albeit far less than Andrew's!

YutaoLiu said:

In the noise summary report, noise contribution is listed in a unit "V^2/Hz" or "V^2". What does it mean for the device in a clock buffer?

The noise summary setup GUI allows for both "spot noise" and "integrated noise". The former is the noise, measured on a per Hz basis, for a particular offsets frequency whereas the latter is an integrated noise over a range of frequencies. Hence, the units of the result will be either in V^2/Hz or V^2 depending if the result is a spot noise or integrated noise respectively.

The options for the noise summary function are defined on page 262 of the Product Version ICADVM20.1, March 2021 Ocean reference manual at the Cadence  On-line support URL:

https://support1.cadence.com/tech-pubs/Docs/oceanref/oceanrefICADVM20.1/oceanref.pdf

and discusses the available summaries and their associated options. This may be helpful to your understanding YutaoLin.

YutaoLiu said:

Does the simulator convert the jitter (in unit of s) contributed by clock buffer to output voltage noise (in unit of V^/Hz)? 

I believe it does as the simulator output noise computation is based on the steady-state output waveform which will be dependent on the clock buffer as it controls the manner in which the track and hold switch introduces noise.

Is there a reason you think it will not be included?

Shawn

The clock buffer is nothing special - it doesn't generate "jitter" and it has to be converted to V/sqrt(Hz) or V^2/Hz. The simulator deals with noise voltages and currents, some component of which is amplitude modulated and some component of which is phase modulated. It computes the transfer function from each noise source to the output (where noise is being measured) and multiplies that by the amount of noise at each noise source, then adds up the noise powers (taking into account any correlations when it does so). You might then choose to visualise that as a conventional noise (in V/sqrt(Hz) or A/sqrt(Hz) or the equivalent powers), or you might convert it to a time-domain metric. The simulator itself is not modelling the noise sources as a jitter quantity though.

Does that help?

Andrew.