Input referred sampled pnoise

Dear Matthew Love,

Matthew Love said:

Is there a way of doing input referred sampled pnoise? When using the jitter measurement form, you can only specify the output. I manually tried to set the iprobe parameter however it said it was incompatible with the sampled mode.

Basically, the input referred noise is found by dividing the noise at the output by the gain of your block. There is an example in the ADC rapid adoption kit (RAK) that is discussed in this Forum post. I hope I understood your question and this helps Matthew!

https://community.cadence.com/cadence_technology_forums/f/custom-ic-design/47380/how-can-i-get-the-input-referred-noise-of-dynamic-comparator-by-pnoise

Shawn

Hi Shawn,

Thanks for the reply. The link is broken in your old post but from reading the replies the gist is to manually calculate the gain instead. I can do that with PXF but as I said, adding PXF increases my sim time from ~9 hours to ~22 hours. I'm wondering if there's any way to avoid that.

If I try and specify iprobe then Spectre prints:

WARNING (SPCRTRF-15471): The parameter iprobe will be ignored because the
parameters trigger and target are used to specify the ports in which
noise will be measured in sampled noise.

I'm just wondering why the use of sampling disables the input referring. Is there something special about sampling that makes traditional input referring invalid?

Hi Shawn,

Thanks for the reply. The link is broken in your old post but from reading the replies the gist is to manually calculate the gain instead. I can do that with PXF but as I said, adding PXF increases my sim time from ~9 hours to ~22 hours. I'm wondering if there's any way to avoid that.

If I try and specify iprobe then Spectre prints:

WARNING (SPCRTRF-15471): The parameter iprobe will be ignored because the
parameters trigger and target are used to specify the ports in which
noise will be measured in sampled noise.

I'm just wondering why the use of sampling disables the input referring. Is there something special about sampling that makes traditional input referring invalid?

Dear Matthew Love,

Matthew Love said:

The link is broken in your old post but from reading the replies the gist is to manually calculate the gain instead.

Rats! Sorry about that! Unfortunately, they have removed my access to internal documents and hence I could not check to verify the links were still valid. The link was pointing to a 2018 version of Cadence's rapid adoption kit for analog-to-digital converters entitled "ADC Verification Rapid Adoption Kit (RAK)". There is a more recent version and perhaps that is why the link is no longer functional.

Matthew Love said:

If I try and specify iprobe then Spectre prints:

WARNING (SPCRTRF-15471): The parameter iprobe will be ignored because the
parameters trigger and target are used to specify the ports in which
noise will be measured in sampled noise.

I'm just wondering why the use of sampling disables the input referring. Is there something special about sampling that makes traditional input referring invalid?

Well, I can give you my guess, but I can not speak for Cadence. I think the error message when trying to specify an iprobe is a clue. When doing a sampled jitter analysis, the trigger is defined as the reference and hence serves as the reference for the measured jitter at the time of interest which is expressed relative to the trigger). Hence, the trigger is serving as the single input from which to refer the noise. 

However, I would rely on Cadence Support to provide a firm answer as I am not a Cadence employee and not privy to their design decisions.

Shawn