Simple kT/C Noise simulation in Cadence - V/sqrt(hz) instead of V^2/Hz

[pifouille]

Hi everybody,

I'm learning how to perform noise simulation in Cadence/Spectre.
I've designed an simple RC network (R=1k, C=1p) and then performed a noise simulation between
1 and 100GHz and set as the output the voltage around C.

I would expect the famous shaped DSP with 4kTR (16.5e-18 V^2/Hz @300K) at low frequency.
And spectre actually shows this almost exact value (I guess it uses 27ºC instead of 300K)
but my confusion is that it gives this value in V/sqrt(Hz)...

Shouldn't it be sqrt(16.5e-18) instead of 16.5e-18??

thanks!

 

[Dominik Przyborowski]

What you are expecting is power while simulator can give you amplitude as well. Both are available in results.

 

[pifouille]

Thanks a lot for your answer, even if I didn't really get it

I'm doing a very simple noise analysis on the circuit attached (rc_noise.png).
I set output noise as voltage between out and gnd and runned the simulation.
Going thru Direct Plot -> Main form, I can choose either V^2/Hz or V/√Hz
in this case, I got the result I expect: 16.5e-18 V^2/Hz or 4 nV/√Hz depending on my choice.

However, if instead of using Direct Plot -> Main form, I symply plot the out voltage (selecting plot from schematics),
I get the graph shown in rc_noise_dsp.png which shows in 16.5e-18 V/√Hz.
This is the result I would expect in V^2/Hz.

I understand it might not make a lot of sense to plot an output voltage in a noise simulation,
I just don't clearly see why...
As I said, I'm trying to understand..

Thanks again anyway for helping.

--- Updated Mar 6, 2021 ---

I actually thinks it some kind of bug cause it's clearly plotting VN2().. it should show V^2/Hz on the y-axis..
If I define my simulation outputs as 2 expressions VN() and VN2(), again, I get the right results with the right dimensions.
anyway. I was just curious.