Hi,
I want to use the resistor as a thermal noise source.
I use spectres to simulate the thermal noise of the resistor, and make a noise source V²=4KTR serial with the ideal resistor to act as the thermal noise as showed below.

RES is the noise source. And it is verilog-A code is:

// VerilogA for random, rnoise, veriloga

`include "constants.vams"
`include "disciplines.vams"

module rnoise(p, n);

inout p, n;
electrical p, n;
parameter real r = 50k;
analog begin
V(p, n) <+ white_noise(4 * `P_Q * $temperature * r, "thermal");
end

endmodule

But when I use the tran simulation, the noise can not come out, also I add a amp the it the make the bandwidth big, but it could not come out.

What is the matter, please?

Is the white_noise funtion only used in .AC simulation? So it can not work in .tran simulation?





Please help me.

Thank you.
Best regards!

Hi, Stefan. Thank you for your reply.
I think I can not get the latest release version of cadence tools. My cadence version is 5141 usr4. I don't know does it have the transient noise feature. And I will try it.

And does any other simulator have the same feature, such as the latest version hspice?


Thank you.
Best regards.

Thank you very much!
I will try it.

4kTR is the power spectral density, in some sense the variance of the Gaussian distribution of the voltage samples.  Transient noise will generate a new sample from the Gaussian distribution for each timepoint -- and parameters to the analysis determine how many timepoints are needed to accurately represent noise at a particular bandwidth.  Because it's  a Gaussian, there's a non-zero probability of a voltage (sample) greater than 1V.

You should probably read the white paper on transient noise.

I'd set the noisefmax to be at least a decade above the 3dB bandwidth, because something just above that point could still have a significant effect on your output.

Of course, if you've got noise folding/mixing/frequency translation, then you have to account for all of that, also.

It's probably not a bad idea to increase the bandwidth by a factor of 2 or 10 and see if it makes a difference before accepting the results as golden.