Spectre pnoise settings

[prestonee]

I am trying to perform noise analysis on a chopping integrator.
can i get a better explanation on when to use the noise type source over the noise type timedomain?
also does relharmnum and refsideband matter?
and for chopping since I have a dc signal being integrated at a sample freq and chopping occurring at 1/2 the sample freq, should i use pnoise at the period of the chopping period, or use something more like a qpnoise and specify both the chopping freq and the signal freq? other question is how to make the chopped integrator periodic? is it ok to place a periodic reset in to reset the Cint, or would this impact the noise? if I cant, then what if feed the outs back to the input so it negatively integrates the output. or would this also affect noise since it has cancelling feedback(output would be = out - previous out)

-Preston

 

[D.A.(Tony)Stewart]

an integrator will amplify DC offset error over time, so a chopper converts it to AC so the DC offset eliminated.
Any harmonics must be eliminated by 40 dB to 80dB depending on what your tolerance for DC offset error such as 1% to 0.01% .
If you do not filter enough it will create a convolution to DC if synchronous and degrade the benefit of a chopper. (meaning if the harmonic ripple is not perfectly symmetrical, it will create a DC offset.)

A harmonic integrator will create alias noise further adding error, unless it is pre-filtered with high Q narrow bandpass.

A harmonic mixer used in PLLs may be better as then the BPF becomes a LPF which is easier to control/ design.

 

[prestonee]

Sorry but I feel you are providing design recommendations, I am looking for spectre setup variable descriptions.

When I place the chopped integrator in a closed loop situation the noise drops significantly, and results seem to be very dependent on the settings mentioned above plus noisetimepoints. when in open loop the noise is much higher and seems to be less impacted by the various settings.

 

[D.A.(Tony)Stewart]

Perhaps show your work would be helpful to explain what you have done and what you want help with..
Naturally any integrator will reduce noise by 1st order effects.

for example you want to know how to use The iinteg function with pnoise function?

see pages 70 ~78
ftp://ftp.sitsemi.ru/pub/Cadence/calc.pdf

Nyquist criteria for sampling may show you can sample at 2x the maximum freq but this imposes harsh requirements for harmonic filtering so sample at 3x the maximum f and use a steep 5th order LPF, not JUSt an integrator