White Noise simulation for MOS transistor

It depends on the model you're using, but I assume you're looking for thermal noise contributors, which are are usually called things like rs, rd for the source and drain resistance thermal noise.

Andrew.

It depends on the model you're using, but I assume you're looking for thermal noise contributors, which are are usually called things like rs, rd for the source and drain resistance thermal noise.

Andrew.

Hello Andrew,

Thank you for your prompt response.

I am using a PSP103p5 model located here:

http://www.cea.fr/cea-tech/leti/pspsupport/Documents/psp103p5_summary.pdf

On page 144, the white noise is given as sqrtsfw which is given as the Input-referred RMS white noise

voltage spectral density and I couldn't find the parameter from the simulation. 

Thanks a lot

Dear wgtkan,

wgtkan said:

ow is it possible to determine the white noise of transistors from simulation?

Since the white noise components, which are mainly due to the series resistances to the device terminals, are not a function of frequency (other than though the frequency response of your device, and the shot and flicker noise are frequency dependent, is there a reason you cannot examine the noise components above the flicker noise corner frequency (f_flicker). The flicker noise corner will be larger than the shot noise corner frequency and hence, for f >> f_flicker, the noise will be white noise component.

An added comment is that the trace resistances due to routing to the device terminals may dominate the rs, rd, and rg components of your model. If you are interested in performance in an actual application, I might add you utilize a layout based extracted view to include these routing resistances in your result.

Just a thought based on my understanding of your question.

Shawn

Hello Shawn,

Thank you for your response. What I want to do is plot the Input-referred RMS white noise

voltage spectral density at 1 kHz. As indicated in the attached copy the PSP103.5 noise model. I have plotted the flicker (1/f) noise. Do I have to divide it by the gm of the transistor?

Thank you Shawn.

Dear wgtkan,

wgtkan said:

Do I have to divide it by the gm of the transistor?

Is it my eyes, but it appears the sqrtff term for input referred white noise contains the Sfl - the flicker noise ccurrent omponent at 1 Hz? If so, this is not white noise as the two are distinctly different types of noise. It seems the term should not include th flicker noise. Perhaps I am not understanding the noise model?

In any case, the only reason that expression divides by the gm is the expression is using current noise (Sfl) in A/root Hz. To convert it to the units of sqrtsff input Voltage noise source, the expression is divided by gm. If you plotted a voltage noise term, then there is no need to divide by gm as the units will no longer be in V^2/Hz of V/root Hz.

Shawn

Hello Shawn,

sfl is the flicker noise spectral density in Amps squared /Hz so when divided by gm squared, It should be Voltage squared Per Hz and after taking the square root, it becomes Volts/square root Hz.
according to the manual it should be white noise Spectral density.

I just want to confirm if this is true.

Here's what I see (note, my models are made up examples, and I don't have any sheet resistance for the source, drain etc hence the zero values, but it's illustrative of the outputs you'd expect:

Device    % Of Total    Inp Ref Noise    Param           Noise Contribution

/M0       26.95         1.18704e-15      total           1.28543e-15       
                                         Sfl             1.26775e-15       
                                         Sthd            1.76708e-17       
                                         Shotgd          0                 
                                         Shotgs          0                 
                                         Sjnoise         0                 
                                         Djnoise         0                 
                                         Sfledge         0                 
                                         ididedge        0                 
                                         Rgatenoise      0                 
                                         Rbulknoise      0                 
                                         Rjunsnoise      0                 
                                         Rjundnoise      0                 
                                         Rwellnoise      0                 
                                         Rsourcenoise    0                 
                                         Rdrainnoise     0                

Bear in mind that the noise analysis output gives you the output referred noise, so it will be in amps or volts depending on whether the output is a current or a voltage. You're not getting the noise values at the noise source - so dividing by gm makes no sense (to me).

I didn't read the model documentation, but hopefully the names correlate with what you're expecting - al those "R" noise contributors will be thermal noise from the corresponding resistances.

If you need more detail, you'll need to contact customer support as this would need R&D input as (unfortunately) we don't document the noise outputs in the model equations (I have asked for this in the past, but it's not happened).

Regards,

Andrew.