The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Simulators >> Circuit Simulators >> Noise floor in Spectrum
https://designers-guide.org/forum/YaBB.pl?num=1575155703

Message started by Ali1990 on Nov 30th, 2019, 3:15pm

Title: Noise floor in Spectrum
Post by Ali1990 on Nov 30th, 2019, 3:15pm

Hello all,

I have an issue with spectrum in cadence.
I’m trying to see the spectrum of a sinusoidal signal at different frequencies and amplitudes.
Sampling frequency is 9.709037037MHz and the strobe period for transient simulation is equal to sampling time (I activated the conservative mode for simulation as well).
Nfft is 32768 and start and stop times in spectrum measurement is defined based on frequency resolution of fft (Tstop-Tstart=Nfft/Fs).
 In the attached photo, all selected frequencies are on bin. if I change input frequency from f1 to f2, the noise floor will change by 20log(f2/f1). My question is that what is the reason to have higher noise floor for higher input frequencies?
Noise floor also increases when I increase the amplitude of the signal.

The version of virtuoso is 6.1.7-64b and i'm using spectrum tool.

Title: Re: Noise floor in Spectrum
Post by sheldon on Dec 5th, 2019, 4:07pm

Given the available data it is difficult to know for sure what the problem
is, however, I suspect that you are seeing the effect of the windowing
function on the data. If you look at the following two plots you will see
a similar pattern. The Spectrum is effected by the window close in to the
tone. However further out, where the window function is below the FFT
noise floor, the FFT noise floor becomes visible.

Title: Re: Noise floor in Spectrum
Post by sheldon on Dec 5th, 2019, 4:08pm

Example II, the whole spectrum

Title: Re: Noise floor in Spectrum
Post by sheldon on Dec 5th, 2019, 4:11pm

I have tried other experiments with low sideband windows and these produced similar results. Since the input source has very high dynamic
range, the noise floor of the FFT, ~ -300dB, see the PSS plot. The noise
floor is below the noise floor of the FFT is potentially below the level
of the windowing function.

Title: Re: Noise floor in Spectrum
Post by Ali1990 on Dec 7th, 2019, 5:57am

Thanks for your reply Sheldon.

 In the spectrums, I have used just Hanning windowing. Still I didn’t find why we have higher noise floor for signals with higher frequencies which is generated by ideal sinusoidal source and how we can avoid this issue. To clarify my question, I have attached another photo. The two spectrums in the attached photo is for 2 different frequencies generated by ideal sinusoidal source.

 Sampling frequency is 9.7090370370370MHz (this is the sampling frequency of my CT delta sigma modulator) and the strobe period for transient simulation is equal to sampling time (I activated the conservative mode for simulation as well). Nfft is 32768 and start and stop times in spectrum measurement is defined based on frequency resolution of fft (Tstop-Tstart=Nfft/Fs=3.375ms).  

 In the attached photo, the red spectrum is for the case that the input frequency is at  (13/32768)*Fs and the yellow one is for fin=(3375/32768)*Fs. As you can see, the noise floor is higher for higher input frequency. So, I wanted to know the reason for higher noise floor at fin=(3375/32768)*Fs, and possible solution to reduce noise floor.

  Still the noise floor is very low enough and can be considered ideal, but when I apply this source to an amplifier, implemented by transistors, the noise floor becomes around -160dB. Since I have to down convert the amplified signal and then apply it to a sigma delta modulator, I have a concern that I will not be sure that I’m calculating SNR in a correct way or not….

Title: Re: Noise floor in Spectrum
Post by Ali1990 on Dec 10th, 2019, 9:43am

After modifying TOLERANCE OPTIONS, noise floor decreases at the output of the transistor level implemented amplifier and seems fine now.

The Designer's Guide Community Forum » Powered by YaBB 2.2.2!
YaBB © 2000-2008. All Rights Reserved.