sheldon wrote on Oct 17th, 2014, 6:04am:Jacki,
You have not explained a lot about what you are doing or at least I
am missing the point. I think that the noise floor is indeed the total
noise divided by the Nyquist frequency, fsample/2, expressed in dB.
In the case of a data converter, the noise floor should be determined
by the quantization noise, unless you are doing a transient noise
analysis. So there is not much information that you can extract from
this results, that is, you can just directly calculate the noise floor
from the quantization noise and Nyquist frequency. No need to
simulate.
If you are really trying to simulating the noise floor, then you
probably should re-consider your approach:
1) You need to be using transient noise analysis
2) You should use the power spectral density function, psd
3) Convert from power to voltage
4) Use the IC616 spectrum function
In IC616, the SNR can be directly calculated from the frequency
spectrum. So in IC616, you can calculate the SINAD/SNR/...
from the Fourier Integral or psd results.
Also be aware that the transient simulation time needs to be
much longer for power spectral analysis.
Note:
Spectrum is a Calculator function, not a Spectre function, and it
has been upgraded in IC616.
Finally, you need to consider the transient simulation accuracy.
Numerical noise may limit the noise floor unless you set the
error preset appropriately.
Sheldon
Hi, what I wonder a bit is how the Viva spectrum assistant is really working behind the scenery? E.g. is SNR only taking noise into account or also harmonic distortion, actually calculation SNDR? And e.g. in which freq ranges?
How is psd doing the averaging in detail? I played with some parameters like "smooth" but see no impact!
Bye Stephan