1
$\begingroup$

I am working with a new Digital-to-Analog Converter (DAC) design in simulation and I'm trying to analyse the output. The device takes in an ideal 14-bit digital representation of a sine wave and outputs through an ideal Butterworth filter (a̶n̶t̶i̶-̶a̶l̶i̶a̶s̶i̶n̶g̶ anti-imaging).

In the simple analyses, I have been setting my input sinusoidal frequency such that it is coherent, and my sample capture length set up to collect 8196 points. I'm then running this through a simple MATLAB script that windows the simulation data and calculates the Welch periodogram using pwelch(), to reduce spectral leakage and variance respectively. Once I've done that, I'm measuring the characteristics of the device such as the SNR, SINAD (SNDR), THD and so on.

The next step in my analysis is slightly more complicated. I have run 5 transient noise simulations which result in 5 sets of output data, the only differences being the noise seen on the transient signal. How can I combine these spectra to get a more accurate picture of where the noise floor is in the device?

Assuming the noise is random, can I approach the situation similar to Bartlett's method but averaging across all 5 FFTs and creating a periodogram of the result?

Improve this question
$\endgroup$
4
  • $\begingroup$ I'm pretty sure you mean an anti-imaging filter, not anti-aliasing. $\endgroup$
    – Ben
    Commented Mar 17, 2021 at 19:19
  • 1
    $\begingroup$ If you average the FFTs, you'll run into the central limit theorem. It might be better to apply the calculatios on each spectrum, then average the results. $\endgroup$
    – a concerned citizen
    Commented Mar 17, 2021 at 19:42
  • 1
    $\begingroup$ @Ben thanks you're right, more experience with ADCs so lazily applied the term, edited. $\endgroup$
    – Alex Steven
    Commented Mar 17, 2021 at 20:55
  • $\begingroup$ @aconcernedcitizen thanks, I'll give that a go! I think I might have been overcomplicating it.... $\endgroup$
    – Alex Steven
    Commented Mar 18, 2021 at 10:10

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.