Gary and Scott; I seem to be unable to deliver to the Labview Forum, so I'm attempting to contact you directly I hope this is alright. (BTW, I am reposting to the forum also, in case it get revived, so I apologize if for the multiple copies). Please see message below. Thanks Rick
-----Original Message----- From: Mahoney, Richard C Sent: Monday, February 23, 2004 1:45 PM To: [EMAIL PROTECTED] Subject: RE: Filter question Gary,Scott, et al. Thanks for the responses Scott, I am well aware of the fact that "that is what windowing does...". My question was, if I'm measuring the Self noise of a sigma delta A/D converter, which if any windowing function should be used? Sorry, I should have worded it better. Gary, This "Hann or Hamming is recommended" is what I was looking for, but even these have a difference in output Although not really that considerable. Dies this recommendation come from the "LabVIEW Signal Processing" source you cited in your message? The fact that the values can change by so much based on the windowing chosen, speaks volumes about the importance of choosing the correct window. My actual requirements reads similar to: "The self-noise at the output of the A/D converter over the band of interest shall not have an amplitude greater than -XX dBV and have a total energy that is no greater than -XX dBV within the band." For the first measurements I've implemented an FFT Power Spectrum function, and the 2nd an FFT Power Spectral Density function, but The windowing function based on what I select has the ability to change the Pass/Fail status of the measurement. This is why I'm asking. I could just code to PASS, but I want it to be right! Thanks for any additional input Rick M. -----Original Message----- From: Scott Hannahs [mailto:[EMAIL PROTECTED] Sent: Monday, February 23, 2004 10:48 AM To: Mahoney, Richard C; Info LabVIEW Subject: Re: Filter question At 07:15 -0500 02/20/2004, Mahoney, Richard C wrote: >Is there a summarized list somewhere of why one particular windowing function would be used over another >in DSP measurements. I know if I select Hanning vs Hamming vs Exact Blackman vs any other type >that my returned values can vary by as much as 3dB! > >I am particularly interested in the accuracy of measuring a VME A/D converter Board (of the sigma delta variety) >Self Noise parameter (with differential inputs shorted together). I've written a meas VI that implements the FFT >Power Spectrum and FFT Power Spectral Density VI's but the windowing function can change the results dramatically. Well, that is what windowing does. It is a problem of quantization in frequency space. If you have a pure tone that is not one of the frequency points but lies between two points, you will get amplitude in neighboring frequency bins of sin(w-wn)/(w-wn) where wn is the freqency of the nth bin and w is the pure frequncy. This also changes the amplitude in the various bins. Different windowing functions will changed the way that this signal bleeds into different bins. Some preserve signal height, some signal width and some integrated power. What windowing function you need depends on what you are trying to measure. Get a good book on FFTs. (Bendat and Peirsol, Brigham, or Bracewell are ones I used long, long ago, but the results haven't changed). The other reason for windowing is that your signal is assumed to be infinite and repetative at the signal length (ie all the values between the frequency points are zero). Thus unless your last point is the same as the first point, there will be a big discontinuity in the repetative signal. To get rid of this the windowing function will make the signal go to zero at the ends. (actually these two reasons are mathematically identical). -Scott -----Original Message----- From: Gary W. Johnson [mailto:[EMAIL PROTECTED] Sent: Monday, February 23, 2004 1:09 PM To: Mahoney, Richard C; [EMAIL PROTECTED] Subject: Re: Filter question >snip For noise measurements, Hann or Hamming are recommended. Always use the highest-level VIs for your measurements. They incorporate the correct scaling for effective noise bandwidth so that amplitude measurements are consistent and accurate. VIs from the Waveform Measurements palette in LV 6/7 are great, such as FFT Power Spectrum, which includes a selector for the window type. The new Express "Spectral Measurements" VI will also do very nicely. -- Gary W. Johnson Electronics Engineer [EMAIL PROTECTED] Lawrence Livermore National Laboratory (925) 423-0156 7000 East Ave., P.O. Box 808, L-183 Fax (925) 423-5804 Livermore, CA 94551-0808
