FFT Reference Curves

1. Normal FFT display I can set the number of spectral lines and therefore indirectly the FFT window length; then dewesoft tells me the resulting frequency resolution: eg at 32kS/s and 16k lines, I get a resolution (df) = 1Hz. So this implies I have a window length of 1s = 32k Samples. However, the spectrum on the screen updates at a much faster rate than once every 1s. Why is this? Is there a lot of window overlap applied (out of my control) or are you interploating the results to make a smoothly changing spectrum? 2. FFT Reference Curves Again I can set the number of spectral lines and hence the frequency resolution and windown length. I can also set the overlap percentage. eg at 32kS/s and 16k lines, I get a resolution (df) = 1Hz; hence a window length of 1s = 32k Samples. With 50% overlap I get a new spectrum every 0.5s However, when we plot the results of the FFT reference curve in a recorder trace, there are many more points than one every 0.5s; changing the overlap percentage alters the result, but not the rate at which state changes can occur .... also the state changes in the FFT reference curve plot don't occur at the same time as the "normal" fft (ie 1 above) crosses the limit curve. What is going on? Is there some kind of interpolation going on here as well? Can I turn it off?

Nick, The FFT is calculated for each refresh of the display. Therefore you will always get the latest data no matter what the FFT size is. If the computer doesn't have enough power, Dewesoft will slow down the display not to loose the data. Version 7 have a feature to create FFT like array math where you have a full control of the speed of the update.

Dear Nick, regarding your second question. The answer is the same as for the first question. Always the last block is calculated also for the refence curve. That is why you will get non-equidistant points (points are related to refresh rate of Dewesoft Displays). If you need more control you can use <> math (under Math section in Dewesoft) like Jure metioned above. For the reference curve you can use <> math, although it is different than FFT refrence in some ways. If you have any problems you can contant me through support mail. Best regards, Tilen.

Thanks for your reply .. and apologies for taking so long to notice it! This explains why the FFT results are asynchronous with the block timings, but I still don't understand why the FFT reference curve results are different from those of a "normal" FFT - if both are synched to screen refresh, then surely both should give the same result. Also, if FFT reference curve results are synched to screen refresh .. how can the specified block overlap be achieved? Meanwhile, I have had a play with the Spectral Analysis functions in the Math setup and found the kind of analysis I want. this is how I'd like the normal and fft reference curve FFTs to work; but a) I can't find a way to compare the results to a reference curve b) I don't fully understand the STFT function (what is "block length" and "FFT length"?) A tutorial would be good. Thanks for your help.

Hello Nick, a) You need to create math channel: FFT reference curve > click "Get current data". Then in Measure click on FFT and Curve tab, check FFTReference. See screenshot: http://i40.tinypic.com/2uz5u6f.png b) From Dewesoft manual ( http://www.dewesoft.com/download?file=UserManual_703.pdf ): Short time Fourier transformation is the procedure which calculates more lines than the normal FFT. This is achieved by having smaller real block size of data and larger FFT size. Real data is windowed and zero padded and then FFT is calculated. With this procedure we can calculate more FFTs for the same time base. It is nice to be used with fast transients. Block size defines the number of real data samples to be taken for the calculating FFT. FFT size defines the number of resulting lines and with that the ratio between real and zero padded lines. Best regards, Marjan.

Thanks for the info: I've no problem creating a comparison between FFT and reference curve using the FFT reference Curve function - but as this refreshes on screen update, it doesn't give the result I want. For the future: It would be good to have an option within the FFT reference curve setup to "refresh FFT on screen update" or "refresh display on FFT update". At present there is no way to achieve a real time display (including comparison to a reference curve) where the FFT block timings are fixed by the user. So it's not possible to say "for how many FFT windows was the signal greater than the limit curve" thanks