From: Randall Peters PETERS_RD@..........

Date: Mon, 12 Feb 2007 10:13:04 -0500

John, All too often people want to take the FFT of a full, long record of length T. Of course with most real systems the spectral lines that provide information concerning system dynamics evolve with time. Thus one must use an n-segmented record set of the total, where n = integer to make the most sense of what is happening. The length of each of these n segments is T/n. This total set of n records comprises the so-called short time Fourier transform (STFT). A common practice is to let the segments be overlapped in time by 50%; i.e., half of the data in an interior segment 1 < i < n being shared with those in segment i+1. The Windaq code Bob has mentioned is really nice for doing the STFT manually (but without overlap), since one can step in time by constant amounts by simply clicking the left mouse button in the appropriate 'bar' area. With Amaseis, I have been doing the steps by highlighting each of the segmnets sequentially and saving them as a new file. Then later I go back and load these segmented records into Amaseis to read the peak values. For a free-decay, if one plots the dB values versus the shift time, then a straight line results if the decay is exponential. By doing a trendline fit to this line, one can estimate the Q easily and accurately. The process is described in more detail than most would want in Chapter 21 that I wrote for CRC's "Vibration and Shock Handbook". The Q value is given (Eq. 21-4) by 27.29 times the frequency divided by the absolute value of the slope indicated by the trendline fit. About the the PSD. I will be glad to write an Excel file to do the PSD for your AS-1 there in Corvallis--after I get some free time in the hopefully not-too-distant future. Perhaps you would want to post it to your website? If you could send me a known-by-USGS earthquake record readable by Amaseis or WinQuake, that would be a great help. Randall > > .------ ------ ------ ------ ------ ------ ------ ------ ------ ------. > | Message 5 | > '------ ------ ------ ------ ------ ------ ------ ------ ------ ------' > Subject: STFT and PSD > From: John or Jan Lahr> Date: Sat, 10 Feb 2007 10:48:31 -0800 > > Randall, > > The STFT sounds useful, but I'm not sure what it is. > > Also, could you share your spread sheet for computing the PSD? That > quantity has always left me scratching my head. Maybe if I played around > with it a bit, I would be able to appreciate/understand it better. > > Thanks, > John >