Path: utzoo!utgpu!news-server.csri.toronto.edu!rutgers!apple!kchen From: kchen@Apple.COM (Kok Chen) Newsgroups: sci.electronics Subject: Re: 90 degree phase shift Keywords: phase shift Message-ID: <45732@apple.Apple.COM> Date: 17 Oct 90 17:16:53 GMT References: <3899@osc.COM> Distribution: usa Organization: Apple Computer Inc., Cupertino, CA Lines: 52 jgk@osc.COM (Joe Keane) writes: >I have what seems to be a fairly simple problem. I want to buld a circuit to >phase shift an audio signal by 90 degrees. I also want this to work over a >range of frequencies and have fairly flat amplitude response. Take a look at the Phase-Sequence filter that is described in both the 2nd edition of Horowitz and Hill ("Art of Electronics," McGraw-Hill, I believe) and the 1990 Radio Amateur Handbook that is published by the American Radio Relay League. It also appeared in some IEEE Transactions a while back. The ARRL Handbook didn't actually call it a Phase sequence filter, so you won't find that in the index. Just leaf through the SSB transmitter section of the Handbook and you can't miss it. The P-S filter starts with the original signal and the 180 degree phase shift (i.e. inverted signal) of it. These are passed through 4 intercoupled R-C networks in what appears to be a succesive approximation of the desired result. The output of the P-S filter are 4 signals that are at 4 successive 90 degree phase shifts from one another. The P-S filter looks far superior to the old time phase-shift networks that comes in octal tube sockets (remember those? I think Miller made one of them). And the R-C values are *far* less critical. Be aware that although the output phases of these filters are in quadrature, they do *not* have a nice flat phase response w.r.t. the original input signal. If that is what you want (i.e., absolute 90 degree shifter rather than one that provides a pair of in-phase and quadrature outputs) you would have to find an approximation to the Hilbert Transform. The Pade' approximation would be one way -- e.g., take the Euler expansion of exp(-i.T) and approximate as many terms as you can afford with RC filters, and summing their outputs. In answer to your comment about SSB generation - the method that is more in favour nowadays seems to be the one that generates a DSB suppressed carrier signal and pass that through a really sharp cutoff crystal filter. Good crystal filters in the 5 - 10 MHz range are relatively inexpensive nowadays and are far less "tweeky" than SSB exciters that use phase shift networks. You need phase accuracies in the sub-degrees to get sideband suppresion that approaches that of even a pretty sloppy filter (just try subtracting sin(x) from sin(x+delta) to see how small delta has to be so that the resultant amplitude is, say, 40 dB down -- I would offhand guess about 1/100 of a radian). Kok Chen, AA6TY kchen@apple.com Apple Computer, Inc. Disclaimer: the last time I even attempted the phasing method of SSB excitation was for an undergrad lab over two decades ago :-).