[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: digital filter design, Lee Jackson, etc.

At 04:26 AM 1/30/2005, Ramdas Kumaresan wrote:
>Also take a look at L.B.Jackson's text book "Digital filters and Signal
>processing" Kluwer publishing, Ch.4. It  has a unity gain resonator which
>can be  tuned by varying just one coefficient. Incidentally that was the
>first digital filter ever built (with J.F.kaiser, and J.McDonald) some time in
>1964?? RK

That's a very nice design that I hadn't remembered.  The pole are still parameterized in the "direct" way by real part and radius in the z-plane, but the gain coefficient automatically adjusts to keep unity gain at the resonator peak, at least approximately as Leland points out.  It's probably possible to combine it with my structure that lets you provide frequency and damping coefficients instead of real part and radius, in case that's of interest.

Speaking of Lee Jackson, Jim Kaiser, and Hank McDonald and the first digital filter, I have a few stories to relate to that history.  I had the good fortune to have a summer job in 1973 in Hank McDonald's section, and got to know Jackson and Kaiser that summer as well.  I got a patent (or Bell Labs did, really) on extending their "bit-serial approach to the implementation of digital filters" to work directly with two's-complement signed bit-serial data in the multipliers.  Using the bit-serial approach, my department (Stan Freeny, Bruce Kieburtz, Stuart Tewksbury, and Kent Mina) had built a complete channel bank transcoder (FDM 24-channel voice group to two T1 TDM channels) using ECL small-scale ICs implementing complex high-order filters before I got there; an amazing piece of work.  I think I spent way too much time after that on bit-serial signal processing architectures, until Moore's law made that approach obsolete about 10 years or so later.

I'm not sure it's quite accurate to say that these guys made the "first digital filter ever built".  Apparently by "ever built" you mean in hardware, not programmed on a computer.  It's possible, I suppose, but I haven't seen that claim before.  I don't find a clear answer in the oral history transcripts of Jim Kaiser and Ben Gold, who give each other a lot of credit for the early work:

Jim Kaiser oral history transcript:
Ben Gold oral history transcript:

It seems likely that digital filters were already built in hardware in sampled-data control systems before that, e.g. in anti-aircraft artillery aiming systems, since that where the theory had been developed.  If you have more info on this history, I'd love to hear it.

By the way, Lee Jackson managed to misspell my name as "Lyons" in a couple of references in the back of his book, in at least the first two editions.  I haven't checked the third yet, but he said he would fix it.