[DSP-10] EZKit-Lite

[dsp-10@mailman.qth.net]

>One thing to consider is that if you do that, you tie up a reasonably large,
>expensive, and power-hungry Pentium box.  With an external DSP processor,
>almost any cheap, small, battery-operated laptop can be used as the
>"operator's panel."  This can be a problem for anyone doing portable/mobile
>operations (as is planned here on uWave, using the DSP-10 as the i.f.).

This really isn't a problem anymore. Even low-end Pentium laptops (150
MHz, say) can do quite a bit of DSP. I see them on Ebay for a few
hundred bucks.

All but the very earliest Pentiums have the MMX instruction set.  MMX
(Multi Media Extensions) is the first of three SIMD (single
instruction, multiple data) instruction sets specifically designed for
DSP. MMX was followed by SSE (Streaming SIMD Extensions) on the
Pentium III and SSE2 on the Pentium 4.

SSE2 on a Pentium4 makes an awesome DSP engine that outstrips many
(most?) dedicated DSP chips. I've done a k=7 Viterbi decoder in SSE2
that runs at 13 megabits on a 1.8 MHz P4.

But even MMX is quite powerful. Several years ago I wrote a
demodulator and decoder for the ACE spacecraft. It did symbol timing
extraction, matched filtering, Manchester decoding, Viterbi decoding
and Reed-Solomon decoding on an audio bandwidth signal. It ran with
plenty of real time to spare on my old HP 5700 laptop, a 150 MHz
Pentium.

>A second issue may be that there are many flavors of "sound card."  The
>"standard" seems to be "SoundBlaster," but there's no guarantee that
>motherboards will continue to provide a "100% clone."  If/when that's the
>case, which one do you write for ?

Try Linux. It already has drivers for nearly every sound card on the
market, and they present a uniform interface to the application. In my
case I didn't even have to talk directly to the sound device. I simply
used the supplied Linux program "brec" to set up the device and pipe
the audio samples into my program. For output, I use "bplay". With
them, my DSP programs merely read and write raw 16 bit linear audio
samples.

>Distributed processing DOES have a place in the world, and this is one of
>them.

I think you mean "special-purpose processing", e.g., with dedicated
DSP chips. If so, I think the window of opportunity is rapidly
closing, if it hasn't already closed. This is especially true in the
ham market, which is just too small to benefit from the economies of
scale present in the general personal computer market.

Phil