Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!ncar!boulder!unicads!les From: les@unicads.UUCP (Les Milash) Newsgroups: comp.arch Subject: Re: DSP56001 query Message-ID: <513@unicads.UUCP> Date: 27 Jun 89 20:32:16 GMT References: <8765@chinet.chi.il.us> Reply-To: les@unicads.UUCP (Les Milash) Distribution: usa Organization: Unicad Boulder, CO Lines: 56 In article <8765@chinet.chi.il.us> edlee@chinet.chi.il.us (Edward Lee) writes: >>[is this BEGINRUMOR >>BUT, a new player has emerged - the Motorola DSP56001 - it's a digital signal >>processor, basically a very fast A/D + D/A converter that has appeared in some ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ that's not a good description of a DSP. DSP's do what you want done after you get it from A to D, and before you turn it back to A. they tend to be "primitive" processors; technically RISC (hardwired decode logic, 1 cycle ops) but with various uglyisms in 'em; they're hard to hand-code for. often no wait states; you get to shop for small fast SRAM chips. they tend to have 1-cycle 16x16=32 multipliers with >32 bit accumulators on their back ends. they tend to have index registers with post-increment. they are optimized for signed integer convolution. i say "primitive" but we mean balls-to-the-walls cycle time AND a bunch of memory AND one-cycle-mpy AND ... i think this guy's claim to fame is 24 bit wide datapaths with a 56 bit accumulator. i believe that the 56001 is a followon to the 56000 (probably more ram or fewer nS or something). i'm kind of surprised that the 56001 was used in the nExt. it's way wide enough (even 16 bit is plenty wide enough), but the AT&T DSP-25C has single prec floating point. some (mostly old cheaper worse) DSPs did have analog i/o, but i bet this one doesn't. i'm checking; i'll let you know either way. >>for this application), and the price is $56. In single qtys yet. ^^^^^^^^^^^^^^^^^^^ that would be astonishing for such a powerful chip. you can bet i'm checking that one too. >>ENDRUMOR true?] these are particularly interesting to me, so i'm checking it out. if these were true, even just the $56/qty 1 part, that'd change the face of the dig synth industry overnight; a startup could compete with yamaha's custom VLSI and win in terms of !/$. or maybe i've just been in the dark; it seems distressingly common. i'll either refute this claim or start looking for venture capital within 48 hours, i promise :-) the original poster was talking about some kind of modem. I can well imagine that this chip is fast enough for that, i bet the A/D&D/A will be external, there'll be 36 sq. inches of board, and it'll take some macho software development; it ain't gonna cost $56. i did see a TI TMS320 being a 1200 bps modem (40 MHz = 10 "MIPS", 16 bit, integer) once, probably some less exhalted type of modem. btw these chips are not synthesizers, they are about enough synth to keep one finger happy. i calculate that a yamaha DX-7 is about a dozen of 'em. Led Z. Milash McCabe's Law: just cause it's wrapped in lycra, doesn't mean it's good.