DSP etc

[Terry Fox]

Happy holidays to you, Doug.  I am sorry that I did not make it to Frank's 
party, and missed you.  It sounds like your interests are just what the 
amateur community needs!

I agree with a lot of what you have written.  It appears there are some 
things that a dedicated DSP would be best to use, while other things could 
use a generic PC.  I just bought a new 900MHz CPU, motherboard,  & 128MB of 
RAM for under $200.00.  That's about the same as what I paid for the old 
Motorola, TI, and Analog Devices DSP development boards.  I was just 
thinking over the weekend that a bare-bones 900MHZ PC has got to have more 
horsepower than a 56003 EVM.... which I used to program.

There are still places where a fast DSP would be better, however.  My 
latest interest is faster VHF/UHF/Microwave modems.  Having a dedicated DSP 
chip is certainly smaller and easier to implement, after design is 
completed.  I want to send digitized MPEG video over these links.

Fast serial interfaces are non-trivial, as you suggest.  For years, I have 
used the parallel port for just that reason.  I recently started 
investigating USB.  I am presently playing with the DLP Design USB 
interfaces.  I haven't had time to really get to know them yet, but they 
look like a good solution, if USB is available.  The speed should certainly 
be high enough.  Right now, I am trying to interface a homebrew HF spectrum 
analyzer to a PC using PIC chips, DDS, and the DLP USB.  I am also 
relearning Visual C.  I use it so rarely that I need to re-aquaint myself 
to it on every large project.  I am still comfortable in the DOS/TurboC 
world of ten years ago.

Anyway, I think your ideas are great!  We could use the help in several 
areas you mentioned.  Glad to see you also have an LF interest.  I hope the 
gang talked to you about possibly putting up a remote receiver.

Thanks for the interest, and the Email.
Terry Fox
WB4JFI
wb4jfi@cox.rr.com


At 05:29 PM 12/31/01 -0500, C-Lab wrote:
>Hello,
>
>I recently "caught the bug" after visiting with Frank G
>this Thanksgiving, and am interested in participating/helping,
>that is, if you guys will tolerate a non-ham.  I suppose I
>could be called a pro, instead, because I did indeed design
>interesting DSP-radio things for "the government" some
>20+ years ago -- back when one did this with wire-wrapped
>TRW multipliers and homebrew cpu's in general.  It's a lot
>easier these days!  Since then, I formed a company that
>does DSP consulting work.
>
>We do embedded DSP (mainly using the TI stuff) and windoze
>programming (from drivers on up, we _DO_ Windows) for
>a living here at our solar powered facility in Floyd, VA.
>We also do general embedded uP stuff, mainly using PIC
>uP's for lots of interesting things.
>
>Here, we actually have enough room to make a for-real antenna
>at LF if we wanted to do it, and there's no powerlines for
>at least 1/3 mile in any direction...could be interesting!
>I can of course turn off any locally generated electrical noise,
>since *I* am the power company for these 30 acres.
>Last time I hooked up a SW receiver it was pretty nice.
>
>I was impressed enough with Frank's setup to get started
>duplicating it, all the stuff is on order and on the way,
>or will be very soon.  (Someone needs to make that
>cool antenna a kit item and sell it!!!  What a pain to
>chase all that stuff down!)
>
>It did seem to me that the DSP on the tentec was in need
>of assistance.  There are a number of possibilities I'm looking
>into for helping it along.  Most of them would include tapping
>the signal off at the low IF before the builtin DSP sees it,
>since there you have some room to work and do more interesing
>stuff.  I note that the higher quality pc soundcards (the only
>type worth it, the normal SB clones have serious built-in birdie
>troubles, due to "nearest neighbor" builtin resampling and
>other bad design jokes) now support 48k sample rates....
>
>My company also makes a DSP card that will support just about
>any serial a/d out there as long as you provide the timing signals --
>the board just has a sync serial port that can go 10+mbits.
>
>One approach would be simply to ditch their dsp card and replace it
>either with code on the PC/soundcard, or another dsp card
>with some for-real horsepower.  This latter could be done
>such that the radio remained back-compatable with existing
>control programs, but would allow for smarter control
>programs to tell it to do "more interesting" things.  By adding
>autobaud software it would be easy to tell if the control
>program was original (1200 baud) or something "new",
>(some faster baud rate, gheesh) for example.
>
>Plans include adding modes (costas loop "am stereo", nbfm
>fsk psk etc etc etc) and some *good* noise removal using some
>DSP tricks we've learned over the last 20 years of "practicing".
>Better (and maybe adaptive) filters are of course obvious
>and also easy to do.
>
>For those PC-averse, or who want to be able to use laptops
>which aren't all that fast, we could roll a for-real DSP board
>to replace the existing one in the ten-tec.  In fact, we have
>at the moment a "DSP core" design that we use for proto work
>that would make it easy to test embedded code.  It lives in
>an ISA slot, and has "everything" you could think of for doing
>DSP -- 4 ch of a/d,d/a...half a meg of dword-wide static ram,
>64megs flash, 5 serial ports (one is 5 mbit) and so forth and
>so on.  Oh yeah, it has a 150 mhz TI 'C33 on it, which we
>find is a bit faster than a top of the line pentium for DSP -- it
>does several things per cycle instead of the other way 'round.
>And it runs my opsys, not windoze, so nothing is wasted.
>
>Our main customer rolled this up to our specs and we use
>it for prototyping for their designs which can then be put
>into production fast by simply removing the stuff that project
>didn't need.  Yes, there are windoze drivers (and all types of
>windoze too, not just the 9x/dos variants) and plenty of
>debugging tools we've written over the years.
>
>I'm not sure what your main needs/desires are, but I'm planning
>on doing this anyway, and would of course be glad to share
>it with all.  (It gets boring doing things like Voice over IP etc
>for a living...telephones!)  I also have some interest in spread
>spectrum (hopping)...if the tentec can't switch fast enough,
>do you suppose guys would be willing to get "a couple"
>and hook them up so the switching time could be hidden?
>
>(Some of my other hobbies include astronomy and benchrest
>shooting, and there, people will spend the dough to get
>what they want, no question at all.)
>
>Not sure what resume I'd give, but you might look up my book,
>Digital Audio Processing (Doug Coulter) on Amazon.com
>for a hint.  It might even be worth buying for the "free" code
>which includes a complete windows wavefile editor with
>lots of nice windoze programming examples and plenty of
>reusable DSP code (C++ and asm).  Heck, if about 10,000
>more people buy it, the publisher might even have to write
>me a check someday...but I doubt it.  Or if you can stand
>a HUGE download, I can send you the code.  But you'd
>want the text to know how to get in there and tinker
>with it I think.
>
>What I'll probably do first is write some "shell" software
>that just gets the pc soundcard going and
>
>// "your dsp code goes here"
>
>more or less.  (most of the initial dsp code can be swiped
>from waved, the book's "example" program)  I can wrap this in an
>appwizard for devstudio so as to make it easier to share.
>
>(Most experienced programmers know better than to cut of their
>own nose to spite Bill's face - the M$ tools are the best out
>there, like it or not, and most computers run windoze, again,
>like it or not.  Heck, guess I'll just never be politically correct.)
>
>Since the details of getting control of and using a soundcard
>are arcane enough to be daunting for most, I think this
>might be best?  DSP and dealing with "chunked" streams
>are tough enough for most people without that other overhead.
>
>Oops, another thing I should do is put in some for-real serial
>port(s) support, another non-trivial windoze thing if you need
>any speed and/or control.
>
>Comments?  I'd love to hear from you guys.  The phone number
>is good most times, as I both live and work here, but of course
>during normal business hours, I'm "at work" and can't talk long.
>
>
>Doug Coulter, owner
>C-Lab
><mailto:c_lab@msn.com>c_lab@msn.com
>540-763-3753