On Wed, 25 Mar 1998 20:49:05 +0000 Danjel McGougan
>I've been thinking about DTMF detection in software using a PIC.
>Has anyone tried to implement this using one bit samples and a DFT
>calculation?
>
>I've implemented a little program in C to simulate the calculations
>that could be performed by a PIC. The DFT routine is very simple to
>implement because you don't need any multiplication when the samples
>are only one bit.
>
>The DFT calculation is a bit simplified since it does not perform a
>square and squareroot in the last addition. I use absolute values and
>skip the root. This should work fine though.
>
>The results are pretty good. I was surprised to see that the DFT
>calculated amplitude was about 5-10 times higher than other frequency
>components with a noise amplitude that's half the DTMF tone
>amplitude.

Yes !  I did exactly this about a year ago.  I initially did simulations
in mathcad and then duplicated the results in quickbasic.  I was also
somewhat suprised by the peaks.  I wrote up some code for a 16C84 and got
it running in Parallax's PSIM. For the PIC simulation, a wrote another
quickbasic program that generated the bit patterns I expected to capture
for different DTMF characters and would then just hardcode the pattern
into the PIC program.  The simulations said it should work, but I was
darned suprised that the first programmed chip actually did work !

I took 80 1-bit samples at around 4Khz.  The PIC would transmit each DTMF
character serially to a PC running any garden variety terminal program.
The minor problem I had (on a 4Mhz 16C84 running at 3.58Mhz) was it took
a little over 0.25 sec to actually decode a character and transmit it
serially (at 9600 baud).  I _could_ press buttons fast enough that it
would miss characters.  I only used about half the PICs program memory,
but nearly all the RAM.  I always meant to get a 10Mhz 16C84 and try it
on it. About that time a got busy with other things and am JUST getting
back into playing with the PICs (just a hobbiest).  I would opt to go
with a 16CF84 now, although my programmer doesn't support it without an
upgrade :(.  I also thought about using an FFT to speed things up, but
that looked pretty hairy !

BTW, I graduated with a B.S. in electrical engineering about 10 years ago
and learned more about the DFT on this project than I ever learned in
college.

Doug Boulware <dougrb@juno.com>

_____________________________________________________________________
You don't need to buy Internet access to use free Internet e-mail.
Get completely free e-mail from Juno at http://www.juno.com
Or call Juno at (800) 654-JUNO [654-5866]