Actually, it's an interesting problem.  You're not, by chance,
looking at phase controlled AC sine waves (like outta light dimmers)?
        The trick, of course, is that you need the square root of the
mean of the squares of the instantaneous voltages.  In continuous time
(ie, analog devices), this involves squaring the sample voltage,
integrating it, dividing by the integration time, then taking the square
root.  The "integrating and dividing by the integration time" is pretty
much just running it through an appropriate low pass filter where you
trade ripple in the output for response time.
        Although it COULD be done with a bunch of op-amps and diodes or
transistors to do the squaring and rooting, it seems that the single chip
solution from AD is a lot simpler and more accurate.
        I SEEM to recall Maxim or Linear Technology having an RMS to DC
chip that actually had a heater on it and then measured temperature rise
due to current throguh the heater.  I think it was accurate to 100 MHz or
so.  If you want, I can try to find info on it.  I don't know how
accuracy would compare with the pure analog technique described above.
It seems that we would need two temperature readings along a "thermal
resistance", then determine the "thermal current", which would be
proportional to the power dissipated in the resistance.  This is pretty
much how a thermocouple ammeter works.  The voltage out of the
thermocouple is approximately proportional to the temperature difference
between the two junctions, so we are using temperature difference to
measure rate of heat flow (power) out of the hot wire.  Of course, the
thermocouple voltage is then approximately proportional to power, which
is proportional to the square of the current, so the thermocouple ammeter
has a "square law" meter face.
        Then... there's the DSP approach!  I don't know if the PIC is
fast enough to get enough samples (especially sampling four lines).  If
you're not in a hurry, maybe you could use some sort of sub-sampling
where you move a little further into the cycle to sample on each
successive cycle, reducing sample rate to 60 (or 50) times per second,
but it would take a bunch of cycles to get a result (hoping that the
waveform is relatively stable during that time).  However you get the
samples, square each one, add them up, divide by the number of samples,
then take the square root.  RMS voltage!  Moving complexity from hardware
to software at its best!
        Sounds like something fun to play with.  Lemme know your
thoughts!

Harold


On Sat, 2 Aug 1997 14:07:11 -0400 Mike <erazmus@WANTREE.COM.AU> writes:
>Hi all,
>
>Sorry for slightly off topic - but it might be of use to others also.
>
>I need to read 4 voltages at 60Hz and I need the true RMS value. I
>know
>that I could use a true RMS detector/converter by Analog devices, but
>these
>are just too expensive - any others around in small DIP packages ?
>
>OR
>Is there some 'appropriate arrangement' of LM324 or similar opamps
>that
>might provide this conversion for frequencies around 50 to 60Hz ?
>
>Any ideas ?
>
>Rdgs
>
>mike
>Perth, Western Australia
>