How to find the magnitude of a 0-10VAC signal

value.

I've used my version (with fast OpAmps) at 400KHz.

...Jim Thompson

--
|  James E.Thompson, P.E.                           |    mens     |
|  Analog Innovations, Inc.                         |     et      |
|  Analog/Mixed-Signal ASIC\'s and Discrete Systems  |    manus    |
|  Phoenix, Arizona            Voice:(480)460-2350  |             |
|  E-mail Address at Website     Fax:(480)460-2142  |  Brass Rat  |
|       http://www.analog-innovations.com           |    1962     |
             
         America: Land of the Free, Because of the Brave

Since you state constant frequency and want "instantaneous" response, then do quadrature sampling. Each sample set will give you a proper voltage reading. This can be done with analog circuitry, but it's a lot easier and cheaper to do this digitally. There are approximations to convert to magnitude with simple arithmetic if you can handle a few percent accuracy.

Or, a peak detector with a zero cross reset.

--- start: clear registers vnew, vold, and vpeak.

uploop: do an A to D on the input signal store the result in vnew compare the magnitude of vnew and vold if vnew vold goto uploop move the contents of vnew to vold goto dnloop

vpeak will contain the value of the last peak and will update on every peak.

-- JF

AM detectors are surprisingly hard to get right--which is really annoying the first few times, because it seems like such a simple thing to want to do.

For 1%-type accuracy, you can use a comparator driving an analog gate followed by an RC lowpass and an op amp buffer--connect the comparator to measure the difference between the signal input and the op amp output, and keep it open whenever the output is too low. Delays and switching transients make this a 1%-type solution, but it works up to pretty high frequency. One drawback is that the RC time constant effectively slows down as you get close to equilibrium, because the duty cycle goes down, but you generally want the RC to be fast compared to the signal slew rate anyway. Reset it with a leak resistor or another switch.

Perfect rectifiers work great up to hundreds of kilohertz, but get fairly bad after that (though as Jim points out, there's a two-amplifier version where the rectifying amp only winds up by +- a diode drop instead of saturating, which is much better). I haven't had much luck just using faster op amps for this, though I haven't tried it in awhile. As others have pointed out, you still have to find the peak of the rectified waveform, or use a lowpass filter and settle for the time average of the rectified signal.

You can get better accuracy by digital methods, though you have to test for digitizer artifacts--just piecing together the specs of the parts won't necessarily tell you the real story.

If the amplitude range isn't too wide, and if the area under the curve is an acceptable substitute for a real amplitude measurement, I like to use a gated integrator, switched at the zero crossings of the input signal. For a wider amplitude range, I usually make two complete detectors with different gains, and use the insensitive one when the sensitive one rails. It isn't too hard to stitch them together--if you're digitizing, you can do it with post-processing, and if not, you can use a spare analog switch and a pot.

Cheers,

Phil Hobbs

How about two peak detector circuits, A and B. When A peaks, it resets B, steers the input to B and gates A peak to Vout; when B peaks, it resets A, steers the input to A and gates B peak to Vout.

Ed

--
That\'s what mine does except it doesn\'t reset, it just follows the
input and, whenever it peaks, loads that value into a register and
holds it until the next peak.

Responding instantly to a decreasing the peak voltage is the tricky bit. You can do it if the wave shapes isnt too ugly and if the voltage range isn't to wide. You can detect that the voltage has stopped rising to trip the switches.

()*^$ I missed yours. (more deleted expletives) I hate it when that happens. :-(

Yup - found it and it's sweet. :-)

Ed

Yes. As you can see from the variety of answers the first suggestion is to learn to frame your problem in detail. You have to know what you are trying to find before you have a chance at finding it. Think of questions like what is the waveform of the "AC"? What do you want the DC representative signal to correspond to? Examples might be peak rectified voltage, average voltage, rms voltage, or rectified average voltage. Think about this. Do you want the average voltage (corresponds to the DC offset of the AC signal) or the rectified average voltage (positive and negative values all converted to positive values) which corresponds roughly to heating ability of the signal. Or do you want the rms value which corresponds to the EXACT heating value of the signal. ETc.

And then it gets still better. You need to put some specs on your DC siganl. How accurately do you want it to correspond to the AC parameter you chose to measure? And how FAST should it follow that parameter? All of these things are important questions. If you want to have the most accurate description of your AC signal then digitize it and calculate whatever you want. Complex. On the other hand a simple diode bridge and capacitor can do wonders to give a signal that more or less tracks "something".

I hope you realize by now the problem isn't your problem, the problem is YOU!

--
Thanks! :-)

"Benj"

** Very well said.

..... Phil

Ok so far.

There is no need to get insulting. Good designs and great inventions initially start with a very poorly thought through question. Think of the FM radio case.

You stirred up some really old memories. I remember doing something about 30 years ago. I filtered the signal

4th order butterworth low pass, digitized it, used lsttl and cmos logic to do peak hold and zero crossing detection, fed the data to a 6502 single board controller, displayed various kinds of results in real time. Detected both negative and positive peaks and reported on both zero crossings, captured time on that to get RPM. It was fun. Fed that uC from 300 to 1000 interrupts / second.

Which invention related to FM radio do you have in mind? Got a reference or two?