Circuit elements I'd like to find...

Trying to measure an AC amplitude, one would ideally like a V**2 square-law component (yes, I know about thermal converters). But, what about other U-shaped I-V transfer characteristics? Is there a good way to make an AC level detector that's accurate to one percent or better, over a wide bandwidth (10 to 10**7 Hz)?

Following such a detector with an integrator would make AC level-crossing detection easy: just use an op amp integrator with the (+) terminal at some set-voltage point.

One example would be an absolute-value circuit (but those mainly use op amps, and high frequency accuracy suffers). Another is two comparators in a window-detector configuration, but THOSE have unwanted high gain and switching transients. A third is a diode bridge, but the forward voltage is temperature-dependent and the output is differential, not ground-referenced (mixers using transformer coupling don't suit the lowest frequencies).

So, what OTHER detection schemes should I consider?

Reply to
whit3rd
Loading thread data ...

Analog multiplier? I'm using the AD743 or AD734, or something like that. $20 each, good to 10MHz.

George H.

Reply to
George Herold

Have you see Jim Williams' "High Speed Adaptive Trigger Circuit" (e.g. DN185 or AN47 p44)? There are probably some other applicable ideas in these references.

--

John Devereux
Reply to
John Devereux

AD734 is the multiplier, AD743 is/was a FET-input opamp I think.

Reply to
Stephan Goldstein

It's worth thinking about, but a tad pricey for casual uses. Certainly, that does the V**2 function well. I'm currently considering a pair of differential amplifiers, with the outputs AND-ed, in window comparator connection. This amounts to a single LM13700, under $1 and also good to 10 MHz. There's an analog-multiplier circuit that uses half the LM13700...

Multipliers have the virtue of true-RMS measurement of any waveform, a nifty addition to the original goal.

Reply to
whit3rd

whit3rd Inscribed thus:

Bolometer !

--
Best Regards:
                     Baron.
Reply to
Baron

Sample it with an ADC and do the math. Works great.

John

Reply to
John Larkin

The problem with the square-filter-sqrt approach is dynamic range. Squaring a signal is a losing proposition almost always--to get 12 bits (74 dB) of dynamic range, you need 148 dB dynamic range at the multiplier output. Unless you really need it to be all analogue for some reason, you're very much better off digitizing and doing the rest in software.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

Hmmm, or you could use the nice Linear LTC1966-7-8, with a pre-S/H stage that'll fold the high frequency energy spectrum into the LF region the RMS/DC circuit can cope with. Of course that'll shift the burden on the S/H portion that should not inject any energy you can't filter out. Oh well...

--
Thanks,
Fred.
Reply to
Fred Bartoli

Doing that at 10^7 is a bit hard.

Reply to
MooseFET

A) Not with a pipeline ADC and an FPGA. Not hard at all. A modern FPGA has dozens of MAC cores that will each square-and-accumulate at 100 MHz or so.

B) There's no requirement that the ADC and math run at anywhere near

1e7 Hz to measure the RMS. The sample-and-hold needs to be fast enough to track the signal, but the sample rate only needs to be compatible with the desired filtering and display rates. You could sample and crunch at 5K events per second and do a nice RMS display. A little sample time dithering would avoid aliasing artifacts.

I've done some thousands of utility survey electric meters that sampled each channel at 27 Hz. Nyquist doesn't apply, because the samples are just gathering statistics on the waveforms, not trying to reconstruct them.

John

Reply to
John Larkin

The intent wasn't to get accurate wide-range output; my target system has an AGC amplifier, this is to fine-tune the AGC for an exact output value, THEN the amplifier's control voltage is a good measure of the logarithm of input signal strength, in decibels. The AGC amplifier has to be a fleapower unit, so the self-heating is negligible, because the control voltage vs. gain depends on the chip temperature. That means the as-detected signal wants to be relatively small, maybe 0.1V RMS into a few-kilohm load. That is why a simple diode is problematic.

A detector that has good stability at ONE point on its in/out curve is the only real requirement; dynamic range can be truly bad, it won't matter after the system balances the AGC . The integrator and setpoint comparison is a single slow, low-performance op amp, but the detector has to be as wideband as possible.

I had several applications in mind, like grid-dip metering and Wein bridge oscillator stabilization. Digitizing and computing don't fit these tasks particularly well, of course; it's an analog output that's required to operate the AGC, or the stabilizing FET, or twitch the detector needle (moving-parts needle is much handier than a digital readout) on a grid-dip meter.

My current thought is that a Gilbert cell multiplier might be the thing to use.

Reply to
whit3rd

If your swing is small and well known, you can use a back diode, or apply a bit of DC forward bias to your signal diode.

OTOH if all you want is a low-level AM detector, the easy and cheap method is to use a comparator and a diode or Gilbert cell mixer. Something along the lines of an LM311 driving an MC1496 would be under a buck.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
 Click to see the full signature
Reply to
Phil Hobbs

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.