Any thoughts on AC input level detection

[throbscottle]

I've been working on designing the rms part of my little bench auto-ranging dmm project. It's based on a LTC1968 true-rms converter. This device takes a 1v peak differential input, and outputs up to 1v DC, where 1vrms in = 1vdc out, if I've understood everything correctly.

The auto-range up-switching was going to be triggered using the adc's over-range indicator, and I was going to amplify the rms converter's output so it would be able to reach the same level, and adjust the gain in the amp before the converter for the AC ranges, so all is tidy.

It occurred to me, after much head scratching, that I should be creating an auto-range trigger based on peak voltage, rather than the dc output from the rms converter, since voltage peaks could appear which are out of range, but don't develop enough rms value to show up higher than 1v on the converters output.

Creating a full-wave peak voltage detector (consisting of a full wave precision rectifier, comparator and voltage reference) takes quite a few extra components though, and I'm wondering if it's really worth doing.

So I'd really appreciate other people's thoughts on the matter, and what is the best way to go.

The project so far is a 6 digit, auto-ranging bench dmm, using a 24bit adc with a differential input range of 2.5v. Input is through a permanent divide-by-ten for protection, compensated by an amp to bring the level back up. I've got quite a lot of it up to a whole first version now!

Hope all that makes sense
TIA

 

[crutschow]

Yes, to avoid inaccuracy when measuring the voltage of waveforms with a high crest factor you should use a peak detector at the input for range changing to limit the peak to 0.5V (±1V). Since the peaks of most waveforms are symmetrical you could likely get by with just a half-wave peak detector.

Alternately limiting the output to about 200mV rms should take care of most waveforms with a high crest factor.

 

[throbscottle]

Hmmm, I'm glad you agree with my reasoning! I did wonder about symmetry - I'm just a bit zealous to catch every case

Just a question of which is the best approach, then.

 

[throbscottle]

Sometimes I am so dumb - of course it would be equal + and - peaks, if it's ac coupled. duh. Thanks.

Think I am going to use a peak detector - that way it could also do duty as an AC Peak Voltage function on the meter, useful for non-sinusoidal waves, since I only have a very ancient and fairly inaccurate scope.

Oooh, that's almost the entire front end I have now!

 

[KeepItSimpleStupid]

I did a software autorange for a 4-terminal DC ammeter that I built. It's a little tricky. You basically have 10 V FS to deal with and you need to know when to range up and range down. I may have used 10% or a decade since that was my gain steps. 90% up range, > 100% down-range. <10% go to a lower range. Bouncing can also be an issue.

 

[throbscottle]

Interesting. My range thresholds are closer together than that, so I'll increase the overlap. I've provided manual range override and range lock too, so should be be able to prevent issues with "bouncing" if that means toggling between adjacent ranges.

I've never heard of a 4 terminal ammeter, and google is not my friend on this. I'm providing for 4 wire resistance measurement, but I take it this is something different. Please enlighten me!

 

[KeepItSimpleStupid]

They generally don't exist. That's why I had to make one.

It is basically a lopsided analog version of an SMU: https://en.wikipedia.org/wiki/Source–measurement_unit

It had the ability to source a voltage from -10 to 10 V and measure currents in 4 Full scale ranges ( 10V out for 100mA, 10V/10 mA, 10V/1 mA and 10V/0.1 mA). I should convert all of this to me in mA/Volt and not Full scale/Range since that's what the output was. If the bias was confined to +-5V, it could suppress 50 mA. The typical range it was used at was +-1V and 20 mA max. It had an Open Circuit Voltage mode, A 2 and 4 terminal mode, a Zero Check mode and a Zero correct mode, and a suppress mode. It also had a negative/positive peak detector which blinked a light for 1 sec green or red if the 10 V limit was exceeded. It had 2 system voltmeters (IEEE-488), one for voltage and the other for current and a simultaneous analog output, A system A/D converter and 8 bit output port (IEEE-488 based)

So basically, it could source a voltage and measure a current using 4 terminals (Measure and Force) with < 1 mV voltage drop. Lead resistance and contact resistance was compensated for. It was used as a front end for a lock-in amplifier.
Nothing really exists like it. Try measuring AC currents at 25 mA and voltages from -1 to +0.8 Volts with no voltage drop.
The currents were caused by a chopped beam of monochromatic light at 70 Hz or so and this could be in the presence of a high intensity white light.

I wasn't given enough time to get the DC performance better and the Zero circuitry never worked since the plan was to do it electronically and the A/D converter did not output Zero when commanded to do so. Another nearly identical instrument apparently did. It was a Gotcha. So, I could not null out a few mV or so of offset. This did result in picoamps of current flowing because 0 Volts wasn't zero volts. The resulting in-house product was way more than what was needed.

The Lock-in used was an SRS SR830: https://www.thinksrs.com/products/SR810830.htm

 

[Roff]

Sometimes I am so dumb - of course it would be equal + and - peaks, if it's ac coupled. duh. Thanks.

Not true. Consider any waveform with even harmonic distortion, like a clipped sine wave, or any pulse waveform with non-50% duty cycle.