converting a keithley 175-AV from averaging to TRMS

[catjuice]

Hello, first post here.
I was wondering if it would be possible to modify my keithely 175-av multimeter to be true rms instead of average responding. it seems to have an rms to dc converter chip (AD637JQ) that claims to have true rms functionality yet when I probe a 1v @2khz square wave i get a voltage reading of 0.5505 v instead of 0.5v.

so after consulting the datasheet of the AD637 I noticed that the averaging capacitor within the keithley was a measley 22 pf compared to a recommended value between 0.47 - 6.8 uf in the datasheet. so i soldered a 4.7 uf in parallel to it just to see what would happen and nothing changed :/.
it is not exactly in the same configuration as the datasheet specifies since its input doesnt incorporate the internal buffer and so c2 isnt represnted in the same way as in the datasheet.
im just curious why keithley would choose to make an average responding meter when they have the parts to make a trms one?
I don't understand exactly why it isnt acting like a trms meter in this configuration.

also any tips on this modification would be super appreciated! (assuming that it is possible at all)

[catjuice]

here is what the keithley 175-av's configuration looks like according to the service manual. u102 is ad637jq

[Conrad Hoffman]

Need to think about this. It's absolutely bizarre that Keithley would have used that chip and not given the meter TRMS capability. It seems like a mod would be possible.

[mawyatt]

Didn't Dave have a video about this with a Fluke DMM?

Best,

[alm]

I guess this was the eighties equivalent of a software bandwidth limit on a scope? Or maybe they had a specific (military) contract that called for an average responding meter because it was in a procedure for maintaining a some expensive piece of kit like an airplane, and this was Keithley's solution that required a minimum amount of engineering effort?

They added average responding AC measurements to the military version of the Keithley 2001 (the 2001M) for this reason of the military still needing those measurements for something that's more expensive to replace than a multimeter

[catjuice]

I imagine this would be a really expensive solution for the amount of engineering effort required to replace that chip. that chip is still being sold on digikey and other places for 40$ each. for that price I would have expected them to at least make trms an option?

[alm]

Say you are a Keithley who already has the 175A as a current product. Then a large customers comes to you and says "Can you sell me a 4.5 digit meter with average responding AC measurements? We are not very cost sensitive". Why would you spend engineering time making the RMS feature optional if the customer specifically asked for an average-responding meter? Just in case the US air force changes their mind? And why would you spend time and money to respin boards to optimize BOM costs for a customer that is not very cost-sensitive?

I checked some paper Keithley catalogs, and although I find no mention of the 175AV in the main catalog from 1989, 1991, 1993, 1995 and 2002, I did find it in the 1993 and 1995 price lists. It's listed for the same price as the 175A (US$ 575 in 1993). The 175AV with the battery option 1758 has NSN 7Z6625-01-310-6232, as far as I can see it's the only DMM in that price list with an NSN number. This is consistent with my theory that this model was specifically created for the US government / military.

Based on this, I doubt the meter was ever marketed or sold to consumers and small businesses. Only to customers with a very specific need. Customers who needed True RMS would have bought the normal 175A that would have had much better availability.

[catjuice]

well that seems like a rather reasonable theory and also gives me thean idea. if the 175-av has revisions like the 175A and 175 with a trms functionality. Ill i might need to do is just copy whatever changes they made in their schematics. notably the op-amp configuration into the ad637 seems to be different. maybe the secret lies there. ill simulate them in ltspice and get back with the results.

[catjuice]

the differences between the 175 and 175-av's ac conversion section are

c133 (1pf 175-av) + R139(100Meg 175-av) - these seem pretty inconsequential in this op-amp configuration. I made them in spice and i couldn't see a difference at an input of 3v @1khz.
I did desolder the lead of c133 to observe it's behaviour and it did seem to make a difference for my 1v 2khz square wave but not by a large amount it read 0.5487 v instead of the usual 0.5505 v.

C115(22pf in 175-av & 2uf in 175) - this is the Cav capacitor. this directly affects the averaging error based on it's value. - unfortuntely I wasn't able to see any difference in behaviour even after adding a 4.7 uf capacitor in parallel to my 175-av.

R116(76.8k in 175-av & 64k in 175) - this is just a voltage divider. this definetly would effect the value of the ac output. I don't happen to have a 64k 0.1% resistor on hand and I suspect that this would just divide the output of the ad637

C132(1uf 175-av)- this is right beside another 1uf cap... doesn't seem super concequential. but I could be wrong.





 

[catjuice]

UPDATE:
success!
the modification I have mentioned before works! the issue I was running into was my varification. since a square wave has a crest factor of 1 the rms output should match the average value before any sine rms correcting factors I think. so before modifying the voltage divider in the rms to dc converter circuit I couldn't see a difference before or after the modification so I assumed I wasn't doing anything. NOPE I was just being dumb.
I only had a square wave calibration signal from my osciliscope as a referance. so to varify the conversion I went out to my cegep to test and do a rough calibration with their function generators and multimeters.

I had replaced the 4.7 uf electrolytic averaging cap (C115) with 2 orange drop 1uf polyproplene capacitors (for worries about leakage currents) and the 100k (R115) resistor with a 120k resistor to make an equivialant voltage divider to the 100k 64k one in the 175.
I wouldn't worry about modifying the c133, r139, or c132 I mentioned in my previous post.

after a rough calibration where I compared a signal generators output with three aglient 34450a multimeters and adjusted my keithely. I was able to get ac voltage values within a 0.5-4% error of the other multimeters readings across various waveforms.

I'm fairly happy about this mod I hope any other keithley 175-av owners can take something from this.

[mawyatt]

Nice work!!

A couple easy waveforms to verify the RMS to peak value besides the squarewave (1 to 1) and sinewave (1/rt(2)) are the triangle and 1/x^3 waveforms. These have RMS values of 1/rt(3) and 1/rt(7) wrt the peak value. These waveforms, including the 1/x^3 waveform, are available on some modern quality AWGs.

Best,