Fluke 887AB Differential Voltmeter Chasing Zero

[Cooler Than I Look]

I'm new here and I've just acquired a Fluke 887AB Differential Voltmeter.   Its in good condition, very clean inside, and seems to work well except for zeroing.  Its has one of the older mechanical choppers which I've read are the usual failure point.   Sometimes I can successfully zero it but other times not, running out of range on the zero pot,  drifting around over maybe 5-10 minutes or so.

What I've done so far.

Visually inside it looks fine, none of the caps seem to be bulging or leaking and those that I've measured in circuit seem OK.  Resisters too, even the carbon comp - the ones I can measure in circuit meet their spec.  The power rails are within 0.1 of ± 18 volts as required and I can't see any ripple on the scope.  The 6v rail has more than a volt of ripple but with a single diode rectifier and 500uf cap thats probably to be expected.  Hanging another 1000uf over it greatly reduced the ripple - you could hear the change in the chopper but otherwise had no effect.   It seems to only feed the chopper driver circuit.

I've run through the relevant sections of the 'Performance Evaluation' in section 4-9 of the manual and apart from the difficultly getting it to zero and stay there everything else would seem to be well.  Even the null detector noise and offset are OK which surprises me.

A Plan

I've seen many suggestions from replacing the chopper with FETs to replacing the entire null detector amp with a modern opamp.  I'd like to keep it as original as possible so I'm tempted to just replace the chopper itself with a solid state replacement.  I'm looking at using a 74HC4066 'Bilateral Analog Switch' (because I have some).  The datasheet even lists chopper as an application!  I'll derive the clock from the chopper drive, replacing the chopper coil with a suitable resistor if the driver needs it and use a 74HC04 inverter to give me the two switches 180 degrees out of phase.  This can fit on a little board and I only need to disconnect the wires to the chopper so its easy to return to original later.

Alternatively whats wrong with a conventional reed relay?  I suspect there must be issues with contact closure time or symmetry or something because it seems an obvious solution (to me) and no one even mentions it as an option. 

 
Praise? Derision? Suggestions?


Thanks
Murray

[Cooler Than I Look]

I've made a little test board and lashed it up:




and it seems to work.  Its noisy but I'm not surprised with tall the long leads and a CFL just above it.  It does have some offset though, about 50uV.  If I swap the switches around and take the other side of the chopper drive for the clock the offset inverts so I'm pretty sure that implies its coming from this 4066 switch.     I could add a corresponding offset to the existing zero circuit but I'm not sure what the consequences of that may be. 50uV is probably more than the current zero's full range.  Perhaps I need a better switch?

Murray

[Cooler Than I Look]

Hmm, there should be an image in the post above but I can't get it to work. I'll try attaching it here...

[Kleinstein]

The 74HC4066 is in principle OK and can reach low offset. The problem is more with charge injection and external capacitive coupling. An example is capacitive coupling of the control signal to the output. This can add some extra current error and depending on the circuit this can have the same effect as an offset voltage. Ground routing can also be an issue. Another point may be the delay from the inverter.

[Cooler Than I Look]

Thanks Kleinstein.    I'd be surprised if the inverter delay at 84Hz would be an issue, especially compared to a mechanical relay but as I enter this world of uV its not as easy as it first seems...

I have the board mounted on top of the old chopper with much shorter leads which has got rid of the noise but I have significant offset, 50-60uV, half scale on the most sensitive setting with the input shorted.  I've tried adding a correction voltage but now the offset changes between shorted and open.  So from what your saying I'm looking at charge injection.   Whats a strategy for dealing with that?  Is there a better device I should be trying?

[Kleinstein]

The 4066 CMOS switch is not that bad, when it comes to charge injection. Like with most such switches the charge injection depends on the voltage level. Besides the charge injection inside the chip external capacitive coupling can behave similar.
The cheap HC4066 may have some leakage current too - at that price level one can not expect testing to the pA level, though chances are that many of the chip are actually not that bad. The large spread from typical to limit values is to a large part to skipping tests. 

For the switch delay the critical part can be a short time when 2 switches are on at the same time when they are no supposed to be. Even though only a few ns this can have a relatively large effect in some cases. It depends on the circuit used.

A 50 µV offset looks like quite a lot too me, so there may be other contributions (e.g. from the part on the old board that generates the 84 Hz clock), than just the switches. Due to filtering and protection parts the short may not be all the way to the chopper part - so even the voltage with a shorted input may be due to a current.

AFIAK the older CD4016 switch should behave a little better with charge injection.  However with just 84 Hz I think the 4066 should be good enough. The old CMOS instead of 74 HC version might also be a little better, though higher resistance.

[Cooler Than I Look]

Heres the original circuit (if I can get these images inline!):



and this is my mod:



You set me on the right track. I set out to see if I could measure the delay through the inverter but instead I've found the 84Hz clock is very unstable with a lot of jitter so I guess those caps are not as pristine as they look.  I'll pull out that board and get that multivibrator stable then see what my offset looks like.

thanks!
Murray

[2N3055]

I would suggest H11F Photo FET as switches. Speed is not a problem here and charge injection should be zero.

[Cooler Than I Look]

Some success 

I noticed the symmetry of the chopper waveform wasn't great but the other side (the chopper is 'across' the multivibrator) looked perhaps a bit better.  I had to swap the connections of the switches over to match the inverted clock and the offset was much better!   Is it the other side of the clock or have I lucked upon a better order of the switches; cancelling rather than adding errors?  Who knows..

I was gladdened to see Fluke needed to add a little offset even with optofets.  https://www.eevblog.com/forum/metrology/teardown-fluke-845aabar-tweaks-and-mods-(and-repairs)/msg926393/#msg926393  (the blue trimmer by the two H11Fs) so I've tacked a 10Meg resister onto the 'zero' switch input and with a trimmer I'm adding 0.85 volts to zero the meter.  The 'zero' leg has 200R to earth (R236 in the schematic above) so I calculate that to be 17uV I'm adding.  Seems to work.  The needle is fairly quiet, doesn't seem to drift over 10's of minutes as it was with the mechanical chopper and with one of those cheap AD854L references from ebay I'm reading 4.99982 volts.  Considering the unlikely accuracy of the reference and the stated 0.0025% accuracy of the 50 year old Fluke meter the fact its that close is pretty amazing.  Or a coincidence (a fluke!).

[dgminala]

There's a simpler solution to the chopper problem.  The idea is to replace the entire null detector amplifier with a single chopper stabilized opamp.  See the attached PDF file for details.
I have an 887AB that had a bad mechanical chopper.  I replaced everything with the modification and you just have to set the zero adjustment one time, and then forget about it.  It's completely stable and noise free ( no needle jitter).
The majority of the time required is the time to remove all the unnecessary components.
Good luck with your instrument.

On another issue is a much newer 887AB (s/n 2513) that I acquired a few months ago that has several germanium transistors missing from the reference supply board.  Finding suitable replacements might be an issue, but maybe not a show stopper.
This instrument has a (factory installed) null detector board that uses a single opamp, a Burr-Brown OPA-111.
I bought a manual for this instrument from Artek Manuals that covers the serial number of my instrument, but there's nothing in the manual about the new null detector board (887A-4022, p/n 812610).  If anyone has any information (schematic would be wonderful), I'd love to get a copy. 

Thanks for any info
Dave M

[Cooler Than I Look]

Wow thanks Dave.   

Others did suggest a radical null amplifier replacement but I don't like doing extensive mods to equipment as fine as this.  My mod has a little offset but it doesn't bother me and its a 10 minute job to restore the voltmeter to original.  Still I've filed that PDF away in case other parts of the amp start to die.

I've read on this forum and others that the germanium transistors can be replaced by silicon in most cases.  I'd say especially so in the DC sections.

Murray

[Cooler Than I Look]

I couldn't leave it alone   

The offset (difference between shorted and open inputs) was > 26uV (Fluke spec is 2uV) and the zero point did still tend to drift a little.

I've added Randy Evans' opamp mod that dgminala brought to my attention but I didn't do the wholesale removing of the existing null amplifier, just lifted a couple of wires to bypass it and unplugged the transistors for the chopper driver (socketed transistors!)

I did change his opamp circuit though. There was no input protection on the LTC1050 so I've got jfets with their source and drains tied together as low leakage diodes to Vdd and Vss.  They'll allow the input to exceed the power rails by 0.6 volts and the 'absolute maximum' in the datasheet is 0.3. Oh well, maybe theres some tolerance there.  Low leakage input protection is hard!  Dave had a video using two BJTs with their bases and collector all joined. I tried every bjt I could find around here and none would clamp lowered than 9 volts. 

My other concern with Randy's mod was the amount of current available to the meter.  Its a 100uA - 0 - 100uA meter and on overload (the above jfets conducting) the meter could get nearly a mA through it.  Not sure abut how much these meters can take but I through that was a bit much.  Theres a 3.9k resistor on the output (R225) which I split into two, 1.2k and 2.7k.  At the junction of those two I've got two 1N4148 diodes clamping to earth. That limits the meter current to 130uA or so.

So its working again.  The meter is very stable and my offset is about 4uV.  Thats still more than Fluke's spec'ed 2uV but I suspect that might be the tiny leakage through the jfets.   Waving your hand past the input connector can cause the needle to deflect nearly full scale (the opamp input is completely floating) so I'm not too worried. 

I put 30 volts into it on the 1 volt range with the null selector in the most sensitive position and the opamp survived as did the meter.

Murray