HP 3470 System Multimeter - Ohms mode (FIXED)

[GregDunn]

The last of the trilogy of cheap equipment I picked up at the hamfest last Saturday.  While waiting for parts on some other stuff, I decided to dive into this.  It's a modular multimeter, with one of the old dot-LED displays and a choice of several plug-on modules such as a DCV unit, an AC/DC/Ω unit, and a BCD output (before the days of HPIB).  This one had the standard multimeter module and the 5-digit display; for $10 I couldn't pass it up.

Quick testing shows DCV and ACV to be accurate on all ranges.  I had some flaky readings at first, so I hit the switches with Deoxit and that seemed to tame them.  Ohms is good on the 200Ω and 1KΩ ranges (short and open behave as expected), then gives nonsense readings on the higher steps.  The PSU gives good ±5 and ±12 values, so I dived into the troubleshooting tree.  Seems like the voltages around Q15 are a bit off, though only about 10-20%.  The guide says U2, Q15, Q16, and CR5 - but looking at the schematic, CR5 is in the ACV section, so I think they really mean CR15.  They're all common parts except for Q15 which is the HP 1855-0308 dual FET found in several meters.     So I can probably scrounge up the other stuff, but this one seems to be a problem - there's no readily known substitution for it that I can find.  Used parts are fairly readily available around $15 or so, but I'm not too excited about sticking a used part of unknown quality in something I'm trying to troubleshoot.  Also, I hate to make multiple orders for a repair just to find out the fault wasn't where I thought it was; I've done that enough as it is.   

Does anyone have a suggestion about an appropriate replacement part for this?  It's a dual matched depletion mode JFET, probably 30-40V rated.

[coromonadalix]

My guess would be to find the closest gain / hfe specs  and glue them together to match the temperature / vs thermal coupling,   it worked for me in the past, but not for high end instruments or measurements 

[GregDunn]

That could work and I would consider it - unfortunately, the specs for this device are not even available on the web.  There's not even a cross reference that I can find; it's missing from the xdevs and bitsavers listings, for example.     The only hint I have is that the Vds max should be more than about 24V because of the power supply voltages.  Dual JFETS with that voltage rating aren't easy to find in modern parts catalogs; in fact they usually only give the Vgs.  I have come up pretty much empty when searching for a similar-looking part that is readily available - except for the used HP parts on the Bay.

I've done a little more troubleshooting on this unit since the other day.  One of the things I tried is cooling the FET and the surrounding op amps/resistors/zeners to see if it might be a leakage issue.  The Ω readings don't change at all, on any range.  They do seem to diverge from the correct values as you switch to higher ranges, so maybe that's a hint. I've cleaned the PC traces with no effect, so it's not contamination.

It's odd that this meter behaves very much like my new Fluke 8600 on higher ohms ranges, and that's a totally different issue - there's no range PROM on this HP meter and the range switches are now pretty stable.

[MadTux]

Had a similar problem on a HP-3456A where I accidentally smoked the AC-section badly . A HP 1854-0753 dual NPN transistor deep within on the RMS converter was fried too. Also no obvious replacement available.

My usual procedure in such cases, go to my Tek 7CT1N curve tracer and compare transistors. Picoamp-Meter/MegOhm-Meter is also useful to check for leakage, where it's needed, I have a Radiomete IM6 for that.   At first tried some expensive LM 394 transistors that I had lying around. But in turned out, these have such weird characteristics (they have very large saturation zone and need a lot of Vce to get into active zone)  that they didn't work there. Finally bought some cheap Tesla KC811 from Czech republic and they worked fine.

I guess at least 1 of those JFETs is fine in your instrument. So go to curve tracer and compare that JFET with others that you have lying around (Compare Vgs with original one, Vds breakdown of both). Probably something like a 2N5911/5912 will just work fine.

[GregDunn]

I don't have access to any sort of curve tracer, so I guess using one of my highest impedance ohmmeters is what it will have to be.  I will say that the gate and drain voltages for each half of the package are pretty close to each other (the source pins are tied together), so I'm not too sure about one of them being bad.  Worth testing, I guess.  It actually looks to me as though both of the zeners are off by 0.5V or more, and they're super inexpensive, so I'll grab a couple of those whileI'm ordering parts.

I looked up the 2N5911/5912 and they are only rated at 10V Vdss; during quiescent operation each JFET has about 11VDC across it.  IFN5199 looks a little better(20V), and it doesn't seem like there's a higher rated part readily available.  I'll put one of those in the cart.

Thankfully all the parts in the circuit are easy to find except for that dual FET.

[duak]

How about replacing U2 with a FET input opamp such as a TL071 and bypassing the discrete FETs entirely?  Based on just the attached schematic excerpt, it seems the FETs realize a differential high impedance input stage, with some sort of high frequency rolloff network.  If you do follow up on this, please note that the FET diff amp inverts the phase, so the + and - inputs of the opamp would have to be swapped.

I think hp introduced these meters in the early 70's before monolithic FET input opamps were available.

Cheers,

[Kleinstein]

The dual JFET should not see more than about 15 V DS (the two zener diodes + 1 BE junction). So JFETs with a 20 V DS rating would be sufficient. Still not that many FET pairs available, especially in the right case.

Replacing U2 seems like a real option.

[GregDunn]

U2 is one of the components listed in the troubleshooting tree when the gate voltages don't meet the spec, so it is definitely on my short list.     I'm ordering an LM201 which looks like it has slightly better offset and noise than the 741 that is cross-referenced - I know I don't want much higher bandwidth lest it be prone to oscillation.

If it turns out to be the FET, I'm going to try the 5199 before altering the circuit topology.  I have worked a lot with TL07x op amps, so if it comes to that I'll be comfortable with substituting one.  With my luck I'd still pick the wrong polarity when wiring it up.   

I know the FET shouldn't have more than about 14-15V across it, but that's assuming neither of the zeners fail open, and assumptions like that make me nervous.  I just got finished rebuilding an HP6920B voltage/current standard which had a 50VDC rated electrolytic on a PSU secondary which put nearly 55VDC across it.  Another forum member confirmed this on his copy, so I replaced the HP spec cap with a higher rated one.  Possibly why the unit was defective when I got it...

Anyway, parts ordered and hopefully I'll have something happy to report in a few days.

[duak]

GregDunn,

The LM201 isn't a typo, is it?  That opamp is uncompensated and requires a 30 pF compensation capacitor for unity gain stability.  However, I don't have the whole circuit so I can't say if it is needed.  It also occured to me that the RC network between the opamp inputs may be there to stabilize the circuit by altering the noise gain.  This is an old and not often seen trick to stabilize opamp circuits.

Cheers,

[GregDunn]

Ha; I had a cut/paste error.  No, what I bought is an internally compensated op amp since the circuit does not include a compensating cap.  There was an LMx01 in the cart when I was looking at various options and realized it wasn't the right model.

[GregDunn]

When waiting for parts on one meter, I frequently cycle around and have a repeat look at another one.  The parts came in for this one, so I decided to exchange the recommended components one by one as recommended in the manual.

The results: replacing U2, Q15 and Q16 made absolutely no difference whatever.  I neglected to pick up a JFET to replace Q11 with so that will have to wait till the next order.  With my luck, it'll be the problem.

What's weird about this circuit is that all the PSU voltages and references are dead on but every indicated measuring point in the circuit reads noticeably lower than it is supposed to.  Not by much; only say 0.5 to 0.8V, but it's strange that they're all low, including all the zeners.  And the gates of Q15 are still unbalanced by about 0.7V no matter what I do.  The voltages on U3 are fine, and U2 is good except for the absolute voltage on pin 2 / gate of Q15B.  I've checked most of the passive components as well, and they all are reasonable looking    

And for those who don't read the first post, ACV and DCV work perfectly on all ranges so it has to be the ohms section at fault.

[GregDunn]

Fixed!  All along the problem was Q11, the JFET which controls the current source.  The original part was probably OVed or ESD damaged, because it essentially looked like a soft 1-2K resistor between any 2 pins.  What slowed me down was that I had no clue about the device parameters and initially chose a JFET with too high a Vgp.  It wasn't getting shut off, so the current ran too high.  It also looked quite a lot like the damaged part in-circuit, so when changing it made no difference, I went down the garden path looking at other components.  Someone on the HPAK group provided the correct device specs (oh, you have to love a group where HP design engineers contribute!).  Putting an appropriate part in the circuit did the job.

[Shock]

Nice work!

[GregDunn]

Oh, I guess I should also report on what I used to fix the unit.  Pretty much any good JFET with 20V or so Vds and a max of 4V Vgp will function and seems to give accurate results.  I actually tacked a cheap NTE 469 in to test my assumptions and it worked too.  Something like a 2N4858 or PN4118 would be fine.

And for the dual FET input, I used an IFN5199 and it worked great - although I restored the original part after isolating Q11 as the real fault.  So far the unit has cooked for several hours and seems rock solid.  I think I'll run a quick cal check on it and add it to the list of working DMMs.