As promised, the followup to my HP3490A repair.? I found this 34702A/34740A at a hamfest for $12 and thought it deserved a home.? Upon hooking it up, the DCV and ACV seemed to measure pretty accurately but the Ohms measurements were way off.? Not too bad on the low ranges, but it eventually went OL at or above the 100K range.? This one has been frustrating.

I initially suspected a leaky Q15 (the dual FET) but replacing it and Q16 its current source made no difference.? Maybe U2, the op amp that controls the current?? Nope.? The manual suggests that out of spec on the highest range could be caused by Q11 (JFET) and so I pulled it and it tested bad.? Aha!? Replaced it with a known good part (though not an exact replacement), and it only made a slight improvement.? Now the readings are 2-3x what they should be up to about 100KΩ (steadily getting worse at higher resistances) but it still reads OL on a 1MΩ resistor.

I pulled Q13 and Q14, and they tested OK and nearly identical so I didn't replace them.? The voltages look reasonable; Q14 was questionable in-circuit but it may just be because of the defect disturbing its bias.? It seems to me that the current source is not pushing enough current through the unknown, and that's why it's seeing higher resistances than it should.? I'm trying hard not to just go on replacing parts until it starts working, but I'm about halfway there already.? ;-)

Hello Greg, it's nice to see you back with a new problem ;-D

Just a quick check, just to make sure the problem is not being caused by some leakage in the mainframe (the 34740): When switched to DC volts, on the 1V range and the input open, does it read 0.000 ? Or how much ? We're checking for leakage in the 10 nanoamps range here.
Cheers,

Joel

On 2/7/19 12:06 AM, gregdunn@... wrote:

Replaced it with a known good part (though not an exact replacement), and it only made a slight improvement.? Now the readings are 2-3x what they should be up to about 100KΩ (steadily getting worse at higher resistances) but it still reads OL on a 1MΩ resistor.

Have you tried checking the volts at output of U3 and the pots R59 R64 and the rotary switch contact next to R59?
Was it dirty at all? If so, is there conductive dirt on rotary switch?

Are R53 R52 good value?

They're all pushbutton switches, and I have (at least twice) cleaned the contacts with Deoxit.? It helped with some flaky behavior on the voltage ranges but did nothing for Ohms.? The upper 4 ohms ranges read garbage before I replaced Q11; now they at least are somewhat consistent, though way too high.? U3 actually measures good voltages at all the marked locations; but despite that, the gates of Q15 read different values.? The manual suggests shorting C1 and then checking gate voltages - it didn't make any difference.? I haven't changed the pot settings since replacing Q11, but I will go back in and measure the test points again before proceeding. If necessary I'll pull the pots and measure them.

The divider resistors R52-56 all measure proper values; you can force the switches to "all open" which essentially takes them all out of circuit and makes measurements possible without removing them.

I checked DCV on the 1V range, and it reads 0.03 whether open or short.? So, 30mV?? That seems quite a bit too high.

I'll take another look at some of the test points today and report back if anything changed markedly after swapping Q11.

On 2/7/19 9:46 AM, gregdunn@... wrote:

I checked DCV on the 1V range, and it reads 0.03 whether open or short.? So, 30mV?? That seems quite a bit too high.

That's suspicious, and if A and B volts of Q15 are different, the feedback is not working. (have not looked at rest of schematic though). Does the manual say what I ref should be, and is it? (can calculate from volts across those measured
and good R52 or R53 resistors.)

The ref current through R55 is ~1mA, R54 100?A, R53 10?A and R52 around 1.4 ?A - so the current is noticeably higher on the 1MΩ range.? Since the ohms converter output goes directly to the display (and not through the DC trimmer), I suspect it's supposed to be powers of 10 by range in order to not require any further gain shift.? R52 measures 900.0 kΩ, incidentally, and the others are similarly exact.

Time to take a closer look at the passives around U3, I guess?

Incidentally, DCV goes through the range trimmers and then straight to the display - there are no active components in the path.? And the display has an auto-zero circuit so it shouldn't be dependent on the analog stuff preceding it.? There may be issues in the display module too...

I was checking component values and noticed that R64 is a 10K pot instead of the marked 1K on the schematic.? Also, R69 is totally missing.? Apparently this unit is really early (s/n 00134 of the 1212A sequence) and later units have the 3480Ω resistor and a 1K pot.? Actual measured resistance is in the mid 3K range so that's believable.? Even the backdating info in my copies of the manual doesn't cover this very early rev.?

Just a follow-up: having swapped out all the active components in the Ohms section, the problem remains - 2-3x the actual resistance value on all ranges except any value over a few hundred kΩ reads OL.? I'm pretty sure I checked all the diodes already and found reasonable looking voltages.? So I guess it's time to start spot-checking some of the non precision resistors and seeing if any of them look iffy.

开云体育

Maybe a dumb question have you checked that the current being to applied to the test resistor is actually what it should be? ?A whacked out current source could drive bogus readings

Oh, there's no doubt that the current source is generating the wrong output.? Measuring the voltage across the test resistor gives a value which matches the erroneous display reading.? What we're trying to figure out is why the current source isn't working properly.? As mentioned, I've replaced all the active components in the ohms converter and am in the process of checking all the passives.? So far, nothing raises a red flag and yet the voltages are wrong.

Hi Greg,
Trying to find some logic process here... Here's my attempt:

- Let's first forget about the high-megohm problem, and focus on intermediate ranges for the time being.
- On the 1K ohms range, the current through the unknown resistor should be exactly 1mA. In other words, connect a milliammeter as the unknown resistance, the milliammeter should read exactly 1mA. Does this check ?
- This means that the voltage drop across the reference resistor (combined R55+R56) should be 1V. Check ?
- This 1V drop is equal to the voltage difference between U3 pin 3 (minus 6.2V) and U2 pin 6 (minus 7.2V). Does this check ?

Then we'll take it on from there.

I hope this helps !

Joel Setton

Joel, your thoughts are always illuminating!

1K range, loaded by my Fluke 8600A in mA mode - 2.08 mA

R55+R56 = 2.1V

U3-3 to U2-6 = 2.06V

Completely consistent, though the wrong values. ;-)? I checked the zeners last night; all are pretty close to their rated voltage except for CR14 which I think is part of the protection circuit and shouldn't be breaking down in normal operation.? I actually tested Q13-14 out of circuit and they seem OK, so something appears to be turning that Darlington pair on too hard.? Replacing Q11 improved the situation on the lower ranges a little, but it's clearly not the major fault.?

The voltages on Q12 look OK, though B-E is a little low (probably because the protection circuit isn't being activated?).? I tested it out of circuit and it looks fine.? Not sure where to go next...

If you're measuring about 2 mA test current in the 1 K range, it's about twice what it should be, so focus on the opamp U3 circuit. The key is to have the 1 volt drop on the reference resistor string, indicated by U3's output being -7.3 V nominal, as shown on the schematic. If it isn't, then there's the problem. If you've already tried swapping all the active parts, then look at the resistors around U3. You may want to especially check the pot R64, and jiggle it around a bit and rotate the wiper a few times. Normally, it should have only a few percent total effect, but if it opens up, then U3's output should tend to stick at -6.3 V instead. Now, this condition should cause the current source to shut off or be a very low value instead - the opposite of what seems to be going on. But, perhaps if the string voltage is lost, it can cause something in the overall circuit to go out of its normal range and lock up or oscillate.

So, first make sure the -7.3 V voltage reference circuit is right, by measuring the voltages all around its feedback loop. U3's circuit operation is independent of the rest of the system, so should be straightforward to diagnose. The reason I always suspect trim pots is that they sometimes take a lot of abuse during their life, being over-torqued and such, even to the point of destruction. Small board mounted plastic ones especially can have their screwdriver slot (and wiper) mangled, or even have their leads broken loose from the board, by overzealous tweaking.

Once you get the right reference voltage, check it while operating through the various ranges and some unknown resistors, to make sure it stays put with different loads.

Ed

I see that the output of U3 is at -7.4V and the input is at -6.4V, so off from the specified value by 0.1 at each end.? The gate voltages of Q15 are finally balanced, which they weren't before, but the reading across R56-R55 remains at ~2V.? The voltage at the end of R52 starts to drop as you switch ranges, going from -5.4V to about -4.1V by the time you go to the 10M range; this agrees with the display error increasing at higher ranges.? That seems to point to U2, which I've already replaced.

R64 is fine - I tweaked it earlier to be sure that the value was in range since it's a 10K and there is no R69 on this board; it did change the displayed reading over a small set of values as you would expect.? Following the voltage from U3-6 back to U3-2 the voltage changes in steps as you check each resistor junction R58, R60, R61.? They don't seem to be a problem.

开云体育

Hello Greg,

I think we're getting close to the root cause... but I thhink I need some more data.

First, temporarily unsolder the collector lead of Q12 and one lead of CR13, which will disable the overvoltage protection.

Then, with a 1K resistor across the Ohms input, and when switched on the 1K range:
- We know that the voltages at both inputs of U3 are -6.3V, which is OK.
- The output of U3 is -7.3V which is OK, too.
- If the two gates of Q15 are balanced at -6.3V, then there should be a 1V drop across R55-R56.
- But you have measured the output current to be 2mA, whereas it should be 1mA.

Then please let me know...

Also, I'd like to try a little bit of lateral thinking. Maybe we're dealing with some spurious oscillation, could you connect a 1K resistor across the Ohms inputs, and look at the outputs of U3, then of U2 with an oscilloscope ?

Joel

With the protection circuitry disabled, the 1K resistor reads 1.87K.

U3's inputs are both at -6.4V.? Its output is at -7.4V.

Q15 seems to depend on what I'm measuring; for this test, its inputs are at A:-5.5V B:-6.4V.? I was getting a fair match yesterday when testing the output current.? There is still a 1.9V drop across R55-R56.

U2 pin 6 (its output) is at -10.9V now.?

U2 and U3 outputs are both quite clean with the 1K resistor on the unit's inputs:

The problem is obviously revolving around the op-amp which is made up of Q15 and U2.
The two gates of Q15 should be balanced at all times, otherwise the output of U2 would be stuck at the +12 or the -12V supplies (because U2 is supposed to have infinite gain).
So you could do this: in a situation where the two gates of Q15 are NOT balanced, look at the output voltage of U2, does it saturate ? What are its two input voltages ?
Does it control the output current ? Are you seeing any voltage drop across R43 (in series with U2's output) ?
I think we're getting close to the "Aha !" moment !!!
Joel

> I think we're getting close to the "Aha !" moment !!!

I hope so.? ;-)

In this setup (Q12 / CR13 out of circuit), the gates of Q15 are offset just a bit - A: -5.0V B: -5.8V? The drains are of course then far apart - A: 2.8V B: 8.46V? and the sources are tied together at -4.9V.

This means the inputs to U2 are both way off from the quiescent value of 5.7V.? U2 is not quite at the rail, but close: -10.9V and there's no drop across R43.

This should be trying to reduce the current from the current source by turning off the FET/darlington circuit, right?? Yet it isn't.? I can't quite believe that the replacement Q11 is also defective, but maybe...

If U2 is railed negative, then Q11 should cut off. Are the parts in sockets? If so, try pulling Q13, which should drop the gain and the current to a very low value. If there's no effect, then try Q14. If Q13 or Q14 are bad, the current will be out of control. These two are exposed to the outside world in ohms range, so could have been damaged by over-voltaging, even though the Q12 circuitry is supposed to protect them somewhat (SOA-wise, but not much beyond Vcbo). They should both be fairly high Vcbo types (I'd guess 2N3440 or similar - an HP favorite), able to withstand momentary? specified fault conditions. They may test OK at low voltage, but if Q13 in particular is leaky, it could throw the current way off. For testing and diagnosis, you can replace them both with ordinary NPNs like 2N3904, then put the right ones in after you figure it out.

Also, you may want to recheck Q11 to be sure it's OK, and the of right type. It's easy to quick-check with a DMM - the D-S should look like a resistance with no gate connection, and the gate to D or S should look like a diode.

Ed

I was just looking back at the schematic and OP - I see that you reported Q11 was bad and replaced, but not with the same type. This could be the problem, since JFETs all work more or less the same, but can vary greatly in characteristics. Check the Idss spec of your replacement versus the original type. It could be that Q11 simply can't cut off far enough with the available negative gate drive from U2. If you don't have the right one, or one close enough, you can help it out a bit by adding a shunt resistor from the gate to -12V, to drop the gate more, then U2 can lift off the rail and get control of the situation. Another option is to temporarily add a diode or two in the source lead to lift it, and help cut off. You should put the same or a proper equivalent part in once it's figured out.

Ed