HP3478A redux

[jeffjmr]

So, after slogging through most of the posts related to various troubleshooting of the 3478A, I have concluded that the demonic U102 on one of my units is bad.

I have found a company that claims to have a few thousand NOS of these. Price?

$2399 USD

I’m sure there are many who love this meter, but do you think these are flying off the shelf at that price?

Surely the considerable brain trust on this forum could develop a fix for this part. I am routinely amazed at the level of expertise and creativity exemplified here.  Lots of us own this meter. At least a few of us have bad U102s, and since it appears to be a case of “when” it fails, not “if” it fails, I’m guessing others have a high level of interest in a fix. I know,  but…..

Who’s up for the challenge?

Jeff

[Kleinstein]

The 3478 is a 5.5 digit meter that sold relatively cheap as used. The main nice point is that it has a 30 mV range and can thus resulve 0.1 µV. The LCD display can be a bit hard to read. So one would not spend more than $100 on a spare part. The 3468 would be a similar cheaper meter as a possible donor (lacking GPIB but otherwise very similar).

The U102 input switching hybrid contains quite some switches and precision resistors. For the U102 part there is still the possibility to be only partially dead, as it's likely a hybrid and include multiple functions. So one may not have to replace all.

The circuit looks somewhat similar (but not identical) to the input hybrid in the 3457 and 34401, which are more valuable ones. Chances are a bit higher to find a replacement project with one of these 2 meters, but I don't know of any.
So while it may be possible to build a replacement circuit on a PCB, it would be quite a large one. Not sure if the info in the service manual are really detailed enough to get all the functions. The AC part would be quite tricky.

[Conrad Hoffman]

I love my 3478A but since you can get one on eBay for <$200 (usually) I wouldn't spend too much to fix it. I've had mine for a very long time with, knocks on wood, no problems.

[ch_scr]

The same hybrid is used in the 44701a, which can be had for reasonable money. Cheapest & easiest option for repair.
Edit: Upon closer inspection, "the same" is not quite true - but it is at least similar. A cursory look on the pinout, it seems not too different - I guess there is only one way to find out 
Edit2: I've added a comparison picture - looks like the biggest problem might be a difference in how the shift register wants to be talked to? Aside that, only pin 8/9 would have to be worked around?

[AVGresponding]

The internal configuration of pin 40 seems to be slightly different.

[Kleinstein]

They are similar, but not the same. The AC part (e.g. pins 34-39) of the 3478 also look different. So I don't think it would be worth the hassle.

[wolfy007]

Before you go looking for the hybrid, in case you havent checked the COTO relays, check them. I had one with input issues, and at first I thought it was U102, but ended up checking the relays and one was faulty. I used the schematic to follow the input path to find the relay related to the input problems, set up the meter in that mode and tried taking a measurement after shorting the suspect relay to bypass it and the meter took correct readings.

After replacing the relay the meter was good, just needed to check calibration.

[jeffjmr]

Thanks. I’ll try that when work allows me to return to my bench. Only the 30M range is belly up so worth a try.

Jeff

[Kleinstein]

With only the highest resistor range not working, there could be many different problems, starting with extra leakage currents, e.g. from contamination.
A point to check would be the input bias current in the voltage ranges. A high bias current may not be obvious when testing the voltage modes, but could effect the high ohms measurement.

[jeffjmr]

Many of the troubleshooting steps that point to the U102 Hybrid also state that “Before replacing Ul02, make sure the A/D Controller (U462) is good.”

I have an HP 5004 on the way to play with that will check the A/D controller, but I am left wondering if that IC controller is any easier to replace than U102.

It is an Intel P8049AH. They seem plentiful enough but there also seem to be various sub-versions. And don’t they need programming? If mine is bad, is an appropriate replacement feasible?

Thanks,
Jeff

[Kleinstein]

The 8049 would is a µC and thus digital. A failure there is not likely to effect only 1 range. If the ADC does not work, other ranges would likely also not work, at last not correct.

If an ohms mode fails, is the current source working right ?

For me the suspect parts are U101, Q205 (diode used JFET at the ohms protection), U203 and as general suspects some electrolytic caps at the supply.
The protection of the ohms mode is in many cases one of the less protected paths.

[jeffjmr]

Thanks, Kleinstein.

I was trying to avoid boring everyone with the specifics but your detailed analysis would certainly benefit from them. I am deeply appreciative of your time and attention.

All electrolytics, Rifas and battery have been replaced. All power supply voltages are within spec.

All DC troubleshooting tests check good.

All ohms ranges check good and measurements match my K2015 very, very closely, with one exception. On the 30M ohm range, a 10M .01% resistor measures 10.00960 on my Keithley measures noisily between 10.0157 and 10.0179 on the 3478a. A 20M .01% measures 20.00851 on the Keithley, 20.0305 to 20.0292 on the 3478a.

30M .01% shows 30.0056 on the Keithley, and bounces rapidly between OVLD and 27M on the 3478A. Same result 2W and 4W. I know the 3478A can only be calibrated on one or the other, and I don’t yet have the setup to determine which (I am assuming that when I have my GPIB working soon, I can download the files and determine which.) It does make me wonder why you don’t get a UNCALIBRATED message on the one not calibrated!

I will redo the ohms troubleshooting process as I did not record the good and bad readings. I also must admit I have not yet determined which of the four COTO relays is used for the 30M range. Still squinting at the schematics to try to figure it out.

With gratitude,
Jeff

[free_electron]

It is an Intel P8049AH. They seem plentiful enough but there also seem to be various sub-versions. And don’t they need programming? If mine is bad, is an appropriate replacement feasible?

yeah, that is the mask rom version. AFAIK you can't dump those. ( there is no programming interface in those chips)

[bd139]

I’ve read a mask ROM 8048 out in the distant past. It may be doable.

See https://www.sbprojects.net/projects/8049spy/index.php

[jeffjmr]

I’ve read a mask ROM 8048 out in the distant past. It may be doable.

See https://www.sbprojects.net/projects/8049spy/index.php

Well that would be half the battle. But how would one go about programming a new chip?

Thanks,
Jeff

[bd139]

Think 8748 is pin compatible and comes in a UV erasable and programmable version.

Found a disclaimer elsewhere that only some 8048 parts had readable ROMs. YMMV.

[free_electron]

It does make me wonder why you don’t get a UNCALIBRATED message on the one not calibrated!

if the backup battery was fully depleted the calibration is lost. that what that message means.
when swapping these batteries you need to provide a 3v supply to prevent this from happening.

[Kleinstein]

The difference in the "calibration" of 2 wire and 4 wire ohms should be only in the zero offset. This is relevant for thelower resistors, but not for large resistors like 1 M and mode.
A seprate offset constant for 2 and 4 wire ohms mode would indeed make sense - the 4 wire case may even have a fixed zero / the same as the voltage case.

The high resistors don't really need 4 wire mode and te 2 wire mode would be better, as less leakage and less hum pic-up is involved. It may help to use really short cables.  One may even consider using the non AZ mode for the 30 M range, as the AZ switching may introduce some trouble with a high source impedance.

The obsoerved difference is quite small. I think this should still be in spec. The 30 M range has quite some tolerance (0.078 %).

[jeffjmr]

I kept 3v on the pins during battery replacement. I just thought logic would dictate that since only one ohms mode can be calibrated, that a message would be generated when trying to use the uncalibrated mode.

I am not concerned about calibration or small differences in measurements at the moment. But the behavior with a 30M ohm resistor seems to indicate a serious problem.

I forgot to mention that AZ off made no discernible difference.

Is the couple k-ohms of noise on the 30M range when measuring the 10M and 20M resistors normal?

Jeff

[Kleinstein]

The high ohms range may be quite noisy. It is also one of the ranges most sensitive to mains hum. So it really helps to keep the cables short and maybe have a shilded around the DUT. Also make sure to use the slow reading rate.

Just in case it may help to check if the mains frequency is set / reading right. I don't expect it to be a problem, but a wrong value could cause more effect of hum. I would expect mains hum to be the bigger problem than true noise.

The meter seem to actuall use AZ mode even with the 30 M range. Getting similar or maybe even less noise in the non AZ mode is no surprise, though the specs call for an extra possible error.

[jeffjmr]

While I have some hard data, I have realized a significant handicap.

Both of my 3478As are 2911 series Serial numbers, very late production. The only service manuals I can find apparently don’t cover these late models. For example the manual I have contains schematics for circuit board part number 03478-66501 Revision B and C circuit boards. My units have board 03478-66511 Revision A. My boards do not have JM403, so the signature analyzer I have coming will be useless for examining the A/D controller. Hard to tell what other changes were made to the later board, but not having a suitable schematic is troublesome.

For what it’s worth, here is what I have so far. Ohms troubleshooting instructions:

7-C-8. Inaccurate Readings on All or Some Ranges

7-C-9. This failure is normally caused when the ohms current changes value due to a load change. Do the following procedure.

a. Set the 3478A to the 2-Wire Ohms Function and the 3K ohm Range.

b. Connect a 3K ohm Resistor to the 3478A's INPUT Terminals.

c. Using a high impedance Digital Voltmeter (like the 3456A), measure the voltage across R205. Since the ohms current on the 3K ohm Range is 1mA, the voltage should be approximately .47V dc.

d. If the voltage is radically wrong, the ohms current is incorrect. Use the overload troubleshooting procedure (in paragraph 7-C-6) to determine the faulty circuitry.

 I have several microvolts only across R205.  So I go to the overload section:

7-C·6. Overload Readings on All Ranges

7-C-7. An Overload is normally caused by a high ohms current (or an open between the INPUT Terminals and the Input Circuitry). Make sure the DC Volts Function is operating correctly, before troubleshooting for an ohms failure. Do the following procedure.

a. Set the 3478A to the 2-Wire Ohms Function and the 3K ohm Range.

b. Connect a 3K ohm Resistor to the 3478A's INPUT Terminals.

c. Using a high impedance Digital Voltmeter (like the 3456A), measure the voltage between pin 17 and pin 19 of Ul02. Connect the low input of the voltmeter to pin 17 and the high input to pin 19.

d. If the reading on the test voltmeter is + 4V dc and the 3478A displays an overload, the Range Resistors in Ul02 may be too low. Replace Ul02 .

•

e. If the reading is other than + 4V dc, the Ohms Current Source is defective. Go to paragraph 7-C-20 for troubleshooting.

I have 4.1Vdc but there is no OVLD display.  Frustrated, I decided to check current directly. I am getting the appropriate current on each range, within the range of accuracy and noise floor at milliamp to microamp values afforded by my K2015 with 24 inch leads. I checked current with my K2015 in series with the 3478 inputs to the 3K resistor, as well as voltage drop across a .1 ohm, 1% resistor and the numbers are correct.

Lastly, the state of COTO relays K101, 102, 103 and 104 do not change on any ohms ranges. K101 and 103 are closed and K102 and 104 are open. The only thing that changes between ohms ranges is the voltage to the contacts.

So, unless someone knows if a service manual or schematics exist for later units with the 03478-66511 circuit board please share. My service manual is number 03478-90008. I have found one 03478-99008 but it has the same publication date so I doubt it is any different. I am assuming others I have seen available, 03478-90004, 90005 and 90006 are even earlier. If someone knows differently, please advise.

Thanks all for your feedback. Maybe some of this data will be useful to others with 3478As as well.

Jeff

[Kleinstein]

At least for the plan I have (..008 revision), R205 should not have much voltage, as it is in series with a small capacitor and only for dampening, avoiding oscillation with a highly inductive DUT. So chances are there is a version conflict / other resistor to be measured.  If the test current is about right, chances are this first part of the test would pass.

A possible suspect to check would be if the ohms test current  changes with voltage. So maybe measure the current with an external meter and compare the current with only the amp meter and a resistor near the full scale range in series. While the absolute current has quite some tolerance, it should not change with the extra resistor. For the lower test current (e.g. 300 K or 3 M range) one could use something like 100 K/ 1 M to measure the current to get a good resolution also for low currents.

With 2 meters to compare, one could use the K2015 to measure the open circuit voltage for the higher ohms ranges. Excessive leakage is expected to reduce this voltage.

[jeffjmr]

A possible suspect to check would be if the ohms test current changes with voltage.

Not sure I follow how I would go about changing the voltage to conduct that test. Could you please explain?

I checked the open voltage on all ranges on both of my 3478As. The voltage is consistent 6.48vdc on 30, 300, 3k and 30k on the meter in question. It drops to 5.14v at 300k, 4.09v at 3M and 3.93v at 30m. My other 3478A follows nearly identically, just .01v lower.

These happen to be the same voltages I measured on the closed K relays.

Does this sound like a problem?

Jeff

[Kleinstein]

The open circuit voltages look OK, about what one can expect with clamping diodes insider U102 and the series resistors from the protection.
So the main function of the current source is OK. If at all the problem seems to be a small extra leakage current in the nA range. The 30 M range may also very well just have limited performance. After all the 3478 is not a high performance meter, but more like a cost effective version

Changing the voltage for measuring the ohms current would be by changing the DUT resistor as this changes the voltage at the ohms drive side.
So one idea is to measure the current with additional series resistance. Ideally the current should not change.

A current depending on the voltage could come from the current source (measurarble in 4 wire ohms mode) or from the voltage reading part (also causing more input current in voltage mode).