HP 3455A Repair

[m2circuits]

Hi all,

I purchased this HP3455A a while back in non working condition. The problem is it just hangs on powerup. All (well almost all) the button LEDs light up and the display is blank. Pressing buttons does nothing.

I have already changed the caps in PSU and verified that my rails are present. I also made sure that both the in-guard and out-guard clocks are present and ~4V in amplitude, as per the service manual. I have the startup pulse at U26, pin 9.

The one thing, so far, that stands out to me is that the IN ENA pin on the out-guard nano processor is low, when the manual says it should be held high. This leads me to believe that there is an issue in the interrupt circuit. Every step after this tells you to use the signature tracer, but I don't have one - and looking at prices online I probably won't be buying one.

I'm kind of just stuck at this point, and without a second unit I have nothing to reference. I'm wondering if anyone else has had this issue and can lend a hand in the troubleshooting process.

Here is the link for the service manual: http://bama.edebris.com/manuals/hp/3455a/

Thanks,

M.

[wn1fju]

My experience with these (I've fixed three of them) is as follows.  Maybe they can give you some ideas.

1) When the machine hangs on startup, it is usually caused by the nanoprocessor not getting a "measurement complete" signal back from the inguard circuitry.
It is rarely caused by failures on the outgard circuitry.  But, reseat the processor daughterboard just to make sure.

2) You say you replaced electrolytic capacitors.  You need to check the caps on the various local power supplies on the inguard board.  Check all of the DC
voltages with an oscilloscope, not just a voltmeter.  Excessive ripple will play havoc with the inguard processor and the measurement will never complete.

3) Check the reference voltage coming out of the (removable) plug-in module (available at one of the inguard test points).  This has to be correct or the
inguard will be confused.

4) There are a number of LM399 comparators.  These proved very problematic in my units.  One of them is on the A/D daughterboard and is used to determine
the end of the measurement.  There are also some used on the main inguard board to control all of the FET switches.  Look at the two inputs to each
comparator and make sure the comparator is acting like it should.  Again, any discrepancy can cause the inguard measurement to never finish correctly.

5) If everything on the inguard looks OK, check the optoisolators that are passing signals back and forth between the inguard and outgard circuitry.

I have found that the service manual flowcharts for diagnosing problems is a bit inadequate.  They sort of assume you have a single point of failure which may
or may not be the case especially if, for instance, a couple of FETs are being turned on simultaneously when only one should be (an LM399 fault).  I have seen
many of these inguard problems cause the outgard processor to hang in such a way that you are fooled into thinking it is some sort of outguard interrupt
problem.  I have been down that erroneous path myself.

Good luck!

And by the way, I have several HP signature analyzers.  I don't think I've ever actually found a digital fault with them.  All I've ever seen was confirmation that
the digital circuitry IS working correctly.  And, of course, if the program ROMs don't match the version in your service manual, the published signatures will
be wrong.  So you are not missing much without one.

[m2circuits]

Hi wn1fju,

Thank you for your response! I appreciate you taking the time to help me.

I believe at one point at scoped all the rails, but that was a while ago.

I went ahead and looked at them again, and there are small (<20mV) pulses on the 5V rail. I'm not sure if this is enough to cause the problem I am having. I have no idea what the source is of the pulses either. The other rails have no ripple whatsoever. And yes, I made sure to measure them in reference to the in-guard ground using a differential plugin. It doesn't have to be a floating measurement, but I figured I can - so why not.

The +/-10V references look good.

On another note, I was looking around and there are some test points (+V and -V) on the A12 board (ohms converter) and I'm not measuring anything. The board is transformer isolated, so I took at differential measurement of the secondary of the transformer, and there's just nothing there. I could be barking up the wrong tree, but that just seemed odd to me.

I also took a look at the optocouplers. On the in-guard side I see the handshake pin go high on power up and then it drops low. There is no activity on any other pins of either one other than that.

Let me know your thoughts.

[wn1fju]

The 20 mV pulsing on the 5V rail shouldn't be of too much concern.  The inguard processor is controlling a lot of TTL chips so there will be some of that.  If
the "ripple" was something like 0.5V, then you would have problems!

At least on my unit, the ohms board V+ and V- measure zero until the 3455A is placed in ohms mode - then you get voltage.  Since your unit is locked up, you
can't select ohms from the front panel.  So I wouldn't conclude the ohms board is at fault (yet).

All your described symptoms sound very much like the inguard circuitry is not completing any measurements (whether the 3455A is running tests at startup, or
whether it is in the default DCV mode after startup).  You can try working through the flowcharts for the A/D daughterboard first.  Again, I would also look at
the bank of FET switches in the front end and the LM399 comparators that control them.  On one of my units, an LM399 failed that was controlling two FETs,
only one of which was supposed to be turned on at a time.  The failed LM399 turned on both FETs and created a lot of havoc.  You could also have a leaking
FET, but that condition seems more rare.

It's a tough debug since you sort of need measurements to be running so you can see what is wrong.  Yet if the measurements hang, then nothing will run.  It
is sort of circular reasoning.

[m2circuits]

Ok so i went ahead and started the ADC troubleshooting. I set up the meter as per the instructions, removing the interconnect to the out-guard board, disconnecting the analog jumper, and feeding in -10vDCc to the input terminals.

I am only getting one waveform coming out of the ADC board, so I followed the troubleshooting guide further. I am not getting -10v at A10TP4, there is just 0.3v sitting there.

I found out the front/rear input switch had the HI contact bent and not making contact. I bent it back and now I have -10v on the input to the first resistor. Something is loading down the multiplexer node after A10Q1. A10Q1 is being turned on by the comparator, which is receiving 5v on pin 7 (+ input) of A10U8. The theory of operation says that no attenuation should be active from 0.1v-10v, and that it is directly connected to the input of the DC preamplifier. Q2 has its gate pulled low, so it should be off. Perhaps there is a shorted FET somewhere, or another comparator is shot.

I think we're on the right track, and that the problem might be in the input amplifier.

Also, yes, the ohms converter is switched on by a latch controlled by the in-guard processor. The divide-by-4 that supplies the switching to the DC-DC converter has the CLEAR pin tied to a latch.

Let me know what you think.

[wn1fju]

First of all, I apologize for keep referring to the comparator as an LM399 - it's an LM339.  Must have a stuck bit in my brain.

Sounds like you are making progress.  I don't have much more to add without studying the schematics in detail, which I haven't done since the last time I fixed
a 3455A.  Keep attacking it.  I do know that if the FETs aren't switched right, the thing can't fully do a measurement (because it is switching references in and out)
and will generally hang.  But you already knew that!

[m2circuits]

Aha! I found that A10U6, an LM339, has two outputs stuck high, even though the +input is at 0v. That means that two of those multiplexing FETs are turned on - when they shouldn't be.

I have ordered some LM339s and they should be here this week. I will update once I replace A10U6!

Thanks again for all your help.

[m2circuits]

I changed A10U6 and I now have the -10V being fed in to the inputs appearing on A10TP4. 3 of the 4 waveforms for the ADC look correct now, the only one I'm missing is the Zero Detect signal. I reconnected everything and the unit still hangs on startup.

Looks like I'll need to trouble shoot the zero detect circuit, yay.

Making progress!

[wn1fju]

For what it's worth, the very first 3455A I fixed had a defective LM339 on the A/D daughterboard (which is the zero-detect).  Some of the others had defective comparators controlling the FETs in the front-end as you've already experienced.  Between your unit and mine, one wonders if HP bought a few thousand marginal parts back in the day.

[m2circuits]

It turns out that there is a typo in the manual, where it says that the zero detect is on pin 5 of the daughter card, when really, if you read further, it's on pin 'S'. The other side of the PCB.

The ADC waveforms are all correct now. But the thing still hangs.

I'm getting all activity on the data lines of the in-guard processor, but the out-guard seems pretty silent.

I noticed previously, that there is a missing waveform on pin 7 of A1U48. I've attached the screenshot of the schematic area in question. I had already replaced A1U48 and capacitor A1C29, as per the troubleshooting guide, about a year ago. Didn't make a difference.

I'm really hoping it's not the MOSTEK ROM, I have no idea where I'd get one if it is the cause behind it.

[Kleinstein]

For what it's worth, the very first 3455A I fixed had a defective LM339 on the A/D daughterboard (which is the zero-detect).  Some of the others had defective comparators controlling the FETs in the front-end as you've already experienced.  Between your unit and mine, one wonders if HP bought a few thousand marginal parts back in the day.

For some reason LM339 comparators seem to fail relatively frequent. It could be a general problem with those chips. It could be some aging problem (e.g. due to relatively high voltage)  or a common failure mode in the use as a JFET driver (e.g. pulses during start up) - though I don't think they are especially susceptible to latch-up.   


If the ROMs are bad, there may be a chance to replace it with an EPROM+adapter.

[wn1fju]

I opened up my unit this morning and took a quick look.  As the attached scope picture shows, the signal on U48 doesn't appear until 2.34 seconds after the
unit is turned on.  It goes through whatever start-up tests and procedures it does, then you hear a reed relay chirp, then the signal appears.  The scope was
set to 1 sec/div, so you can't see the detail on the waveform - but I did verify it is the correct ~1.4 ms period.

I would surmise that your 3455A is still stuck on start-up due to inguard problems.  Therefore, the fact that you aren't seeing U48 outputs isn't necessarily
a problem with U48 or associated circuitry. 

If you are 100% certain everything on the inguard side has been debugged (and who is ever 100% sure anyway), you might take a look at the optocouplers.