Repair: HP 3325A function generator - advice sought

[Lorenzo_1]

I have a 3325A that functioned perfectly until recently after being brought out of one-year storage.  It started up with a brief single flash of front panel lights and display segments then went blank and unresponsive to key inputs. Has stayed that way since. I have no signature analyser, so manual provides limited guidance.  Power supply checks out fine and supplies are correct and stable at all points tested, including A5 keyboard/display and A6 control board. Disconnecting ribbon cables to secondary boards one by one and all together per manual makes no difference. Main 1.2MHz clock on control board is present at all points expected. A6U9 nanoprocessor has stable supply voltages (+9.4V, +5.0V & -4.8V). A5U8 chip is sending expected -ve pulses on keyboard presses and no stuck keys are evident (per manual p.8-A-1).

The one thing I've been able to find is that there's no 1kHz clock that the manual indicates should be at Test Points 1, 2 and 3 on A5 keyboard/display board (also per p.8-A-1) - just looks like noisy lines that are often HI and drop LO intermittently - and the signals are poorly formed/decaying square waves (unlike the 1.2MHz clock).  TPs 1-3 are fed directly by U17, U25 and U16 decoders(R/W, R, and W control respectively) on the A6 control board.  If main data bus cable is disconnected from A5 board and control board is started up in standby mode, same results apply - suggests A5 board is likely not the source of the fault.  Logic analyser applied to data bus indicates repeating and consistent patterns of data are appearing on the bus data lines and the noisy feeds to TPs 1-3 nonetheless have a distinct repetitive structure to them (i.e. they're not random noise). Haven't had time to delve further into that aspect.

I wonder whether anyone else has encountered this start-up problem or can offer suggestions as to possible causes.  Is it possible/likely the bad clock signal would cause the startup behaviour?  I don't have access to a signature analyser so I can make only limited use of the manual's test regime. I'm keen to bring this machine back to life if possible. Suggestions appreciated.

[rf+tech]

Hello Lorenzo_1,

Yes, I experienced the same failure mode on my HP3325A, almost one year ago. As I recall, the symptoms were the same. By the time I got the 3325 fixed, I had replaced more than twenty ICs, about half of those were on the A5 front panel with the rest on the main board.

Here are my scant notes from the A5:
 - U8 pins 2,5,9,12 all show high until keys pressed
 - U8 outputs do not reflect inputs
 - Clock signals on TP1,2 and 3 look strange, no waveforms shown in manual
 - Power-on reset holds low for about 100 mSec
 - Shotgunned U, now comes up consistently with displays 1,5,9 lit indicating an issue with the front panel bus multiplexing.

I did not record exactly which IC U numbers were replaced, but it was "many." I am fairly confident the problem is with the multiplexing of data to/from the A5 and my feeling right after repair was that there were multiple IC failures, seeing how three digits began working after "some" ICs were replaced.

I'm sorry I cannot be more precise but do hope this information will be of some benefit.

Please do post back with your progress.

RF+ Tech

[Lorenzo_1]

Thanks for the feedback - sounds a bit ominous! Will advise how I go with it - suspect it will take a while.  Do you recall if the TP1-3 clocking looked liked a regular 1kHz clock signal after the repair or did it remain 'strange'?

[rf+tech]

Okay, I have pulled the printed manual out of the closet and taken a fresh look. And jogged my memory a bit. The A5 ICs replaced would have been U3, U6, U8 and U9. The majority replaced were on the A6. I may still have some of the printouts of the A6 board where I marked up in color which ICs were replaced as I worked through the repair.

I don't recall the clock waveform being erratic as you describe. They may have been erratic, but there were so many signals to try and make sense of at the time last year that I resorted to the shotgun approach.

That in itself is a clue. BTW, by all means do sync your scope externally from the 1.2 MHz SA clock. Every signal to/from the display/keyboard is referenced from this clock and using external sync will enable one to make far better sense of the waveforms.

The clue I'm thinking of with the 1 kHz clock(s) is to check each with the A5 W27 ribbon cable disconnected, to see if the clock is being loaded/corrupted by the A5.

Look at the three decoders, U16 pin 9, U17 pin 15 and U25 pin 9 that source LWKD, LCSR and LRKB, respectively. U16 and U25 also provide signals to A1 and A4 - check each output to see if the same symptoms are present as for the three A5 clocks. Then progress to the input signals for these three ICs.

Everything to/from the A5 is multiplexed so without the clocks, there can be no decoding/encoding, no display and no keyboard response.

RF+ Tech

[Lorenzo_1]

Thanks for the additional information.  I did check the clock with the W27 cable disconnected - and it makes no difference, so I think the A5 board may well be OK.  Also the clock signal at the three decoders is messy - as it is on the A5 board.  So I think the problem is on A6.  Waiting for some grabbers to arrive for my logic analyser so I can get a better look at what's happening on the inputs to the decoders and why the clock signals look awry - hopefully they'll turn up in the next week or so.  Will let you know how I go.  Thks for the confirmation that the 1.2MHz clock is the right reference - I haven't used it for external synchronising of the CRO yet.

[Lorenzo_1]

Progress - I think!  After much investigation, I ended up replacing 1kHz clock timer (A6/U8) that drives the keyboard/display interrupt cycle and a couple of its capacitors. It was a bit flaky on probing but after replacement seems to be running nicely and generating a nice square wave at 850-935Hz.  The wobble is driven by the TRIG signal input.

Didn't fix the display turn-on problem but I was able to confirm the turn-on circuitry is working as expected - a voltage comparator (U44) and NPN transistor (Q1) feeding into the main processor Interrupt Request pin. The display flash on startup is created by the initial HI on the collector of Q1 - then the base starts conducting as U44 comes up and the collector goes LO.

After much debugging on the A6 board that was inconclusive (except showing a lot of it is working fine) I came back to rf+tech's suggestion of checking LWKD, LCSR, & LRKB that previously were very noisy.  Turns out I now have a pretty clean set of 1kHz signals at tests points 1, 2 and 3 on the A5 board - corresponding to these three signals.  They're clocking at the same rate as 1kHz timer - around 0.9kHz.  They're offset from each other - and LWKD gets two negative pulses separated by 65uS, so not a simple 1kHz as the manuals suggests.  But looks as though this bit might be working OK now. 

I've attached a photo of the logic analyser output showing TPs 1,2 and 3 (P1180560) and also the input to TP5 which is Interrupt Request on the main A6/U9 nanoprocessor.  Not sure yet why this is showing constant HI when the TP1-3 data indicates the 1kHz is now driving keyboard/display interrupts correctly. Also a CRO shot of the timing signal on TP1 which is a 410ns -ve pulse and quite clean (the CRO is set up with the raw signal at the top and the differential signal from circuit ground below it - I don't have a differential probe). TP1 does show a consistent small +ve excursion before going LO (see photo) while TP2 and TP3 are clean of this.  Not sure of the cause of this. Have also attached extracts of the A6 circuit in a document - the original online is not scanned well so readability suffers.

Still not getting display to turn on but seem to be making progress, so will now check out the A5 board further. All the other suggested tests on U6 and the nanoprocessor that I can do without a signature analyser check out OK.  That includes turning off the ROM switch on S1G to disable the U10 buffer (per p8-C-1 par 3) - which lights up the whole display and suggests the basics of the display drive system are functional.  I'm still a bit suspicious about the lack on an evident Interrupt signal at TP5 but will debug A5 a bit further first.

Any suggestions welcome of course...

[Lorenzo_1]

Looks like those small positive spikes on the LWKD clock supply line may be part of the problem. Checking out A5/U9 that carries the machine data bus lines to/from the display board, it seems the Q outputs are firing on the positive spikes and not on the rising part of the ~1kHz clock signal.  The CRO picture attached is synchronised to the U9 1kHz -ve clock pulse (on the lower graph) and shows the 4Q line on U9 on the upper graph. The wide 15uS +ve excursion on the upper graph has been triggered by the glitch and not by the timing pulse. There's also a 2nd one just visible on the RHS screen - clearly glitch driven.

As I'm not experienced looking at bus traffic, would appreciate advice as to whether my interpretation here looks right or not.  The bus must be carrying traffic for other bits of the system but I wouldn't have expected that traffic to be transmitting through as CLK line glitches.  And as U9 should only be transferring bus data to Q outputs on U9 CLK signals, I shouldn't be seeing these non-synchronous spikes on its Q lines. Would appreciate advice as to whether I'm missing something here.

As far as I can see, the various supply voltages look good across the boards I've checked.  But is it possible small deviations from calibration values will create significant problems?  Not sure how much tolerance there is in this system.

[rf+tech]

Lorenzo_1

Interesting find with these erroneous clock pulses and glitches. Are these seen on the Vcc pin of U16? Or on the Vcc pin of adjacent chips?

Looking at the 74LS273 datasheet and noting that Vih is 2 Volts minimum, the clock line is already well above this, at 3.71 Volts, if I read the display info correctly.

Vil must be less than 0.8 Volts (at the highest) to be considered a valid low. From 3.71 Volts, the clock line would need to drop by 2.91 Volts to meet the Vil upper limit, prior to the next positive going edge. The amplitude differential of these glitches look to be about a few tenths of a Volt, nowhere near the required hysteresis to be valid TTL.

U16 is displaying virtually no hysteresis on the clock input, making U16 look like a candidate for replacement.

rf+tech

[Lorenzo_1]

Hi rf+tech.  Really appreciate yr help with this. Haven't been able to do more for a few days and will be unable to get to it for a week or so.  However, had a quick look at Vcc on U9 and noticed it was showing Vmax ca 5.4V and Vmin ca 4.6V on DC coupling.  Had a closer look at that and attached photo shows glitches on the U9 5V supply (top display is raw signal and bottom differential - AC coupled probe).  They look to be swinging +/- 1.45V.  Had a quick look at the +5V test point on the A6 processor board (next to the bus connector servicing keyboard and display) and the same glitches are there.  Similar glitches also show up at Vcc on U16 but only swinging +/- 0.6V.  I presume they're real and not measurement artifacts.  I'll try and correlate them with the false triggers showing up on the U9 input lines.  They might be part of the problem.  Looks like I may need to go back to the power circuits and make sure they're stable and on spec.

[Lorenzo_1]

A quick look at these power supply glitches at Vcc on A5/U9 shows they are tightly linked with the false triggers on the U9 input lines but start consistently at the falling edge of the positive glitch, not in the lead-up to it, so there's a cause and effect issue to resolve.  Also checked the three 5V supply lines at the power supply output with all other boards disconnected - the signal is very clean with ca 50-80mV noise p-p.  Connecting the A6 processor board with the A5 keyboard/display bus disconnected causes the glitches to recur.  So it looks like something on A6 is causing them.  Regrettably my Tek485 died on me - I think the heat and humidity of the wet season is the culprit.  So I'm stuck with the HP digital scope to try and track this down.   

[rf+tech]

Hi Lorenzo_1,

My very first reaction to these glitches is probing error. Forum member W2AEW has an excellent video on the effects of ground lead length on signal integrity. Definite;y worth viewing.

1.45 Volts peak to peak seems too severe to be a decoupling problem.

Are these glitches showing up (at a lower level) on the power supply itself, on the supply side of the interconnect to the A6?

It would be worth having a look at the Vcc pin of other A6 chips to see if there is any glitch amplitude variation that can aid localization.

RF+ Tech

[Lorenzo_1]

Quite right - turns out the big 1V+ swings are indeed artifacts. In differential mode with shortest ground link I have available (about 2") between probes, the 5V supply is very stable at the power supply and rock steady +/-65mV noise.  However, glitches still show up on downstream side of 5V supply filter caps on A6 board at 0.4V p-p. Also present on Vcc of A5U9 (against its Grd pin) at up to 0.9V p-p.  I think it sensible to replace the filter caps in any event then look at glitches on U6 Vcc pins more closely. Will be a bit of a challenge telling what's signal and what's artifact without a differential probe. 

[rf+tech]

Lorenzo_1,

If you have a 5.5 digit voltmeter, measure Vcc at each IC against each IC's own ground pin, on the A6. Even with the relatively low DC resistance of the copper tracks, there will be a small measurable voltage drop as one works away from P5 Vcc point of entry. Make note of any place where Vcc drops to (or below) 4.75 Volts as this may imply a defective chip drawing a bit too much current. I've seen first-hand some weird TTL behavior when Vcc dropped to 4.2 Volts locally.

Just as the copper tracks present some small DC resistance, they also present some inductance. As current pulses on Vcc with gate transitions, there will be some additional amplitude imparted by the track inductance charging and discharging. The ceramic bypass capacitors located at each chip act as small charge reservoirs such that the Vcc current pulses of each chip are drawn from these bypass caps, rather than the larger filter caps on the board edge. In the unlikely event that multiple bypass caps have failed open-circuit, it is conceivable that Vcc would show lots of glitches downstream from the P5 Vcc point of entry.

I enjoy good challenges and view them as a means of continually honing my skills. Glad to be of assistance.

RF+ Tech

[Johnny10]

Just to let you know... that rf+tech knows his stuff !!

[Lorenzo_1]

Looks like it's not even this subtle.  A quick check indicates I have a whole bank of chips with no Vcc (0.2V) - at the end of the board I've not been working on. Hard to think that's not a major cause of the problem.  Feel now as if I should have checked this first, but I presumed after fixing a power supply fault and getting 5V Vcc everywhere I looked that this bit of the supply was OK.   Will have to wait a few days before I can check it out properly but hopefully this is getting closer to the root cause.

[rf+tech]

Just a heads-up to spare you some grief in case you have not noted - the HPIB section at the rear of the A6, shown in gray crosshatch in the manual, does not share Vcc and Gnd with the front part of the A6. It floats on its own isolated supply. Caused me a bit of head scratching at first 

[Lorenzo_1]

Bingo!  Replaced the fried bridge rectifier (A6/U74) and it came up good straight away.  That is indeed the 5V supply for the HPIB segment of the main board, that I hadn't been looking at - there's a lesson for me there! 

In the meantime I'd replaced several of the Al caps on the main board - 1000uF/16V, 470uF/6.3V and 220uF/35V as some were quite off spec.  Could only get low ESR caps in the right values locally.  Is that likely to cause a problem and should I replace them with regular caps?

In slower time I'll try and have another look at the glitches I was finding and see whether they've reduced or disappeared now that it's up and running.

Thanks to rf+tech and others for lots of helpful advice.  Now I just need to get my Tek485 going again.  Had it working beautifully just a few weeks ago after cleaning the attenuator switches.  It's the 4th time it's failed, but I've been able to resurrect it each time.   It's a very nice scope to work with.

[rf+tech]

Lorenzo_1,

Replaced the fried bridge rectifier (A6/U74) and it came up good straight away.

I am guessing that the display is now working?  I had not even considered that the HPIB section could cause loss of the display. That was not the case with my 3325A. 

In the meantime I'd replaced several of the Al caps on the main board - 1000uF/16V, 470uF/6.3V and 220uF/35V as some were quite off spec.  Could only get low ESR caps in the right values locally.  Is that likely to cause a problem and should I replace them with regular caps?

For bypassing applications, low ESR capacitors are actually preferable. To reason this out, think in terms of the ESR as a source of internal heating from the internal I*R drop. Low ESR and low internal heating implies a longer service life. That is, if the capacitors purchased are trusted name brands like Panasonic or Nichicon as opposed to WunHungLo... 

Glad to have been of assistance,

RF+ Tech

[Lorenzo_1]

Yes - display came back up fine.  Self-Cal passed completely and unit now back together and fully functional.  I didn't find anywhere in the manual that suggested checking the HPIB voltage regulator in the case of display failure and it just didn't occur to me to look at that when the other half of the board appeared to have good supply and to be largely functional, despite some odd looking signals on the lines.  The manual is much focused on signature analysis and that doesn't help me absent an analyser.  I guess also I was coming off the back of finding a dodgy counter chip in my logic analyser and so targeted this problem somewhat the same way.  Anyway, great to have it going again. 

A big problem is the poor scan quality of the only service e-manuals I've been able to find online (including the HP-Agilent archive).  The circuits are hand-annotated and much of the annotation very hard to decipher even magnified (some is just blobs).  I also couldn't find all the signal mnemonics that are on the circuit diagrams in the dictionary on p.8-30 (it goes from HA.. to HS..) yet the next page is labelled 8-31 so there doesn't appear to be a page missing.  Can't be sure whether this was just a problem reading bad circuit diagram scans or a deficiency in the dictionary.  Makes for slow work.  I might have to invest in a hard copy manual unless I can find a high-res scan somewhere.

[rf+tech]

Hello Lorenzo_1,

Turned on my HP 3325A a week ago and was greeted by a blank display - for the *second* time in about a year and a half.  Worked through the initial A6 troubleshooting steps from the manual to no avail. The three clock signals to the A5 keyboard/display looked good and could confirm there wasn't any data being sent to the display latches. Changed a few tri-state buffers here and there thinking one might be the culprit, still no success.

Following up on the clue from your own experience of a shorted bridge rectifier in the isolated 5 Volt supply for the HPIB section, disconnecting the LV AC input to that supply made no difference. Methodically checked the input and output sides of every optocoupler to/from the HPIB and found one half of U55 with an output stuck at +5 Volts. Replaced U55 with an HCPL2631 and the 3325A is back in service. 

For the benefit of yourself and anyone else who may experience an optocoupler failure, that can cause a blank display at turn-on, here are the original and replacement part numbers:

U51          6N136       6N136 (still available)
U52,U53   5082-4355   HCPL2530
U54,U55   5082-4364   HCPL2631

RF+ Tech

[FlyingHacker]

Nice job, both of you!

Nice scope there, Lorenzo. Always cool when a CRT is tied to digital. I have a Tek 2232, but it isn't nearly as advanced in terms of those reports you show in the pics.

[Lorenzo_1]

Hi rf+tech. Thanks for the details on another cause of blank display. Hope I won't need it.  Good to hear u got it going again. I do rather like this fine old gear that it's feasible to work on still. A lot easier than debugging my current iPod Classic charging problems.

Your suggestions on voltage checking convinced me to finally invest in a bench multimeter. I bought an HP34401A multimeter - freshly calibrated. A nice instrument!  Used this to check performance of the 3325A and my Tek485. Frequency cal on 3325A is spot on and absolutely stable across 34401 counter range (~300kHz).  And the 485 reads the 3325A signal spot on from ~10Hz to 60MHz within limits of display. So it looks like all three units cross-check well. Not worth paying local prices for calibration of either, so this will have to do for now. Does look as though they're all performing pretty well. I did tweak the 3325A's -15V supply by about 20mV to bring +5V supply back into spec. Output amplitude continues to run a little above calibration 10.8V rather than 10V. Undecided whether to try and adjust amplitude as it's probably not critical for anything I do - perhaps better to leave well alone...and fix the iPod.

[flanneltuba]

Though this thread is now almost 3 years old, I wanted to append my little bit of an HP 3325A success story to it so that then next intrepid traveller who Googles their way to this topic with their busted 3325A will have one more breadcrumb to add to their path.

I received a intermittently functional 3325A from an ebay "as is" purchase, where it would only occasionally power up into a functional state, but more often than not power up with a fan-only, or fan with a random button lit or scrambled LED display hieroglyphics. If I let the unit sit powered on for anywhere between 5 to 20 minutes, it would then typically power up successfully and mostly behave itself.

I tried being all technical about troubleshooting this thing, which among other things involved Google searching every which-way I could think of one of those which ways found me here on the EEVblog on this thread. Though my particular problem was not exactly like Lorenzo_1 or rf+tech described, it gave me some good information for thought and generally led me in the right direction.

To cut ot the chase, the offending component in my case turned out to be u14 on A6, a 74LS174.

To locate this part as the faulty one, instead of getting all super technical like Lorenzo_1 and rf+tech, and since my unit's faults seemed to be warm up related, I employed the old Freeze Spray and Hot Air station to divide and conquer my way across A6, until I singled out u14. Freeze it, and the 3325 will fail to start up; warm it up, and bingo, working like a champ.

For those of you in a similar fix with a complex machine like this, perhaps also like myself lacking in the specialized equipment and for that matter the specialized experience, good old freeze spray and heat might be your saviour too.

Best of luck.

Moo.

[Oh. And P.S. I too happened to have used a Tek 485 in my efforts. Faithful old friend of a scope, it is; long may its little screen flicker...]

[martinihenry]

I too would like to share my experiences with the 3325A, as I've had the same symptom with mine on multple occasions. In each instance, the miserable "Bill West" switches beneath the pushbuttons were to blame. They're NOTORIOUS for the flat spring binding up if they get dirty, and causing a "click" or "snap" when the button is depressed. When the buttons start snapping, it's possible for the spring to pop out of its plastic retainers, in which case the button usually remains in the stuck "on" position, eventhough it may not FEEL like it's stuck on. This will put the machine into the 'comatose' condition on powerup, as the OP described. In the 5+ years I've owned my 3325, I've had to disassemble the front panel 4 separate times to reinstall flat springs that have shot free from their Bill West Switches.
I feel a contributing factor to this is due to mine being located in a garage which sees pretty extreme temp swings between the seasons, which no doubt exacerbates the problem.

"Mark's Late Night Lab" over on Facebook told me how to fix the problem, and by jove, he's right, it works perfectly. The cure is to pull the flat spring from each switch, and clean it and give it a light coating of oil. I used PB Blaster. I also hit the trough that the flat spring lays in, as well as the slot in the vertical plunger that the spring engages. Obviously, you don't want to apply so much oil that it runs onto the circuit board, use common discretion. Once cleaned and lubed, test the pushbutton, and you should have a nice, smooth snap and click-free depression.

I hope this helps!

[SilverSolder]

Thanks for the hints about the Bill West switches @martinihenry.  I've been wondering what actually causes them to behave like that.  I have seen problems with those switches even when the spring looks to be in good condition, with no corrosion whatsoever.