Request for help with repair of 3456A board A10 (03456-66510)

[DefinitelyNotLarryEllison]

I've had a 3456A for a while that I love, but it's lonely. So I got a good deal on a second one on ebay that I was hoping to repair. I've spent some quality time with the A10 Inguard Power Supply board over the last few days but I've come up against a wall. I'm betting that several of you guys know the insides of this thing pretty well and I'm hoping that someone might take pity on me and help me (a relative beginner) out.

I took the seller at their word ("burning smell," etc.) so when it arrived I didn't bother plugging in or turning on. Upon disassembly, the obvious problems were on board A10, Inguard Power Supply:

• C19 was 100% dead short and bulging. But not leaking (yay!).
• Scorching on PCB around diodes CR9 and CR10.

I swapped out C19 (as well as all other electrolytics). As for the (possibly) burned diodes, other than the darkness on the PCB, they both actually looked OK. I tested both out of circuit with diode test function of a handheld as well as my 34401A and both showed a forward drop of .5 to .6V and open circuit when reversed. I still wanted to replace them, but based on their size, they look to be fairly beefy components and none of the rectifier diodes I had sitting around look like they have enough capacity. So I put them both back in, then checked continuity to make sure that both CR9 and CR10 were connected properly to other components on the board - they were, so the PCB wasn't burned as badly as I originally feared and the traces were OK. In addition, there was NO continuity where there shouldn't have been from, for ex., the burning causing carbonization.

I then plugged A10 back into the transformer, disconnected from everything else, and powered up. No smoke, magic or otherwise. I tested A10 output voltages: +15, -15 and -18V rails were all perfectly fine. But the +33V rail, shown as "+33V to +46V" on the schematic, *started* at about 47V and steadily climbed to at least 49V. The 5V rail was similar - it started at about 5.4V, but climbed to at least 6. It was about this time that I "smelled heat" and noticed that CR5 (1N5366B, 5W/39V Zener) was incredibly hot, so I shut it down.

I then did a bunch of A/B testing with the A10 board from my good meter - so far, the only difference I've seen is that the good board has a stable max of 47V on the +33V rail (say at JMPR2 relative to the TP1 ground test point). But the broken board starts at 47 and just goes up from there.

Because the +5 and +33 rails are the only problems, I've generally ignored the VRegs on the +15, -15 and -18V rails - in particular, VRegs U2, U3 and U4. But since that 5V output was high, and since the +33V rail is (apparently) partially derived from it, I thought I'd take a shot and put in a new 7805 in a TO-220 pkg since I happened to have one. This made no difference at all.

I've really tried testing for shorted components everywhere. I've also tested all diodes in-circuit (fwd and rvs), including the bridge rectifiers. So I believe I've eliminated all the REALLY stupid stuff. The problem Zener CR5 has also been tested out of circuit and normal fwd/reverse tests are normal, although I did not try testing the 39V breakdown voltage. Is that the place to start? Should I perhaps just obtain an exact replacement for that item and go from there?

Is there someone that might be kind enough to give me a bit of guidance on where to go next?

[Chris56000]

Hi!

First of all, I recommend you buy a small packet of ceramic spacers (about 1.5 mm hole dia and 2 mm high, from the bay, to support any hot–running zeners or rectifiers off the board about 4mm to prevent scorching the pcb more than it already is!

The next step is to look at your voltage–tap selector switches S5/S6 and ensure these are set in the '120V' position, as your mains could be on the high side if you're in a rural area or near to an Electricity Substation–transformer!

After you've checked these points, obtain a 2k7 5W resistor and solder it across CR5 to provide the specified loading across the +33 to +46V circuit as given on the diagram. Switch on and test again, if the voltage across CR5 is still too high, have a look at capacitors C7 & C8 – the e.s.r. of modern replacements may be much lower than what H.P. designed the voltage–doubler to give, and reducing them to 33uF for C7 & 68uF for C8, both @ 100V wkg. each may prove helpful – if you can get n.o.s. ones in good condition about the same date/year as those originally fitted to the board, this may help as well – this voltage–doubler is a typical example of where a modern very low e.s.r. electrolytic may be unsuitable, as the +33–46V line relies on the series impedance of C7 & C8 for correct regulation!

Aim for a voltage of about +36V to +40V max. across CR5 when you're selecting replacement caps for C7 & C8 with the temp. 2k7 load in place!

Once you're happier with the voltage measured across CR5, I suggest buying a new 39V 5W zener and fitting it to the PCB on ceramic–spacers, about 5–10 mm above the pcb, as mentioned in my first paragraph. I think the original type or an equivalent should be available from the bay, Mouser, etc.

The same remarks about the e.s.r. of capacitors C16 and C17 in the negative rail doubler circuit apply as well.

Regarding the +5V line from  U1, obtain a small 12V 5W car sidelight–type bulb and connect it across the +5V rail from U1 o/p (right–hand leg) to chassis, and retest. If the 5V line is still higher than +5.25V with this lamp–load connected, make sure the centre pin (GND) of U1 is properly returned back to the bridge–rec/reservoir cap. Negative, i.e., not dry–jointed or cracked print, etc., and if this tests fine, get another '7805 from Digikey or Mouser of a reputable make – many of the modern replacements being supplied these days have a very thin mounting–tab, much thinner than the standard TO–220 specification and I suggest not buying the thin ones!

If you can get the "L7805CT" 1.5A SGS–Thomson type, so much the better!

My experience with three–terminal regulator devices, from many years of doing industrials, is their o/ps tend to be 10–50mV on the low side with reputable quality devices, so if yours is greater than 5.1V, turf it in the bin!

Another possibility is that U1 may be oscillating, so please ensure C9 on it's o/p pin isn't o/c – whoever drew this schematic didn't look at it carefully enough as the capacitor symbol is missing – there's just a gap in the connecting–lines where the symbol should be!

Fit a 1uF 16V tantalum bead with it's + lead connected to the right–hand lead of U1 and the – lead to chassis–earth 0V line, if you're not sure what to use, or it's not clear from the parts list – I presume you have the whole 3456A S.M. to refer to!

If you're still stuck with excessive voltage outputs from the transformer despite all the above, and you're CERTAIN it's not running hot due to possible s/c turns in the primary, you might have to add a metal–clad wirewound resistor of about 22 to 47R at about 15W in series with one of the primary leads, taking care not to bypass it by accidentally wiring it after the tapping–switches – you need the additional resistor working in both the 100/120V and 200/240V selections!

Please come back if you're still stuck after investigating all the above!

Chris Williams

[DefinitelyNotLarryEllison]

Thank you very much for your detailed reply!

I am in the US as you appear to have guessed, so yeah - 120V AC. The transformer was the very first thing I looked at when I opened the box because I thought there was a decent chance that someone else had NOT done that. But it's definitely on the 120V settings.

I do have my working meter (and its A10 board) to compare against. When I put the non-working board into my working meter, the behavior is identical (so again, I believe transformer is fine); also, putting the good board into bad meter gives a positive result (so xform in bad meter should also be OK). On the working board, the "+33V to 46V" rail stays right at 46V and CR5 is cool as a cucumber, even with no load. The non-working board starts at about +47V and just keeps steadily climbing and that CR5 diode is INCREDIBLY hot right from the start. It's a small sample size (I have only one working board to compare with), but I do not believe the problem is that HP totally screwed up the thermals so badly that I should expect CR5 to put lovely grill marks on my fingers under normal use.

I kept suspecting that the CR5 39V Zener had a problem. It is a 5 Watt part, which I don't have, but I did temporarily swap it with a "normal" 39V Zener (small, glass, etc. - I think it's probably 1/4 Watt). It was basically the same problem, except the voltage on that rail started at 49 or 50 V (not 47) and started climbing from there. In addition, I've tested CR5 out of circuit and I believe its behavior is just about perfect - I tested it with a 4k resistor to get about 13 mA at 50V and the zener breakdown is just about perfect at 39 V.

Even so, I still don't trust it - I've ordered some replacement 39V 5W zeners.

But the fact that the voltage on that 33V rail starts at about the right place and then just slowly increases has me baffled...

Again, I can't thank you enough. I'll be trying out more of your suggestions over the next few days and reporting back.

[Chris56000]

Hi!

HP intended the 39V 5W zener CR5 to be used as an overvoltage–clamp rather than a voltage stabilising device, and designed their voltage – doubler with the capacitors available to them at the time, which were considerably higher e.s.r. than the modern ones, typlcally the blue–&–gold sleeved ones, that are sold today!

If you test the circuit without a load connected, C7 & C8 will charge up to the peak of the unloaded r.m.s. value of the mains–transformer secondary voltage, which may typically be 20V off load, so the peak o/p from C7 could be 28.4V added to the same from C8, giving a possible max. of 56.5V, which, as you've found, overshoots that even, if you have the transformer off load!

This is the reason I said "use a 2k7 5W load" – in fact a 2 or 3W power–oxide type would be adequate for the voltage across it – the calculated top worst–case dissipation from it is only about 0.75W.

If you have this resistor connected across CR5 and use a 12V bulb on U1's output as I suggest, this will drop the output–voltage from the transformer enough to give you time to experiment with the values of C7 and C8, and the much more efficent modern types will do the same job for a lower capacitance value – try reducing the series capacitance C7's value first to 33uF first, then the parallel capacitor C8's value to 68uF as a next step.

That generation of H.P. equipment used generously–sized (and specified!) mains transformers, and like all such manufactured items, are made to tolerances, as well as that, the permeability of the lams varies from one item to the next, so the o/p voltage quoted on a diagram, parts list or supplier's datasheet is only a nominal figure!

This is why I also said "you might need a 22R to 47R limiting resistor in series with the primary" – it would be totally uneconomical to try and buy another 3456A just because you're unlucky enough to get one with a mains–transformer on top tolerance – a series primary resistor will help compensate for this!

Chris Williams

[DefinitelyNotLarryEllison]

Are you saying, at least in part, that the amount of capacitance on C7 and C8 would have an effect on the voltage present on that 33V rail? Hopefully it seems like I'm not COMPLETELY clueless -- I think I do know "a lot" for someone that's actually software engineer by trade -- but those caps affecting the circuit in that way is something that I wouldn't expect to be true based on my limited knowledge.

I did replace all the electrolytics on that board. I didn't have exact replacements, so I just found something that at least met or exceeded the capacitance and voltage rating of the originals and had the same pin spacing. This means that for both C7 and C8 I used Nichicon VZM-series 470uF/63V caps. C7 is shown as 47uF on the schematic and C8 is 100uF, so obviously what I have on there now is WAY over the capacity of the originals.

Could that be at least part of the issue?

[Chris56000]

Hi!

Yes – electrolytic capacitors cannot be made to a close tolerance, –20% +50% is specified for most of the modern replacements available on the market today, and as I explained in detail in my previous replies, the output voltage from the doubler is determined by the series impedance of these capacitors as well as their capacitance value!

The only way to get this circuit to perform as HP designed it is precisely as I explained in detail earlier, by connecting the rated full–load to it and then be prepared to select lower values for C7 and C8 – this is what HP did in their design laboratory before the units went into full production and service documentation was prepared for them!

The American manufacturers of electrolytic capacitors were able to specify much closer control over their manufacturing processes than modern Far Eastern manufacturers are today – this is why I suggested looking for n.o.s. replacements ("new old stock") – as these would have a capacitance value and e.s.r along the lines of what HP had available when they designed the instrument – many Sprague electrolytics could be obtained to ±20% tolerance!

Have a look at your parts list in the service manual – you will find a 4 or 5 digit number–code with each part listed in many HP and Tek service manuals of that period, and a key table giving the number to the original manufacturer is at the beginning of the parts list!

This experience you have is PRECISELY why you must be careful about general "recapping" of power supply and other analogue circuits in test–equipment – their operating tolerances are much tighter than general consumer equipment circuitry, and if a parts list or circuit–diagram in a S.M. for a piece of test gear says "selected" or only gives a maker's part no rather than a generic type for any particular component – it means it, and "ad–hoc" replacement of components can easily put things outside their working design parameters leading to failure to calibrate or measure correctly, or even put bias voltages and currents out enough to cause damage to semiconductors elsewhere – many, many people on this Forum have found this when attempting repairs on test–gear!

If you have the original capacitors for C7, C8 and C19 then see if you can get the exact types used, or failing that, the exact same types of electrolytic that are working correctly in your good 03456–66510 A10 board!

Chris Williams

[DefinitelyNotLarryEllison]

Well, why didn't you just say all that in the first place?



I'm sorry if I seemed like I was questioning you or didn't believe you - that was not my intent at all. It was mostly my own lack of knowledge, coupled with the fact that I hadn't yet had time while at home to actually WORK on the thing.

At least initially, I decided to take the lazy route based on your help. I grabbed some cheap-ass caps that I had from a terrible ebay "capacitor collection" and threw them in for C7 and C8. I was able to match the 47 and 100 uF capacitance "exactly" (to the extent that they match what's on their labels) but could only get 63V rated, not 50V like the originals. I figured that it was almost impossible for the voltage rating to cause any problems (except maybe to the extent that the different voltage ratings may impact the ESR??). I fired 'er up and, lo and behold... the voltages were all either really close to -- or within -- spec.

LIKE YOU SAID!!!

In particular, the 33V rail that was now holding steady at barely over 46V as the HP engineers intended. However, that CR5 zener was still acting odd - it was cycling every 5 to 10 seconds between nice and cool and very warm. On the other hand, it was still not anywhere close to as hot as it was getting before, and this was also with no load at all... which, as you said multiple times, is very likely to throw things off a bit.

The 5V rail was now a bit high and showing a steady 6V, but I still decided to risk it and hooked everything up and switched it on. In short, it seems to be working perfectly, and AC and DC V and resistance measurements are all damn near spot-on in all the ranges I could test.

The caps on the board from my working meter are all Nippon Chemicon "SL" series, which (of course) were discontinued years ago. Their web site says that the replacement is the SMG series, so I've ordered a couple full sets of replacements from Mouser. I'll have to get my hands on a decent LCR meter and try to measure the ESR on the caps on my working board vs the replacements, but maybe I'll get lucky and just by matching all the capacitances, everything will magically fall into place...

I really can't thank you enough for all your help!

[Chris56000]

Hi!

Glad you sorted it - if you're still doubtful about the +5V rail being high, you might be able to add a 3A schottky diode in series with U1's output, either by cutting the print and adding it on the copper side, or if there's a convenient link wire on the PCB, you can remove the link and connect the diode in it's place - that will drop about 0.25V at typical working current!

(connect anode to U1 o/p and cathode (band-end) to the +5V rail)

Good luck!

Chris Williams

[KK6IL]

The 5V supply is likely high from the 39V circuitry acting through CR8 to pull it higher.  Three terminal regulators only shut off the input to output path to regulate, they cannot reduce the output when it is being pulled high from another source (a possible problem when op amp inputs are protected by diode clamps to power supply lines - excess input voltages can travel through the protection diodes and increase the power supply lines above spec).

I would suspect CR8 is acting like a zener and breaking down on + voltage peaks at the C7, CR6, Cr8 junction.

I don't know what CR6 & 8 are, but 1N4004's would be good replacements.

I would be tempted to run the CR8 anode to ground instead of to the +5 supply if the 39V supply continues to be too high.

[dietert1]

In our two 3456A i put 22R resistors in series with CR8. This avoids direct action of the transformer secondary J14/5 to the 5V regulated output via C7 and CR8. In my eyes the original design is unlucky. With this resistor you can also measure the direction of current through CR8. Maybe it is leaky.
In our two 3456A i could not find the zener CR5 and the voltage on JMPR2 runs above 45 V, too.

Another recommendation: The original design lacks two more diodes to avoid reverse polarity on the voltage doubler capacitors C7 and C16 during power-on, so they will last longer. One diode is Cathode=C16 + and Anode = C19 -, the other one Cathode=C8 + and Anode=C7 -.

Regards, Dieter

PS: I also found those coolers of the voltage regulators are not mounted very well. In one of my 3456A the 5 V regulator U1 was not sitting flat on the cooler and got very hot, while the cooler remained cool. I pulled the 5V regulator and put it back it in a better position, including a little grease.