Tek 7704A Power supply repair

[gmilliorn]

I have a Tek 7704A with a -15V issue, fortunately it is on the low voltage regulator and not in the switcher part (no 'ticking').  I found two shorted tantalums on the interface board (the most inaccessible location) and replaced them.  The impedance went from ~4 mOhm to 40 ohms, so I think
that is OK now.

However, it still is not working.  I took a bunch of measurements and nearly everything is off, making it hard to pinpoint the problem.
If anyone sees anything jump out from the picture I'd appreciate it. 



But also, I was wondering about this debug strategy:
- disconnect the 470 ohm between -15V and -15V_sense (a safety, I suppose)
- break the -15V sense connection from the interface board
- inject -15V from a bench supply into the feedback node

This should put error amplifier Q3275 in the normal spot, allowing other measurements to
taken and hopefully pinpoint something.  Any reason not to do that?

[David Hess]

Those measurements show that the base or emitter of output transistor Q3294 is open.  There is no way the base-emitter junction can sustain a forward voltage of 9 volts.

[gmilliorn]

interesting - I had replaced both Q3290 and Q3294, Q3290 lost a pin to rust and Q3294 was frying due to the short. 
Maybe the NOS MJE2801s are not working.

[gmilliorn]

You're quite correct - thanks!  I removed Q94 and B-E is open-circuit.  It was NOS, either it was DOA or died on power up.  I
reinstalled the original transistor, since I replaced it before finding Q94.  It starting working after finding yet another
shorted tantalum on the display board.

I think I've replaced all the tantalums in the acq. unit by now, other than those in the switching power supply which I've
been reluctant to open up.  The display is a bit jittery so there may be more to find there or in the display half.

[David Hess]

You're quite correct - thanks!  I removed Q94 and B-E is open-circuit.  It was NOS, either it was DOA or died on power up.  I
reinstalled the original transistor, since I replaced it before finding Q94.  It starting working after finding yet another
shorted tantalum on the display board.

I have not had any problems with the linear regulator circuits in my 7000 mainframes yet, so I have not qualified suitable modern replacement transistors beyond calculating the foldback current limits in some cases.  I would not use NOS parts.

Q3283 implements fast foldback current limiting to protect the output transistor but of course it could have been DOA or unreliable because of age.

I think I've replaced all the tantalums in the acq. unit by now, other than those in the switching power supply which I've
been reluctant to open up.  The display is a bit jittery so there may be more to find there or in the display half.

The various metal cased wet and dry axial tantalum capacitors are very reliable, except of course when the wet ones leak sulfuric acid which should be very apparent.  The epoxy dipped tantalum capacitors are the suspect ones and I have changed several in 7000 plug-ins.

[gmilliorn]

Here's the cap buried in the interface board for reference.
The one in the display board had no visible damage but was equally ~0 ohms.

[David Hess]

That would be an epoxy dipped solid tantalum capacitor.  I have seen a few of those but not that failed yet.

[med6753]

As an owner of 2 - 465B's, 2 - 485's, 1 - 475A, and 1 - 7904 shorted tantalum capacitors come with the territory. All except the 475A have had at least one or more go short. By far it's the + / - 15V bus with the 20V/25V tantalum where you'll most likely have the issues. Back when the instruments were new this represented an acceptable voltage headroom and for the expected lifespan would remain so. But we're now talking 40+ years of aging and degradation and things aren't so rosy. Obviously there are other factors that come into play such as the amount of power on hours, storage conditions, etc. The biggest offenders in my experience are values 22uf and above with 47uf being the most troublesome. I replace them with caps rated at 35V.

These scopes are completely loaded with tantalum of various values/voltages and to replace them all would be too time consuming and costly. And probably not necessary. But the values/voltages I mentioned above are potentially ticking time bombs and should be addressed.   

     

[David Hess]

Sorry, I mean I have not seen the blue ones from Tektronix short yet.

By far it's the + / - 15V bus with the 20V/25V tantalum where you'll most likely have the issues.

Those are the ones I see fail, but they have been the striped and yellow/orange ones and never the blue ones.

Back when the instruments were new this represented an acceptable voltage headroom and for the expected lifespan would remain so. But we're now talking 40+ years of aging and degradation and things aren't so rosy. Obviously there are other factors that come into play such as the amount of power on hours, storage conditions, etc.

It is not real clear what is going on.  They are not failing because of surge current, and NASA says crystallization around flaws only occurs in high voltage parts.  I suspect they are failing for the same reason that newly soldered epoxy dipped parts fail; thermal stress causes cracks in the dielectric.  But in this case, I think age and repeated heat and humidity cycles are causing the stress which damages the dielectric.  Hermetic dry solid tantalum capacitors seem to be immune to this.

The recommended solution covers whatever is going on.  Higher voltage parts have a thicker dielectric which can absorb the heating from the self healing mechanism which clears shorts without entering thermal runaway.

[med6753]

Sorry, I mean I have not seen the blue ones from Tektronix short yet.

By far it's the + / - 15V bus with the 20V/25V tantalum where you'll most likely have the issues.

Those are the ones I see fail, but they have been the striped and yellow/orange ones and never the blue ones.

Back when the instruments were new this represented an acceptable voltage headroom and for the expected lifespan would remain so. But we're now talking 40+ years of aging and degradation and things aren't so rosy. Obviously there are other factors that come into play such as the amount of power on hours, storage conditions, etc.

It is not real clear what is going on.  They are not failing because of surge current, and NASA says crystallization around flaws only occurs in high voltage parts.  I suspect they are failing for the same reason that newly soldered epoxy dipped parts fail; thermal stress causes cracks in the dielectric.  But in this case, I think age and repeated heat and humidity cycles are causing the stress which damages the dielectric.  Hermetic dry solid tantalum capacitors seem to be immune to this.

The recommended solution covers whatever is going on.  Higher voltage parts have a thicker dielectric which can absorb the heating from the self healing mechanism which clears shorts without entering thermal runaway.

I've seen many of the blue tants short. They tend to be the 47uf which I have found to be the most troublesome.

As far as what mechanism is causing the fails it seems to be a total crap shoot.

[David Hess]

As far as what mechanism is causing the fails it seems to be a total crap shoot.

We do know some things:

1. Proper voltage derating prevents failures by allowing the self healing mechanism to work without thermal runaway.
2. Long term failures are not limited to high surge current applications.
3. Long term failures happen even on reformed parts not subject to thermal cycling during the soldering operation.
4. Failures caused by strain inducted cracks during thermal cycling from soldering can be prevented by reforming the capacitor dielectric which is normally only done during production.
5. Hermetic tantalums do not fail this way.

3 and 5 above makes me think it has something to do with temperature and humidity cycling of the epoxy package straining the tantalum pellet inside.  If it was defect growth, then it should affect hermetic tantalums as well.

[gmilliorn]

Temperature is an interesting possibility - on the '7704A those tantalums are sandwiched between several layers of boards and connectors.

I wonder if a 105C-rated lytic at higher voltage might be more robust, assuming an equivalent ESR was available.   I used 25V
tantalums since 35V ones were unavailable or insanely expensive.

[David Hess]

Temperature is an interesting possibility - on the '7704A those tantalums are sandwiched between several layers of boards and connectors.

I wonder if a 105C-rated lytic at higher voltage might be more robust, assuming an equivalent ESR was available.   I used 25V
tantalums since 35V ones were unavailable or insanely expensive.

An aluminum electrolytic would need many times the capacitance to get the same low impedance at the operating frequency so might not even fit, but I would certainly consider it.