[solved] Help on HP 8116A short at +/-24V supply

[pelule]

A "miracle" problem at my HP 8116A function generator (ser# 2124G02272)
At start-up it showed E21 - thus I started to investigate...

test 1: measure the supplies (all boards connected)
result: +24V (after F1/W15, fuse) and -24V (after F2/W16) are 0.00V - both fuses are blown
   all other voltage levels are within given tolerances (correct)
note: change #64 applies: fuses instead of Zero-Ohm-Bridges, F1 replaces W15, F2 replaces W16
   mmhmm.. shall be 125V, 5A fuses - 0.5A may probably correct value

test 2: visible check for suspicious parts on the board (dark resistors, leaking caps etc.)
result: nothing suspicios found

test 3: disconnect all boards, test main board A1 only, measure current with ammeter instead of fuses.
result: main board sink ~0.6A @ +24V, ~0.8A @-24V - full low impedance short
note: max. current limited by the +/-24V regulation circuit

- solder two 12k resistors over the burned fuses, to be able to measure in the circuit without excessive current.
   voltage drops down to ~30mV behind the resistors => short impedance less than 25 Ohms

- analyse the schematics for potential locations creating the low impedance shorts
  (transistros, diodes capacitors, theoutput amplifieer is normally the first candidate)
  schematic 1B is power supply (until fuses W15 and W16)
  schematics 1C, 1E and 1F do not use +/-24V
  schematic 1D use +24V, but not -24V
    +24V at U200A (pin 4) OpAmp positive supply, GND is negative supply => may cause at +24V, but not at -24V
    +24V at R211 (1 MOhm) and R218 (1 MOhm) to Ramp Current Source - can't sink such high current (1 MOhm)

  schematic 1G use +/-24V (ouput amplifier, based on tranisistor, would expect the problem there)
    +24V at R516 (11k) - can't sink such high current
    +24V at R558 (75R) to capacitor - can't sink such high current, would kill resistor (>10W)
    +24V at R524 (100R) - can't sink such high current, would kill the resistor (>10W)
    +24V at R533 (19.8R) - can't sink such high current, would kill the resistor (>10W)
    +24V at CR503 - potential defekt, may have current via Q503, Q505, L500, Q507, L501, Q508, Q506, Q504, CR504 to -24V
       would be a very complex (unrealistic) current path, but need to be tested
    +24V at L504 - can't sink such high current via Q510 and/or Q512, would kill R543 and/or R545 (>10W)
       verified by cutting connection - no change
    -24V at L505 - potential defekt, current via Q511
       verified by cutting connection - no change
    -24V at R527 (100R) - can't sink such high current, would kill the resistor (>10W)
    -24V at CR504 - can't sink such high current, would kill the resistor R537 (>10W)

SO WHAT NEXT?
Are there some know issues like Tantal capacitors?

Any tip or help is highly welcome.

PeLuLe

[richnormand]

Over the years I have found many instances of tantalum caps going dead short. The teardrop types in particular.
In some cases, even taking the power supply with them (Canberra MCA and Sonoscan acoustic microscope).

Many issues were on large logic boards using several caps distributed all along the traces near the ICs. There were even some designs with NIM boxes that would insert a small resistor in series, voiding the low ESR argument but insuring the resistor would open in case of a dead short, instead of bringing the whole unit down. For some long term nuclear measurement 24/7 reliability was more important than a few missing decoupling caps.

[MarkL]

First, I would verify if you're dealing with an actual short, or if some other part of the circuit which is still under power is influencing the behavior of the +/-24V rails.

What do you get when you measure the actual resistance of the short?  (I'm talking power off, with an ohmmeter.)  From your previous investigation, it would imply it's not a dead short.  Perhaps it could point you to a possible path through a low value resistor.  Does the polarity of your ohmmeter make a difference?

How about measuring the resistance individually from +24V to GND and -24V to GND?  Again, does polarity matter?

When you say "would kill resistor", keep in mind the fuse blew.  The current would likely not have a chance to kill the resistor.

Here's some other ideas:

- You could turn up the current little by little until you can feel what's getting hot.

- If you're getting a drop of ~30mV, start measuring voltages across the resistors connected to the +/-24V supply.  Find out which resistor(s) account for that drop.  I would still bump the test current up a little, maybe 10 or 20mA, but keep the rails under 0.6V to prevent junctions from turning on and throwing you off the path.

- If the short resistance is low enough and you have enough sensitivity on your meter, you might be able to track the short by following the trace resistance to the short.  Although since these are power supply traces they are probably thick (low resistance) and this may be difficult to do accurately.

- Start measuring the PN junction voltages on transistors and diodes with low impedance paths to the +/-24V rails with your meter diode test.

While I wouldn't rule out a bad capacitor, I don't see any across the full +/-24V rails.  They are always between one of the rails and GND.  It doesn't seem likely you would get two capacitors failing at the same time to cause a rail-to-rail short.

A current tracer probe would be handy about now.

[free_electron]

dead output amplifier.
desolder the six last output transsitors.

[MarkL]

dead output amplifier.
desolder the six last output transsitors.

OP says +/-24V power to the last stage (all 6 transistors) was isolated by cutting the traces to L504/L505 with no change.

I would at least try checking all 6 with diode test before going shotgun.  Those, and going backwards into the driver stage starting with Q501/Q502.  Something could also be wrong with the differential drive if it's not a dead short on the rails.

[pelule]

First let me say - thank you all for your ideas and tips. Have started to do the further investigations...

What do you get when you measure the actual resistance of the short?  (I'm talking power off, with an ohmmeter.)  From your previous investigation, it would imply it's not a dead short.  Perhaps it could point you to a possible path through a low value resistor.  Does the polarity of your ohmmeter make a difference?

How about measuring the resistance individually from +24V to GND and -24V to GND?  Again, does polarity matter?

I measured resisitance with a DMM in both directions - DMM+ at high rail, DMM- to GND and the opposite polarity. Used 2k-Ohm range first (Diode testing) and 200 Ohm Range. Thats the result:
+24V rail: ~5.1 Ohms in both polarities/directions
-24V rail: ~4.9 Ohms in both polarities/directions
The values are starting at a higher value (~10 Ohm) but decreasing fast down to ~5.0 Ohms (like discarge of a capacitor, I assume thats the caps in the rails).

When you say "would kill resistor", keep in mind the fuse blew.  The current would likely not have a chance to kill the resistor.

Yes correct - also the supply circuit has a kind of current limit. I have tested using two current-Meters instead of the W15/16 bridges, results to ~500mA current, but none of the Low-Ohm resistors heated up (0.5A, 10 Ohm == 2.5W).

While I wouldn't rule out a bad capacitor, I don't see any across the full +/-24V rails.  They are always between one of the rails and GND.  It doesn't seem likely you would get two capacitors failing at the same time to cause a rail-to-rail short.

There are no Tantalum caps used. Only ceramic types and aluminium electrolytics. I checked meanwhile that caps, which are directly connected to the +/-24V rails. All are ok.

Here's some other ideas:
:

Next I will measure (in-circuit) the semiconductors of the output amplifier from the end backwards to the driver stage.
If issue not found, I will do the "heat-up" test using an external +/-24V tracking power supply to have better control on voltage and current limits.
Keep you posted

PeLuLe

[pelule]

Problem solved.
Thanks for the help to all. As in the past, MarkL had the best tips to search.
Problem was caused by two ceramic caps, creating a low impedance short.
  C19 (0.47uF) @ -24V rail
  C521 (0.47uF) @ +24V rail
As this is an unusual problem, I have noted, how I did the testing.
1. disconnect all boards (only base board A1)
2. as MarkL recommended, I connected a dual tracking power supply to the +/-24V rails (+24V/-24V, 0A..2A)
   set to +/-200mV, increased current limit in 100mA steps until supply went from CC mode to CV mode
   at each current step I checked the board for heated parts
   currents at CV mode +200mV/1.3A, -200mV/0.7A
   as voltage is below silicon conducting level (~0.45mV), it must caused by the short.
3. now I measured the voltage drops of the rails, following the PCB traces, starting at W15/W16, where the power supply is connected, following the traces in direction to the amplifier.
At that currents the impedance of the PCB traces generate voltage drops, voltage should decrease more and more, as more far from connection point, should not drop anymore behind the location of the short.
@ -24V rail at C19 down to ~180mV, behind C19 there is no voltage drop anymore.
   cut one leg of C19 -> current went down to 0.0A, voltage drop gone.
@ +24V rail drop increase as more far I measured from the connectiion, ~140mV at C521, behind C521 there is no voltage drop anymore.
   cut the leg of C521 -> current went down to 0.0A, voltage drop gone
4. replaced the defect fuses (W15/W16) with 0-Ohm resistors (a wire) and replaced C19/C521 both 0.47uF/50V ceramic caps.
5. measured all voltages (+24V/+15V/+5V...-24V) = all voltages are at specified levels.
6. connected all boards, checked the supply voltages again and did a complete function test.
HP 8116A operates perfect again.

[The Soulman]

Congrats and nice method, going to use that one day. 

[MarkL]

Great, glad you found the problem!

And not one, but TWO ceramic caps gone bad at the same time - highly unusual!

Interestingly, there was another eevblog user who had a bad ceramic a few months ago in an old Tek scope from the same era.  It measured like a resistor of a few hundred ohms.  And it was the same AVX axial SA series.  Hmmm....

Perhaps MLC manufacturing technology was not so great back then.  I'm definitely bumping those MLCs up on my suspects list from now on.

[pelule]

Hi MarkL

And not one, but TWO ceramic caps gone bad at the same time - highly unusual!

Not just two ceramics at the same time, also at different voltages rails.
Quite curious - I would have expected a crack, but not a low impedance short of less than 0.3 ohms.
I assume, both are from the same batch, which had a bad long term quality.
BR
PeLuLe