Fused Drains

[jjoonathan]

Ok, this is the third time in as many months that I've taken apart an old HP module and tracked a problem to a dead amplifier with a fused-open drain bondwire. Amp 1 was a narrowband VCO buffer in an 8360 low-band module, and Amp 2 and Amp 3 were TC702 distributed amplifiers in doubler modules.

In all three cases the amplifiers sort of came back to life after rebonding the drains and readjusting the bias, but in two of the cases the bias to obtain the recommended current was outside the recommended voltage range and in one case most of the amplifier's high frequency gain never returned. I'm trying to figure out what happened to cause the problem in the first place.

1. Input signal too high. Seems unlikely. How often do sealed VCOs have amplitude surges? How high would the mod/amp stage need to surge the signal to kill an amplifier on the other side of a YIG filter, PIN switch, and a bunch of path loss?

2. If you fix gate bias via trimpot, drain current will slowly "cook up" over the years as the amplifier ages, eventually killing it.

3. Drain voltage surge? Seems unlikely -- all three cases would have been protected by crowbar OVP.

4. Something else.

If #2 is a thing I should probably make a project of re-biasing the old amplifiers in my test gear, but it would be a big production so I thought I'd ask around first. Does anyone know what's up?

[Dulus]

Not necessarily have a solution, but I'm definitely intrigued by the subject.

Always lovely to see a microwave assembly, no matter what it is.

1. While not knowing the amplifier part numbers, these dont seem like higher than couple of Watt output amps. And those can usually be driven up to Psat. So even an input surge high enough would drive the amp to Psat, and drain current would stop increasing there (if it increases... it may go down too).
And an even higher RF input surge would damage the amp itself i assume. You've mentioned that the gain came back. So its unlikely.
Considering the preceding stage losses, i think option.1 is most unlikely.

2. If i had these amps, i would drive them with an active bias. Thinking of the whole system, there would be bias sequencing too.
Can you trace the module bias towards the power source? And you can measure it also.

3. Voltage surge also goes to the option.2. Active bias circuit & sequencing.
And the internal voltages would be monitored i believe.
This is possible, but a spike so long that can burn a gold bondwire? Other parts on the system would have been damaged too.

4. "... old HP module ..." is the keyword i think.
Kinked bondwire at production.... Poor bond interface...

Years of vibration & shock & thermal cycle could fatigue the wire and increase resistance.
And the active bias would push the same current from that node no matter what's the resistance of bondwire became.
We can guess the rest...

Do you have the before pictures of fused wires? Or the attached are the before photos and i cant see the splits 
Where did it split?

Would love to hear more.

Regards,
Deniz

[jjoonathan]

Yeah, these are beautiful! They've inspired me to start learning macro photography. I have one of those Canon EF-M macro lenses with the ring light, I've started to grok the exposure triangle, and it feels really great to have understanding and control vs just trusting phone magic. I still need to figure out how to get decent images out of my microscope, but that's another story.

I didn't think I had a "before" pic, but it turns out I do! See attached. The bond wire fused in the middle, and this was the case for all three dead amplifiers. My guess is that fusing in the middle means a thermal mechanism, whereas mechanical excitement would tend to effect the ends, but that's all just guesswork.

HP didn't do active biasing in any of these designs, but they did do sequencing. Perhaps they should have done active current control. The gate bias is fixed voltage, adjusted manually with those little blue trimpots, one per amp. Sequencing is accomplished by watching the -10 gate bias supply with an op amp. The +8V drain supply, which provides power and cooks the amp if applied without the gate bias, is switched on with a FET and only gets switched on when the op-amp sees the negative rail. This is the case in all three modules. On the 8360 I have access to, I've scoped the drains and the mechanism appears to work. The +8V is never up when -10 isn't.

Err, is it a Doherty pair if the splitter is a Wilkinson and not a hybrid? Eh, I'll just (ab)use the term.

Dead Amp A:
* Narrowband 5.4GHz VCO buffer from low band module
* Found with fused bond wire
* Re-bonded
* Came back to life, hasn't keeled over yet.

Dead Amp B:
* TC702 traveling wave amplifier (see attached doc) from 50GHz doubler
* Found with fused drain wire
* One half of a Doherty pair (other amp wasn't fused)
* Re-bonded
* Zombie: degraded frequnecy response, gate bias needed big adjustment, outside of typical voltage range

Dead Amp C:
* TC724 traveling wave amplifier
* Found with fused drain wire
* One half of a Doherty pair (other amp wasn't fused)
* Re-bonded
* Came back to life, bias needed small adjustment, inside typical range


Notes here: https://www.dropbox.com/s/2682992emftu6rm/8360%20Doubler.pdf?dl=0

TC702 Amp: https://www.dropbox.com/s/f6ix9seghzt1k9l/TC700%20TC702%20TC900%20TWA%20Amplifier.pdf?dl=0

TC724 Amp: https://www.dropbox.com/s/9kt70tm2vlaw27f/TC724%20TWA%20Amplifier.pdf?dl=0

[jjoonathan]

Oh, context: obviously since doublers are expensive I want to get this one working. To that end, I bought a sacrificial donor module with 4x TC702 amplifiers. I bought it in uncertain condition, hoping that at least one would work, but all four work, so I'm getting cold feet about the operation, you know how it goes    I'll get over it, but in the mean time I am procrastinating by trying to figure out the failure mechanism, which I still don't have a convincing theory for. So far, the one I like the most is "over years, drain current slowly cooks up at constant gate bias" -- but I don't like it very much at all. How does it explain two dead amplifiers in series, but only one in parallel?

Maybe I'm overthinking this. Maybe a single amplifier died and a buffoon killed the other while blindly turning trimpots, and then sold his mess to me through fleabay. 

In any case, it'll be my first time wedge bonding to a chip. I've only ever ball bonded chips, but I don't have a ball bonder. I need to practice first. Advice would be welcome.

[coppercone2]

it could just be poorly manufactured parts. You never know if the assembly lines got screwed up one day. Maybe the test machine for the modules got messed up and it was damaging stuff. That's possible. Maybe someone borrowed some stuff from production at 4pm to prototype something in engineering for some reason and it appeared back on the shelf at 11pm after it got stressed accidentally because they forgot a attenuator or pressed the wrong button.