HP 6920B adventures part 2 – why recapping is not repairing. Lets start with “spot the differences” game:
Why did I recap this device even though I said I won’t? Two reasons: I noticed something odd on capacitor C6, and I had some issues with DC output wandering a bit too much and I suspected another degraded cap. Curiously C6 tested OK on all parameters (capacity, ESR and leakage) but it looks suspicious to me:
That almost seems like it started to vent at some point, possibly because it sits right next to two 7W power resistors that get over 100C hot.
Did the recap help? Nope, actually things got worse afterwards. But before I venture into that – a small digression. I’ve seen some photos of recapped PCBs, including one 6920B, that had capacitors looking like that replaced with modern electrolytics:
Please don’t do that. These are not aluminum electrolytics, these are high stability hermetically sealed tantalum capacitors with temperature range up to 125C. Military and aerospace use those, you really can’t find anything much better – look up KEMET T140 series. Doesn’t matter if old or not, these don’t really age. Do not change them unless you suspect an issue, and even then you should replace them with good quality tantalums as well – that’s what I did. In fact none of the ones I’ve replaced measured out of spec and these are ’76 caps.
So what was the issue? Well, I noticed the DC output tends to wander up/down and it’s not just down to ripple. After the recap, once the device got properly warm (after 2-3 hours) it got so bad the voltage started to go up well outside the spec and on 1V range I was unable to even turn the output on, without significant load being present, and not hit overvoltage condition right away.
Look at the current sensing shunts in the inverter when output is switched on:
See the ringing? There’s clearly some sort of resonance around 500Hz, and then also continuous oscillations at about 3.3kHz. But when cold these tend to die down on their own – this is output-on steady state:
Watch what happens when temperature starts to rise:
Yup, it gets worse, but still just manages to die down. And when you reach high enough temperature inside the device:
On 1V range these oscillations never die down and pretty much take over as the primary operating mode – over 1A of load is needed for the device to have any chance of regulating the output.
After some more head scratching and poking around I concluded the issue must be with the feedback loop and voltage regulation in SERIES REGULATOR module (see schematic in previous post). HP eventually replaced Q10 transistor (still germanium on my 6920B) with a different one and recommended adding a ferrite bead to emitter lead. Unfortunately they only gave HP replacement part number and I wasn’t very keen on replacing Q10 anyway.
Eventually I came up with another mod, a bypass capacitor of exactly 4.7uF (for best result, 1uF or 10uF give worse performance) for VR2 keeps collector voltage on Q11 clean enough to prevent Q10 from oscillating on it’s own. And boy, does that make a difference for both start-up and steady state inverter currents:
No more DC output wandering, hot or not. No need for output load to keep things stable. 1V range starts and stays clean every time. I’m pretty proud of that one. So here’s both mods I’ve made to my 6920B, first one prevents DC stage in AC REFERENCE from oscillating, the second cures the ringing in INVERTER when switching phases:
Ignore the weird spots, must be the side effect of spraying 3M cleaner nearby. You can’t even see that, it’s only visible using the camera flash. You’ll notice I reused the tantalum capacitor I removed from the PCB, after all there’s nothing wrong with it. Why is it needed at all here? Well, HP did add another cap nearby too, in change #19, but that didn’t work for me. Was this the issue they were trying to fix, or was it something else? Is this inverter ringing something common in 6920B devices – perhaps so, I’ve seen (current) owners mention this DC instability more than once. Was it there from the start or does it happen with age, which is also likely considering all the germanium transistor used and these like to get noisy. Who knows, it’s fixed now, hopefully for good.
Also, the output indicator bulb got replaced and I’ve also changed the power cord socket, it was broken on the top and then also cracked on the side. That actually wasn’t easy at all, the original socket is riveted to the case.
And one last photo:
This is 1V DC output as measured by HP 3465A – which I might describe a bit more in another post someday. The 3 at the end is well within spec (+/- 10) and I could dial it to zero but this way I get slightly lower offsets in 10V and 100V ranges, so it’s a good spot.