Metal can TO3 failure mode

[BurningTantalum]

I have a heatsink with 6 TO3 transistors of unknown spec. They are connected in parallel with 0R56 emitter resistors (For full story see my post on 'Olympic-sized ultrasonic bath...) and drive 12 parallel connected ultrasonic transducers.
The bath and circuit board have a fair bit of corrosion, but nothing too fatal, possibly caused by it being stored outside on the coast or due to the aggressive cleaning fluids used (not known what)
I fired up the bath for the first time today and it ran for 20 mins, nothing getting too hot and the bath working well enough to turn some baking foil into silver snow. Then the main fuse ruptured. One TO3 had gone short, so I replaced it with a spare. The same happened to a transistor in a different position after about 15 mins. The two failed components were the worst corroded of all the 6 plus the 5 spares I have, with patches of rust on the bases and the 'domes', see pic attached.
I cannot see any electronic reason for the failure, but the transistors both failed 'short circuit all ways round'.
Does anyone have any experience/views on metal can TO3 reliability when corroded or operated for a long period in corrosive conditions (30+ years)?
Does the seal integrity fail?
I have still found no spec on the 2H0U8 transistor, but it must be similar to horizontal output CRT type ie High voltage 1500 etc
Thanks,  BT

[Seekonk]

No, but I have thought about purity issues of the die causing failure over time.  I recently sold 37 NEW germanium transistors from the early 70's and the buyer tested all of them and reported five of them were bad.  They had all been in my possession that entire time and never used. 

[amyk]

Open them up and see how the die looks.

[T3sl4co1l]

Creeping rust around the base seal leading to catastrophic shorting?

Tim

[flynwill]

I agree, cut them open and see what happened inside.  If you find corrosion then you might want to just replace all 6 with modern parts.

[Anks]

if there running in parallel then being unmatched could be a issue but the emitter resistor is there to balance that out to some degree.

[mikerj]

It could easily be a hermeticity problem if the pins have rotted up to the glass seals.  Has rust crept under the bottom of the transistor?  If so this would likely have increased the thermal impedance to the heatsink.

Possibly these were remarked with a custom code for the manufacturer which is why there's no trace of that number anywhere.  A BU208 shares many of the characters in the part number and might even be suitable

[TimFox]

Are the metal base flanges (collector) connected directly to the heat sink or through an insulating washer?
Direct case:  is there corrosion between the transistor and the heat sink to increase the thermal resistance, overheating the transistor?
Insulated case:  has anything happened to mess up the contact pressure from flange through grease to washer through grease to heat sink?
In either case, the grease is needed to cope with (hopefully minute) gaps between two nominally flat surfaces. 

[BurningTantalum]

Many thanks for the suggestions and info.

Close inspection of one of the failed components shows rust around the edge of the 'dome' where it meets the base metal. There is rust under the base but only where it extends a millimetre over the washer. I will chop the top off one later and see if anything is obvious although I do not have access to a microscope.
Incidentally - is this type of package evacuated or filled with an inert gas? I suppose that seeping of non-dry air could cause failure over a period - I am aware of failure modes in plastic packages, and resistors 'ageing' out of spec.

The devices are mounted in sockets, which is handy for swapping them plus removal of one screw isolates the case/collector from the rest for easy ident of the dead one. The socket is one side of the alloy plate and the 3 mm thick plastic spacer is on the other side so a long way from ideal for thermal transfer.  The devices appears to have been mounted with some type of clear silicone grease - I have only ever used the thick white type so was a bit suspicious and intended cleaning it all off and reapplying the correct type.
Mikerj -  I hadn't noticed the part number similarity ! How sneaky of them; you must be a crossword buff.
Thanks all,  BT

[BurningTantalum]

I lopped the top off one of the dead devices - the damage on the chip was evident; it was burned on one edge, once I had peeled off a blob of silicone rubber (which didn't seem to be bonded very well at all.)
The thick insulating spacers are a ceramic material that I have not seen before - I cut my teeth on mica spacers !
I removed all of the transistors, cleaned up the plate and spacers etc then reassembled with lots of white thermal grease. The tank ran for over an hour with no problems.
The heatsink is not a brilliant design, and the holes under the device for b, e, and screws take up about 25% of the area of the device contact patch.
 I recall when I was employed making test rigs - I was told to "give each device the same access to the fins"... Ah, the 2N3055...
On some of the devices it was evident where the die was inside the package by a visible 'spot' on the case and a corresponding 'spot' on the spacer. This must have been due to a hot spot.
Thanks everyone once again.
BT

[mikerj]

I cut my teeth on mica spacers !

They aren't good for eating to be fair

[SeanB]

Those insulators are aluminia and are actually better than the thin mica washer, especially at high voltage. The transistors that failed probably were loose on the heatsink, or had corrosion on the pins in the sockets ( not surprising in something 30 years old and exposed to an ionic solution), so the cleaning up and new heatsink paste, along with the correct mounting torque ( it is actually critical for long life, it needs to be correct, not just tighten till it creaks) with the right screws and hardware will help the devices to stay cool.