Germanium transistor, connections reversed?

[agardiner]

So this is a new mystery to me. At first, I thought I must have got it wrong and shelved the unit(s).

I have now come across a few transistor sets where a germanium transistor has 'reversed' its collector/emitter connections internally. Most of the faulty transistors have been AC126, although one was a TO3 cased AL100.

When testing on a peak transistor analyzer the problem was confirmed, along with the fact that the transistors had gone almost unity gain. I tried out of interest reconnecting the device with the 'new' connections, and the sets did work although obviously low gain and a lot of hiss. Replacing the defective device did of course fix the problem.

So, has anyone else come across this phenomenon? We are all aware of the AF117 and the lockfit issues; now it seems we might have an AC126 problem! I presume something must actually happen to the germanium crystal structure.

[paulsherwin]

AC126s do develop internal shorts sometimes, and component analysers aren't very good at dealing with Ge devices. I think this transistor is just faulty.

[Maarten]

Germanium transistors are a bit more symmetrical by nature, I think.

Three possibilities as I see it:
- worn/defective: transistor getting more symmetrical due to deterioration
- component tester not really good at symmetrical devices
- whiskers

[Craig Sawyers]

Germanium transistors were always a difficult thing, even back in the day. I recall when I was still at school (early 70's) trying to get to grips with a detailed book on transistors. Understanding h parameters etc. But it was written when Ge transistors were the only show in town, and there was a raft of stuff on bias point stability given the fact that leakage current was a strong function of temperature.

But many of them, over the decades, seem to have degraded. Not just the whisker shorts in things like like the AC126 and others, but power transistors that have seen no, or light use, losing beta and exhibiting high leakage.

At least silicon transistors *usually* either work perfectly, or are totally dead.

Craig

[Radio Wrangler]

The early manufacturing processes coincided with the germanium era and were prone to ageing effects like dopant migration. Transistors were a lot more symmetrical in those days. Silicon, the planar process and much better control of doping all came along about the same time. No wonder the little testers can't tell E from C.

If replacing AF117s, I take the silicon route where supply voltages permit as it's likely to be more stable and doesn't use up a limited resource.

David

[barrymagrec]

Some Germanium transistors such as the OC139 series (admittedly NPN) were specifically specified as symmetrical.

[TonyDuell]

I think I read somewhere that at one time when a germanium transistor was made, the gain was measured 'both ways round' (swapping the emitter and collector connections) and the thing was packaged/labelled the way that gave the most gain (!). It is perhaps not too surprising that dopant migration, etc can affect the junctons sufficiently that the gain is greater with the emitter and collector the 'wrong way round'

I am pretty sure that these little component testers work out which is emitter and collector by a similar method. They find the base lead by doing junction tests, then measure the current gain with the emitter and collector assigned to the other 2 leads both ways round. The way that gives the most gain is the one it displays.

I don't have such a tester but I've used a similar method with my cheap DMM to do it by hand. Use the diode test range to find the base and the polarity (NPN or PNP) and then use the transistor test range to display the current gain with the base in the right socket and the other 2 transistor leads both ways round in the other 2 sockets. The way that gives the higher reading is the right one. At least if the transistor is good...

[Craig Sawyers]

In Spice models, beta(f) and beta(r) are included so that "inverse active" mode can be set up, where emitter and collector are swapped. Not surprisingly beta(f)>beta(r), usually >>

Craig

[G6Tanuki]

Some Silicon transistors were also 'symmetrical' - things like the BCY3x series. They were generally rather low-gain [Maximum Beta of 30, typically 18-25 according to Texas Instruments!] and were often used in pairs to simulate a SCR as relay-drivers in telephone exchanges.

I've got a bunch of these underachievers here - they're 'good enough' to work in a simple 2-transistor LED-flasher-multivibrator but I wouldn't use them for anything serious.

I've also heard it said that JFETs like the 2N3819 are reversible and that sometimes lower internal-noise-generation is found when the source and drain are interchanged.

And yes I've had AF-series TO-1 transistors with whiskers - the AC127/128, AC132, AC153 and AC187/188 audio-types sometimes have 'issues' when fitted to earthed heatsinks because of this.

[JonSnell]

When I was a test Engineer at Weyrad Electronics back in the late 60s, here in Weymouth, we started using transistors instead of valves. I recall the main one was an XB113. It was a pnp type Germanium, but if put in a multivibrator circuit upside down, it still worked and didn't seem to mind. We found the gain to be about 15 the correct polarity and around 8 (depending on temperature) when reversed the polarity. We are talking an unusual to us 10volt supply rail. We were used to 250volts on ECC81s acting as flip flops in the counter and my Model 7 was having trouble reading the collector voltages for some reason.
I still prefer valves.

[ortek_service]

Originally Posted by G6Tanuki

>>
>>
I've also heard it said that JFETs like the 2N3819 are reversible and that sometimes lower internal-noise-generation is found when the source and drain are interchanged.
>>
>>

Yes, JFET's are completely-symmetrical, with the gate in the middle on the symbol and so can be used to make 'transmission-gate' analogue-switches etc.
. Whereas MOSFET's aren't, with they 'body-diode' and gate being at source-end (Although sometimes a P-Ch etc. one is deliberately used with drain & source swapped over, to provide a low-drop reverse-polarity protection switch that the body diode won't bypass)

[Andrew2]

Ah, this is interesting. Back in my schooldays, Ge transistors were still current and we were messing about with OC71's and OC81's (and the wonderful ASZ21 as well, but that's another story). I had an 81 that was working perfectly normally but suddenly went a bit low gain and noisy. For some reason I turned it round the wrong way and tried it that way instead and found it worked both ways. My mates wouldn't believe me but there it was....

[Keith956]

Old germanium transistors like the OC71 were made from a slice of N type germanium, on each side a small pellet of indium was placed then heated so it diffused slightly, forming P type doped junctions. The collector pellet was larger than the emitter pellet but in many early devices that was the only difference - so the reverse gain could approach the forward gain. Some devices were sufficiently symmetrical to be used with collector and emitter swapped.

With silicon planar transistors the doping profiles could be tweaked to be asymmetric and gain in the forward direction became much higher than in the reverse direction.

[Craig Sawyers]

A currently manufactured 2N5551 has a forward beta of 608 and a reverse beta of 5 (according to the Spice model).

But you're right - the diffused junction Ge transistors had much greater symmetry.

One of the first things I built was an RF signal generator (out of PW) that used Ge transistors (NKT something or others). I blew it up after I bought a Hartley 13A oscilloscope, and tried out using the cathode follower probe. That is how I found the the input coupling capacitor on the probe was leaky. Lesson learnt.

Craig

Craig

[kalee20]

Of course the silicon planar process is highly asymmetric, at least as used for commodity transistors anyway... Start with n-type silicon (which is the collector); diffuse acceptor dopant into it to turn a region into p-type material (which becomes the base) and then diffuse donor dopant into a smaller region of that, turning it back into n-type (which becomes the emitter).

It's worth noting that many germanium alloy transistors have far better reverse blocking of the emitter-base junction than silicon planar (Vbe reverse is often limited to 5V and in practice doesn't greatly exceed this). Very useful if you like operating your transistors backwards, but also important if you are making a classic Eccles-Jordan multivibrator operating from more than about 6V supply.

[cmjones01]

It doesn't surprise me that the characteristics of a germanium transistor can change a lot over time. I have here this 2N1183 which was regulating the current to the crystal oven heater in my HP counter. After the best part of half a century of running warm, it's very leaky. So leaky, in fact, that the oven overheated and destroyed its heating element, the wiring in the oven, and the transformer that fed it all. I repaired the damage and replaced the transistor with a silicon device (no other circuit changes needed) and the crystal oven works fine.

The old 2N1183's characteristics are very odd. I wouldn't be surprised if it had more gain when connected "backwards". I replaced it Apparently it's possible to "repair" these transistors by operating them the other way round for a long time and the leakage should reduce. That's why I haven't thrown it away yet, since germanium power transistors, even though they're not very good, are a dying breed.

Chris

[m0cemdave]

I have a Ge PNP transistor here labelled OC44M, in the usual black glass enclosure.
It has two (yes two) red dots to indicate what is usually the collector teminal, but no manufacturer's code.
On my curve tracer, it tests as perfectly symmetrical.

[emeritus]

I recall one of the "In the workshop" articles cica 1980, where Smithy explains to his mate how to make a bistable oscillator of very low power consumption and operating from a low voltage suply, by using a pair of cross-coupled inverse-operated transistors. I don't recall any other details or whether they were Si or Ge types.

[Maarten]

Semi off-topic, but Philips actually used low power silicium transistors (think BC337) as mute transistors in CD players. Apparently they're slightly more symmetrical than small signal transistors.

[Craig Sawyers]

Slight on and off topic simultaneously...

I few months ago, I bought a(nother) Gould JB3 audio signal generator. I measured the distortion, and found it to be twice spec, which is unusual for this workhorse. I have another that exceeds spec.

Off came the lids. The power rail was very high, caused by the series pass transistor in the discrete regulator being hard on. There was evidence that it had been replaced already, so I pulled it and tested it - it was just fine.

Took a while to find that the driver (it was a Darlington with the series pass) was a BC327 (PNP) and it should have been a BC337 (NPN). So in repair, someone had replaced the transistor with the wrong polarity, so it was working completely backwards. At which point they just put the covers on. And stuck a CAL label over the lid screws. You have to admire a repair and calibration technician who pulls that stunt.

Once replaced by a vanilla NPN from the spares boxes, all was sweetness and light again.

Craig