Stabilizing a germanium transistor with a germanium diode -- matching them?

[mordechai]

Hi -- quite a while back, I double checked on the "trick" of using a (reverse biased) germanium diode to offset the leakage on a germanium transistor.

RG chimed in and noted that for it to work in an ideal way, the leakage from the transistor had to be matched to the leakage from the diode.   

I have a few 1N34a diodes (from Tayda) and two germanium trannies, each leaking at about 270ma...how do I determine if the diode, reverse biased, would be properly matched to compensate for the transistors' leakage?

[Cozybuilder]

Why not use the second tranny as your diode?

[mordechai]

I would if this was a single transistor circuit, but I'm making (....wait for it...) a Fuzz Face, so I need 'em both.  Hoping that the 1N34a's will be a suitable match for stabilizing them.

[R.G.]

The collector-base junction of a bipolar is a reverse-biased diode junction. It leaks, as any reverse biased junction does. In bipolars, this acts just like current inserted into the base region from the base lead, and is amplified by the current gain of the transistor. Germanium has problems with leakage because it inherently leaks about 1000 times as much as silicon.

A transistor measured on most leakage testers, mine included, just bias the collector to emitter, leaving the base open. So 100% of the current leaking in the collector-to-base junction is amplified into collector and emitter current. Testers measure this current in most cases, not the current in the collector-base junction that starts this.

If one could attach a suction cup to the base lead and suck out exactly as many charge carriers from the base region as the collector leakage is putting it, the remaining near-zero (it's never going to be perfect!) current in the base would result in a much smaller overall leakage.

That's where the external diode comes in. It can theoretically be set up to leak as much current out of the base as the collector leaks in. That's why I mentioned "matching". But the diode leakage has to be matched to the collector-base leakage, not the overall collector current.

So you have a measurement problem. You're not measuring the collector current with the base open, you're measuring the collector current with the EMITTER open. That's what you're trying to match with a diode. Or another transistor connected as a diode. You want the corrective leaking device to be as close to identical to the transistor it's compensating as possible, in both the electrical and thermal/mechanical sense. Ideally, it would be another transistor junction on the same basic die; but they didn't make germanium multiple transistors that I know of, and if they did you couldn't get any of them.

A 1N34 is the same material as the transistor, but that's about it. You'd need to match for same current when reverse biased, or just Easter-egg in diodes until the collector current on the transistor was minimized.

It's a tricky measurement problem. You might do as well with a computing-type solution where one germanium junction was set up as a "typical leaker" for the whole circuit and a opamp arrangement and silicon transistors the sucked at nominal corrective current out of all the germaniums. that would at least offer you the ability to tune the correction.

[mordechai]

So, it seems like that 1N34A diode fix could **help**, but not completely solve the problem of leakage.  If I have that right, then the reverse biased diode might offer suitable stabilizing assistance for most practical applications, but the nature of the beast (germanium) demands a far more complex array of circuitry to fully compensate and nullify leakage.  Correct? 

[DDD]

Ge p-n junction leakage gets 2 times higher with every 10 Centigrades up.
Also, one should take into account the thermal base-to-emitter voltage coefficient that is about -2.5mV/C.
So, it's virually impossible to compensate for the thermal drifts with simple solutions like Ge or Si diode etc.
But, there's a lot of very simple circuits using OpAmp or couple trannies for such a purpose.

[mordechai]

Can you point me to a thread out here that discusses some of these methods?  I did a search but I don't know specifics, so my search parameters were too wide and I can't spot the thread or threads dealing with this...

[Gus]

Used the wayback machine
[url]https://web.archive.org/web/20090122112519/http://www.geocities.com/SunsetStrip/Studio/2987/britface.html#tempcomp[/url