Lower Leakage Germanium Diodes.

[Kesh]

Obviously there's a lot of circuits that use diodes for clipping in feedback, and germaniums are known for killing gain in this use. I'm thinking this may be the reverse leakage current as much as the lower forward voltage clipping threshold. If a negative feedback has effectively a few hundred kohms paralleling it from the diode leakage, then this is going to kill some of the gain. And it's not clipping (good) it's plain old gain loss (bad)

Data sheets seem to quote leakage at 50V, which isn't much use. So I've been roughly testing a few at 3v and 20C. It's only three or four of each.

1n34a is about 6-8uA
1n270 about 1.5-2uA
0A90 about 3-4uA

All these figures nearly double when the diode is held between fingers.

So that old standby the 1n34a is effectively halving the unclipped gain in circuits like the Big Muff or Tube Screamer, even before you factor in the clipping. And worse if your pedal is warm.

Does anyone know if there are less leaky Ge diodes than the 1n270?

[R.G.]

That's a reasonable thing to worry about.

One complication is that leaky diodes don't act like resistors. They act resistive for only the very smallest reverse voltages, then they convert over to acting like constant current sources. That is, over X volts, where X is probably a bit different for every batch out of the oven, the current leakage flattens out to a constant reverse current no matter what the voltage across the diode does, until voltage stress causes reverse breakover.

I'll have to do some thinking to figure what this does to a clipping stage.

[Kesh]

germanium diode leakage isn't very flat, at least according to this graph, and looks somewhat ohmic at the voltages we're talking about. my testing seems to confirm this.



though i strongly dispute that 2uA flattening off line.

the main effect on the sound in my TS is gating, quite conspicuous, and less gain compared to a schottky of similar Vf (and quite a bit less leakage, even for a schottky)

[R.G.]

As I said, it's something to consider.

I have run both silicon and germaniums in a curve tracer to produce much the same result as that chart. It's pretty much what I'd have drawn if I had to make up one freehand for illustration.

As will transistors, the germaniums have a wide range of leakages and curve shapes, from type number to type number and from unit to unit. The 1-2uA number on the chart is representative from what I found. Note the change of voltage and current scales in the third quadrant.

I ran into this two times. One was back in my days of designing power supplies, and the other was in 1999 when I was looking for a "constant current source" to run the bias on the Millenium Bypass. The germaniums I had tended to run about 1uA when tested, although I didn't have a wide range of types to test. That was too much for the sensitivity I wanted for the Millenium.

Out of curiousity, have you computed an equivalent junction resistance in forward and reverse conduction? Actual plots of forward and reverse voltage and current from -0.5V to +0.5V would be interesting. I'd go out and plot them on the curve tracer - except my curve tracer plugin for the Tektronix died and I haven't had the need to fix it until now.

[PRR]

> testing a few at 3v

Well, but if one side is clipped at 0.2V, then the other side is not likely to feel over 0.2V-0.4V, right? We have good reason to doubt "perfect resistance", so a 3V reading may not be much clue to part-volt action.

What you want for the Theory is a Curve Tracer. And for a Diode, this need not be R.G.'s fancy Tek. But it really is better to have an oscilloscope. It can be pretty crude. Got one?

And for your premise that leakage sucks gain, all you need is signal generator and audio meter.

> plain old gain loss (bad)

The essence of post-1938 electronics is that you can always buy more gain. Since the 1970s, gain is really cheap.

I'm thinking there is a knob for Gain, and you just over-boost to get your desired gain with the specific diodes installed.

[Kesh]

I'm making my own personal fuzz lab, and I'm playing with asymmetrical clipping, so the germanium may clip at ~0.3V and the other side much higher or not at all. As my rail-to-rail op-amp can go to +/-4.5 I figured -3V was a good ballpark to start with, but I'm now looking at other voltages, though it's getting close to the limits of my DMM. And no, I have no scope.

I'm also trying to get individual control over each side of the clipping, and switching in a germanium means that is lost to some extent.

According to the graph, if it's to be believed, things are fairly ohmic at 0 to -4.5V (straightish line through the origin) though the model in LTSpice has a 1n34a acting like a resistor only up to -0.03V. We'll see.