Lowering a transistor's gain

[Paul Perry (Frostwave)]

I can't remember if this has been mentioned here before, but I suspect that putting a diode in parallel with the base/emitter of a transistor would lower the gain. As to what exactly would happen in particular cases (eg a 2N3904 with a Ge diode), I have no idea, but I'm optimistic. (in fact, using a 1N34A diode, gain goes from 125 to 74.)

[aloupos]

Interesting.  Do you think the same could be done to a jfet?

[petemoore]

I can't remember if this has been mentioned here before,
  I'd be willing to bet a RP temp stabilization diode got put in the wrong way and somebody wondered what that'd do...
  but I suspect that putting a diode in parallel with the base/emitter of a transistor would lower the gain. As to what exactly would happen in particular cases (eg a 2N3904 with a Ge diode), I have no idea, but I'm optimistic.
  It would almost necessarily have to do 'something, yours is more than a good suspicion.
  (in fact, using a 1N34A diode, gain goes from 125 to 74.)
  It appears you've already done some testing !, I can think of another way to test those numbers...any FF'n suspicion what it is?

[amz-fx]

I can't remember if this has been mentioned here before, but I suspect that putting a diode in parallel with the base/emitter of a transistor would lower the gain. As to what exactly would happen in particular cases (eg a 2N3904 with a Ge diode), I have no idea, but I'm optimistic. (in fact, using a 1N34A diode, gain goes from 125 to 74.)

At a high enough input voltage, the Ge diode will begin to conduct and produce asymmetrical clipping.  You may or may not like the sound, which will have some fuzzy octave-up character.

regards, Jack

[Paul Perry (Frostwave)]

I suspect the base/emitter junction is already pretty asymmetric!
My hope is, (handwaving, rather than actual physical analysis) is that some of the current that would otherwise be injected to the base, would be drained off by the external diode & thus the gain drop.
As to the utility, the proof will be in the pudding, so I hope someone will try out some "modded" transistors in their favorite circuits & report!

[amz-fx]

I suspect the base/emitter junction is already pretty asymmetric!

But usually doesn't come into play....   the lower threshold Ge will give that more opportunity to happen but I suspect it will be more distortion than gain reduction.

I'll do some tests if I get the chance but today is kinda busy...

regards, Jack

[Meanderthal]

 I can't remember who it was exactly, but I read an article in one of the popular DIY sites about FF tech, and they mention this method to halve the gain of Q1, as well as running a Ge diode in the other direction for temperature stability.

[R.G.]

Congratulations! You've rediscovered piggybacking.

This is exactly the chain of reasoning that led me to think up piggybacking transistors for lower gain a few years back.

The idea is sound, but the actual implementation is complicated.

That's because the V-I characteristic of the BE junction is exponential. Tiny, tiny changes in forward voltage near the knee have huge effects in how much current flows into the BE junction. If you put a junction in parallel with a BE junction that has a lot lower knee, the effective gain drops to nearly 0 because the paralleled junction - which is also exponential - is stealing nearly all of the available current. If the paralleled junction is much higher, it never gets any of the available current, so it has no significant effect. The just-right value is forced to be nearly identical to the base-emitter you are trying to lower the gain on.

I mentally went through diodes for a few minutes and then realized that you need a junction that is nearly identical to the the junction you're trying to steal from, and also you need a similar current. So the simplest answer that avoids trying to hand-match different junctions is - another transistor just like the one you're trying to reduce the gain on.

My speculation was that this would reduce gain by something like half.

Wrong-O! It reduces it by huge amounts, much more than half because the internal Shockley resistance in an active transistor makes the BE voltage bigger than a simple junction with no transistor action. Brett did a bunch of experimentation and the easy way to dial in a gain reduction is by inserting a resistor between the emitters of a piggy backed transistor set. The active transistor can have its apparent gain dialed from almost unaffected down to about 10-15 by changing that resistor.

You can pick up a lot more background by searching for "piggyback" in the forums here.

[amz-fx]

I can't remember who it was exactly, but I read an article in one of the popular DIY sites about FF tech, and they mention this method to halve the gain of Q1, as well as running a Ge diode in the other direction for temperature stability.

That was for a Ge diode with a Ge transistor...  I read the question in the first post as a silicon transistor (2N3904) with a Ge diode (1N34a), which will behave much differently. RG's response is good but deals with silicon-on-silicon techniques.

The lower voltage threshold of the Ge diode across the silicon b-e junction will behave much differently because of the radical differences in voltages and because of the gradual turn-on of Ge didoes... 

regards, Jack

[R.G.]

Actually, R.G.'s response deals with all transistors.

Both silicon and germanium have base emitter currents which are an exponential function of the voltage across the junction. This is of the form I = k*e^(m*Vbe) where k and m are constants which are characteristics of the material. As you're aware, semiconductor junction conduction begins at some small voltage and increases quite slowly until m*Vbe hits 1.0; after that, it heads for the sky. That sudden change in conduction is why we say that a junction turns on - it is so abrupt compared to many other things that it's almost instantaneous.

As I said, if your sucker-diode that is pulling base current out is not very, very closely matched to the junction you're trying to abuse, nothing happens in one of two ways. If the sucker-diode is a much smaller voltage than the base-emitter, it eats substantially all the current and the transistor looks like it's gain is almost if not truly nothing.  If the sucker diode has a voltage that is much above the base-emitter, the base-emitter gets it all and the sucker diode might as well not be there. It's only when the sucker-diode is very close to the junction voltage of the transistor Vbe that the sucker-diode can bleed off some of the current that would otherwise go into the base, making the transistor look like it has a lower current gain than it really does.

You can do this any way you like, by using a similar Vbe, or by stacking germaniums or by stacking unicorns, as long as whatever you put in parallel with the Vbe under question gets close enough to the voltage across the base-emitter at the operating current. If it's too far away, the added device(s) either eat too much or not enough of the prospective base current. You could even use resistors - except that resistors, being almost linear, will only eat the base current at one particular voltage. You also need the tracking that another exponential-law junction provides; the sensitivity of an exponential function guarantees that this will be touchy.

Using devices with non-matched exponentials might be the source of some interesting new distortions if you could use devices that came into conduction more or less abruptly than the Vbe being abused. The distortion would be as you can imagine suddenly adding/subtracting and exponential function current to/from the base would be.

But for modifying gain, you need not only good matching, but tracking.

It's really touchy.

The foregoing reasoning applies not to silicon on silicon but Ge on Si, Schottky on Ge, Schotty on Si, GaAsP on InN, C on Ge, and so on.

[Gus]

Remember this site?  It has been up for YEARS

http://www.geocities.com/sunsetstrip/studio/2987/

[Paul Perry (Frostwave)]

Thanks Gus! I knew I'd seen it SOMEWHERE....
and R.G.'s "nothing happens in one of two ways" is going to enter the language around here
I'm finding as I get older that I'm forgetting about as much as I'm learning. A very good argument for a proper filing system - outside my head!
As for the diode across the junction idea in general, it does seem a useful reminder that very simple mods can have massive musical applications. Which is very healthy for DIY.

[amz-fx]

Actually, R.G.'s response deals with all transistors.

From a practical standpoint, because of the wide variations in gain, leakage and doping characteristics, getting matching Ge transistors would be quite difficult unless one had dozens of samples to do testing on. It can be done but not nearly as practical as with the more uniform silicon devices. 

regards, Jack

[R.G.]

From a practical standpoint, because of the wide variations in gain, leakage and doping characteristics, getting matching Ge transistors would be quite difficult unless one had dozens of samples to do testing on. It can be done but not nearly as practical as with the more uniform silicon devices.

Yep, I believe that's yet one more way of saying what I said again, all right.

[Sir H C]

Using a diode connected transistor across the base-emitter junction is called a "controlled-gain" or "controlled-beta" transistor.  Grebene talks about it in his book on analog integrated circuit designs.  Pretty much if you want it stable it needs to have all transistors on the same substrate.

[brett]

It's been 2 or 3 years since I tried out many conbinations of active transistors with b-e junctions paralleled with ....just about anything.  (I'm a scientist by day, so I like not to think to hard about circuits at night).
In this post, I'll refer to emitters, as most of the work I did was with transistors and base-emitter junctions, not transistor-diode pairs.
For the technically minded, somebody (apologies for not remembering you) posted a graph of hFE versus inter-emitter resistance for piggybacked devices, about a year ago.  That is very handy for people chasing 70 and 130 for Q1 and Q2 for a fuzzface.

However, from a practical point of view for all circuits, it is simple to add a 25k trimpot between the emitters and dial in the best tones.  My first piggyback circuit built by ear was a FF that had much lower hFEs than 70 and 130.  It sounded like a cross between a FF and a tubescreamer with the gain at 12 o'clock (ie not very wild at all).

Concerning mixing transistors and diodes and silicon and germanium, I couldn't find many (or any?) combinations that improved the few simple circuits I was looking at (FF, Joe's OD, something else...).  Instead, I realised that by varying the piggybacking transistors (2N3906, BD139) and the piggyback resistor (2.2k to 10k), and perhaps their Ic, a great many new combinations of tones were available from the simpler circuits that tend to show off the characteristics of their transistors.

PS although not shown in many spec sheets, the hFE of transistors is quite low when Ic is a microamp or less (10 uA or less for power devices like the BD139).  For example, the hFE of a BC107 peaks at about 250 at 10mA.  At 10uA, the hFE is about 120.  But the collector cutoff current is less than 0.015uA, so there's a range of Ic "room" to explore where transistors are struggling to amplify as well as at higher Ic.  On a previous post, RG also reminded us that the signal (not DC) will dominate Ic under these circumstances, and may dynamically alter gain and produce harmonically-rich distortion.