"Piggybacking" trannies for lower gain

[brett]

This is about getting running 2 transistors in parallel to get low gain.  It turns out I got VERY low gain.

I've been building an all silicon version of Aron's Hornet, which is a fuzzface variant.  Along the lines of RG Keen's suggestions in this forum, I have been looking at "piggybacking" transistors to get different hFEs out of them.  e.g. A while back Javacody wrote "I also read something where RG suggested running two trannies in parallel, but leave the collector of the second tranny unconnected, thereby cutting gain in half. "

Simple.  Or so I thought....  Well, my Hornet had Q1 and Q2 hFEs of 130 and 340, so I piggybacked then with 2 extra tyrannies of hFE = 200 and 340, respectively.  The Hornet was now more of a Wasp, and sounded great.  I bundled it up and sent it off to the guy I was building it for.  Then I started to wonder what I had done to the hFE of those transistors.  A few checks revealed some things that STUNNED the *^%@ out of me.

I found that piggybacking two similar transistors (PN100s, hFE=350 and 430) to those I'd used for Q2 in the Hornet gave a hFE of either 6 or 12, depending on which transistor piggybacked on which.   So piggybacking DIDN'T HALVE THE GAIN.  Also, the Hornet worked great with a hFE at Q2 of about 10. (What!!??)

I thought I must be measuring something wrong, so I used RG's transistor testing circuit to measure the hFEs.  Same result.  I also tried different transistors and got variable, though similar results.  2 BC549Bs with hFEs of 280 gave hFE = 5 when piggybacked, and 2 BC547Bs with hFEs of 450 and 425 gave hFE = 25 when piggybacked.  

RG, if you see this, do you have an explanation?  It seems to me that the current must be preferentially flowing through the base-emitter juction that has the disconnected collector.  Also, if people are always recommending hFEs of between 70 and 130 for fuzzfaces, why did mine sound so good (it sounded REALLY good) when the hFEs were presumably as low as 5 to 25 ?  I'll definately try to get that Hornet back and measure the hFEs.

Anybody got any suggestions about this?  I welcome any/all comments, and encourage other folk to have a look at this and see if they get similar results.

[Paul Perry (Frostwave)]

Maybe the HFe test on a multimeter doesn't work right when somethng so unorthodox is put to it? Note that the HFe given by a multimeter isn't really the "small signal AC gain", it is something else, which is USUALLY an indicator. maybe one of the EEs here can comment.

[brett]

As well as the multimeter, I used RG Keen's transistor testor, which simply puts 4uA through the base-emitter and measures the resultant collector current.  The results were essentially the same.

[R.G.]

(... oops... don't know ma own strenth...)

I thought the hfe *would* be about halved. The idea is that an identical junction ought to eat about the same current, stealing base bias and signal current in equal amounts to the active transistor.

I'll do some speculating, then go do some research to find the real answer.

======= Speculation======
A bipolar base-emitter junction behaves as though there is a phantom resistor in series with the emitter. This so-called Shockley resistance shows up as added base-emitter voltage and as a gain determining component in transistor circuits. This much is fact.

Since the dummy transistor is getting only base current, not the base-current-multiplied-by-hfe current in the active transistor, the dummy transistor base-emitter may have a slightly lower junction voltage because the current through its Shockley resistance is much lower and it can therefore steal most of the base current, causing a much lower apparent hfe.
====== End Speculation =====

I'll go do some reading, maybe I can figure out what is really happening.

I'm glad it sounded good. There may be a real pony in there.

[petemoore]

I tried this in the Yaff I built this morning
 With 2n2222 in Q1 and NTE103 in Q2, Q2 read 4.76v at collector
 With 2n2222 and 5088 in Q1 and in Q2 an NTE 103, Q2's collector read 5.11v
  The sound of the Fuzz became more refined....lol....not exactly a 'labratory' test, as above noted the two configurations were only ear tested [since this morning] and we're not rebiased between takes.
  I refrain further comment, yet to say, to me
  it seems like a Neat Trick, Very Nice yaFF.
  Sure was easy to get bias on this thing, I put a trimpot on Q2 but haven't even had to move it  the whole time, even with 3904 in Q2 and the dual transistor in Q1 !!!
  It's stuff like this that makes it hard for me to stop building for a minute.

[aron]

Brett,

Thanks for doing so much experimenting on the Hornet.

I need to try what you did. Sounds interesting.

[idlefaction]

yeah man, that sounds REALLY interesting.

[brett]

I did some more experimenting last night, following up RG's idea that there's extra base-emitter resistance in the transistor with all leads connected.

What I did was add an external resistor to the "piggybacking" base-emitter, to "lighten" the piggyback.  I found that if I added about 4k ohms, the hFE of two similar PN100s would be halved when doing the piggyback thing.  I was quite surprised that it was such a high resistance.  I don't know why, but I was expecting maybe 100ohms (perhaps because I was thinking of the base-emitter as a freely conducting diode, but it's obviously not).

Next, I thought I'd see if the hFE-reduction effect depends on the amount of current flowing through the junction.  Sure enough, it does.  I modified RG's transistor tester to do this for a piggybacked pair with a 2k resistor.  With 4uA base current I got hFE=50, with 40uA I got 150 and with 80uA I got 220.  So the hFE is dependent on the base (and/or collector) current.  Next thing I think I'll try is the base-emitter plus resistor (1k?) piggyback method in a fuzzface and see how it sounds.  The variable hFE might have some interesting effects.

That's it for now.  Cheers.

[Jason Stout]

I love this forum!

[R.G.]

Neat stuff! It sounds like we have the beginnings of a DIY low-gain transistor which includes all the dirty stuff that modern high gain, high linearity devices don't.

I can't get my simulator to work right, so I'm falling back on reading solid state semiconductor physics texts to try to get a more formal explanation of the phenomena.

What I did was add an external resistor to the "piggybacking" base-emitter, to "lighten" the piggyback. I found that if I added about 4k ohms, the hFE of two similar PN100s would be halved when doing the piggyback thing.

Did you add it to the base side or the emitter side? I would expect same results whether base or emitter as long as the collector remains open, but different results (much more effect when it's in the emitter) if the collector of the dummy transistor gets current to contribute.

With 4uA base current I got hFE=50, with 40uA I got 150 and with 80uA I got 220. So the hFE is dependent on the base (and/or collector) current.

That's true of all transistors - hfe is an exponential function of the collector current (neglecting losses and parasitics). That's how a differential multiplier cell works - the two transistors have a lower gain at low collector currents than they do at high ones. I'm surprised at how fast your hfe changes though. Maybe you happened to test right on the knee of the exponential curve for that transistor.

Changing gain with collector current is *good*. That's a fundamental means of generating even-harmonic distortion. Bad old transistors (like early germanium ones) did this a lot more than today's devices, and it is worse at low collector currents - that's why the Rangemaster is sitting near cutoff, it compresses the low-current side of things and expands the high-current side.

Next thing I think I'll try is the base-emitter plus resistor (1k?) piggyback method in a fuzzface and see how it sounds.

I'll be very interested. If this can be tweaked in and we can deduce the principles behind it to make it predictable, we may have made searching for dinosaur-egg germanium devices a thing of the past.

...OK, OK, I'm getting carried away...

[brett]

Did you add it to the base side or the emitter side? I would expect same results whether base or emitter as long as the collector remains open...

Yes, I left the collector open on the piggybacker (Q2) and added the resistor on the emitter side as follows:
Base of Q1----baseQ2|emitterQ2----2k resistor----Emitter of Q1

[javacody]

Wow, we are experiencing history in the making.  

[javacody]

I tried two bc109's, one with a gain of 510 and the other with a gain of 486. Combined they gave me gain of 8 and they sound awesome!

[b_rogers]

i tried 2 bc547c's left the collector and everything in there and noticed a little more "amplike" attack and resonance with no loss of distortion ...nice..did the same to q2, a 2n7000 and it just got a little better. not a real huge improvement but noticable. mismatched a bc547 with a bc338 (low gain) and the sound got kinda honky..had to tweak the trim pot almost every addition or substitution..

[Gus]

R.G. how is this different from using say a 1n4148?

 In a Si rangemaster type circuit I built I picked a tip29 from looking at its very curved Hfe vs Ic.

[brett]

Good question about the 1N4148 Gus.  I'll experiment with one tonight.  Concerning the slope of the hFE va Ic relationship, using the setup mentioned above (base-emitter and 2k resistor piggybacking on Q1) I measured the following: hFe = 50 @ 0.2mA, 150 @ 6mA and 220 @ 18mA.  The hFE of a TIP29 varies from about 50 @ 1mA to 90 @ 100mA, so it's hard to tell which is more "curvey".  I obviously "missed" an interesting region between 0.2 and 6mA where the hFE is "jumping" significantly.

When I get to the shop and buy some more transistors, I'll measure some hFE and Ic values with smaller piggyback resistors (ie at lower "synthetic" hFEs more useful for Q1 in fuzzfaces).  I am starting to realise that the piggybacked transistors have lots more curvature in their hFE-Ic relationship, and that the very low multimeter hFE values (mostly under 20) may be simply a reflection of low Ic.

Lots of questions and few answers (so far)

[GFR]

I did something like that some time ago (adding a Ge diode in parallel with b-e junction of a Ge transitor to make it more suitable for a FF).

http://www.geocities.com/gfr.geo/britface.html

[Doug H]

This is good stuff, guys.

After building the Hot Silicon and breadboarding the FT70, (and comparing them to my fuzz face) I decided no more Ge sorting for me thank you when it comes to fuzz boxes. It looks like you may be on the verge of a technique that would make a really nice sounding Si Rangemaster. Maybe I will be able to dump Ge's forever now. Good riddance!!

Thanks for all the hard work.

Doug

[R.G.]

R.G. how is this different from using say a 1n4148?

I'm not sure it's different at all, except that with more similar junctions and doping profiles (that is, more similar than a 7V junction that's deliberately as thin as possible instead of the 100V, gold doped 1N4148 junction) the curve of junction voltage versus junction current is more likely to interact a lot, which is what is causing the effect... I think.

If one junction is a lot lower voltage (say, 25 mv) than the other at some current, it will steal most of the current, you might not get much effect. I'm speculating again.

did the same to q2, a 2n7000 and it just got a little better. not a real huge improvement but noticable.

The 2N7000 is a MOSFET. I would not expect nearly the same results from a MOSFET because of the fundamentally different structure.

Yes, I left the collector open on the piggybacker (Q2) and added the resistor on the emitter side as follows:
Base of Q1----baseQ2|emitterQ2----2k resistor----Emitter of Q1

OK. That's how you want it for experimentation. Try connecting the collector to 9V through a 10K-47K or so and see what changes in the sound.

Hey! Fun stuff!  I think I may have just obsoleted my private reserve of germanium devices  :shock:

[Brett Clark]

======= Speculation======
A bipolar base-emitter junction behaves as though there is a phantom resistor in series with the emitter. This so-called Shockley resistance shows up as added base-emitter voltage and as a gain determining component in transistor circuits. This much is fact.

Since the dummy transistor is getting only base current, not the base-current-multiplied-by-hfe current in the active transistor, the dummy transistor base-emitter may have a slightly lower junction voltage because the current through its Shockley resistance is much lower and it can therefore steal most of the base current, causing a much lower apparent hfe.
====== End Speculation =====


I'll concur on that one, RG. Paralleled transistors are common in audio power amps, and disconnecting the collector of one will cause the Vbe to drop relative to the parallel units and "suck" the drive current (from the previous stage). This often damages the Base-emitter junction of the transistor. Of course, the available currents are very high compared to the stompbox application...

It also follows that if you used some small unbypassed emitter resistors (seperate one on each transistor), the beta would approach the average beta of the transistors (as it does in a power amp).