Protecting BJTs against base-emitter reverse breakage?

[Fancy Lime]

Hi all,

This may be a silly question coming from someone who can no longer claim noob discount but this has really been bugging me lately:

If the base-emitter junction if a BJT is reverse biased beyond it's breaking voltage (Zener breakdown, iirc) even just briefly and only once, the noise characteristics of the transistor can get permanently worsened. In many designs, there are large capacitors connected to the base and/or the emitter. This could allow reverse biasing of the b-e junction from static discharge going into the input. Static voltages just from rubbing synthetic fibers of clothing together can reach thousands of Volts, albeit with tiny charges. And who knows what musicians stick into their stomp box inputs (not THAT, you perverts)? In MOSFET circuits we pretty much always see a protection diode between gate and source. In BJT circuits, we could put a fast switching diode between base and emitter oriented antiparallel to the intrinsic b-e diode to protect against reverse biasing but I have very rarely seen that done despite the negligible cost and effort and seemingly obvious benefits. Sure, in MOSFETs its a different kind of breakdown that doesn't just make it noisier but breaks the device completely, so the analogy is a bit strained, but still.

Why are we not putting in that extra diode? Especially in many fuzzes, it seems to me like a good idea. Yes, it didn't matter in the 60's when the transistors were kinda crap by modern standards anyway and the smart money is always on imitation rather than innovation when it comes to something as deeply soaked in nostalgia as guitar music. But is "if it was good enough for Jimi..." really a good enough reason? After all, the hunt for maximum fuzz with minimum noise is a major theme on this forum. I have a feeling that I might be missing something obvious here.

Cheers,
Andy

[antonis]

Why are we not putting in that extra diode? Especially in many fuzzes, it seems to me like a good idea.

Not at all, especially for Fuzzes..
Actually, that diode makes the difference between fuzz and distortion..

Explain myself: After the BJT is driven negative and starts clipping, that diode kicks in to stop it driven too negative..
Despite it isn't directly in the signal path, it makes the difference between a clear distortion and a fuzz..

P.S.
Another approach to prevent BJT from beeing driven too negative should be a voltage divider to limit the signal from previous stage (like in tube amps) but it's avoided since we want to save as much gain as possible..

[antonis]

In BJT circuits, we could put a fast switching diode between base and emitter oriented antiparallel to the intrinsic b-e diode to protect against reverse biasing but I have very rarely seen that done despite the negligible cost and effort and seemingly obvious benefits.

That diode is always there in power supply output transistors (with or without output capacitors..)
Also, it's reccomended to be placed across Out - In of any voltage regulator (descrete or integrated..)

[Fancy Lime]

Hmm, I see your point. But is that really the reason? The only place where I tested the effect of such a diode was in a fuzz face type circuit with BC550C transistors. There I could not hear any difference between the circuit with or without diodes on both transistors. The input would need to swing down to -(0.6 + bias voltage) to make the diode on Q1 conduct in regular operation, which probably only happens during the initial attack, when the signal is so dynamic that it is difficult to make out any difference anyway. During the decay phase, when things are still plenty fuzzy, the raw guitar signal is way too weak for the diode to conduct. Things are a bit more complicated and gain dependent at Q2 but I could not hear a difference between diode or no diode here either.

If we want to have the protection diode but not have it conduct in normal operation, we could use something with a higher drop, like a blue LED or a ~4V Zener (installed in reverse and with the forward p-n diode blocked by another diode). Ok at that point, it gets a bit more expensive but still not really.

The voltage divider wont help much if we want to cut the potentially thousands of Volts of static discharge to less than the 5V or so that the b-e junction can take. Unless we are ok with loosing 99.999% of the signal, which, as you pointed out we are probably not.

Cheers,
Andy

[R.G.]

I've been advising people to do this for years.; but on the input transistors, not all transistors. A BJT gets a little noisier every time the base-emitter is reverse broken. Over the years, the noise builds up. Amps and effects that are decades old and use BJT inputs often get very hissy. Swapping in a new input transistor usually cures the noise/hiss, and putting in a reverse biased diode keeps it from creeping back in over time. Input bipolars are the ones exposed to static and other transients most, and also have the most effect on noise performance.

Big capacitors on emitters or massive overdrive inside the circuit might be able to break a base-emitter in some circuits. It's possible, but hard to do in a 9V powered effect. The circuit would need to do a diode pumped signal into a base to get more than about half the supply; or have a DC level for a stage biased up near 9V and a low impedance pulldown if it's capacitively coupled; or an emitter biased above about 7V with a big decoupler cap and the base pulled down hard. Most circuits just can't generate the required 5-7V of reverse on a base inside the circuit.

[antonis]

The voltage divider wont help much if we want to cut the potentially thousands of Volts of static discharge to less than the 5V or so that the b-e junction can take. Unless we are ok with loosing 99.999% of the signal, which, as you pointed out we are probably not.

    

I was talking about not going too negative, Andy..
(static discharge is a totally different story..)

As R.G. said, it should be practically impossible for this to occur for a +9V powered circuit but piece of cake for a +/- 15V, say..
(ask me how I know it..)

[Fancy Lime]

I've been advising people to do this for years.; but on the input transistors, not all transistors. A BJT gets a little noisier every time the base-emitter is reverse broken. Over the years, the noise builds up. Amps and effects that are decades old and use BJT inputs often get very hissy. Swapping in a new input transistor usually cures the noise/hiss, and putting in a reverse biased diode keeps it from creeping back in over time. Input bipolars are the ones exposed to static and other transients most, and also have the most effect on noise performance.

Big capacitors on emitters or massive overdrive inside the circuit might be able to break a base-emitter in some circuits. It's possible, but hard to do in a 9V powered effect. The circuit would need to do a diode pumped signal into a base to get more than about half the supply; or have a DC level for a stage biased up near 9V and a low impedance pulldown if it's capacitively coupled; or an emitter biased above about 7V with a big decoupler cap and the base pulled down hard. Most circuits just can't generate the required 5-7V of reverse on a base inside the circuit.

So I'm actually not crazy, that's a relief. I was also mostly worried about the input and thought that getting 5V the wrong way in a circuit with rails only 9V apart may be a stretch. But then I remembered the guitar player of my highschool band. His understanding of electricity was that 1: "if the plug fits in the socket, it must be right" and 2: "bigger number is more louder". I'm not sure if he even believed in voltage and current. I sure wasn't able to convince him that AC and DC aren't the nicknames of the Young brothers. If he found a 36V AC adapter with a Boss-type barrel plug, he would definitely ram that into any pedal. The smell means it's working. Not sure if we can or even want to "Homer Simpson proof" our pedals, though...

Thanks,
Andy

[merlinb]

If the base-emitter junction if a BJT is reverse biased beyond it's breaking voltage (Zener breakdown, iirc) even just briefly and only once, the noise characteristics of the transistor can get permanently worsened.
we could put a fast switching diode between base and emitter

But surely the damage only occurs when the Zener breakdown current is too large? There is usually such a large resistance in series with the base that reverse fault current would probably not exceed the microamp range, in which case is it really damaging the transistor? Do we have any systematic evidence for transistor degradation, or are we just speculating because it 'seems plausible'?

[antonis]

There is usually such a large resistance in series with the base

Sorry Merlin but I'm confused a bit..

Are you talking about physical resistance (item) or r_π (h_fe/g_m) in series with r_bb (Base spreading resistance)..??

[amptramp]

A reverse biased EB junction operated beyond its maximum reverse voltage is often used as an analog noise source for test equipment or in some synthesizer voices where some form of hiss is needed.

[ElectricDruid]

A reverse biased EB junction operated beyond its maximum reverse voltage is often used as an analog noise source for test equipment or in some synthesizer voices where some form of hiss is needed.

You see this a lot in old drum machines too, since they use a lot of noisy sounds. Korg KR-55 is one example. The slightly-later, slightly-fancier Roland TRx0x boxes tended to use digital noise sources though (hardware LFSRs).

[R.G.]

But surely the damage only occurs when the Zener breakdown current is too large? There is usually such a large resistance in series with the base that reverse fault current would probably not exceed the microamp range, in which case is it really damaging the transistor? Do we have any systematic evidence for transistor degradation, or are we just speculating because it 'seems plausible'?

You would think that it's current related, wouldn't you? Well, probably energy- or power- and/or temperature-related? That is in fact probably true, but the sources I found it in say it's semi-permanent and accumulative, each incident making the junction just a bit noisier. Each breakover makes it imperceptibly worse, but in time you get hiss.

I have both a theoretical and practical source for my belief in this. It's probably been over thirty years since I ran into this concept in some tech related notes and a couple of textbooks (I was in my textbook consuming phase. 8-)  ).  Admittedly, this is in the class of things I believe because I was told by an "expert" and work with, but then so is Ohm's law and V=L*di/dt.

I tried just replacing the input transistor in hissier amps and effects, and blam! hiss was no more. This last is anecdotal evidence, the kind I would criticize in other net-expert statements, but it did match the results the textbooks said. Those same amps did not get noisier again over time with the diode fix.

It's easy enough - in theory- to test. Just build a noise testing fixture that can reliably measure the noise spectrum of a bipolar transistor, and one that can adjustably zap a measured amount of breakover voltage/current/energy into the base emitter junction, then run the test over a statistically significant number of bipolar devices (1) of different type numbers, (2) of different manufacturers, (3) different vintages and (4) various numbers of accumulated (5) zaps and (6) power-per-zap. With enough iterations, one could get good statistics on the isolated effects of power per zap, gain/zap relationship, manufacturing process - zap, number of zaps vs energy per zap, etc. Getting some results on a few transistors would be interesting, maybe, but I never had the time. It's probably a good master's thesis project, if schools would approve a thesis on semiconductor use effects these days.

One of the sources I read said that it would be possible to anneal the junction back to lower noise, but it would never get as low as an un-broken junction. I never tried that. Transistors were just too cheap.

At this distance in time, I'd have to do a lot of digging to turn up the source. My brain says (dimly) that it may have been mentioned in Ott's "Low Noise Electronic Design", one of the opamp use books, or notes from Pease or Widlar. That's where I would start looking, anyway.

With me, it's kind of like the last line of the song "Grandma Got Run Over By A Reindeer" - as for me and Grandpa, we believe.    8-)

[antonis]

Just a little thought (definitely without even scratching the surface of semicinductors physics..):

BJT's (as stand alone item) noise is roughly estimated by e_n² = 4kT(r_e/2 + r_bb') + 2qr_bb'(I_C/β₀)(1 + f_ci/f), V²/Hz

Considering identical working conditions for a BJT before and after B-E junction reverse voltage enforcement, the only parameter which could be altered is r_bb' (Base spreading resistance)..
(r_e - or 1/g_m if you prefer - is an intrinsic resistance - not real item - which doesn't exhibit Johnson noise, although it contributes to noise behavior due to Collector current shot-noise through it, considering its value halved - or 2/3 for g_m..)

So, for making a BJT noisier, r_bb' should be made bigger..
(permanently..?? dunno..)

My - purely academic - query is:
1. Is the thin p-type region width reduced..??
2. Is the dopping of the base altered..??
3. Do both of the above occur..??
4. None of the above stands (and I need more coffee)..??