BJT CE breakdown - permanent degradation?

[c4757p]

Basically, the title says it all. Does somebody more familiar with semiconductor physics than I am know whether controlled (low-current, at absolute most a milliamp) C-E breakdown permanently degrades a BJT, the way E-B breakdown does? Or can a BJT be safely used in a position where such breakdown occurs fairly regularly?

I suppose I could ask the same question about small (2N7002-type) MOSFETs. Does forced conduction at the ~70V D-S breakdown limit carry any risk of breaking through the gate oxide, for instance?

[nuno]

At least many of the big FETs are avalanche rated.

[bobcat]

Some BJT's are specifically designed for breakdown operation.

[c4757p]

Sure, of course. But what about normal ones, under gentle conditions?

[IanB]

I think the amount of damage caused by a reverse breakdown will depend on the amount of energy dissipated in the silicon structures during the event. And I suspect that will be a device-dependent parameter. So I don't know if such a question could be answered in generality. I would say even an E-B breakdown could be harmless if the energy dissipated was small enough.

[c4757p]

I was under the impression that any reverse current, even small, through the BE junction could cause permanent gain loss, with the gain loss increasing with the total charge moved across the junction.

Hmm. I might have to just test this. Gentle reverse breakdown in my intended application will be quite extended, not just one event. Perhaps I'll characterize a bunch of transistors, stick a couple mA through them in reverse for a long time and re-characterize them at intervals...

[calexanian]

Just to mention it, a zener diode is simply a junction with a well defined reverse breakdown characteristic. Old low voltage transistors make reasonably stable zeners in a pinch.

[Zero999]

Why not just connect a suitable zener diode in parallel with the transistor?

[c4757p]

Because I'm wondering out loud whether or not I have to. Just experimenting. Obviously Zeners will be used if necessary, but for the sake of shits and giggles, I'd like to find out whether they are.

[Andy Watson]

Doesn't it depend on the design of the junction, i.e. some junctions are designed to withstand avalanche up to a given energy. Like the nano/pico second pulse generators that deliberately ramp up to breakdown voltage.

[Kevin.D]

I was under the impression that any reverse current, even small, through the BE junction could cause permanent gain loss, with the gain loss increasing with the total charge moved across the junction.

Hmm. I might have to just test this. Gentle reverse breakdown in my intended application will be quite extended, not just one event. Perhaps I'll characterize a bunch of transistors, stick a couple mA through them in reverse for a long time and re-characterize them at intervals...

Hi 
On EB reverse junction breakdown I did a few quick measurments some years ago to find effects of different reverse currents and found the beta reduction even with only small reverse currents of 1 mA or so could easily reduce the beta of Very low power BJT's by upto ~80%. The maximum beta reduction I could achieve seemed to be related(exponentially) to EB junction size. i.e the larger the transistor is the smaller was the max beta reduction I could achieve.
Going from  lowish power devices like a bc557 I easily got gain reduction of 50-60%
whilst for larger power to3 device I could not reduce gain by more than ~10% no matter how much reverse current I put through eb.

The gain loss need not  actually be permanent you can recover  the junction by passing a large  current
of 1-2 Amps for a few tenths of a second  forward through the EB junction ( is very easy to overheat junction when attempting this ).
I think the beta reduction you sometimes see in bjt's that are old and have been sitting unused for years maybe the  same physical proccess and if you have ever fixed much old equipment thats being sitting idle for many years you might have already came across faults related to loss of beta due to age/idleness in small signal bjt's. If you do and you can't find a suitable new replacement bjt you could  always try a 'rejuvination' like I did below.
Heres a small selection of the rough measurments I did I found in my old notebook.:-

I found a box of 2sc2320 (npn 50v 200mA) being sitting in a parts bin since about 1990 the beta on most of these was ~ 190 . I passed a 1mA reverse current (breakdown voltage was 8.6v) through the EB junction for 1 minute
Hfe was reduced to 116 . I tried to see how low it would go so then did  10mA reverse current for 1 minute which reduced gain to ~ 100 I could not reduce gain further with higher currents. I repeated this on a few more of these and got a sim response. I then tried to 'restore' the gain by passing 1.25 A for ~ 1 second which then revovered the gain to ~ 170 . I tried to see how high I could get it  so a second attempt using 2 Amps for ~ 500mS the 2sc2320 gave a gain of ~ 240 (so 30 % higher than the original 190 ).This was the max forward eb current I could go before the eb junction started to burn out.
I repeated the above on various bjt's and they where typical figures for similar size
devices so I wont post them all here .On smaller bjt's the larger loss/then boost gain %'s where seen . larger bjt's and saw little beta loss and/or regain where seen .

I also tried power Darlington's heres a small bit from my notes:-

  Darlington bd646 (pnp 8A) initial reverse breakdown was ~9V and it had intitial 
 Ice = 2.13A @ Ibe of .5mA (an Hfe ~ 4050) ,Ice=.95A @ Ibe .25mA (Hfe ~3950) both well above spec given in data sheet.
Found difficult to reduce the Hfe and put upto < 400mA reverse and hardly reduced beta by more than 5% .
Then 500mA for 10 seconds reverse Ibe was applied  beta was reduced to 3600 @ .5mA Ibe (1.8A Ice)
and  2000 @ .25mA Ibe (Ice .54A) .These are both still above the minimun spec for this device in the data sheets.
Then tried 750mA reverse for 10 second (it got hot) this hardly reduced the higher Ice current Beta figure but @ lower Ice current (Ibe=.25mA Ice =.15A). the gain fell to 600 .
Signifying  only some beta  reduction of the smaller first stage bjt of the darlington .

Regards

[c4757p]

Well, a quick test for about five minutes of 2mA through a test PN2222A (Vavalanche = 120V, Vhold=100V) shows no change in gain. I'll keep it running for as long as I reasonably can (not too eager to heat up the room so much running this HV supply constantly, should be able to get a few hours out of it at least). That's probably 10x what it'd reasonably have to handle.

Didn't think to test for leakage before, but I'll check after and make sure it's still reasonable.

It is interesting to avalanche-test a bunch of transistors. Quite a variation among them. The one common thread I see - all my name-brand ones (probably all Fairchild) avalanche at a significant excess of their rated V_CBO (base has 10k to emitter here, so V_CBO will be more accurate than V_CEO), some more than double. Most of the cheapo ones avalanche at only 10 or 20V over rated.

[c4757p]

Gentle reverse breakdown in my intended application will be quite extended

Er, avalanche, not reverse.

[c4757p]

Kevin.D, interesting. I was not expecting this to be recoverable.

I'll have to do some more research on that.

[amyk]

Read this: Degradation of Junction Parameters of an Electrically Stressed NPN Bipolar Transistor

The fact that it seems to be reversible with a high-current pulse in the forward direction suggests some sort of annealing process happens via heating. If this process is reversible for many cycles, it might be interesting to see if one can make a crude form of nonvolatile memory from it... perhaps it could be called transmemristor memory?

[paulie]

A quick browse of internet white papers or IIRC even Wikipedia will reveal all avalanche devices "wear out". It's just a matter of degree and design as hinted above. This includes transistors, glass zeners, and buried type although much less for the last. I suspect this may be a factor in long term drift characteristics of high end $50 voltage references like LTZ1000 which cannot be 100% compensated for.

[T3sl4co1l]

I'm going to guess not, but it's an interesting question.

As mentioned, avalanche damage is usually thermal in nature.  If it's safely below that level, it may not be much.

The mechanisms posited in the Toufik article will still occur for the B-C junction, but the impact would be reversed (inverted) hFE, not forward.  Something else to test?

Tim

[paulie]

I agree. So maybe I should have phrased it "cannot be 100.00000% compensated for".

[Zero999]

I wonder if a cheaper lower voltage transistor could be connected in parallel to absorb the energy? The leakage current should be lower than a normal zener.

[c4757p]

After testing for much longer I still find no loss of gain and no excess leakage. As this is just a (low power) switching application anyway, and a one-off hobby thing, I think I'll go ahead with this.

[T3sl4co1l]

Switching?

Related anecdote;

In this light,
http://seventransistorlabs.com/Images/LED_Light2.png
the power transistor is wired in constant-hFE drive, with positive feedback.  It turns off due to core saturation, and turns on again with a small pulse from the control circuit.

When the transistor turns off, the core demagnetizes of course; this occurs fairly quickly (~100ns), so the spike is tall.

As built, this light was functioning correctly, but had an audible buzzing-hissing sort of sound to it.  Like a chaotic loop problem, but the loop was stable (given the fact that it's adjusting to compensate for supply ripple), and the switching waveform not unusual (but, given that it's frequency modulated, so it's hard to see anything in great detail).

Turns out I forgot to put a suitable ballast/damping resistor across the drive transformer primary; 100 to 470 ohms is about right.  The peak was over 100V, causing the transistor to avalanche -- which would also make some nicely high frequency harmonics.  That is, if they were coherent enough to matter, and if they were allowed to radiate anywhere.  With the resistor, the peak is under 80V, the transistor is happy, and operation is silent.

A bit more robust of a switching circuit than you're talking about, I guess (10s to 100s mA peak here), but an example of high speed avalanche and amplification in a real circuit.  I did not measure the transistor, but as it was still functioning fine (for the circuit, min hFE ~50 ish?), I didn't see a need to check or replace it.

Tim