Max safe voltage tolerance of BJTs ?

[MathWizard]

I really don't like using 50V rated caps w/ over 40V, how close do people get with Vce ratings of BJT's ? I was making a PSU lately and I had to go back and swap a few BJT's that were at or over their 30 and 45V ratings.

So how forgiving are the max Vce ratings? And how much worse is it w/ cheap parts I wonder?

I've been listening to lectures on PN junctions and just getting into BJTs, but I can't answer my own Q, but then what are the ratings, how much safety factor do the makers usually add to their values ?

[David Hess]

It is not like the maximum gate voltage of a MOSFET which will destroy the gate oxide if breakdown occurs.  With some exceptions, PN junctions can survive breakdown if the current is limited.  (1) Parts which are designed and specified for this type of operation include "avalanche" specifications.  Avalanche rated power MOSFETs are also available.

As far as general purpose bipolar transistors, there *is* a decrease in reliability as the maximum voltages are approached.  Low power devices are commonly voltage derated by at least 25% so for instance, the 40 volt 2N3904/2N3906 fit well with +/-15 volt supplies and perhaps +/-18 volt supplies.

I also voltage derate solid tantalum capacitors by 25% to 33%.

There are some old Motorola application notes which discuss bipolar transistor reliability and how they are rated.

(1) Exceptions include JFET gates and low power high gain bipolar transistor base-emitter junctions.  Breakdown does not necessarily destroy them but can seriously degrade their current gain at low current and leakage.

[Manul]

First of all, safety margin depends on a circuit. It is not unusual that some parts in some circuits experience higher voltages (or currents) then expected due to various transients, sometimes caused by parasitics which is hard to predict. Together with overheating it is probably number one cause of failures. So it is important to observe real circuit behaviour, not just calculate on a paper.

Usually BJT breakdown between collector and emitter is not damaging if current is limitted and the junction temperature stays within limits, but generally we avoid that. This means looking at the datasheet and respecting the CE breakdown voltage. There are some BJTs which are avalanche rated, so they can be used in that mode with confidence if needed. Personally, I would not hesitate to use a BJT very near its CE voltage limit (like 90%) if everything else is very well controlled, especially temperature. Normally, maybe 80%. For temperature, I like to stay way below limits.

About manufacturers it is really hard to say, how much margin they add. They specify and guarantee a parameter and outside that lays unknown teritory. If they specify minimum voltage for 1mA of breakdown current, I think it is possible that you will find some samples of transistors which are near that breakdown current at that voltage. Probably unlikely, but it may happen. So just assume their margin to be zero.

For me, good analogy is mechanical engineering. It is not uncommon to push materials very near its limits for performance reasons (think F1 engines, stuff like that), but that requires very high precision and keeping all the factors at absolute control. If succesful, then it is a beauty of efficiency. Yet, any small disturbance is likely to send things into catastrophy. So if you can afford higher derating, that is always a wise idea.

[T3sl4co1l]

It is not like the maximum gate voltage of a MOSFET which will destroy the gate oxide if breakdown occurs.  With some exceptions, PN junctions can survive breakdown if the current is limited.  (1) Parts which are designed and specified for this type of operation include "avalanche" specifications.  Avalanche rated power MOSFETs are also available.

Note that avalanching MOSFETs causes wear, so it is best avoided too.

BJT avalanche varies; in many cases it looks just like an avalanche diode, the tricky case is when base current is modest (typically a B-E resistor of 1-10k sets this up), and avalanche breakdown takes on a latching behavior as the device suddenly snaps from Vcbo to Vceo and overshoots -- in the best-worst case it momentarily shorts out in a fraction of a nanosecond.  This pulse avalanche regime is typically destructive to power transistors, because the breakdown occurs at a point, and discharging its own junction capacitance through that point is enough to destroy it, let alone other elements in circuit (bigass filter caps?).

The resulting failure is also pointlike, i.e. you get some resistance from C-E, a few kohms maybe.  If you apply enough current to overcome that resistance, sure, it still transists -- but it'll never cut off more than that.

(You can get similar failures in MOSFET gate breakdown, too.)

It's probably safest to meet or exceed Vceo limits under worst case conditions (high line, min/max output, whatever).  Certainly for linear circuits.

Switching circuits, you can approach Vcbo more, but the tricky part is turn-off.  Switching BJTs provide an RBSOA, where the permissible Vce depends on time after turn-off, and what the load current was at turn-off.  Vce(max) being reduced at high currents and short times, relaxing to Vceo or Vcbo at low load currents and longer durations.

The mechanism goes something like: the C-B junction depletes in a wavelike manner, effectively the insulating layer (depletion region) widens as time goes on.  So it can't handle as much voltage immediately, but after a few microseconds say, and given a B-E short or reverse bias (often a hard-clamping base drive circuit is used, or transformer coupling, which provides this condition), it will be able to handle Vcbo again.

This is, in part, why you see film caps across horizontal output transistors in old CRT TVs and monitors -- at turn-off, the load current flows into the capacitor, charging it at limited dV/dt, keeping the transistor within its RBSOA.

I haven't read much about lifetime vs. voltage ratings so I don't have much to comment on that, but it sounds likely and a good idea.  Fortunately, BJT voltage ratings are fairly free (you can trivially replace a TIP31A with a TIP31C) so this shouldn't be a big problem.


Hm, I wonder if the same mechanism applies to IGBTs... I've always kind of wondered why 600V ratings are so popular, when 500 and even 400V MOSFETs are more common.

As far as general purpose bipolar transistors, there *is* a decrease in reliability as the maximum voltages are approached.  Low power devices are commonly voltage derated by at least 25% so for instance, the 40 volt 2N3904/2N3906 fit well with +/-15 volt supplies and perhaps +/-18 volt supplies.

Interestingly enough, they avalanche pretty reliably in the 100V range (pulsing effectively with about 4.7k B-E).  Make of that what you will...

I also voltage derate solid tantalum capacitors by 25% to 33%.

Hmm, derate by, or to?  Have always seen 1/3 to 1/4 recommended...

(1) Exceptions include JFET gates and low power high gain bipolar transistor base-emitter junctions.  Breakdown does not necessarily destroy them but can seriously degrade their current gain at low current and leakage.

There's also an appnote, which I forget who wrote the one I'm thinking of, it might've been Motorola again -- which discusses use of E-B breakdown in power switching transistors.  They found yes it does of course reduce hFE, particularly at lower currents; but because it's already so low (like < 20), it's not by much, and no other symptoms were found (such as reliability, off leakage, or switching speed).  I suppose if anything, the reduced hFE at low current suggests it might even switch faster, just, again, not by much.

Which is convenient for transformer coupled drive, as you otherwise have to provide something to clamp flyback.  It seems you can let the transistor do it for you.

But anything with high hFE, the effect is stronger.  Not usually a problem for linear circuits, but is notably common in the astable multivibrator, and may occur in diff pairs (where the maximum diff input is |Vbe + Veb|).

Tim

[David Hess]

BJT avalanche varies; in many cases it looks just like an avalanche diode, the tricky case is when base current is modest (typically a B-E resistor of 1-10k sets this up), and avalanche breakdown takes on a latching behavior as the device suddenly snaps from Vcbo to Vceo and overshoots -- in the best-worst case it momentarily shorts out in a fraction of a nanosecond.  This pulse avalanche regime is typically destructive to power transistors, because the breakdown occurs at a point, and discharging its own junction capacitance through that point is enough to destroy it, let alone other elements in circuit (bigass filter caps?).

I have seen all kinds of behavior on a curve tracer.  It varies especially with old transistors where individual members of the same type and lot can behave completely differently, with some showing identical Vceo and Vcbo and others showing marked avalanche behavior.

I haven't read much about lifetime vs. voltage ratings so I don't have much to comment on that, but it sounds likely and a good idea.  Fortunately, BJT voltage ratings are fairly free (you can trivially replace a TIP31A with a TIP31C) so this shouldn't be a big problem.

You could also grade them yourself for high voltage operation.  The specified ratings are only the minimum guaranteed value where hopefully the part was originally tested.

I also voltage derate solid tantalum capacitors by 25% to 33%.

Hmm, derate by, or to?  Have always seen 1/3 to 1/4 recommended...

By.  The parts which I have seen fail were hardly or not derated at all, and there is published reliability data from testing showing essentially zero failures at those levels of derating under realistic conditions.

(1) Exceptions include JFET gates and low power high gain bipolar transistor base-emitter junctions.  Breakdown does not necessarily destroy them but can seriously degrade their current gain at low current and leakage.

There's also an appnote, which I forget who wrote the one I'm thinking of, it might've been Motorola again -- which discusses use of E-B breakdown in power switching transistors.  They found yes it does of course reduce hFE, particularly at lower currents; but because it's already so low (like < 20), it's not by much, and no other symptoms were found (such as reliability, off leakage, or switching speed).  I suppose if anything, the reduced hFE at low current suggests it might even switch faster, just, again, not by much.

Even though it is apparently a surface effect, I think of it as hot carriers producing dislocations which act as recombination sites, so carrier lifetime is decreased, which should result in lower gain and lower switching delay as with gold doping.

But anything with high hFE, the effect is stronger.  Not usually a problem for linear circuits, but is notably common in the astable multivibrator, and may occur in diff pairs (where the maximum diff input is |Vbe + Veb|).

Radiation damage also produces the same effects.

[T3sl4co1l]

Yup, also used intentionally in high speed diodes, some of which are beta irradiated.  Pt doping also popular.

Hmm, I'm not aware of irradiation being used with BJTs offhand, but gold doping certainly has.



Tim

[MathWizard]

So it's not a big worry then and they aren't going to explode. Later I should hook up a BJT and resistor to my PSU and see what happens

[Marco]

Most datasheets only give Vceo, but that's not all that relevant in most circuits.

The max linear voltage at small base currents is somewhere between Vceo and Vces.