PNP_turn_OFF circuit in isolated FET drive?

that is about the maximum, it acts like a zener diode at these reverse voltages - and you can easily damage the transistor if the current is not limited ...

Good designed gate driver circuits don't need to drive transistors in Vbe breakdown.

Thanks, i wonder if the reason these turn off circuits are "allowed" to overvoltage the V(EB) of the PNP is because the gate drive current spike is so very short? (for a FET being driven).

Can you show a circuit that involves reverse base-emitter voltage > 7V?

BVebo depends on E-B doping. You can find lateral PNP devices with 40V E-B breakdowns in all of the old timey "standard linear" LM-series parts. On the same die you will find NPNs with 7V-ish breakdowns and 5-6V rec max ratings (set so, to keep hot carrier drift out of the Vio / Iib picture).

A wide-base lightly doped device could tolerate higher Vbe(reverse) if emitter doping is likewise reduced. The norm is about 100:1 on vertical BJTs in technologies I've used, this supports an acceptable alpha and beta (subject to nonideal base current and transport actors).

Selecting such a device maybe be a chore or adventure, they will be found mostly in old crusty PNPs that hardly anyone wants anymore and so a likely future obsolescence victim.

reverse diodes are usually used to keep Veb to 1.1 volt or so -- this provides plenty of turn off except for the most esoteric/demanding ckts ...

Can you show a circuit that involves reverse base-emitter voltage > 7V?

Click to expand...

Hi,
The attached is a bog standard half bridge gate driver, with PNP turn off. You can see that the ringed BJTs suffer more than 5V across their V(EB)'s.
The NPN's suffer V(EB) of 9.2V.
It seems that PNP turn-off circuit BJTs are exempt from normal rules on V(EB) maximum?
PDF schem and LTspice sim attached.

(BTW the RF60 diode with "n=6" is actually a 3v9 zener, just that LTspice doesnt have 3v9 zeners)

If, in the sim, you put a diode across the EB of the PNP, then you can see that the turn off is not as good as without this diode. It seems that with PNP turn-off, there is a deliberate effort to put a good few volts across the V(EB) at turn off time, in order to really quickly sweep out the carriers from the junction , and turn the PNP definetely off)

You can put a reverse-biased diode across the base-emitter junction to protect it.
That will still apply a reverse voltage of about 0.7V across the junction to help sweep the carriers out.

I must confess that I don't understand why the circuit has been exactly designed this way, I would specifically avoid to drop a considerable amount of available driver voltage.

Regarding reverse Vbe operation, I expect that current limiting resistors will prevent transistor damage.

all the xtors shown in the pdf are sufficiently protected - even if it of not the most elegant driver ckt in the world ...

AFAIK reverse biasing the base emitter junction won't harm the device so long as current is limited appropriately (though the datasheet may not give guidance on proper limits).

Yes, the large Veb helps speed up turn-off (something related to storage time). If you were to clamp Veb with a standard diode, this would slow down switching. In theory you could instead clamp it with a zener with Vz less than Veb max (also with a series schottky to prevent that zener from being forward biased). But the Vbe junction of the BJT generally performs the same function by istelf. Just need to make sure it stays within its SOA.

A baker clamp circuit also another technique for faster turn-off of BJTs. Probably not applicable here though.

AFAIK reverse biasing the base emitter junction won't harm the device so long as current is limited appropriately (though the datasheet may not give guidance on proper limits).

Click to expand...

Thanks, i think thats exctly my concern, ..how do you know what is the safe current limit....

There's "layers" to this.

At moderate, sustained reverse current charge trapping
from the hot carriers can degrade (particularly) low-current
beta and shift Vbe slightly. For a switching application
maybe neither of these matter much. But you'd like to be
within a decade or two (Ic) of peak beta operating point
in the "on" condition.

There will come a point where you do thermal damage
(like zener-zap programming) and this depends some
on current crowding, defects that break down early
and hog all the heat.

Vendors are going to sandbag any info they give, which
usually isn't much - especially for an operating region
which "should never happen".

You might need to test-to-fail some parts and apply a
safety factor that appeals to you, after you have a grasp
of populations and variations.