Reverse overvoltage on diode

I'd guess that the output is that 300 ohm resistor
and if so, a "flyback" event on cut-source-current
might push back through the upper TVS to the
diode in question. If that diode conducts in reverse
it hits Q1 C-B forward and then potentially B-E will
go reverse biased over the 20V rail and maybe break
down too (a vertical-BJT damage mode).

Now maybe shunt instead of block, or other schemes
could eliminate the need for such high voltage blocking
(and, be more resilient should the diode someday fail).

But it all looks like an "external fault patch" (or safety
strap against insane load inductance values) to me.
No value to normal operation.

cupoftea said:

Hi,
Do you know why the ringed diode would ever need to be 400V rated? Its a general purpose, SMA, 400V diode.
Why ever do you think they used such a high voltage part?
LTspice and PNG attached.

Click to expand...

Layout inductance on Q3 even in 100 nano henries will create a flyback spike of twice the switched Vce on the collector of Q3 upper part of half-bridge from "circled" diode anode D3. To prevent Q1 damage this must be suppressed.

I believe this is a design oversight. Perhaps it came from the belief that a higher PIV diode will have a lower C than Q1 and thus may conduct possible damaging reverse voltage to Q1 Veb > 6V. An effective prevention is a reverse signal diode across Q1-base-emitter and collector-base becomes forward-biased.

Another design flaw could be not having a current limiting base resistor for U1 driver which also seems to have a flaw in LTspice producing a max current of about 2A or about 10 times what it is capable of. Even so, driving the base harder than your 300 ohm load makes little sense.
Actually, this schematic brings up many more questionable details but I guess it this is your responsibility.

This is an XY question.

Thanks, also i was thinking of RC snubbing D3, but theres 8 of these circuits on the PCB and adding 16 new components isnt going to be popular.
As such, we are thinking of just making D3 a 27V zener, but wonder how quickly they "reverse recover" and start blocking? Are BZX84- type zeners just like gen purpose diodes when in the forward direction is concerned?

Its interesting because you can stuff stray L's all over the circuit in LTspice and it never shows any overvoltage to any component.
Though the use of a 27V Zener for the blocking diode seems like a good plan. (as attached).
Any thoughts appreciated?

My guess, given the relatively low voltage ratings of the other components, is that a zillion of these are used elsewhere and adding a new part to the inventory is more expensive than the diode itself.

Brian.

cupoftea said:

Thanks, also i was thinking of RC snubbing D3, but theres 8 of these circuits on the PCB and adding 16 new components isnt going to be popular.
As such, we are thinking of just making D3 a 27V zener, but wonder how quickly they "reverse recover" and start blocking? Are BZX84- type zeners just like gen purpose diodes when in the forward direction is concerned?

Its interesting because you can stuff stray L's all over the circuit in LTspice and it never shows any overvoltage to any component.
Though the use of a 27V Zener for the blocking diode seems like a good plan. (as attached).
Any thoughts appreciated?

Click to expand...

You missed the only overvoltage mentioned in my previous comment Veb =? 10V with 100 nH on Q3C to diode.

There is no diode PIV >32V

With resistive load, I don't even see the purpose of D3. Can you enlighten me?

D3serves no purpose yet Q1 still gets a parasitic flyback on Q1c which drives PNP Q1b higher than Q1e >6V thus requiring a reverse clamp signal diode across Q1be because you need to add a current limiting Rb to Q1b which raises the Vb from the flyback in Q3c. Although Q3c parasitic ESL can be reduced, this only shortens the L/R pulse width and not the amplitude of Ic(Q3)*Rb conducted via Q1cb. At least thats what I see.

Without these changes the PNP Ibe(Q1) gets excessive max current from IC1 Op Amp and excessive reverse Vbe(Q1) from Q3.

A B-E back-shunt on Q1 would protect the emitter better. Downstream consequences of that change in the original or other fault conditions, have not looked at

Searching for the "LT1006" max output current is troublesome as the model I had in the library doesn't seem to match the approx 40 mA shown in this steady state nominal plot.

LT1006 output current would be sourcing, expecting
a load to GND. There appears to be no sink path in
the output, according to the DS schematic


This also says to me that if output is taken over VCC+7V
there's a likelihood of the output NPNs getting hurt instead.
These are necessarily low-BVebo vertical NPNs.

But the PNPs here, are obviously in some cases lateral;
multicollector PNPs are a dead giveaway. The question
then becomes, how are those multiemitter PNPs at the
front end made? Lateral is a possibility and lateral PNPs
have a far higher BVebo (= BVcbo). Field plate asymmetry
might make breakdown outcomes diverge, B-E vs B-C.

With no attempt (shown) at cross-clamping at any of the
first 3 transistor stages, I'm inclined to assume all are
LPNPs. The lousy UGBW (not even spec'd, looks like
1MHz (typ) for 0dB) doesn't make me think a vertical
PNP is used and wouldn't have existed at time of
part creation, VPNPs arrived later in offshoots of the
"standard linear bipolar" flows.

This op amp sure looks like a pig - hundreds of pF of
comp cap distributed, to cover up the front end lag
(from the lateral PNPs) and a lot of transistors compared
to (say) LM158 SSOA. Which does have, at least a wimpy
pulldown leg (but this feature differs at every vendor,
a sign of never-been-quite-right, in the detail layout &
schematic - and DS schematics are flat wrong). Wonder
why such a sloppy circuit needs a high DC precision
(and why you'd want an op amp with what looks like
150-200us risetime in a protection function). Maybe
(going by the gross filter around the op amp) it's
desired to be way slow and the slow op amp is just
"buried" under that filtering?

With resistive load, I don't even see the purpose of D3. Can you enlighten me?

Click to expand...

Thanks, the D3 is needed in case someone puts a voltage source at the load.
This is a possibility since its a DALI circuit.

Pg 11 of the following shows the diode..but they use a bridge there instead...

In the actual cct, Q3 is PBSS4540Z, so is Q1, and Q2 is PBSS5540Z.

PBSS4540Z

PBSS5540Z

You missed the only overvoltage mentioned in my previous comment Veb =? 10V with 100 nH on Q3C to diode.

Click to expand...

Thanks, sorry i put a stray L on Q3C to diode, but saw no overvoltage on Q1...as attached LTspice and PNG

The more I look at this design, the more problems I see.
The LT1006 seems very problematic with anti-reversal design yet, it reverses and oscillates every 80 us with a pulse. I think that is because the current sense is outside the Vcm range. So I dropped in a substitute that is protected R2R 30V. So it is not from parasitic ESL but from a misbehaving LT1006 that is oscillating.

Scrap this design and use the recommended circuit from Microchip with a PN2222A and a dual diode clamp.

Below is proof of the Veb problem with Q1

and fix

Note how I use [net] labels which replace the node #'s so plots are easier to understand.
Beware starting supplies from t=0 produces massive currents since LTspice supplies are not current limited.
The C coupled U1 toi Q1 was acting as a charge pump and boosting the voltage to 32V using the Vbe as a negative clamp causing the -12V on Veb and from any parasitics as I previously thought.

OK, thanks, i will take a look. You mean the diode's (D3) capacitance is part of a charge pump?. Also, i thought Vbe's were like "6v zeners" and so they didnt get bothered by this kind of thing? There's a Dave Jones video which tells of this. (the video is about some kind of measuring instrument which uses BJTs as 6V zeners to protect the inputs.)

Also, thanks for your advice ref changing the current clamp circuit....so i changed it as attached...do you think there's any problem with this circuit now, in terms of overvoltages to D3 or Q1? (LTspice and PNG attached).

I was thinking of D3 conducting then suddenly the current in it stopping....then D3 current slightly reverses (due to stray oscillations)
due to reverse recovery time...then the D3 diode finally snaps off and that current , which had been traversing from cathode to anode in the (recovering) diode...
this causes an inductive spike? (ie an inductive current has been suddenly broken)

Its just that the person who did this circuit used a 400V SMA (general purpose) diode for D3....and so i was wondering what they had been thinking