Diode has IFSM=1A but experiences 90A in circuit (for 2us)

Hello,

The DA2JF8100L diode datasheet says that its IFSM parameter is just 1A.
In our circuit, this diode would experience a current of 90A for just 2us. This is due to a capacitor getting charged through this diode when the mains comes on at mains peak.
(as in the attached LTspice simulation and the jpeg schematic)

Is this current peak OK? After all, it only lasts for 2us and isn’t going to overheat the diode?

DA2JF8100L diode datasheet:
**broken link removed**

Which answer do you expect? Simply consider that the diode specification is incomplete.

I'd be wanting to know bond wire material and diameter
at minimum, to calculate whether the bond wires are
your inline fuse. With so low a rating it might be a
single bond wire, and 90A*2uS is way beyond what
1.25 mil aluminum could take, as I recall. Though at
extremely short pulses the bond wire can take a lot
more.

I'd be wanting to know bond wire material and diameter
at minimum, to calculate whether the bond wires are
your inline fuse. With so low a rating it might be a
single bond wire, and 90A*2uS is way beyond what
1.25 mil aluminum could take, as I recall. Though at
extremely short pulses the bond wire can take a lot
more.

Click to expand...

Thanks, I am grateful as you have pointed out that its not the thermal mass of the whole diode that counts here, but the thermal mass of the titchy little bond wire.....the weakest link. This definetely makes me want to avoid the part and go for something with an IFSM of a few tens of Amps....even if not 90A....something like an SMA package.

The clue is near the top of the front page:
"For high speed switching circuits"

It isn't intended for power usage and it's package doesn't suggest it could dissipate much continuously. As stated, the bond wire and welds would behave like a fuse rather than the whole device being overloaded.

For that application, I can't see why a conventional 1N4007 or SMD equivalent wouldn't work and it would probably be cheaper.

Brian.

Thanks, we are severely space constrained so we wanted to consider the smaller part.
The inrush current that this diode is exposed to is as attached.
Its 90A in peak but lasts for jusr 2us......when you tell people that terrible things are going to happen in only 2us, they look at you in disbelief.

Another point to consider is that "inrush" may not
be a one-time event. For a LED lamp the light may
be turned on several times a day and roll up to
maybe 10K cycles over the advertised service life.
That makes it not a one-time survivability thing,
but a wearout thing (with a lower current*time
limit).

Thanks, very true, but our product really does switch on just once per day. The continuous power dissipated in this diode is virtually nothing after the 2us inrush event has happened.

The discussion tends to become useless. You should simply concede that you don't know if the diode can handle 90A for 2 µs without permanent degradation or even "blowing" a bond wire or chip pad.

There are 1N400x variants in small packages like SOD-123 (probably even smaller) with specified surge capability of several 10 A for a mains half cycle.

Thanks, it has been pointed out here that the first page of the datasheet does say that this diode can handle 1A for the 10ms of a 50Hz half cycle, so surely if it can handle 1A for 10ms then it can handle 90A for 2us?

Not "surely" until you get some relevant bond wire,
and behind that chip interconnect, pulsed power-
to-blow data.

From the datasheet it looks like the series R of the
diode is about 0.5 ohms. The data does not extend
to even 1A. Vf is 1.6V at 0.2A. At 1A would expect
2.0V Vf and at 90A, 46V.

1A*2V*10ms=20mJ
90A*46V*2us=8mJ

deposited primarily in the semiconductor active
volume (forward junction and access resistances).

But the problem is in the quick-time thermal energy
removal (which is likely nil in the 2us quasi-adiabatic
case and can be a lot of it, via lead conduction and
spreading to adjacent material volumes, in the 10ms
case). That, and the question of the non-silicon
features such as bond wires - where I have seen data
saying no way is 1.25 mil aluminum good for 90A even
at 1u but more than a factor of 10 less (and with so
low a rated current it's unlikely that bond wires used
are anything above minimum (1mil Au or 1.25 mil Al).

Since you cannot indicate what the failure mode is,
you can't say which feature dominates abuse mode
reliability. Don't assume it is the silicon volume or that
all mechanisms are simply linear in time.

I could add that in the time this thread has been simmering
somebody could have acquired samples and pulled data.
I would not depend on internet contributed guesstimates to
replace that.