What minimum rating for short protection diode on 3 terminal regulator?

[741]

It is sometimes recommended to protect a 3-terminal regulator from shorts across output or input, which can cause large currents to flow in the device.

We can find the energy in the capacitor, 0.5CV², but what matters is the diode's response profile and rating.

Is there a sensible hand-calculator type way to analyse this? Can LTSpice help if not? (Would you just guess wildly?  )

Example: LM317L, 10u output capacitor, 22V output. Do I need (a) 1N41418 (b) 1N4001 ?

[T3sl4co1l]

You'd need the capacitor's value and ESR (and ESL and strays and other attached bypass, as applicable), and the diode's surge rating for time scales on the order of ESR*C.

The surge is not hard to calculate or estimate by hand, but SPICE can be used if you like.  Helpful for complicated PDN (power distribution network)s.

The diode should also have Vf less than 1-2V under the surge condition, I suppose.  That likely rules out 1N4148.  May also rule out a lot of fast-recovery diodes, which tend to have higher Vf.

Ideally, you would check against the short-duration surge ratings for the diode.  These are rarely given, let alone at the arbitrary duration and waveform a particular circuit would give.

In the absence of such ratings, it's not obvious whether a diode will survive such abuse (though it likely will).

The Vf constraint is probably a good start, and current, surge, energy or power ratings may not necessarily be exceeded if that one is met.  Point is, there are multiple dimensions to constrain here, which cover slightly different directions, and which may have some overlap, but not total.  YMMV.

It is probably okay to assume a diode of comparable size (i.e., a 1A diode like 1N4001 to complement a ~1A regulator) is adequate.

You could make a power supply shorting jig and test reliability, I suppose.  If it passes 1k or 10k or however many cycles, at high and low temperatures, with the maximum expected load capacitance, it's probably fine.

(Note that a failure condition includes the diode or regulator going short or open.  Who knows!)

Tim

[duak]

741, All the LM317 app notes I've ever seen use a 1N4001 rectifier.  I don't know what 3 term regulator you are using.

The attached data sheet shows that IFSM is 30 A peak for a 8.3 ms half cycle sine wave.  What is your peak current and duration?

One parameter that isn't shown is the forward recovery time, ie. the turn on time of the rectifier.  I understand that even slow rectifier diodes turn on quite quickly because the applied voltage just rises until the diode conducts.  I'll bet it's in the tens of ns range and probably limited by circuit parasitics.  The reverse recovery time is long but that doesn't matter here.

The fusing (shorting or destruction) rating, I^2t is 3.7 A^2s.  This value has units of energy or Joules.  A 10u cap charged to 35 V stores 6.1 mJ.  Unless the repetition rate is in the hundreds of Hz, I think the rectifier can handle the energy safely.

[741]

Thank you. Looking again, I realise the TI LM317L DS does show a 1N4002, but this question was intended generically. I still need to look carefully at that diode DS (I had to look up "IFSM").

Regarding the the ESR mentioned earlier, that surely is a factor, although it might be the diode's slope resistance dominates (?)

In my experience, a capacitor of 10u..100u which is not advertised as "low ESR" will have about 0.5...1.5R ESR.

I do not really know how to sketch the discharge height/width in 'back-of-envelope' form, given a few diode parameters and assuming a given lowest ESR.

[T3sl4co1l]

I would be impressed to see a power diode drop more than maybe 3 or 6 volts, for more than a fraction of a microsecond (or for less than a dozen milliseconds, if you know what I mean).  But that would probably be a dangerous voltage drop for the regulator, hence the concern for Vf.

And that would be the case even for surge currents 100A+, which is potentially in the range we're looking at here.

A 30A surge rating at 8.3ms implies at least as much for shorter durations, and we can find Vf at those currents from the plot (if it goes that high, at least).  If your output is, say, 5V and has 100uF of 0.1Ω ESR on it, then expect a worst case of 5 / 0.1 = 50A peak for a span on the order of 100u * 0.1 = 10us.

Note the fusing rating is not energy, because the voltage drop or resistance is undefined.  This makes sense, as a diode or fuse is a series element, and does not care what the total supply voltage is in the circuit.  All that matters is the device's internal resistance (how much the diode can withstand without failure, and how much the fuse can withstand while guaranteeing failure).

This is also why fuses can let through tremendous energy, while absorbing comparatively little themselves.  A fast automotive fuse (quickly clearing 12 or 28V say) might be crowbarred with a modest SCR (let-thru: < 1J say), while a 250V cartridge fuse, subject to swell or surge, will easily destroy a MOV in the process (let-thru >300J?).  High voltage distribution fuses can let through enough energy to physically destroy equipment (arc flash)!

Tim

[duak]

Tim is correct regarding the I squared t rating.  I looked up the term and it's something to the effect of "...is a measure of the energy..."  If my quick back of the envelope derivation is correct, adding an R makes it dimensionally correct.  ie., I^2tR is in Joules.

[741]

Hmm...
E(J) = Q.V = (I.t).V = (I.t).(I.R) = I²tR