Clamping of the inductive spikes

At these speeds, PN diodes will display "forward recovery", the time delay before they start to conduct. That will look a lot like series inductance. I've driven fairly ordinary diodes at +50 forward bias, and they take tens of ns to start conducting serious current.

What's the voltage involved here? You could maybe use a number of paralleled schottky diodes, or SiC schottkies for higher voltages.

Your dI/dT into a fast diode would be on the order of 100 A/ns, so you'll need ballpark 1 nH to keep the spike down. You might parallel a bunch of small power schottkies on a low-inductance PC board. I doubt that any single diode will give you the sub-nH parasitic L you want.

You could easily have many nH parasitics in pcb traces and wiring. A simulation would be interesting, to see how the voltage slew interacts with the diode turn-on. This isn't simple.

John

You are using surface mount diodes?

Jim

-- "It is the mark of an educated mind to be able to entertain a thought without accepting it." --Aristotle

...

Hope This Helps! Rich

You forgot to mention what an acceptable voltage overshoot/duration was, and where this was needed. Obviously you're not trying to protect the choke - is there a particular switch or diode you're worried about? Some are less finicky than others, and providing that the exposed area of parts carrying the transient is small, the vertical capacitive radiation effects need not be overwhelming.

Forward overvoltage, as mentioned, is characteristic of most non-schottky diodes, but will also occur across a bare wire's stray inductance - 1nH is sufficient in this case, as pointed out. The measured stray inductance value of of typical SMD TO220 ( TO252 SC63 SOT428 DPAK LFPAK ), between lead and substrate mounting plate is, in fact, 1.1nH. Most of this will occur in the anode lead (reduced in some vendors versions). SMD capacitors can exhibit fairly low stray inductance, by comparison.

Forward overvoltage can be reduced by intentionally forward biasing the conventional rectifier at light current, if physically possible (ie where capacitively coupled to the transient, as in a current snubber). Larger diodes tend to have reduced overvoltage for the same current transient.

The dV/dT measured could also occur across an ideal 200nF capacitor - chances are there isn't such a beast present, and that adding a local current snubber could have minimal detrimental effect on switching efficiency, in this case. A conventional schottky can be expected to exhibit >200pF for every 10A continuous current rating, below 15V reverse bias - but this is still decoupled by the lead inductance.

Inductance can be reduced in layout by paralleling parts on uncorrelated (or better still - intentionally canceling) loops of current flow. Anything reducing the flux intensity around the wiring structure will result in reduced inductive effects.

Note that suppression should be applied to the intended node - it will tend to be less effective for any distant point in the original 100A branch path.

dV/dT measurements can be crippled by scope probe capability and measurement technique. Make sure your equipment can actually see incremental improvements ( possibly by making it intentionally worse...?) first. Watch the ground contact and confirm a quiet baseline in the live measurement set-up. Sometimes non-contact coupling into a spectrum analyser can be a useful incremental indicator of alterations in branch functional characteristics.

RL

Thanks for the detailed reply, John. Your observations are quite right, and the simulation shows the similar behavior if the parasitics is taken into the account. Unfortunately, they usually don't provide the turn-on related parameters in the datasheets. So, it is not very obvious which diode has better performance. Finally I found that the B360 shottky works satisfactory for this application.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

We use lots of the B360's, mostly as switcher catch diodes and reverse protection. You could probably lay out a small pc board with a solid plane on each side, a bunch of B360's in parallel, and a few clever via/pour connections to the diode pads. That could be a very low-L structure, and would keep the diodes cool too, if that mattered.

Did you determine the effective inductance of a B360? That would be layout dependent, of course. Maybe I can TDR one.

John

According to what I measured at the pads, the effective inductance of a sigle B360 over the ground plane should be at the order of 2.5..3nH. The typical value for a power diode seems to be about 5 to 10nH. That can result in the turn on time as long as 100ns or so.

Vladimir Vassilevsky DSP and Mixed Signal Design Consultant

In simple terms, think very carefully about why and how you place each increment of capacitance, resistance, inductance, and transient energy / voltage deflection / absorption. Which breakdowns where are at issue.

As a regular here says: Engineering is making what you need out of what you have.