TO220--thin leads--high current--How?

I've never been "bitten" by IR specs, because they looked absurd to me at first glance.

Do I sound shy to you?

John

What package are you using? At how much continuous current?

Soldering or clamping the leads to something massive, very close to the plastic body, will obviously help keep the leads from fusing.

John

With decent heatsinking, it appears to me that power MOSFETs should reliably take more than half the true maximum current indicated by careful reading of the datasheet (continuous silicon limit or continuous package limit, whichever is lower). I usually think 75%, though I think more like

50-60% for the extreme example that John Larkin mentioned. Given the extreme current figures being mentioned lately, I would like to say 60% instead of 75% when the numbers are eye-catching big ones.

In the above example, I would say 60% of 160A, or 96A.

Then again, I am having trouble imagining a D2PAK's leads handling 96 amps continuous without some significant heatsinking or parallelling with additional copper. 80 amps sounds like a need to solder thick copper wire to the leads close to the package. The heatsink needed will probably be very substantial, with power dssipation likely in the 10's of watts.

One more thing about MOSFETs, especially used for switching: I would design for maximum junction temperature of 125C. Consider the high positive temperature coefficient of Rds(on). Maximum current to avoid exceeding the maximum rated junction temperature is only a little more than the maximum current to avoid exceeding 125 C in any given usual heatsinking arrangement and ambient temperature.

- Don Klipstein ( snipped-for-privacy@misty.com)

Yup, I always design for double Rds(on) too. That covers thermal and manufacturing variability.

Funny how datasheets seem to over/understate things by almost exactly a factor of two!

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

The sil-pad folks use a factor of 2 or 3 for theta.

John

I would consider that sufficient when designing for 125 C maximum junction temperature. (I have noticed how "they" like to state typical and not worst-case variation of Rds(on) with junction temperature.)

That means beyond doubling Rds(on), allow for the worst ambient temperature where you would rather avoid getting a phone call for failing to handle as opposed to saying that the customer was wrong, and include sufficient heatsinking for junction temperature to not exceed

125 C (or at least neither exceed 125 C by a lot nor spend a lot of time exceeding 125 C).

I probably said enough at this late hour in the day...

--
 - Don Klipstein (don@misty.com)

My father told me many times that he likes to divide (or multiply if appropriately making the story worse) datasheet figures for power handling, and all figures said by salesmen, by pi. (I have a preference over pi in favor of 3.2, 50/16, or square root of 10, whichever of these makes easier for me to do calculations / "calculations" with my "Brainiac" personal calculator that is with me everywhere I go when I am awake.

As for ambient temperature - I seem to think that Newark NJ USA gets some "hot flashes" and even now has some of its population lacking A/C, meaning 40 degrees C in homes to deal with occaisionally, even when not on top-floor of a multi-story residential building. The record high in Phoenix AZ USA appears to me to be about 50 degrees C, and even-lower elevations in S and SW AZ and SE CA get a few degrees hotter still for alltime high temperature record.

Automotive electronics appear to me to have extra need to survive heat, although it may be "somewhat unreasonable" for someone to drive a sun-baked hot car in a hot area on a hot day with A/C malfunctioning and the car's interior temperature in the 60's C (with electronics actually operating).

Probably "enough said" at this late hour...

--
 - Don Klipstein (don@misty.com)

t

of.

eed

There is some poorly documented impedance between the gate leg and the actual controlling gate element.

The reverse transfer capacitance is not a constant with voltage or time. It conspires to stop the gate electrode from moving just at exactly the time when the conditions are the worst from a power point of view.

The body diode is lossy and slowish. If the body diode is being used even if it is just to pass a small spike from ringing, the losses are increased.

Body diodes can also snap on reverse recovery, which can blow out gates.

John

In other words, gate spreading resistance. And maybe some inductance, and other squirrelies, depending on how detailed you want to get. More important also is the source lead inductance (which is sometimes specified at 5nH or so).

I get the impression that spreading resistance has dropped over the years. Back in the day, you didn't see FETs spec'd for very impressive rise/fall times. But they also used fairly large Rg's, so is it just that they didn't try?

Cgd is fun.

DCM PFC circuit, somewhere in the middle of a half cycle (well, 4.3ms in, about the middle of a 8.33ms half cycle), 10 ohm Rg, transistor FDH44N50.

When the inductor current drops to zero, voltage sinks down from +boost (410V) to Vin (maybe 160V at this point). So the gate voltage drops by about 0.2V. Then the PFC chip turns on, current rises, then off it goes. (Ringing is turn-on trash from the catch diode, which needs a snubber.)

Tim

--
Deep Friar: a very philosophical monk.
Website: http://webpages.charter.net/dawill/tmoranwms

I generally assume that mosfets are infinitely fast; you just have to drive the gates hard enough. You can get fets to switch in a fraction of the datasheet values if you whack them hard enough.

Here's a 100 volt pulse into 50 ohms, transformer coupled.

The output stage uses a couple of BSS123s, SOT23 fets which literally cost 2 cents each. We're actually getting under 1 ns rise and fall at

100 volts.

How can anybody sell a mosfet for 2 cents?

John

--
8nm prop delay?

JF

Yeah... they transpose units of measure to keep folks on their toes.

Good catch. I'll have it fixed.

John

But aren't those pulses pretty?

John

How did you do it? It does look nice.

Tom

and

d .

idn't

Way back when there was little point in going below a 20 Ohm gate driver. Today that is very much no longer the case.

The funny thing is that the Siliconix parts of the past had low gate resistance can worked as very nice class D RF stages at 16MHz. I could get a couple of watts out of a VN88 with good efficiency.

I've had it eat a driver stage for lunch

a=20

it's=20

In the mean time they can be interesting high rpm fractional hp VFDs. 1/2 hp at 24 V is non-trivial otherwise.

IF you can bother to read the datasheet and app notes correctly they use an _infinite heat sink_ powered by large flow of LN2. They know the spec is bullshit, but the marketeers (who will never know anything of reality) insist on the specmanship games.