Low Rds(on) tempco

Insane. 254 watts times 0.59 K/w is a 150K rise, and that's just the silicon to the case. The DC SOAR curve sure looks like 160 watts max to me.

IR publishes absurd specs on their mosfets, too.

It's hard to find a TO-247 that will dissipate an honest 250 watts.

ftp://jjlarkin.lmi.net/ExFets.jpg

John

Even Lower from TI their new CICLON line.

Full line up is here.

They sent me an email about them. They are nice for LV only 25VDS max; easy drive requirements as well.

It's mainly the lower voltage ones that have the best tempco's. When you get up in higher voltage they are all usually about 1.8 to 2 x

Anything over 50VDS I just go at 2x 25C rdson to be safe.

On a sunny day (Sun, 24 Jan 2010 12:23:13 -0800) it happened John Larkin wrote in :

How cruel :-)

Fig 11 shows 1ADC @30V limit, with Tc @25degC, so obviously something is not quite up to snuff, where data sheet accuracy is concerned here.

RL

Switching-type fets are happiests at low drain voltages, and tend to detonate if they dissipate a lot of power at higher voltages. The corner of the DC SOAR curve is about 160 watts (2 volts, 80 amps) but you're right, that tapers off to 30 watts on the other end. So it's not a constant-power curve.

John

I want to avoid SMT since it's nonobvious to heatsink (I don't have PTH facilities at home), and at 100A, 3mohm is 30W of PCB-melting toastiness. The duh answer, use four in parallel, but then gate charge is back where I started. So I figure heatsinking is best.

I have been more and more tempted to buy SMT though. So far, the only thing I've built with SMT didn't use a PCB:

With solder paste, gobs of flux and a heat gun or toaster oven, I'd probably enjoy it a lot, but I don't quite feel like investing in yet another parts line, made of things I can hardly see and certainly can't tell apart (99% unmarked capacitors. seriously. wtf.).

Mmm, nice price.

I have some IRG4PC50UD's that clock in at some 270nC. They were expensive, too (~$8 each).

Fairchild's Field Stop IGBTs are like popcorn nowadays.

IGBTs rule above 300V. They're not even bad down around 100V:

I put a bunch of those together, way cheaper than any IGBT or FET module you'll find for a motor drive:

Tim

Looks like some alien insect about to jump on my face.

Use all 0.33 uF and that's not a problem. But seriously, it would be nice if the caps were marked somehow, even color dots or something.

Switches about 3 kilowatts per dollar.

John

There are too many good parts out there not to use SMD particularly FET's in the sub 60VDS DPAK. You don't really need any special equipment for most packages. I use a $10 solder iron from Canadian tire and toner transfer method for PCB and I do fine. Just keep track of the parts in labelled coin envelopes.

Agreed on the caps some marking scheme would be nice ;)

Yea for kW SMPS you would be looking at IGBT's they do have some good high speed capable FET's but they are slim pickings and they are pricey.

Usually you don't want to be banging around double digit amps at rectified line voltages at hundreds of kHz anyways.

I picked those FQA24N50 mainly for the price for a buck a pop I could always parallel them and still be ahead $. They show 1.3uS worst case getting eaten up on transitions using a 25ohm signal generator for gate drive. I'll have to test a couple of them with a beefy driver.