Fet (power) specs

Mosfet specmanship. IR rates power dissipation of some of their parts while immersed on a boiling solvent.

A SOT89 can dissipate about 3 watts if soldered to a lot of pcb copper. That SC70 can't.

On a sunny day (Mon, 17 Feb 2014 10:06:12 -0800 (PST)) it happened George Herold wrote in :

I use a IRLZ34N for heating. the surface mount version IRLZ34NS can do 2W, the TO220 version a lot more, and there is also something inbetween IRLZ34NL low profile through hole, like a TO220 without tab. It is logic level, a few volts (2 to 3) will switch it 100 % on. It my standard MOSFET for projects. Even did a switchmode with it.

ter.

Do you know how big (or small) is the sc70 package? I would be surprise if it can handle couple of watts continuously. With proper heat sink and coo ling, the IXFH36N60P can handle over 40A and 600W in TO-220 housing. TO-22

I believe that's a mistake. The 300usec pulse should be for something like Rds(on) where the die temperature from self heating changes the parameter in a direction considered unfavorable by the marketing department.

You can probably actually get 3.3W out of an SC-70 if you could find

marketing).

If you're trying to heat up a PCB, maybe something with a thermal pad or tab would be sensible.

Best regards, Spehro Pefhany

ater.

com

com

Thanks guys, No TO-220 packs! JL they don't make many P-fets in sot89 (at least according to digikey) There are a number in SOT223 and I'm just going to pick one of those. I want to run it in the linear region, rather than switching... is there so mething I should look out for? Something like this looks nice,

Thanks again.. mostly I was just whinning about the power specs.

George H.

I am sure someone with more knowledge than me would jump in. MOSFET are not very linear. They would turn on and off over a narrow gate voltage range. I would use regular bipolar transistors for linear control.

Hi Edward, this is for a heater application (see other threads by me.) Well BJT's aren't very linear either :^) But I'll put the FET inside an opamp loop and feedback to keep the current constant. I'm thinking (but not knowing at all) that the fet will work better at 77K.

George H.

Even in a same-sized package I see significant differences in junction-to- tab thermal conductivity, and significanter differences in junction temperature.

If you need to operate at 85C, a junction temperature of 175C will let you dissipate considerably more power than a junction temperature of 125C.

It all becomes an exercise in studying the data sheets, playing with layouts, running the thermal numbers, and hoping that the manufacturer was just playing spec games and not outright lying when they put down the numbers.

Snsip>

Hi Tim, I've spent most of the morning searching digikey and browsing spec sheets. (One thing that would be nice in DK's search engine is if they could just l ist one instance of a chip, rather than three different times, per piece co st, bulk and cut tape... but I digress) Not everyone lists the junction to lead thermal resistance, but for those t hat do I notice that a higher voltage device (Vce or Vds) tends to have hig her resistance. Going out on a limb I'd guess that higher voltage devices have lower doping concentrations, less electrical conductivity and concurr ently more thermal resistance. (I'd love for some device guy to comment an d/ or correct me.)

George h.

Well, LN2-- 77K, you can get as much as ~10W out of that SC-70, in theory anyway. (448K rise at 45K/W).

Best regards, Spehro Pefhany

You can select only CT, tube and bulk (maybe tray), which gets rid of all that nonsense.

Best regards, Spehro Pefhany

You're assuming that the thermal resistance goes as that of copper between room temperature, right? Because an SC-70 is nowhere near 45 K/W at room temperature--it's more like 300 K/W on a good day.

Cheers

Phil Hobbs

I'm just reading the datasheet. They wouldn't lie, would they?

assuming the foot could be kept right at 77K (attached to a big heatsink immersed in L2).

You'd expect thermal conductivity of Cu to be a bit higher at 77K, but not enormously, and most of the leg will be a lot hotter than 77K.

Not very practical, but getting 10W out of an SC70 isn't easy.

Best regards, Spehro Pefhany

t list one instance of a chip, rather than three different times, per piece cost, bulk and cut tape... but I digress)

Ahh, Great. I thought there might be something. OK is there anyway to kill some of the columns?

George H.

e that do I notice that a higher voltage device (Vce or Vds) tends to have higher resistance. Going out on a limb I'd guess that higher voltage devi ces have lower doping concentrations, less electrical conductivity and conc urrently more thermal resistance. (I'd love for some device guy to comment and/ or correct me.)

com

com

No, but they are not telling the whole story straight.

conducting through the legs.

If the OP insists on SC70 foot print, at least get one with a center heat p ad. Not all SC70 are created equal.

I don't see how you come to that conclusion.

Never seen one with a thermal pad. It would have to be awfully tiny. Got a part number?

The claims for this particular transistor seem to be real outliers.

Part of the reason is that they've gone to a copper leadframe rather than Alloy 42 Nickel-Iron low CTE, which provides a claimed 4:1 improvement in junction-to-foot thermal resistance (that's about the difference in the bulk metals).

One might expect that if you used an Al-core PCB, you could get some serious power dissipation out of that little thing.

Best regards, Spehro Pefhany

That's a leadless package, not an SC-70. Guess I still haven't seen one.

Best regards, Spehro Pefhany

It fits in the SC70 foot print. However, you won't get the full power rating without the corresponding PCB heat pad. If you need to build big enough PCB heat pad, To-220 class package is not so bad (big) after all.