IR is insane

It is also a good media, if chilled, for cooling a PC for the overclocker crowd.

I would like to put a PC motherboard in a fish tank full of it to freak people out with.

*Every* time I've seen it used there was a phase change somewhere in the process. There are a *lot* of cheaper and better alternatives if a lowish temper phase change isn't needed. Fluorinert doesn't have a stellar specific heat.

--
For IR, about 4K/W.

I got into a mood to try the math on this.

Back to the datasheet,

My eyeball-estimate of the length of the source lead, between the solder joint and the bonding wire, is 8 mm, about half with a narrow portion (datasheet says .51 by .38 mm minimum), and about half with a wide portion (datasheet says 1.14 by .38 mm minimum).

Wikipedia says the resistivity of copper at 20 C is 1.72E-8 ohm-meter, which is 1.72E-5 ohm-mm. I would like a 75 C figure, which is 19.5% higher, and that is 2.055E-5 ohm-mm. From here, I get .425 milliohm for the narrow half and .19 milliohm for the wide half, for a total of .515 milliohm.

.000515 times square of 195 amps is 19.6 watts. That sounds unreasonable to me. I seem to think that the narrow part of the source lead will melt.

Try for the wide half alone: .00019 ohm times square of 195 amps is 7.2 watts. Repeat for 170 amps, and that is 5.5 watts. That sounds to me able to be heatsunk from the source lead by soldering big fat wire to the wide part of the source lead, but it also sounds to me sort of insane to package such a big MOSFET with such requirements in such a small package.

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

I certainly might, though companies that use specsmanship that radically=20 do not inspire my trust. Plus i hate fighting through datasheet = gibberish=20 to get realistic properties; which gibberish, when detected, will almost=20 always make me change vendors. I do not want to spend the time to get=20 past obvious baloney to find out what the part can do in normal use.

magic?

once was=20

baths.

freak

=46ish tank, maybe $250, chiller (repurposed small refrigerator) maybe = $200=20

15 gallons Flourinert $600. Will you get $1050 plus time modifying the=20 refrigerator of entertainment out of it? Hiding all the mods and holding= =20 the Flourinert at 3 C or less is easily and extra $300 in materials. If=20 you are wiling to put some effort in moisture protection for the tank,=20 you could drop the fluid temperature to about -40 C and OC the hell out=20 the computer.

magic?

once was=20

baths.

lowish

specific

At which point i have to ask, was it Flourinert or just some refrigerant. See:

=20

of=20

solder=20

portion=20

=20

=20

7.2=20

the=20

=20

package.

As if you have not caught on yet, the ratings are for big heat sink in a=20 tank full of LN2 (liquid nitrogen). I doubt you need to review "infinite= =20 heat sink".

The idiot was thinking of a CFC.

Fluorinert. FC-86, was used in the project I'm most familiar with. It was chosen for its boiling point for a direct-chip nucleate cooling system (10W/chip 121 chips per 5" module, IIRC). The project was eventually scrapped because it was impossible to get Fluorinert pure enough. Boiling = distillation and all the crap got left on the chips which eventually poisoned them ("black plague").

AlwaysWrong is *ALWAYS* wrong.

was

You've gotta admire him for his dedication.

John

was

baths.

specific

You certainly did a good job of naming.

--
I don't disagree, But in their already cited application note IR
addressed the "specsmanship" issue rather candidly, I thought, and also
gave their reasons for using their approach to determining the
balls-to-the-wall capabilities of the MOSFETs tested that way.

Being aware of that, all the data required to operate the devices in
more conventional ways is presented in the data sheet, so I don't see
their "SuperSpec" being a problem, especially since the application note
is referenced in the data sheet.

Plus, I like IR. :-)

JF

How about using the 100 or 125 C relevant junction temperature figure of Rds(on), preferably the 125C one that I have previously cited (and have difficulty remembering better than 4.2 milliohms or whatever), and considering that such case temperature (close enough to junction temperature in this case) is good for about 3-4 watts in a reasonable FR4 PCB design with 1 square inch of heatsinking copper on both sides of a

2-layer PCB with a "reasonable" wide-spreading heatsinking copper layout with reasonable deployment of themal vias.

At this moment, I am getting into a mood to calculate 4 watts from 4 milli-ohms. That is amps being square root of 1,000, as in 31.6. Although I am in a good mood to approve 35-36 amps for a fairly aggressive version of a FR4/"conventional" PCB with wide copper heatsinking area and

*heavy* deployment of appropriate thermal vias. I give a slight chance that 40-50 amps can be designed for reasonably on such a scheme.

With metal core PCB, I don't have a problem with 80 to possibly 100-plus amps as long as *everything gets worked out*. I seem to think maximum of ballpark of 70-80 amps RMS even on MCPCB unless "beefing up" the high current path every couple millimeters of the way, including the narrow portion of the "source lead" of the MOSFET. I would hope that 4-5 mm trace width on MCPCB is sufficient for survivability and conductivity and sufficiently low voltage drop when conducting near or over 50 amps!

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