eddy currents in metal core PCB's

John Larkin a écrit :

There are some nice really fast mosfets, but goes only to 200V.

Probably cascoding one with a kV mosfet will do what you want.

Oh, and to Joerg's delight, DK carries them :-)

--
Thanks,
Fred.

ditional

core

the power

is in

in the

high

like to make

don't seem

would

m? It

e

he heat

harm.

about

which

me as 15

opper to

layers

ter

like a

s 4 with

ork with

the

eat

eed

t

It doesn't seem to be a problem in practice. One gram of water can absorb 4.18 Joule if heated through 1K. The same gram if water can absorb 2,257 Joule if you evporate it - that's 540 times more heat. If you allow your circulating water to heat up by 10K, rather than 1K, the ratio drops to 54 to one.

Heat pipes don't need a rapid flow of water.

Presumably oxygen and nitrogen diffuse into the sealed pipe-work eventually. We went over to the heat pipe because circulating water with a pump wasn't reliable - you eventually got a bubble of air somewhere in the pipe work, and the pump wouldn't be able to get water past it. The service engineers had to check out the machine every six months and purge the air from the system.

Since the heat pipe doesn't have any moving parts (except the water being evaporated and condensed), there isn't a lot to go wrong.

A gallon a minute - eight pounds of water (in your country) - is a lot of water to move. Evaporate a couple of ounces of water per minute and you will absorb the same amount of heat as you would by heating your gallon through 10 degrees Kelvin.

A US gallon is 3.632 litres of water. They'd certainly begrudge throwing away 132 gallons per eight-hour day, and if they don't throw it away they have to pump it up to a cooling tank, which is pipe runs and plumbing. It can be done, but it isn't cheap or convenient.

You'd be more popular if you found a more compact and self-contained solution.

-- Bill Sloman, Nijmegen

That last one is pretty beasty. Ls of that package is around 0.5 nH, according to the Spice model on the first datasheet. That's awfully good: 10 amps per ns only costs 5 volts of gate drive.

The ST part actually has better specs, specifically lower Cout, but the package isn't as nice. I wonder what the source inductance is on a DPAK. I'll try to find out.

IXYS acquired DEI some time ago.

John

additional

l core

of the power

B is in

ts in the

he high

d like to make

s don't seem

o

k would

num? It

uce

the heat

at

y harm.

s about

, which

same as 15

copper to

CU layers

etter

t like a

its

FR4 with

work with

f the

heat

ir

need

0

or

hat

, t I

ck

the wick shouldn't really be needed if you only need it to work in vertical position

it's a refrigerator with out a compressor it should last forever

those who make water heating elements generally say you need to stay under

10W/cm^2

I think it is safe to say that if the have 1MW laser they already have plenty of cooling water

-Lasse

Here's some stuff on Ls...

DPAK datasheets seem to mention Ls like 7.5 nH, sort of high for fast switching. Using several smaller fets in parallel might be better, for DPAKs.

John

1k$ that's it? :)

Jamie

What orientation will the board be in? You might want to turn the switchers through 90 degrees so one doesn't bake the other with convection drafts!

If skyline permits, you could perhaps add a soldered-in heatsink between the switchers to help them dump heat into the air. It could conveniently be formed from another PCB, with 2 ounce copper, soldered into a slot in this board.

If you use vias for thermal transmission, suppress the "tenting" effect many CAD packages assume you want. In other words, there is no gap between the plating round the via, and the copper plane. On some packages, if the option does not seem to be offered, you can force this by plonking a solid region of copper over a via.

the buck at 350kHz (30~90V to 13.5V at 400Watts).

Could you buy an off the shelf mains --> 13.5V supply which fits in this volume? Avoids monster surging currents through those backplane type connectors, all the testing and certification is done for you...

snipped-for-privacy@netfront.net ---

--
> 
> 
> I think your board size (and lousy connector density) is going to
> waste a lot of money, in AL core. Better channel the heat into
> heatsinks.
> 
> RL
Damn, for a moment it looked like "AL gore"

Jamie

I don't know why a IGBT transistor can't be used? they make them to handle very high voltages.

Jamie

power

make

seem

About $120 each. My engineer working on it told me he'd blown up a bunch, and said "if I just run it to 80% of rated voltage, the fet will die. Want to see?"

I declined.

They claim to have worked out the problems.

These parts have stunning Gm/Co ratios, fast as hell, and need small, ECL-type, gate swings to turn them on and off. If another project comes up, I'd be happy to give them another try. The technology is GaN on silicon.

John

There are apparently a number of failure mechanisms. MTBFs in the

15-20 year region are apparently possible if everything is done right.

Moving water is probably better.

That's what I was thinking.

John

Hi,

It goes into an ATX PC case, so depends on the orientation of the case itself, normally it would be horizontal though.

Interesting idea! I guess you could also solder on some SMT cooling headers onto this heatsink too :)

I haven't seen an off the shelf mains --> 13.5V 800watt+ supply that has active PFC for a price less than this board will cost when paired with an ATX PC case and 800watt+ ATX active PFC powersupply. Also I'm adding in programmable output voltage/current and the board also has a buck converter (for a MPPT tracking input) and dual 240VAC/10Amp auto transfer relays. The testing/certification is a whole other issue, but right now I'm just making a prototype.

cheers, Jamie

--- Posted via news://freenews.netfront.net/ - Complaints to snipped-for-privacy@netfront.net

Hi,

I made a test "heatsink PCB" meant to be put on SMT fets, ie. directfets with metal packages:

"

"

Its two layer 2oz or greater copper, and has a SMT dual row pin header on it, the longer the pins, the more cooling area there would be. The bottom of the PCB could be tinned and maybe even soldered to the directFET packages.

The main drawback is the vias are the only significant heat transfer route from the component to the header pins, so the FR4 thickness should be kept to a minimum and lots of vias would be necessary. Could also be made from a metal core PCB I guess.

cheers, Jamie

--- Posted via news://freenews.netfront.net/ - Complaints to snipped-for-privacy@netfront.net

Bill Sloman schrieb:

Hello,

but the density of the water vapour is much lower than the density of liquid water. To shift the same mass of water, you will need pipes with much larger diameter.

Bye

Gaw-Dayum!

How can a man who's so knowledgable about the physics of heat pipes be such a blitherng idiot when it comes to politics?

IOW, how can someone so smart be so stupid?

IOOW, that's exactly how heat pipes work. I even read somewhere that you can make one at home with a piece of copper pipe, a couple of pipe caps, and some copper wool or just plain lamp wick. Solder one of the caps on the pipe, line the pipe with wick, pour some water into it, and heat up the closed end until you see the "steam" come out the open end. Then quickly solder the other cap onto the other end, and Bob's your uncle!

Thanks, Rich

Since Rich's idea of intelligence vis-a-vis politics involves taking Libertarianism seriously, this is less insulting than it appears at first sight. To have Rich endorse you as having intelligent ideas about politics might be damaging, though nobody actually takes Rich serously.

Since Rich's ideas about what is sensible seem to have included drinking and blowing his way through most of his stock of working neurones, this is decidedly ironic.

It really pays to get rid of as much as possible of the non- condensible gases before you seal off the system. I'd be tempted to out-gas the water with a good vacuum pump, then freeze it with liquid nitrogen and pump the assembly down to a hard - milli-torr - vacuum before I sealed off the system, but I always did like over-kill.

-- Bill Sloman, Nijmegen

and

imit.

y

lts

n.

I just saw that they also have a matching driver:

-Lasse

20 amps, 30 volts, 4 ns. Nice.

I have decided that the greatest evil on the planet is probably source inductance.

John

You forget the huge Delta-H of vaporization of the water. Try sticking your hand in front of the spout of a kettle that's on the boil.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058

email: hobbs (atsign) electrooptical (period) net
http://electrooptical.net