HV transistor heatsinking / High Voltage / Forums

Forums

4hv.org :: Forums :: High Voltage

« Previous topic | Next topic »

HV transistor heatsinking

previous 1 2 3 next

Move Thread

LAN_403

Herr Zapp

Wed Nov 18 2009, 07:41AM

Registered Member #480 Joined: Thu Jul 06 2006, 07:08PM
Location: North America
Posts: 644

HH -

I'd also recommend the multiple heatsink approach. Certainly a bit more mechanical complication, but I think in the end it would be more reliable as far as HV isolation.

Will the cooling air be filtered, or is this just ambient air supplied by the rotor system, or ram air? (Free-flight vehicle, or tethered?) You might want to increase your creepage distance if this thing will be stirring up a lot of dust and then directing that dust-laden air across your power semiconductors.

I would NOT recommend the thermal epoxy for a high-vibration environment, at least not without some supplementary mechanical hold-down. Look at the coefficients of thermal expansion for your JFET package, the aluminum nitride insulator, and the actual heatsink (aluminum alloy?).

Regards,
Herr Zapp

GeordieBoy

Wed Nov 18 2009, 03:32PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

That sounds like quite a project you've got there!

I'd go for the aluminium nitride slabs but I wouldn't glue the devices. Just coat both sides of the AlNi slab with a thin coating of heatsink compound then clamp the devices down such that you are applying pressure individually to each device at a point above where the die is situated in the package. That will give the best thermal contact. You can often get seperate spring-steel clips to push down on each device. Or put a metal bar across them all, with little pieces of rubber exerting a force from the bar onto each device above the die.

The relatively think AlNi sheet should also help reduce EMI due to the device's backplate voltages slewing thround 5kV in a very short time! The high dv/dt is apt to drive large common-mode current spikes into the heatsinks, and ultimately back into the ground wiring if you're unlucky!

Another thing that you might consider is whether the system has to withstand something like a locked-rotor test without blowing up. When you mount devices onto those aluminium nitride sheets you are relying on the local thermal mass of the device's own tab (also known as backplate or heat-spreader) to absorb the heat generated by sudden overloads. So when you do something like a locked-rotor test it's possible for the device tabs to overheat quickly before the heat has had chance to be conducted through the thermal paste, aluminium nitride sheet, thermal paste again and finally into the main heatsink!!!

If this turns out to be a problem, I would recommend putting slabs of metal directly underneath each device to add thermal mass. You don't need much material to absorb the heat and allow it to be conducted more leisurely through the AlNi sheets and into the main heatsink. These additional heat-spreaders not only provide thermal inertia for the devices, but can also increase the surface area available for thermal conduction through the electrically insulating AlNi sheet. This is what I have done previously to improve transient overload withstand capability where devices cannot be clamped straight down to a common heatsink.

I hope this info helps.

-Richie,

Proud Mary

Wed Nov 18 2009, 08:33PM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

I'd very much go with Richie here, but will add a jot of commercial engineering to the mix.

What would a perfect heatsink be? An infinitely large silver mass?

So in a sense the commercial manufacturer chops down this infinite mass bit by bit, creates cooling fins &c, until the necessary degree of cooling is available at a certain price.

Convection cooling - as with fins - makes commericial sense, but the hobbyist is not bound by the economics of mass manufacture. She can bolt her semiconductors to a solid aluminium block, and through conduction can dissipate heat hundreds of times faster than any convection system.

As an aside and afterthought, black paint applied to convection-type finned heat sinks actually reduces their capacity to dissipate heat,
the paint acting as thermal insulation. An effective 'black' heatsink is made by first anodizing the part, dying it black (black Dylon will do) and then closing the nano-pores by further simple chemistry.

Stella

Henry H

Wed Nov 18 2009, 10:48PM

Registered Member #2298 Joined: Sat Aug 15 2009, 08:16PM
Location: ex UK, now Santa Cruz, CA
Posts: 35

Thanks very much for the advice everyone. I feared the epoxy approach would be too good to be true. I'll see if I can figure out a good clamping method. I had earlier looked at using heat spreaders and decided against due to the extra space they took up, however that was before I'd discovered how big the heatsink needs to be anyway, so it's less of a penalty now. The JFETs just arrived today, hopefully I'll have time to test them soon..

Sulaiman

Thu Nov 19 2009, 05:46AM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3141

Just came across this thread,
I have opened up failed thyristor blocks (similar

they have some kind of ceramic sheet bonded to the aluminium,
all of the components can be scraped off with a blade
and new ones of my choice glued on with epoxy resin,
I make a surrounding with bits of plastic & fill with the resin,
I end up with a block of resin encapsulating/holding everything.

Higher temperature adhesive would be required for a profesional job

At 60W each I'd go for one heatsink per device with no insulation
and mount the heatsinks on FR4 fiberglass sheet with screws.
For experimental projects I find several small modules more serviceable than one large one, usually prototypes need easy access!

Proud Mary

Thu Nov 19 2009, 08:47AM

Registered Member #543 Joined: Tue Feb 20 2007, 04:26PM
Location: UK
Posts: 4992

With your proposed 30kV iteration, I think it would be a serious mistake not to submerge the entire HV circuitry in a good dielectric oil,
(such as sunflower seed) which will also act as an effective heat-exchange medium.

Vegetable dielectrics have a life-expectancy of several years continuous service before coronal degradation products, and atmospheric water absorption, significantly degrade the dielectric strength.

I would call any 30kV circuit not immersed in oil a hissing, fizzing liability. If you can smell any ozone at all, it means that the device has already set out on the path to destruction, since the nascent oxygen will start to attack most polymers immediately, with UV-rich micro-coronas accelerating the degradation until carbon-tracking starts to occur.

Finally, and at the risk of stating the obvious, it is always good practice to have either a dry schnorkel vent in the outer shell, or to put a handful of compressable polymer into the oil, so the casing doesn't begin to leak when the oil warms up and expands, a surprisingly common blunder.

As for insulators, I will sometimes cut down cheap Be TO-3 types, and drill the appropriate holes in it with diamond. As others have pointed out, oxide of beryl is highly noxious, so it is prudent to wear a gas mask when drilling, and do whatever is necessary to prevent
the spread of the dust.

For rough work, TO-3 insulators can often be used without any cutting at all, since problems arise when an insulator is too small,
rather than when it is too large.*

I use ordinary Electrolube heat sink comound with good results.

*Edit added: I intended that a TO-3 sized insulator can often be used with devices of a smaller footprint with a little improvisation.

Sulaiman

Thu Nov 19 2009, 01:22PM

Registered Member #162 Joined: Mon Feb 13 2006, 10:25AM
Location: United Kingdom
Posts: 3141

Just thought, if each transistor needs it's own snubber and/or trigger with the associated inter-stage voltages, you must consider that as a part of the overall physical layout/construction.

MinorityCarrier

Fri Nov 20 2009, 12:22AM

Registered Member #2123 Joined: Sat May 16 2009, 03:10AM
Location: Bend, Oregon
Posts: 312

Another thing about SiC is that the junction temperature runs quite a bit higher than Silicon, 250C or higher. (SiC die attach metallurgy is a major area of research right now). Your devices are likely to run hotter than what you are used to. If you choose to use thermal grease, look for a type intended for higher than typical temperatures.

wylie

Fri Nov 20 2009, 03:15AM

Registered Member #882 Joined: Sat Jul 07 2007, 04:32AM
Location:
Posts: 103

As an aside and afterthought, black paint applied to convection-type finned heat sinks actually reduces their capacity to dissipate heat

I have a pile of 1" W x 2" H TO220 heat sinks from some SGI Indigo PSUs. They're all painted black....methinks SGI got ripped off by their PSU supplier! Not the kind of mistake i'd like to see in MY really expensive workstation. Luckily i got 4 of em for $200, and after putting all the working parts together and realizing i'd need to solder on a new battery in order to boot...well, i had 4 nice PSU's to take apart. Lots of caps, toroid chokes, and some fun irf810 and irf730 that i totally destroyed (not before getting some results though, hence the fun).

Those purple cases make beautiful replacements for the blocks in cinderblock+boards furniture. Plus a nice stack of big gold plated PGA chips. Should i start a thread about getting that gold off, anyone got some ideas?

(Sorry for all the thread jacking; well not really, but doesn't seem right to start a new thread just to talk about my painted-black heatsinks and where i got them, when i was replying to Harry's insights above about that problem.)

Henry H

Fri Nov 20 2009, 03:52AM

Registered Member #2298 Joined: Sat Aug 15 2009, 08:16PM
Location: ex UK, now Santa Cruz, CA
Posts: 35

I think Mary was pointing out that it's a bad idea to paint a bare-metal heatsink black, not that black heatsinks in general are bad.

WRT junction temperature the speced max for my SiC JFETs is 175 C. Still a bit higher than typical silicon devices but not up at 250C levels. Junction-to-case thermal resistance is 1.1 W/K so at 60 watts the case will need to be at 120C, within the range of normal thermal compounds.

Got the thermal epoxy today. Still waiting on the AlN insulators.

previous 1 2 3 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob