"Big Bad" the CW coil / Projects / Forums

Forums

4hv.org :: Forums :: Projects

« Previous topic | Next topic »

"Big Bad" the CW coil

previous 1 2 3 4 next

Move Thread

LAN_403

...

Sat Sept 04 2010, 06:01PM

Registered Member #56 Joined: Thu Feb 09 2006, 05:02AM
Location: Southern Califorina, USA
Posts: 2445

The die in the ixfx55n50 is about as big as you can get period (at least for a mosfet, the bipolars come in the puck package now), you can get slightly better thermal contact by going to something in a brick package, but usually the benefit from a brick type package is better electrical contacts (as you don't have those tiny leads to deal with) which is something you aren't really having problems with. The solution is to parallel more of them together, the IGBT crowd does this extensively (inside the large brick modules there are a number of normal to-247 sized dies soldered to the same heat spreader), but it is trickier due to the positive temp coefficient associated with MOSFETS.
There is the super-20347 package you noted, but all this gives is a large heat spreader, so as long as you have good thermal contact between your mosfet and the heatsink, it won't help you much. You might want to confirm that you are in fact getting a good thermal bond, it is very possible to have a 50C+ temp drop from a mosfet to heatsink if you screw up the bond. However with the relatively small heat sinks you have I think a layer of artic silver won't be hurting you too much.

I think that at the power levels you are running for this coil, the IGBT will really win out in the end, at 200KHz they do become quite attractive. The good ones are pricy, but something like a IXGK60N60C2D1 should kick the mosfets where it hurts.

Marko

Sat Sept 04 2010, 07:37PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

I realize that for greater power levels, the only sane solution is likely going to be usage of bricks or paralleling devices myself. I'm still optimistic on this one though, and especially with igbt's. I have some large bricks around, but they are just incompatibile with this project - they's basically require complete rebuild of the coil.

I measured the frequency the coil resonates at with small spark, and it's 145kHz - which is likely going to get even lower, perhaps as low as 100kHz with full spark load!

Looking at the ixys igbt's from your post, they at first seemed to suffer higher saturation voltage than 30N60's, but examining the graphs showed otherwise.

I'd love to put some of those puppies into this project - they also seem to come in isoplus package which can come handy for lower power projects.
In short, *I really want them!* As usual digikey has them out-of-stock, and I only seem to find diodeless igbt's on ebay.
Still I always preffered uninsulated devices, because they have only like half J-C resistance of insulated ones. Despite my heatsinks look small, I believe my design thermally is better than any single giant freudian heatsink with ISOTOP devices on it.

I'm definitely considering implanting some NTC chips into the IGBT's as well - this way I'll not only have a safety measure, but I also think I'll learn a lot about all things that affect temperature of the devices - what helps, and what doesn't (increasing gate drive voltage, frequency, thickness of thermal paste layer, fan speeds...)

Marko

Kizmo

Sun Sept 05 2010, 09:02AM

Registered Member #599 Joined: Thu Mar 22 2007, 07:40PM
Location: Northern Finland, Rovaniemi
Posts: 624

CW coils

I personally like these because there is so much metal melting power packed in short plasma flame :)
Way more interesting than interrupted coils.. turn up the variac and see lights dimming

Got one with IRFP460 H-bridge

About 300V heavily filtered DC to bridge and 7 turn primary. Resonates at 300kHz. Secondary and primary wiring were almost smoking after that 20s run at ~4kW

EDIT:
Incomplete irfp460 bridge

Marko

Sun Sept 05 2010, 07:03PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hi kizmo, nice to have you in the club!

Any more pics and details of your coil? I'd love to see a thread about it.

Insulated wire for the primary and PVC tube forms for secondary won't make it for long in CW operation. I built my own fiberglass tube to use for a secondary form, and copper tubing was necessary for the primary coil. Having some air blow around the base of your secondary and over the primary is of great help too -
my idea was to enclose the upper level of my coil in plexi so that all the air blown through the ehatsinks is also blown over the secondary base. This would probably cause some pressure increase and decrease the efficiency of fans, and I still haven't come to getting neither the primary nor seondary so hot actually!

Your bridge design is curious - I could never find a source of sil-pad sheet suitable for this purpose, and CPU heatsinks were cheap and I thought thermal resistance would be the lowest with direct contact to the heatsink anyway.

Marko

GeordieBoy

Tue Sept 07 2010, 02:45PM

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

Ooooh, CW SSTCs! Me likes!

Nice work Marko. It looks good. For increasing the power level i'd make the move to either larger die MOSFETs or parallel up a couple of something more modern than IRFP460's for more current. I found that IGBT's don't fair so well in prolonged CW duty unless they are the latest generation ones with heavily controlled current tailing.

You can get away with quite slow IGBTs in a DRSSTC where the operating frequency is quite low, and the devices can turn off softly as the current goes through zero. The small duty ratio of DRSSTC operation also massively helps keeping average switching losses under control!

In a CW SSTC with link coupled primary there's always quite a lot of inductive current flow from the inverter ("magnetising current") which means that the inverter won't switch at zero current. Even if it switches in phase with the resonator base current it will still have to interrupt at least a few amps of free-wheeling inductive current from the primary winding itself. MOSFETs love inductive loads because it minimises their turn-on losses and causes soft-recovery of the free-wheel diodes. IGBT's don't like it as much because interrupting the relatively large magnetising current causes current tailing and comparatively high turn-off losses.

Kizmo, Nice heat spreaders. Are you going to mount several MOSFETs on each one connected in parallel? It's nice to see someone paying a lot of attention to the thermal management side of the layout. Just remember to keep the devices as close together as you can to minimise stray inductance too. Particularly in the MOSFET, free-wheel diode and supply capacitor loop. This is where stray inductance is most damaging.

Marko, The "sil-pad sheet" is called Warth sheet here in UK. Made by a company called Warth Industries I think? You can also get ceramic material too that has good thermal properties and electrical insulation.

-Richie,

Marko

Tue Sept 07 2010, 05:02PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

Hi Richie,

Yes, I'm aware the IGBT's are going to suffer higher switching losses than mosfets, but I still believe they are going to beat them by far in conduction losses. I'm on college again so this project is suspended for a while.

The main reason I can't put more devices in parallel is lack of space on my heatsinks - I could probably fit two devices if I didn't drill those holes, but still I'd like to parallel at least like 4 or 6 devices to be sure, and cost of such a setup would be prohibitive.

Other problem with mosfets is that there just doesn't seem to be mosfets with significantly bigger die - 55N50's, and that's about it.

The IGBT's I'm using are HGTG30N60's - according to datasheets they actually seem to be faster than those big mosfets, especially when one considers fall time which is critical because of turning off large current...

IGBT's :

Mosfets:

Despite all this, I still want to remain careful, as these igbt's are hard to get and I don't want to blow them - I'll up the power in steps and measure the temperature of the devices during run.

One other thing I was thinking about, was to use series resonant primary circuit - like of a DRSSTC, but much higher impedance, with a primary of many turns and several nF of doorknob caps for the cap. This would yield two advantages:

- Good zero current switching on the igbt's,
- Possibility to lower the primary significantly to decrease the probability of flashover to it.

The biggest hurdle though is building the primary coil of required high inductance in a practical volume - due to very large currents involved, simple windings of insulated wire would pretty surely just melt, while copper pipe in required amount would be both expensive and very problematic to mount in small enough volume. (I may need several tens of turns!).

Marko

ScotchTapeLord

Tue Sept 07 2010, 06:42PM

Registered Member #1875 Joined: Sun Dec 21 2008, 06:36PM
Location:
Posts: 635

CW with a resonant capacitor... Are you going to let the current ultimately be limited by the ohmic resistance? That's a lot of heat and some serious voltage across your resonant capacitors...

I had an idea to drive a sstc with this topology:

which drives the primary at resonance but without resonant voltage rise.
But on the thread I started, people said that it wouldn't improve things much.

Perhaps it would make a difference with the power levels you are using, though.

Marko

Tue Sept 07 2010, 07:25PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

ScotchTapeLord wrote ...

CW with a resonant capacitor... Are you going to let the current ultimately be limited by the ohmic resistance? That's a lot of heat and some serious voltage across your resonant capacitors...

Nah, the current is ultimately going to be limited by loading by the secondary. Though in order to keep the current within the safe area, a much higher impedance tank circuit would be required than in a normal DRSSTC.

I had an idea to drive a sstc with this topology: which drives the primary at resonance but without resonant voltage rise.
But on the thread I started, people said that it wouldn't improve things much.

I don't know how would LSR topology work on a SSTC, I've never hears of such a proposal.

Another possibility would be to use LCLR (another commonly used in induction heaters) topology, although I'm not sure how that would work.

teravolt

Wed Sept 08 2010, 04:58AM

Registered Member #195 Joined: Fri Feb 17 2006, 08:27PM
Location: Berkeley, ca.
Posts: 1111

hi Marko, maby a larger primary with more turns and higher coupling say .5 would help. aslo water cooled fets. Steve Ward uses this aproch in his QCW coil.

have you thought of using minibricks?

That is a tough project with the heating and power diapation but very cool. The spark looks like a class-E type spark

Marko

Wed Sept 08 2010, 12:06PM

Registered Member #89 Joined: Thu Feb 09 2006, 02:40PM
Location: Zadar, Croatia
Posts: 3145

teravolt wrote ...

hi Marko, maby a larger primary with more turns and higher coupling say .5 would help. aslo water cooled fets. Steve Ward uses this aproch in his QCW coil.

Hi teravolt - larger primary, definitely yes. More coupling - no, I already had a fire on this coil and am actually thinking to lower the primary.
I'm a bit frustrated to realize that I'm too short on primary turns - I'm not going to be able to run on full available voltage - the coil just draws way too much current. I thought I have an overkill primary and that I'm going to settle for 3-4 turns, while in the end 6 proves not to be enough.

Copper pipe is not cheap so I'm unsure what to do regarding it.

Regarding steve's primary - his coil is still a pulsed one, not a completely CW coil like this - I would need much more primary turns than he has, and I just won't be able to fit that much copper pipe around the base of the coil. If I could find larger gauge litz wire it would probably work, but am not sure how to mount it.

Watercooling? Sorry, but I'm really not sure how to pull that off now - it would just be way over the budget and require rebuild of the whole coil.
And since I'm already exceeding the capabilities of my mains supply, it probably doesn't make much sense anyway.
If I wanted to push the CW to the end, I'd probably do something like a 50kW generator forthe supply, watercooled CM600 bricks, and series feed topology with 30kHz resonator... but it's not time for that yet.

teravolt wrote ...

have you thought of using minibricks?

Actually, no - apart being expensive and hard to find, they also have nearly twice as high internal thermal resistance compared to plain uninsulated TO247 devices. Hence I decided for them and four separate heatsinks and try to make TO247 devices give their max.

teravolt wrote ...

That is a tough project with the heating and power diapation but very cool. The spark looks like a class-E type spark

Mine or steve's project? :P

Cheers,

Marko

previous 1 2 3 4 next

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