Experimentonomen´s first argon.

[Xytrell]

mosfets... had some popped, had some smoke, but finally achieved the goals

Do you have any pointers to avoid things like that? What sort of things tend to go wrong?

[mixedgas]

Do you have any pointers to avoid things like that? What sort of things tend to go wrong?

Massive lack of heatsinks for the fets

Your more on the right track with teh transistor heatsink, but you need 2 of those tunnels.
The one you have now will suffiece to see if you have regulation.

Steve

Sorry, I dont have net access at the apartment right now, or I could have warned you. *&^%$ bad economy....

Steve

[shrad]

there is a need for heatsinking, yes

I had four of those mosfets running on a small heatsink, (pentium 3 slot 1 type) and it was still hot with two fans over it

anyway, I couldn't point you any direction for your system... did you use any mosfet drivers? I know mosfet capacitance tend to be nasty for them if they are not driven properly (but it applies for PWM driving, not for linear)

[Experimentonomen]

That bipolar transistored heatsink tunnel was a dead short, dident work at all so i chucked it too, and ripped a big igbt module from a big psu board and clamped it to the heatsink from the dead mosfet passbank and used four 100mohm 5W resistors in parallel for the CS shunt and it worked like a charm, all i need now is to make a heatsink for it.

The igbt came from a 3.5kW buck converter used to reduce the 700V of rectified three phase to 320V for a big 3.5kW 48V 73A smps on the same board.

[shrad]

some high power industrial darlingtons would also do the trick, they can sustain tens of amperes at voltages up to 600 volts, and they are available on ebay for cheap

they are also capable of current regulation for big diodes with a well designed shunt (classical current feedback driver = cheap alternative)

[Experimentonomen]

So heres passbank mk3:

The controller:


The psu for the controller:



Main capacitors, they are to be replaced but two 4700µF 200V ones:



The fan that blows air through the wind tunnel/heatsink:


Left to add is a transformer for the control circuitry, a preheat timer and a overcurrent trip.

[krazer]

One word of caution-

The reason your first passbank of fets failed is due to issues relating to current bunching. Most modern fets are designed for smps duty, and are not characterized or optimized for linear duty--and as a result the current isn't always uniform across the die, which leads to 'hot spots' which eventually turn into shorts. Many fets are actually many parallel structures on one die, so it is easy to see that without the necessary current sharing resistors for each device it will stress one of the sub-swithces and go caboom.

Now with the monster IGBT you have the issue is further compounded, because internally it is actually several completely isolated dies (usually from the same wafer if not run at least), so you can run into the same issues with current bunching up on one of the sub-switches.

That said, it is possible to get fets or generic non-insulated gate bipolars (darlingtons and the like) that are rated for analog duty, it will usually be stated in the datesheet that they are rated for such abuse.


Now don't get me wrong, your igbt brick will probably work just fine in your application. But I would at least add a fast fuse so that with it goes short the tube doesn't go caboom along with the igbt!
If you choose to go that route, there are some pretty good schematics a few posts down on this page

Looks good otherwise though, looking forward to seeing this thing lase!

[Experimentonomen]

If anything happens, the overcurrent trip will kick in and disable the psu.

[shrad]

fast enough? it might be useful to put a fuse anyway, in the case your overcurrent detection is slower than a quick blown fuse

[Experimentonomen]

The overcurrent trip is a opamp sensing the voltage over the shunt thats part of the regulator and a latching relay that breaks up the interlock chain and light an "overcurrent" led.

This circuit i think is way faster than any thermal fuse in existance.