IGBT Selection / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

IGBT Selection

previous 1 2 3 next

Move Thread

LAN_403

Aragorn

Fri Oct 11 2013, 08:31PM

Registered Member #18516 Joined: Sat May 18 2013, 09:09AM
Location: Lancashire, UK
Posts: 38

Thanks for the info.

I sort of understand the current side of things. Its my understanding that we should aim to minimise hard switching all together, hence the use of phase lead etc? This is presumably why people manage to operate these bricks at currents way above Ice, because its only trying to SWITCH the high currents that causes a problem? By switching at the zero crossing, you ensure current is as close to zero as possible and thus ensure your switching inside the SOA?

Its more the switching times i'm trying to get my head around. As has been alluded to, theres a relation between switching times and coil frequency.

Taking typical figures from the datasheet, the 72n60c3 has combined on/off delay/rise time of 196ns. The 60n60c2 comes in at 173ns.

Going on these times alone, both those devices could switch at something like 5mhz. Clearly however you couldnt run a 5mhz coil from them, so how can you translate that ~200ns switching speed, into a maximum coil resonant frequency? Most of the guides i read simply say to keep the frequency as low as possible.

Is it the case that these modern parts are fast enough that it doesnt even need worrying about now unless your trying to build a tiny coil? For instance Steve Wards "0.5" model had a Fo over 300khz, and ran on fairchild 40n60 bricks which were slightly slower than the above mentioned bricks...

Cheers

Dr. Dark Current

Fri Oct 11 2013, 10:00PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

One thing is that you must stay within the switching SOA, otherwise an instant destruction of the transistor can happen.
The other thing is related to the silicon's temperature ripple/rise, the pulses in a classic DR coil are very short, so you can afford a much higher peak dissipation than in normal operation.

As for your other question... A very general rule of thumb is to add the switching times together and multiply by 10 to get the shortest practical period of switching. But in reality, everything can be summed into these three advices: Stay within the switching SOA, keep die temperature ripple low and never allow desaturation. Based on those principles, you can see, that eg. for low powers, low ON-times and ZCS, you can afford a much higher switching frequency than the one stated above. On the other hand, for long bursts of high current and without any phase lead (hard switching), you might need to decrease the frequency several times, otherwise you would roast the silicon.

Aragorn

Sat Oct 12 2013, 09:23AM

Registered Member #18516 Joined: Sat May 18 2013, 09:09AM
Location: Lancashire, UK
Posts: 38

Excellent thanks.

On the current handling, Is the current rating on the single IGBT bricks such as the 60n60 or 72n60 mentioned above, per device or per half bridge?

IE if running a full bridge do i end up at Ice x2 or Ice x4?

I think its x2 as only one device is conducting at any time (in a half bridge)?

Dr. Dark Current

Sat Oct 12 2013, 09:31AM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

It is always the same, at any time there is no parallel path, so the rating for one part applies.

Steve Conner

Sat Oct 12 2013, 09:36AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

It's per device.

When dealing with peak current ratings, the per-device rating is the same as the rating of the whole bridge, because the peak current flows in one device at a time (unless you're messing with parallel devices)

As far as average ratings are concerned: MOSFETs are resistive so you use the RMS. A half or full bridge can handle 1.4 times the current of a single device.

IGBTs, BJTs and diodes are better modelled as a constant voltage drop, so it's the average current that's relevant, and a bridge can handle more like 2x the current of a single device. A little less, because the voltage drop is slightly dependent on the current.

Aragorn

Sat Oct 12 2013, 12:12PM

Registered Member #18516 Joined: Sat May 18 2013, 09:09AM
Location: Lancashire, UK
Posts: 38

Oh right. So whats the advantage of a full/H bridge then over a half bridge in DRSSTC applications?

Dr. Dark Current

Sat Oct 12 2013, 12:43PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Output voltage.

Steve Conner

Sat Oct 12 2013, 02:51PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Yes, the full bridge delivers twice the voltage, hence twice the power.

Uspring

Sat Oct 12 2013, 04:50PM

Registered Member #3988 Joined: Thu Jul 07 2011, 03:25PM
Location:
Posts: 711

Dr. DC wrote:

Well, if there is a capacitor in parallel to the collector-emitter, the transistor can turn off at a larger current without latching. But how large - you can only guess, the data sheet will not tell you this.

I'll attempt a guess

The CM600 spec tells me, that parasitic capacitance is 11nF and switching time 300ns. If I switch off at 600A, dV/dt will be about 5e10 V/s or a 15kV rise during switching. Either you need external caps or a really fast IGBT.

Steve Conner

Sat Oct 12 2013, 07:05PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

That sort of calculation works for MOSFETs, though you have to include the effect of Miller capacitance. But in IGBTs the turnoff behaviour is dominated by minority carrier lifetime in the BJT part of the device. After the MOSFET part switches off, the BJT takes some time to "come out of saturation" and then the current through it tails off exponentially.

There are manufacturing tricks to reduce the minority carrier lifetime, like electron beam irradiation. But as far as I know they all decrease the current gain of the BJT and make the device tend more towards being a MOSFET.

This goes for hard switching. In resonant operation, the BJT part of the device can be pretty much recovered before the switching event even happens. Old, slow IGBTs can be pushed to surprisingly high frequencies.

previous 1 2 3 next

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