IGBT drivers

[chab]

I am building a HV PS, using the tried and true SLR topology with an H bridge of IGBT's. I have read "Steve's HV page (including Marco's CCPS).

There seems to be a huge amount of additional complexity (for reliability) while driving the IGBT's while using a high side/low side driver chip; yet very litttle when using aan IGBT driver xformer.

Jon or Carl or anyone who has built one of these drives care to comment on how you drive your IGBT's? Part #'s for the xformer if you used that route? Carl, did you ever post a driver schematic (only need the IGBT drive part)?

Thanks,

Charlie Habekost

[Carl Willis]

Hi Charles,

Use gate-drive transformers. You are absolutely right that they simplify things (and they're the cheaper and more reliable solution). The transformers I use were free engineering samples (shhh!!) from CoilCraft, part number SD250-4L.

>Carl, did you ever post a driver schematic

Nay. The circuit has changed too much and I knew it would from the very beginning. But I have posted details on all three embodiments of my HV step-up transformer on this forum. My IGBTs are 20N60A4Ds from eBay. There is nothing I could really add to Steve Ward's excellent treatment of this topology.

-Carl

[chab]

Carl-

That was my take also after reading all the problems that Marco had. I have considered another option, a quick (but more $$$) solution:

http://www.gopcti.com/IGBT_drive.htm

All the nasty bug tracking is done with. I will get a couple of coils, and give it a go for a less expensive first try.

Thanks for sharing.

Charlie Habekost

[Jon Rosenstiel]

Hi Charles,

I went the home-brew route with my GDT. Core is an Amidon FT-140A-J, tri-filar wound, 25 turns. An old parallel printer cable kindly gave its life for the wire.

Jon R

[Chris Bradley]

I have been building similar circuits recently and have been using an alternative approach to n-polarity transistors as I wanted to push the switching frequencies to several MHz and I found it essentially impossible to push the hi-side circuit up and down through hudreds of volts quick enough.

Using a conventional push-pull circuit of a P-channel and N-channel MOSFET where (either of) the FETs is driven by a charge-pump circuit, requires only one driver per pair.

A disadvantage is that these have lower voltage ratings (as commercially available parts), but the flip-side is that you can switch them much quicker. Another issue is that there is no dead-time so you'll get shoot through at high frequency, which is a mixed blessing as you get more power through but with this shoot-through risk.

However, I have experimented with several dozen transistors and can happily recommend the 500V MOSFETs from Fairchild, the FQP1N/FQP3N and FQP1P/FQP3P. With a TC4422 12A driver, I drive the 1 amp versions through to 10MHz and the 3A will go to about 6MHz without a huge amount of shoot through (though I'm running them to 6MHz/4MHz which appear to avoid essentially all shoot through). Minimising dead time means you get higher power. I have braized these parts directly to a copper heatsink to keep them cool, though it seems most of the cooling is just to scrub off the heat generated due to the Miller/gate charging at such high switching rates.

So, though you only get 500V rather than some of the bigger 1200V IGBT ratings, you could get 10 times the frequency (with fewer parts), so the capacitors in your multiplier can be both smaller in voltage rating and in capacitance. This puts you into the mains rectifier range of electrolytics, so you can then get many cost effective capacitors with (relative to 'HV' capacitors) simply huge capacitances in this voltage range which then should give very high multiplier efficiencies and small ripple voltages (I'm using these circuits for another purpose than feeding a multiplier, but they happily drive to 100's of watts throughput power so I can't see an issue feeding a multiplier stack).

Also, you still have to generate your baseline 1200V if you are running 1200V IGBTs. It's much easier to create/obtain a stable 450V supply!

best regards,

Chris MB.

Attached - 3A mosfets' output at 2MHz and 400V whilst outputting 100W.

Just a final note - don't forget to protect against EMC emissions when building your own PSU. These things can really pump out RFI.

[Richard Hull]

The battle between FETs and IGBT's is not over yet and may never be as, currently, each have strong points. I remember when FET's were introduced back in the 60's. They were the Godsend as they were the true vacuum tube replacement minus the filament supply and soon you would never be able to buy another base drive energy hungry junction transistor which would only be seen in future museum displays.

Never happened.... as the transistor had a giant 10 year developmental leap on the FET and still did a lot of things far better and more efficiently, plus they were getting very cheap.

There are other devices that just screamed at introduction. The tunnel diode is one of these and had a massive flurry of early apps due to its amazing characterisitcs. Today it languishes as a $100.00 + little component mostly used as replacements in existing gear. The old Eberline PNC-1 neutron counter has a tunnel diode in it!!

Tried to purchase a UJT lately? There are only about two part numbers of the unijunction transistor made now. This is basically a replacement for the neon bulb (NE-2) in relaxation oscillators and saw its major use in 70-90's o'scope saw tooth sweep generators.

Right now, for HV power circuitry either the FET of the IGBT are found with a slight edge in new designs drifting towards the IGBT as its initial problems in generation I and II are being solved and other improvements are wooing FET would-be designs over to its cause, especially where a little extra cost is not a factor.

Richard Hull

[Carl Willis]

Hi Chris,

Interesting to see the successful use of push-pull switching. It is an alternative to the half-bridge, and presumably two push-pull pairs could replace the full-bridge while conferring the drive simplification you mentioned.

>Another issue is that there is no dead-time so you'll get shoot through at high frequency, which is a mixed blessing as you get more power through but with this shoot-through risk.

I don't see why exactly this is an issue for push-pull, any more so than it is in a half-bridge or full-bridge. Most PWM driver chips provide some controllable dead time in order to prevent shoot-through. For switching, a bit of dead time is tolerable and probably necessary for this reason. If you are running your push-pull as a linear amplifier, I can understand the need to smoothly transition from the top side conducting to the bottom side conducting, bringing up the problem with shoot-through, but it wouldn't seem to be an issue for switching.

>so I can't see an issue feeding a multiplier stack

What kind of loads are you driving?

I see the multiplier stack and HV transformer as dominant difficulties as the frequency gets higher. The MOSFETs may be cheap, but the HV/RF diodes in the stack will be costly, and the transformer will be a real challenge. Insulation, losses, and stray reactance will be difficult aspects of the transformer at a few MHz. One solution--using air-core resonant coupling to a bunch of independent secondary coils, each bridged with their own resonant caps and rectifiers that are then stacked to multiply the DC voltage--adds a lot of complexity but is used reliably at ~500 kHz by certain electrostatic accelerator companies. Moving up in frequency is a limitation for sure, and mostly because of the multiplier. But it would be nice to do.

I'm very pleased with my IGBT experience, and not so pleased with MOSFETs. Richard points out that the argument is ongoing over which you use when, but among the high-power hobby users (Tesla coiling, quarter shrinkers, railguns), the IGBTs have really excelled. This is not because they are especially compelling from the perspective of losses, but because they just seem to be bulletproof to extreme currents and voltages. I did a bit of Tesla coiling with MOSFETs on a very small scale, and I did go through a lot of FETs. I have yet to blow an IGBT. For push-pull circuits, one is stuck with MOSFETS since as far as I know a p-channel IGBT does not exist.

-Carl

[Chris Bradley]

All points agreed. The comment on the shoot-through was just in reference to when using a charge-pump driver.

Many of the designed-for hi-lo drivers have an inbuilt 400kHz drive circuit and this is becoming common (and thus cheap) for motor drives. So, yes, for many reasons including the robustness you mention and good compatibility with these drivers' frequencies, IGBTs easily have the edge. I was just offering an alternative if anyone wants to experiment.

One thing I think worth commenting on though. What does or does not blow up, and how well it does or does not operate before blowing up, is very individual amongst parts. Data sheets are one thing, and how they work is another! In the 400kHz range of standardised parts you need not worry too much about these things and they are all much-of-a-muchness. Push to MHz and you'll have to experiment with what does or doesn't work. Just because one brand of transistors works doesn't mean another of the same nominal spec will. This applies to the chip drivers aswell. I recommend you buy a few and see which ones work the best for you.

best regards,

Chris MB.

(The load in the screen-shot was a stabilised PTC ohmic resistance (a light bulb+fixed resistor for trimming) I was using for the purpose of terminating a transmission line (one of these HI-FI enthusiasts screened mains cables - seems to be about 500ohm line impedance with what I am using). The actual end objective is only to deliver a few watts into electrodes, but the amount of transmission line reflections due to the wide spectral content of this square wave appears to require 'full ohmic' terminations, which is CONSIDERABLY power hungry at these voltages! gt.200W or so! I can use a serial capacitor (10's of pF's) and resistor as a termination which mostly does the job and only a minor power consumption, with a few left-over ripples. If anyone has any other good ideas for creating an equivalent 500ohm termination without the power consumption, please let me know!!)

[Dustinit]

I also built a fet driven tesla coil which worked surprisingly well considering I was running it at 15Mhz but I used an oscillator topology which makes driving the fets much easier. (gate power consumption increases hugely with increasing frequency)
On that topic, here is a picture of a H-bridge oscillator which runs at 27Mhz of my own design, now patented. The gate currents approach 10 amps at full power (2kw).
A quartz glass tube goes through the centre coil to produce an inductively coupled plasma in argon at atmospheric pressure.
An oscillator is cool because it self matches to the load. Driving with a fixed frequency requires constant measuring / rematching to the load by varying the impedance matching network.
Dustin

[bpaddock]

> Right now, for HV power circuitry either the FET of the IGBT are found with a slight edge in new designs

There are the even more obscure Emitter-Switched Bipolar Transistors:

http://www.st.com/stonline/products/fam ... o/esbt.htm

"These devices have a Power Bipolar structure to withstand high breakdown voltage (up to 2,200V) and a Power MOSFET structure for faster switching speed (up to 150 KHz) in Cascode configuration."

[SteveZ]

You might consider designing a switching supply using one of the IGBT driver chips from companies like International Rectifier (IR) [http://www.irf.com/indexsw.html] These devices are used in the electronically commutated motors (BLDC) and are a fairly mature technology. You can even get integrated power modules that contain a 3-phase IGBT "H" bridge and all the drivers. To use these modules in a single phase transformer, only use two of the phases. Do be careful as most of these modules (if not all) are NOT electrically isolated and present an electrocution hazard. Just like the descrete design would. These modules are designed to use a microcontroller and usually have 5 V and even 3.3 V drive capability. Some of the module contain thermal sensors and overcurrent shut down capability. You can get module with 600V and 1200V ratings and current ratings from 1 A to 30 A or more. The modules can be driven with an optically isolated drive stage so that you do not have to have control circuitry floating on the line.

It should be possible to create a voltage/current regulated supply with a computer interface using a microcontroller with an RS-232, or RS-485 serial port, or a USB or Ethernet interface.

If you are interested I can look into this idea further.