Questions for Bob with no images attached

[eldarion]

Hi Earl,

Yes, I am very well aware of all of that!   If you notice what I wrote earlier, I am still waiting on my PCBs--for these tests, I am still using the old hand-wired MOSFET driver board.  The new PCBs have extensive bypassing capacitors, and the SMD capacitors are physically right up against the SMD UCC27322 chips.  The only issue might come from the physical size of the IGBT bricks (and therefore the lead length required to connect to them), but seeing as they are switching cleanly on these atrocious gate waveforms, I do not think there will be an issue with gate drive.

The bigger issue here is that the IGBT, when a pulse occurs, sets up a small magnetic field in the primary.  Then, when it turns off, it essentially open-circuits that end of the primary.  What is the magnetic flux going to do?  It can't instantly vanish, so it causes those oscillations, and at a much higher voltage amplitude due to the open-ended circuit.  I think a snubber will work wonders here; I just need to get my hands on the diodes I ordered before I can try.

Eldarion

[eldarion]

Hi Bob,

Would it be possible for you to post scope shots of the three PWM3F channels hooked up to the coil primaries?

I am having a terrible time getting this thing to work, and I would like to compare my waveforms against yours to see if I can find the problem faster.  I have already found the scope shots you put over on the WorkingWaterCar group, but those were taken without the coil's primaries hooked up to the PWM3F, as far as I can tell.

I tried adding the snubber diodes, and while they did clean up the drive waveform, they also dramatically increased power input (x10!!) and almost completely shut off the secondary winding's output.  Further poking around revaled the source of the oscillation as the secondary winding itself, coupling back into the primaries.

I am still getting no DC voltage on the longitudinal winding(s), and I have quadruple checked the pulse sequencing and hookups.  Also, adding the HV potential still has no effect, so something is still wrong here.

As always, thank you for all of the invaluable assistance you have been providing to this group!   Hopefully we can figure out what is wrong quickly...

Eldarion

EDIT: Maybe the rise/fall times on my gate driver are actually too fast, thus giving rise to the atrocious ringing.  I went ahead and added 470 ohms of series resistance to the gate driver output in order to slow down the transition a bit and "shape" the pulse, and the ringing all but disappeared on that primary coil.  I have not yet had a chance to try the 470 ohm resistor on the other two primaries.

Am I thinking correctly here?  I am assuming that the ringing is going to destroy the effect far faster than a slightly slower rise/fall time on the pulses.

[Earl]

Hi Earl,

Yes, I am very well aware of all of that! ;)  If you notice what I wrote earlier, I am still waiting on my PCBs--for these tests, I am still using the old hand-wired MOSFET driver board.  The new PCBs have extensive bypassing capacitors, and the SMD capacitors are physically right up against the SMD UCC27322 chips. 

The only issue might come from the physical size of the IGBT bricks (and therefore the lead length required to connect to them),
The source connection on the IGBT is the center of the universe.  Everything revolves around it.  The only correct way of doing things is to physically mount all electronics such that the ground pin of the UCC27322 has zero mm length to the IGBT source.  There must be no wires between the UCC27322 ground and the IGBT source.  That is the correct way of doing things, there are millions of incorrect ways.

.... but seeing as they are switching cleanly on these atrocious gate waveforms, I do not think there will be an issue with gate drive.

The bigger issue here is that the IGBT, when a pulse occurs, sets up a small magnetic field in the primary.  Then, when it turns off, it essentially open-circuits that end of the primary.  What is the magnetic flux going to do?  It can't instantly vanish, so it causes those oscillations, and at a much higher voltage amplitude due to the open-ended circuit.  I think a snubber will work wonders here; I just need to get my hands on the diodes I ordered before I can try.

The voltage across the inductance can, and does, instantly change polarity.  Either the switching device has enough voltage breakdown or
1- it blows up
2- goes into avalanche breakdown and depending upon the pulse energy and avalanche rating either survives or blows up.

Using an R/C snubber is very inefficient and usually generates a lot of heat and increases rise time significantly.
My secret is to use a series ferrite bead & C.  I don't use R/C or diode snubbers.
By selecting the proper ferrite bead from a developers kit and using the proper C value the overshoot can be kept small, rise time only minimally degraded, and no heat.
Start with a 1 nF ceramic HV capacitor and go from there through the various ferrite beads in the kit.  Short lead lengths to drain and source.

In Europe, I obtained a ferrite bead kit from Murata called "Chip Ferrite Bead BLM series, kit no. EKEM11UA.  Of course, if you can convince them you are in R&D and a large series production is planned afterwards, then you are in a better position to obtain a kit.

This does not free you from the obligation to keep the output loop length close to zero.

Eldarion

[Earl]

As always, thank you for all of the invaluable assistance you have been providing to this group! :)  Hopefully we can figure out what is wrong quickly...

Eldarion

EDIT: Maybe the rise/fall times on my gate driver are actually too fast, thus giving rise to the atrocious ringing.

With improper physical layout, it is possible to have too fast transitions.  I was not joking when I said your layout must be as if the circuit is operating at 1 GHz.
The only wires should be ones going to the toroid, the driver & switch electronics should have no wires in the switching circuit path.  Ground bounce is a big problem is PCB layout of modern digital logic and only good layout can help.  If you absolutely must use a wire to connect driver to gate, make it as short as possible and if necessary replace one wire with many insulated ones in parallel (litze).  Bob has his reasons for not using litze in the toroid, but these reasons do not apply for hay-wired PCB connections.

I went ahead and added 470 ohms of series resistance to the gate driver output in order to slow down the transition a bit and "shape" the pulse, and the ringing all but disappeared on that primary coil.  I have not yet had a chance to try the 470 ohm resistor on the other two primaries.

Am I thinking correctly here?  I am assuming that the ringing is going to destroy the effect far faster than a slightly slower rise/fall time on the pulses.

I advise my clients never to use gate resistance, but to combat ringing and RFI legal limits with correct circuit layout.  This implies lead length approaching zero.

[eldarion]

Hi Earl,

I'm not quite sure what I was thinking when I wrote that edit.  Experimenting late at night with no sleep is not a good thing...   You are most right, and the gate resistors are now gone.

I do have some other ideas I will try.

Does anyone know of a place to get Dallas/Maxim onewire parts?  I was going to order a DS2432 OneWire EEPROM from them, and then noticed the lead time: 22 weeks!!!   I am using that EEPROM to store settings, and obviously do not want to redesign that portion of the controller.

Thanks!

Eldarion