Gate resistance

[flyingperson23]

I roughly followed loneocean's drsstc 3, and am having the same problem he described having with the gates of each igbt briefly switching polarity and then ringing while turning on. He described the solution to the problem being to lower the gate resistance. My gate resistors are currently 5.1ohm. Should I lower them, and to what value?

[Mads Barnkob]

Gate ringing is usually fixed with higher gate resistance, as that is the idea, to damp the ringing.

Make sure your gate resistors are right at the gate connection and as short wires as possible, to lower the inductance.

Gate resistance does not have result, its individual to all circuits. Somewhere between 0 to 30 Ohm is common.

Try to switch as fast as slowly possible, meaning that you lower your gate resistance to avoid switching losses, but only to a level of acceptable ringing.

[davekni]

If I'm reading LoneOceans web page correctly, 5R1 is where he ended up after reducing from some higher value.  Not certain, however.
His DRSSTC 3 page has a great description and set of scope images showing Vge and Vce as phase lead is adjusted, including the tripple-transitions of Vce when phase lead is not quite enough.

For your case, it is hard to tell without scope traces showing Vge and Vce.  A ring in Vge can be due to under-damping of GDT parasitic inductance.  If that is the case, then increasing gate resistance helps as Mads mentioned.  The polarity reversal (first cycle of ring perhaps)  could instead be caused by Miller capacitance (gate-to-collector capacitance) as Vce switches.  If Vce is switching due to opposite IGBT turning off, it isn't necessarily a problem.  Images of scope traces of your waveforms will allow a more useful response.

I'm presuming your GDT is constructed similarly to LoneOceans' with tightly twisted leads and all four primary wires paralleled.  If not, increased GDT parasitic (series) inductance could be causing issues.

[flyingperson23]

My gdt is created in that manner. Images 1 and 2 show Vge in yellow and Vce in blue for the top left and bottom left transistors respectively. The top and bottom right ones both look like image 2. Image 3 shows the bridge output referenced to the negative bus, left side in yellow and right in blue. Here, the voltage switch I was talking about is visible. Adjusting phase lead, the bridge output approaches image 4.

[davekni]

Images 1 and 2 show Vge in yellow and Vce in blue for the top left and bottom left transistors respectively. The top and bottom right ones both look like image 2.

Image 1 suggests that upper left gate connection is open or 5R1 resistor is shorted.  Image 2 looks Vge looks great.  All for Vge traces should look like image 2.

Image 3 shows the bridge output referenced to the negative bus, left side in yellow and right in blue. Here, the voltage switch I was talking about is visible.

Right side looks like the triple transition shown in LoneOceans' DRSSTC 3 web page.  However, left side looks like upper left gate resistor is shorted (or diode shorted).

The triple transition does look rather long as in LoneOceans before he reduced gate resistance.  So, lower resistance is likely a good idea.  However, I recommend fixing the upper left gate resistor short and scoping again before changing resistor values.

[flyingperson23]

Yep, a diode had blown. All the gate waveforms now look normal. Adjusting phase lead gets rid of the triple transition, getting Vce and Vge to the image shown, but has no effect on the voltage or current spikes.

[davekni]

Yep, a diode had blown. All the gate waveforms now look normal. Adjusting phase lead gets rid of the triple transition, getting Vce and Vge to the image shown, but has no effect on the voltage or current spikes.

Triple transition occurs when phase lead is a little too small.  Your latest image shows even less phase lead.  Yes, it does avoid triple transition, but causes spikes as you see.  You need to get rid of triple transitions by increased phase lead, not less.

What is the duration of the triple transition when it occurs (time from first transition to third)?  You can reduce dead-time by almost that much time, a little less for margin to handle IGBT speed decrease when hot.

Even the falling edge of Vge looks rather slow.  Wondering if Cge is higher than expected or driver output impedance is high (weak FETs or low driver FET Vgs) and/or high GDT stray inductance.  May not be an issue for your low-frequency coil.

[flyingperson23]

The triple transition time is ~600ns. I think I might have weak FETs as well; the driver output only reaches +16/-20v with known stable 24v driver input. I think I haven't been able to get the phase lead right because my variable inductor is too small; I'll get a bigger one and see if it works.

[davekni]

I think I might have weak FETs as well; the driver output only reaches +16/-20v with known stable 24v driver input.

That sounds like more than just a weak FET.  I recommend fixing your driver before any further testing.
A single weak FET would make only one edge (rising or falling) slow.  If weak FET is just high-resistance, voltage would still reach 24V, just slowly.
Recommend scoping your driver, FET gates and outputs, initially without GDT connected, then with GDT.

The triple transition time is ~600ns.

Probably a bit smaller gate resistors would be good, but nothing for sure until driver is repaired.

[flyingperson23]

Yeah so I scoped the driver with no load, then a few amps resistive load, then the gdt connected, and then the gdt connected with the gdt hooked up to the gate of a spare cm300, and every time it was fine; it showed the +-24v it's supposed to and very quick switching. The only thing I can think of is that the cable connecting the driver to the gdt is fairly long, as there isn't enough space for both the gdt and driver to be right next to each other, but I'd think if anything that'd just slow down the voltage, not entirely stop it from reaching 24v.

[davekni]

Yeah so I scoped the driver with no load, then a few amps resistive load, then the gdt connected, and then the gdt connected with the gdt hooked up to the gate of a spare cm300, and every time it was fine; it showed the +-24v it's supposed to and very quick switching. The only thing I can think of is that the cable connecting the driver to the gdt is fairly long, as there isn't enough space for both the gdt and driver to be right next to each other, but I'd think if anything that'd just slow down the voltage, not entirely stop it from reaching 24v.

Long driver-GDT wires will make ringing worse as resistance is lowered, but not cause the voltage reduction.
(To minimize down-side of long leads, separately twist together each pair of GDT primary wires.  Separate only at the driver connection with minimal separation distance.  The four twisted pairs can be bundled together.  That's best.  This is better than twisting the two 4-wire bundles as LoneOceans did.  Make sure all 4 primary windings are still phased correctly, not reversed.)

Sounds like driver and GDT are all OK.  Then one or more of the IGBTs or gate circuitry must be the issue.  Do you have TVS diodes across gates for protection?  If so, that's the first place to look for a problem.  (Of course, these TVS diodes must be 24V or higher.)  For testing, I'd connect GDT output to only one IGBT gate at a time.  See which one(s) are loading down the voltage.

[flyingperson23]

I have 5kp26ca tvs diodes across the gates. Each transistor is loading the voltage equally. The driver output to the gdt looks like 4t with 4 transistors, 3t with 3 transistors, and so on, regardless of which transistors they are. Could this be caused by a bad gdt ferrite? I reused it from a scrapped welder, but I'm not sure if it's optimal for this application.

[davekni]

I have 5kp26ca tvs diodes across the gates.

Great.  No issue there.

Each transistor is loading the voltage equally.

OK, just realized what I should have thought of earlier:  Gate voltage reduction is caused by capacitive voltage divider of driver output coupling capacitor and total gate capacitance.  It is a problem only if you want more gate voltage to push peak current capability of IGBTs.  If so, then the driver output coupling capacitors need to be larger (higher capacitance).