QCW questions

[davekni]

just curious, how did you test it to ~30A only? id imagine you also 'heat up' the igbt prior to testing

Test setup was nothing like normal IGBT use.  I biased Vge at 15Vdc (permanently on).  Discharged a 10.0uF capacitor through an inductor across collector-emitter.  Measured forward and reverse (diode) Vce during this L/C ring-down waveform.  That way Vce is never large, so doesn't saturate scope input amplifier when set to 1V/division.  Current measured by dV/dt across 10.0uF capacitor (actually two 5uF MKP capacitors in parallel that measure 10.00uF together).  Used four-wire connection to IGBTs to avoid connection resistance.  Testing is with single ring-down waveforms spaced out in time, so very little heating.  Current is adjusted by adjusting initial 10.0uF capacitor voltage.

(I already have circuitry around, both FET and TRIAC based, for discharging L+C circuits.  Mostly use these for testing inductors up to saturation current.  Can be done by manually touching two wires together.  That often requires several tries for each test to get a clean switch event without contact-bounce.)

For gate (input) capacitance, I connected collector and emitter to a 25Vdc supply and measured gate-to-emitter capacitance.  Includes gate-to-collector capacitance too since Vce is fixed.  Spec values are at 25Vce for this part.  Graph shows input capacitance as a function of Vce, so you can see what it is supposed to be (roughly) at 0Vce from the graph.

Your parts are measuring high input capacitance, opposite what my counterfeit parts measured.  There is a slower Magnachip part with higher capacitance, so that may be what your parts are.  Mine are likely some other non-Magnachip smaller die inside.  No idea how fast they switch.  That is harder to test.

[Rafft]

David

thanks for the test setup explanation.

Ive measured BT40T60 with Vce 21vdc(current limited) and with gate discharged.

attached also is screenshot of graph. is this the one to look up to? then 5nF is too high @21Vdc?

edit:
tested spare magna igbt. its about 4nF @21Vdc. so I guess the magna was still a bit better then? is 4nF way off with a genuine 60T65PES Cies=2.3nF?

[davekni]

Ive measured BT40T60 with Vce 21vdc(current limited) and with gate discharged.

Not sure I'm understanding the bit about "current limited".  Usually, with 0Vge DC and a small AC voltage for capacitance measurement, there should be almost no collector current.  Usually <1ua.  Perhaps your meter is applying a larger voltage (and perhaps a voltage with a DC component) for capacitance measurement.  You could try testing with a larger (~0.1 to 1uF) capacitor in series with the meter lead, and a high-value resistor (1meg to 10meg) gate-to-emitter to force average Vge to be 0V.  Otherwise, I'm inclined to believe only the gate to emitter/collector shorted readings.

[Rafft]

current regulated. its my CV CC power supply. I set it to 2mA only.

I have re-checked both igbts 60T65PES and 40T60 with 100nF(94nF actual) and 10meg

[davekni]

Both still well higher than the 2327pF typical specification.  (Mine were ~1200pF, so lower rather than higher.)  That would roughly fit for the similar but slower Magnachip part.

Even 2mA shouldn't limit when feeding Vce at 0Vge. Either these IGBTs have high leakage current or gate swing from cap measurement is large or 2mA current limit is not precise.

[Rafft]

yes I agree, that gate capacitance is high as per datasheet.

the 2mA setting in PS was to somehow protect the LC meter(previous test w/o 100nF&10meg).

anyways, I had a full refund with that 40T60. 1 of the 10 igbt has zero capacitance reading gate/emitter(Vce @21vdc). im guessing factory defect? or because it was sent w/o an anti-static packaging

also, Ive used a bnc cable connection for checking Iprim. its somewhat clean on scope(ch2), but gets spikes when I  make connection with fullbridge output(ch1)

[davekni]

the 2mA setting in PS was to somehow protect the LC meter(previous test w/o 100nF&10meg).

Meter may have internal protection for applied voltage, to handle charged capacitors.  Doesn't hurt to be safe.
Have you verified that your PS is accurate at 2mA?  That is very low for most supplies.  Could easily be the supply not going above 21V by itself at such low current setting.

1 of the 10 igbt has zero capacitance reading gate/emitter(Vce @21vdc). im guessing factory defect? or because it was sent w/o an anti-static packaging

Likely a factory defect - bad bond wire.  Those are really cheap counterfeit parts to have such a gross defect.  Would not be caused by ESD.  Unless your meter reads 0 capacitance into a short circuit or low resistance.  (Most meters read over-range for short circuit capacitors.)  ESD damage is generally gate oxide punch-through causing gate-emitter conductivity.  (For SiC FETs, I've seen ~1ohm G-S after gate over-voltage event.  Still works as a FET if short Vgs pulses are applied with sufficient current to drive 1ohm load.)

also, Ive used a bnc cable connection for checking Iprim. its somewhat clean on scope(ch2), but gets spikes when I  make connection with fullbridge output(ch1)

Sounds like an issue with either internal scope crosstalk due to ground current or effects of imperfect shielding on coax.  Either cause can be improved by ferrites around the coax cable and/or scope probe cable (cable from scope to probe body).  Large ferrites with multiple turns are best.  I often use several turns around a ferrite E55 core set on my scope probes and/or coax cables even when using my fancy (but old) Tektronix TDS754.

[Rafft]

the (kit)power supply I am using:
https://www.qsl.net/z33t/dc_0-30v_0-3A_eng.html

so I just set the 'current' pot to minimum, and checked if there is output voltage.yes there is.

ok my igbts are counterfiet, except the IRG im using in the buck switch.

my lcr meter shows. 0.00 reading with open leads. ill try using ferrites for the coax then. thanks

[Rafft]

Sounds like an issue with either internal scope crosstalk due to ground current or effects of imperfect shielding on coax.   

David, I tried 

[davekni]

David, I tried 

Looks like that is working well.

so I just set the 'current' pot to minimum, and checked if there is output voltage.yes there is.

Current limit accuracy (including temperature drift and change with output voltage) is going to be poor down at 2mA.  True for almost any supply capable of 1amp or more.  That power supply circuit powers U1 and other circuitry after current sense resistor R7 (0.47 ohms).  Also, 2mA is only 1mV across R7, well under sensing opamp U3's worst-case input offset voltage specification (+-6mV).

To test leakage current, I'd suggest feeding collector through 10k resistor.  Measure voltage before and after resistor to see how much current the IGBT is drawing.  If more than 40uA, that is a clear violation of a worst-case specified limit, so proof of counterfeit (or damaged) parts.  (Spec. is 40uA at 650V at 25degreesC.)

[Rafft]

Hi David

kindly check my scope shots. should I be worried about those negative going spikes? and why only on the negative side? and also, they only show when I change to shorter scope timebase. scope error?

Im now using 8pc IGBTs. BT40T60.  is it still ok to use 2R2 gate res for each IGBT? or should I raise it something like 5R since they are now connected as parallel.

[davekni]

kindly check my scope shots. should I be worried about those negative going spikes? and why only on the negative side? and also, they only show when I change to shorter scope timebase. scope error?

Negative spikes are within reason.  Only negative because duty cycle is under 50%, longer low time than high time.  Bit hard to tell without zooming in even farther, but the rising edges appear to have a bit of the triple-transition typical when phase-lead is slightly less than ideal.  Falling edges are earlier (farther before current zero-crossing) due to duty-cycle (shorter high time), so have more phase lead.  I think the spikes are due to slightly too much phase-lead on falling edges.  Too much is better than too little.  Spikes aren't large enough to be an issue.
Is the scope acquisition in hi-res mode?  With long timebase, sample rate is low enough to miss spikes or average them away if in hi-res mode.  If you change to peak-detect mode, the spikes should show up.

Im now using 8pc IGBTs. BT40T60.  is it still ok to use 2R2 gate res for each IGBT? or should I raise it something like 5R since they are now connected as parallel.

2R2 is lower than typical for TO247 IGBTs, whether single or paralleled.  However, if I am seeing triple-transitions correctly, that suggests dead-time is sufficient.  (BTW, are there a diode across each of the eight 2R2 resistors?  It is hard to tell in the picture.)

[Rafft]

David

yes each IGBT has its own (sb) diode. and about the ringing on fullbridge output, ive tried from the current 2R2(each igbt) to using 10R(and SB260 diode) 30v tvs as I ran out of 20v ones.bridge ourput ringing seems  to have gone down. also added more cap decoupling on the fullbridge bus rail(around 13uF).

primary LC is 17uH/11nF. im using a newer secondary 4" x 6.2" #30(about 18mH). I also noticed that using a bigger top load makes a great sounding thump than using smaller topload(harsh sounding).

I have a couple scope shots run on 230Vdc bus and Iprim=100A. I see ringing. im not sure where this comes from(already readjusted duty to 50% on self-osc). Im guessing its the IGBT(?)

spark length around 26" or so 230Vdc@100A. still needs some more tuning

[Rafft]

to be more complete, I redid the scoping on 'peak' acquisition and did some more zooming
------
and as compared to a 'high-res' mode

[davekni]

These latest scope plots show mostly lagging phase (voltage switching after current zero-crossing).  Not sure what changed to cause this.  Previous plots show sufficient phase lead.

The glitches you zoomed into are clearly a problem.  Coil control circuit (modified UD2.1) is getting confused.  Did some ground connection come loose?  Some switching noise from buck converter coupling into coil driver circuit?  IGBTs appear to be surviving these occasional instances of confused gate drive control even though they are causing hard switching.  Don't know if these issues are related to whatever is causing lagging phase instead of desired slight-leading phase.

If nothing else, you are getting good practice in using different scope acquisition modes:)

[Rafft]

Hi David

testing from 4igbt to 8igbt inverter , I did modify the (Gao)buck to use discrete high-side driver, for more repairability than using UCC chip(from free samples few years ago). its all working as expected.

next was the self-osc mod. I added IC socket for adding parallel cap for freq adjustment. and used cheap ceramic ones. they have thin component legs(wiggled quiet easily), Im sure this is the reason why I see those hard-switched outputs. so I just soldered it (parallel cap) direct to board. one thing I also noticed was 'self-osc' should be far from the running freq, else I get dirty Iprim. Ive set my self-osc to 440KHz and my running freq is 370KHz.  my older setup with  self-osc=390KHz and running freq=370KHz resulted in ugly Iprim. anyways, Ive set my ocd limit to around 1.55V(155Apk). Ive tested 250Vdc bus and got good spark length. and there is always that triple transition I have also been trying a few coupling K.

spark length around 26-27" 250Vbus  around 130Apk 370KHz

btw, I have already killed 2 scope probes(open ckt). RIGOL PVP2150 on scoping the bridge output. is there an easy hack to converting this to something like x100 (probe in x10)?

addtl (last 2 attachments):
modified the ramp gen to do this(branching at tip), but only in 115Vdc bus:

[davekni]

and there is always that triple transition

Those look fine.  A slight reversal at the end of the main transition is not problematic.  Indicates you have sufficient dead-time and barely enough phase-lead.

spark length around 26-27" 250Vbus  around 130Apk 370KHz

Looks great!  Nice progress.

btw, I have already killed 2 scope probes(open ckt). RIGOL PVP2150 on scoping the bridge output. is there an easy hack to converting this to something like x100 (probe in x10)?

All scope probes have reduced maximum voltage capability as frequency increases.  Some (especially cheap probes) drop dramatically at TC frequencies.  Good probes (and some cheap ones) include max-voltage vs frequency curves in specifications or user manual.

For many purposes you can get away with a small capacitor (of sufficient voltage rating) in series with the probe tip, value ~1pF.  I often home-make such capacitors with just two twisted insulated wires, or a bit of heat-shrink or tape over the probe tip with a bit of copper foil over that.  This foil version is less capable for high voltages (>1kV), but much better for fidelity.  Avoids most stray signal pickup.  Any random capacitance from other nodes to probe end of ~1pF capacitor will add unwanted signal.

[Rafft]

Those look fine.  A slight reversal at the end of the main transition is not problematic.

Thats great!

Indicates you have sufficient dead-time and barely enough phase-lead.

with "barely enough", does this mean I need a slight component adjustment on the lowpass filter? ~1uH perhaps since my coil works better at lower frequency (370KHz). and perhaps why this is so, because of said IGBT being THAT slow? based on datasheet for 40T60, its calculated speed is said to be ~270KHz. when changing primary MMC from 13nF(Fres=370KHz) to 8.2nF(Fres=430KHz), the triple transitions have higher swing and Iprim more noise. so I take it as a hint that it could either be lacking on the lowpass filter, or slow IGBT (or maybe im wrong?)

Looks great!  Nice progress.

Thanks. goal is 1Meter for this battery-powered

All scope probes have reduced maximum voltage capability as frequency increases.  Some (especially cheap probes) drop dramatically at TC frequencies.

good info! seem to have forgotten about the freq response. but is ~400KHz too much for a 100MHz rated probe?

For many purposes you can get away with a small capacitor (of sufficient voltage rating) in series with the probe tip, value ~1pF.  I often home-make such capacitors with just two twisted insulated wires

I would go this route. gimmick capacitor. but probe switch set to x1 only? how do I set this up on scope?  and could I get an accurate voltage reading (for TC use) of these?

[davekni]

with "barely enough", does this mean I need a slight component adjustment on the lowpass filter? ~1uH perhaps since my coil works better at lower frequency (370KHz). and perhaps why this is so, because of said IGBT being THAT slow? based on datasheet for 40T60, its calculated speed is said to be ~270KHz. when changing primary MMC from 13nF(Fres=370KHz) to 8.2nF(Fres=430KHz), the triple transitions have higher swing and Iprim more noise. so I take it as a hint that it could either be lacking on the lowpass filter, or slow IGBT (or maybe im wrong?)

"Barely enough" means it is OK as is, but not much margin for IGBT temperature to increase or any other change that increases delay.
The standard UD2.7 phase-lead circuit is high-pass, not low-pass.  Is that what you are referring to?  What values do you have for inductance and feedback CT resistance?  If I recall correctly, you were close to the practical limit for amount of phase lead.  May be hard to get much more.  If impedance of inductor at operating frequency is already 2x or more of resistor value, it may not make sense to go farther (higher inductance).  Perhaps a smaller resistance would be better.  I don't know if anyone has explored the practical limit to phase-lead.  Theoretical limit is 0 resistance (90 degrees of phase lead).  Real inductor will have some non-zero resistance.  Might work with only inductance (and it's bit of internal resistance).  Haven't heard of anyone going to that extreme.

good info! seem to have forgotten about the freq response. but is ~400KHz too much for a 100MHz rated probe?

Just looked at specifications for my cheap 100x 2kV 100MHz probe.  It is rated for 300Vpeak at 400kHz, only 15% of it's 2kV low-frequency capability.

I would go this route. gimmick capacitor. but probe switch set to x1 only? how do I set this up on scope?  and could I get an accurate voltage reading (for TC use) of these?

I've always used 10x probes with series capacitors.  Most probes have much better frequency response at 10x than at 1x (if they have a 1x option).  It is difficult to hit exactly 100x, but not too hard to measure actual attenuation.  Measure a somewhat-lower voltage signal directly with another probe and simultaneously with the series-capacitor probe.  Calculate the ratio.

[Rafft]

Thank you for the replies David

regarding UD, I will just leave it then. and Im using counterfeit IGBTs. who knows what the real specs of these are. I expect the worst.

cheaper scope probes voltage rating dimish fast even at 400KHz only . wow.  so I will just have to use x10 and ~2pF for monitoring bridge output. and x10 as well for Iprim scoping