How to measure IGBT bridge?

[Plastic]

Hello, this is my first post here. I'm working on project with high power IGBT bridge and have some questions how to properly measure the high and low side IGBT turn on and turn off times and also a dead-time between them. So the basic "schematics" looks like this: http://www.electronicspoint.com/attachments/igbt_-_1-2_h-bridge_1635-jpg.875/

Is there a good way or any way to measure the things I need without having a floating osciloscope probes? If yes, then how. I want to measure both IGBT's simultaniously.

All I can think of is to connect the "Ground" leads of a probes to middle point of a half bridge. Then connect the one "signal" lead to +HV and other to -HV (to a DC-Bus) and then invert the the lower IGBT channel on osciloscope. Then precharge the DC Bus capacitor bank, disconnect it from power outlet and connect it to IGBT bridge. So I don't know how good this method would be (dispite the fact that connecting a "hot" cap bank is dengerous in it's own way.. )? 

Any thoughts or ideas? Help me to save my osciloscope.   

p.s. English isn't my native tongue..

[T3sl4co1l]

1. With a 4 channel oscilloscope and four probes, measure:
Ch1 = low side gate to emitter/GND
Ch2 = low side collector to GND
Ch3 = high side gate to GND
Ch4 = supply to GND
turn on Math1 = Ch3-Ch2 (high side G-E), Math2 = Ch4-Ch2 (high side C-E).

2. With a proper differential probe (rated for the common mode voltage), probe high side directly.

3. With an isolation transformer on the DUT, probe just one side at a time.

(1) sucks because you need lots of channels (or on a two channel scope, you can only get one measurement at a time -- including analog scopes where you can invert one channel and ADD them for the same result), and the probes have to be perfectly matched, which they probably aren't.  You also can't see much of a signal for the gate measurement, because each channel input has to cover the common mode range (too sensitive and it goes into clipping), most of which you have to subtract to see the actual gate voltage signal.

(2) is great, if you can afford the probe(s)...

(3) is almost always a bad idea, even with a very good isolation transformer, because that transformer still has quite a lot of common mode capacitance, which means you're blasting the full switching edge across the loop between probes, scope, power and return.  The resulting measurement may be reasonable, but it will inevitably have something weird superimposed on it due to that common mode current (bounce and ringing at some odd frequency, perhaps in the 100s of kHz).

Another option is to set Vout = 0 (disconnect the bridge supply voltage, and leave the bridge unpowered, or even short out the rails) and read the high side gate waveform.  This works for gross verification (is it getting gate voltage and swinging fast enough, yep, great), but isn't representative of the operating condition because it specifically excludes Miller effect (there's no collector voltage swing!).  Which you can infer from the low side gate, if the waveforms are symmetrical, but no, you can never know exactly what the high side is without measuring it.

On the upside, it's not usually very important to measure it.  You can tell from other aspects of the circuit whether it's doing its job or not (output rise/fall, AC supply current if you have an amp probe, etc.), and that's all that matters in a switching circuit.

Tim

[DanielS]

(2) is great, if you can afford the probe(s)...

Or 2b) build your own. Depending on performance requirements, a usable non-isolated differential probe could be doable for less than $50 - assuming he is allowed to use homebrew instruments.

[T3sl4co1l]

Do you have an example?  I haven't heard of any projects claiming high CMRR, bandwidth and sufficient accuracy (+/-3% within the bandwidth, say) for that kind of instrument.

An OSHW diff probe on par with a Tek would be very useful though.

Mind that, if the interesting quantity is stuff like gate rise time, plateau, bounce and so on, it will need 20MHz bandwidth, and preferably much more than that!  Many cheap to moderately priced differential dongles don't even go over 5MHz.

Tim

[DanielS]

Here's one DIY probe from around here:
https://www.eevblog.com/forum/projects/oshw-diy-1kv-100mhz-differential-probe-(dilemma-vs-hope)/

Your CMRR would be largely limited by how lucky and patient you are with compensation trimmers.

[David Hess]

Do you have an example?  I haven't heard of any projects claiming high CMRR, bandwidth and sufficient accuracy (+/-3% within the bandwidth, say) for that kind of instrument.

What I have done in the past when other (expensive) methods were not available was to design in measurement circuits which would produce ground referenced outputs for diagnostics.  They always involved some type of transconductance output which supported high voltages which is easy enough to do with a cascode.

An OSHW diff probe on par with a Tek would be very useful though.

It would.  I am always surprised one has not been published.

Mind that, if the interesting quantity is stuff like gate rise time, plateau, bounce and so on, it will need 20MHz bandwidth, and preferably much more than that!  Many cheap to moderately priced differential dongles don't even go over 5MHz.

I use a pair of Tektronix 7A13s for this sort of thing.

[NiHaoMike]

If limited run time is OK, you can use an inverter or unplugged UPS to power the DUT or scope. (Wear insulating gloves if the scope has exposed metal.)

[poorchava]

In professional (corporate) setting, you may want to use multiple low-end scopes powered through a multi winding isolation transformer instead of isolated differential probes. This is motivated by the fact, that good differential probe will quite likely cost the same as 3-4 lower end scopes (DS1054Z, some low end Signelt/Atten, Owon, Hantek  etc)

[David Hess]

In professional (corporate) setting, you may want to use multiple low-end scopes powered through a multi winding isolation transformer instead of isolated differential probes. This is motivated by the fact, that good differential probe will quite likely cost the same as 3-4 lower end scopes (DS1054Z, some low end Signelt/Atten, Owon, Hantek  etc)

Just keep in mind that if you do this, the probe ground connection will have considerable capacitive loading through the oscilloscope chassis and isolation transformer.  This is not a problem if the probe ground can be connected to a low impedance point which is usually the case.

Oscilloscopes like the Tektronix TPS2000 series have isolated vertical inputs with lower probe ground capacitance but are relatively expensive.

[T3sl4co1l]

In professional (corporate) setting, you may want to use multiple low-end scopes powered through a multi winding isolation transformer instead of isolated differential probes. This is motivated by the fact, that good differential probe will quite likely cost the same as 3-4 lower end scopes (DS1054Z, some low end Signelt/Atten, Owon, Hantek  etc)

Just keep in mind that if you do this, the probe ground connection will have considerable capacitive loading through the oscilloscope chassis and isolation transformer.  This is not a problem if the probe ground can be connected to a low impedance point which is usually the case.

It's worse than this:

The current flowing through ground, to charge the oscilloscope and isolation transformer capacitances, puts a huge load on the circuit.  That may not be a big deal to a power switching circuit, but you're putting all that common mode current into the circuit under test (which might screw it up in turn), and into the probe.

The probe isn't made for that kind of current.  The ground strap obviously will drop a substantial voltage, disturbing your measurement in an obvious way (all that common mode current will be predominantly a somewhat slower ringing).  This remains, even if you use a short probe (grounding spring?).  You can reduce but not eliminate it.

Moreover, the probe cable itself is leaky, and will drop voltage along the shield, which doesn't get dropped along the core.  Result, common mode once again.

Even using extremely huge ferrite beads won't help, because your only hope is to absorb the entirety of the switching signal, which requires a lot of flux (big core, lots of turns); and what's worse, you need to absorb all that flux while keeping a high enough impedance that the scope ground stops moving around so much (big core, absurdly high mu_r).

Possibly, a large nanocrystalline core (one of those $50-100 parts from VAC, with >>20uH/t^2 in the spec) with 10-20 turns would be enough for typical switching applications (meaning, over 50kHz), but man... and anyway, you'd use up all your probe cable length trying.

Moral of the story, no matter how you cut it: you really just don't want to be making this measurement.  There's always a better way!

Tim

[poorchava]

Oscilloscopes like the Tektronix TPS2000 series have isolated vertical inputs with lower probe ground capacitance but are relatively expensive.

And they have a major drawback, that voltage difference between channel grounds can't exceed 30V or so or you're risking damage to the scope. A 100:1 probe probably won't help, as the grounds are tied together anyway. Those scopes are very convenient for debugging analog circuits though.

[David Hess]

Oscilloscopes like the Tektronix TPS2000 series have isolated vertical inputs with lower probe ground capacitance but are relatively expensive.

And they have a major drawback, that voltage difference between channel grounds can't exceed 30V or so or you're risking damage to the scope. A 100:1 probe probably won't help, as the grounds are tied together anyway. Those scopes are very convenient for debugging analog circuits though.

That is not correct.  The separate grounds for the BNC inputs are catagory 2 safety rated to 600 Vrms compared to the instrument ground; presumably you could have 1200 Vrms between probe grounds.

The 30 Vrms safety limit applies to typical x10 passive probes which expose their ground connection to the user.  Fluke for instance makes special safety rated x10 passive probes where the entire probe ground including the outside of the BNC connector is insulated.

[jtbili]

Hi,

The best way is to use Rogowski coil and scope. This method is used for high power IGBT and SiC components (for inverters and converters fed from 3kv DC and output powers of MWs)

jtbili

[sarepairman2]

I recommend a caliper

[johansen]

Use a common mode inductor from an smps power supply as a 1:1 transformer. if your switching frequency is above 50khz you should have no problems with the core saturating.

couple the transformer to the igbt gate drive with a capacitor.. yes, you lose the dc information, but you may not need that information. the ac voltage will be preserved and you will be able to observe the dead time.

[pxl]

3. With an isolation transformer on the DUT, probe just one side at a time.

For measuring DUT on mains I use this method. Simply, because the differential probes for mains all use 1:100 at least, which would introduce too much noise on measuring 1 V range (not to mention that the probe itself will modify the measured data). However it could be potentially dangerous and you always think twice what and how to measure.

Working with smaller voltage I would probably use differential probes.