IGBT H-Bridge. Current peaks problems.

Thread Starter

lemariva

Joined May 22, 2012
5
Hello Everybody,

We have a problem with an IGBT-inverter circuit (H-Bridge), and an inductive load.

We have the following load:
R = 0.8 Ohm and L = 80 mH.

The IGBT power module is BSM50GD120DN2

with the driver 2SD106AI.

The input voltage (Uzk) that we usually use is 580V but we made this test with 130V. The maximal current that we have is 25A but this test was made with 2A. The problem remains the same even if you change the operating ranges.

Both IGBT located on the right are controlled by one driver, and the IGBTs located on the left side with another one (see attach file: Picture 1.pdf and the info in Half-Brige Mode with Dead Time (Page 20))

We switch the igbts in the following way: (see attach file: Picture 1.pdf)
  • I- The upper igbts are on (green color = igbt on). The current goes through the upper left igbt, the load and the upper right diode (freewheeling).
  • II-The upper left igbt and the lower right igbt are on, the current goes through these igbts and the inductive load.
  • III-The lower igbts are on. The current goes through the lower right igbt and the lower left diode. (freewheeling)
  • IV-again II
  • V- again I

The problem is that we get during I, III and V current peaks (see attach file: Picture 2.pdf). We have tested with Snubbers, but the problem doesn't seem to be that (We don't get voltage peaks). We think that the problem is the reverse recovery current from the diode that causes a significant overcurrent by switching. But the current peaks that we see in the Picture 2.pdf appears to be too late to be this effect.

Has anyone any idea what it could be? or a solution for it?

More Details:
The signals' colors from Picture 2.pdf correspond to the colors in the schematic diagram from Picture 1.pdf.
Channel 1 - 2 - 4 -> scale 100V:1V
Channel 3 -> scale 1A:100mV

More Pictures:
scope_screenshots.pdf -> They are the same screenshot but the null position / time scale from channels were changed in order to see and compare the curves.

Once again, we tested Snubbers, we changed the Rg resistor (Gate resitor), but these did not improve the problem.

Thank you very much in advance.

Best regards,

Mauro Riva
 

Attachments

praondevou

Joined Jul 9, 2011
2,942
Why is the inductor only charged in one direction?

What current probe did you use?

What is the deadtime between T1 on-->off and T2 off-->on (T4 on-->off and T3 off-->on)?
 

Thread Starter

lemariva

Joined May 22, 2012
5
Hi praondevou,

Thank you for your answer.

I answer your questions:

1- Why is the inductor only charged in one direction?
It's just one test (screenshot), we can change the direction of the current but even the problem remains the same.

2- What current probe did you use?
Here, we made a mistake, we used an Agilent 1146A probe. It can only measure frequencies up to 100kHz. Therefore we had phase shifting and attenuation. That was wrong. Hence we use an Agilent N2792A (up to 200MHz) and we measure voltage drop across shunt resistor. You can see the new measurements in the attachment. For voltage we use an Agilent N2891A probe.

3- What is the deadtime between T1 on-->off and T2 off-->on (T4 on-->off and T3 off-->on)?
Deadtime was 1.1us, now we set it to 2.1us. The problem remains the same.

Extra:
The gate resistor now is 22Ohm, with a resistor of 100Ohm, the problem is reduced but not gone.

If you have still an idea, I will be grateful.

Thank you in advance,

Mauro
 

Attachments

praondevou

Joined Jul 9, 2011
2,942
Ok, now this looks completely different. The current peak/oscillation occurs now while the inductor is charged.

The deadtime seems to be sufficient.

I agree with shortbus, how far is the gate driver from the IGBTs? long gate wires can cause the oscillations we see in the picture.

This looks more like a layout problem. can you take a picture of your setup?

Btw, these pulses are so short, you cannot be sure that you are measuring the correct value.
 
Last edited:

Thread Starter

lemariva

Joined May 22, 2012
5
Thank you very much for your answer.

I send you a picture of the setup and the layout. The maximal distance is 40mm.

We use a 20kHz PWM and we have a resolution of 2000. That means, the pulse width could be up to 25ns.

Thank you in advance.
 

Attachments

praondevou

Joined Jul 9, 2011
2,942
Thank you very much for your answer.

I send you a picture of the setup and the layout. The maximal distance is 40mm.

We use a 20kHz PWM and we have a resolution of 2000. That means, the pulse width could be up to 25ns.

Thank you in advance.
How far are the DC Bus capacitors away from the bridge?

You are trying to measure current peaks/oscillations of a few tens of nanoseconds. Since this high frequency propagates easily through air (radiation) I recommend you make sure that what you measure is real.

Try for example to pass the probe leads a different way and verify that the measured signal doesn't change. If it changes you cannot trust your measurements.
 

Thread Starter

lemariva

Joined May 22, 2012
5
Sorry for my late response, but I was quite busy the last days.

Thank you for your answer.

The distance between (+) from DC capacitors and the IGBT is 30mm and between (-) and the IGBT 60mm. As attachment I send you a picture of the setup.

The current supply is in a closed aluminium box and this box is connected to Safety Ground. I think I shouldn't have problems with EMI emissions since I measure the current with a shunt resistor outside the box. Anyway I changed both the direction of the shunt, and the position of it, the peaks remains the same. I think I can trust the measurements.

As I said, I changed the Rg resistor and the problem is reduced but not gone. Any ideas?

Thank you in advance.
 

Attachments

Top