Arlo's power stage Leaf controller runs and drives page 103

[zombiess]

did you happen to make note of the ringing frequency as you added capacitance?

 

[Arlo1]

did you happen to make note of the ringing frequency as you added capacitance?

Frequency stays the same the size of the ring goes down as I switch slower.

 

[Arlo1]

be carefult with the high gate capacitance, 600nF at 15V, 5 kHz will lead to almost 1W of dissipation in the gate resistors, so make sure they can handle the heat.

I am not going to run more then 200nf added the gate on the igbt is 90nf at 10v so ~ 140 nf at 15v my goal was for a 1:1 ratio or even less. I just wanted to try going that far to see if I could ever make it perfect and I could not.
More then 200nf added to the gate makes it turn on to slow and cause the desat to trip so then I need to slow it down as well.
Where I ended I had 6 0805 100nf caps stacked on a single pad and 4 0805 100pf stacked on a single pad. It takes skill to do that

 

[newbiebiker]

cool.

 

[Arlo1]

I responded to you on your thread newbiebiker. We will try to keep this as clean as we can. And your thread will be a source for info for all You are starting from the basics but I will help as I can.

Now back to business.
I think I found the rest of the ringing is in common mode noise. I put the probe for the gate on the gate with the ground on the gate as well and it make some big amounts of ringing. Should read 0 if the probe was not picking up noise.... I went back to 150nf and run the high side and it didn't look bad so I moved to the low side and tested and it was really good where I have it. I might concider changing the off resistors to 2 ohms instead of 1 which will give 6 ohms total rather then 3 ohms total.

I tested the low side up to 430 amps at 205v and all looks ok so I will try upping the voltage next chance I get.

Here is a could screen shots to compare. Blue is TP- to Phase and it looks good! Switching is just under 500us on and just under 700 us off!

Now question is where do I get good probes to get rid of this? I have 1 differential probe. Should I use 2?

 

[HighHopes]

u should only EVER be using differential probes for anything measured on the gate driver. best probes are made by tektronics, u need 1000V rated min.


common mode noise.. next is to see if it is the probe or true common mode noise. was your dif probe on the yellow? if not, take a shot with diff probe on yellow (nothing on blue since u have only one diff probe). and also both diff probes on gate pin. probably you will see the same as you already showed for yellow trace. make sure to twist the leads. using the diff probe on G/E should be useful to know if it is probe common mode noise or real. if u still see it, chances are its real.

fixing common mode nosie measured G/E. for an upper IGBT, basically what's happening here is you have ground bounce between the "ground" on your gate driver (which is the mid point of your power supply which needs good connection to your IGBT emitter pin) with respect to the phase out node (upper IGBT E, lower IGBT C). using only ONE diff probe (two if you had two diff probes) also eliminates the chance that your scope is bouncing via normal probe's ground clip. its similar for lower IGBT but we're talking more about lower IGBT E, DC Link Neg. if you have an inductance between these two nodes then its not a solid connection and that would account for the ground bounce common mode noise. if this is the case, its not easy to fix.. though there are a couple of things you can try which are easy but if that don't work.. likely u have to scrap the design. i hope it doesn't come to that but this is the whole reason behind learning at lower power levels how to do things properly because at higher power levels you have less margin to success. anyway, we'll see.

its hard to say if any of that common mode noise is contributing to your desat tripping off when it should not. noise can do funny things though.

 

[Arlo1]

u should only EVER be using differential probes for anything measured on the gate driver. best probes are made by tektronics, u need 1000V rated min.


common mode noise.. next is to see if it is the probe or true common mode noise. was your dif probe on the yellow? if not, take a shot with diff probe on yellow (nothing on blue since u have only one diff probe). and also both diff probes on gate pin. probably you will see the same as you already showed for yellow trace. make sure to twist the leads. using the diff probe on G/E should be useful to know if it is probe common mode noise or real. if u still see it, chances are its real.

fixing common mode nosie measured G/E. for an upper IGBT, basically what's happening here is you have ground bounce between the "ground" on your gate driver (which is the mid point of your power supply which needs good connection to your IGBT emitter pin) with respect to the phase out node (upper IGBT E, lower IGBT C). using only ONE diff probe (two if you had two diff probes) also eliminates the chance that your scope is bouncing via normal probe's ground clip. its similar for lower IGBT but we're talking more about lower IGBT E, DC Link Neg. if you have an inductance between these two nodes then its not a solid connection and that would account for the ground bounce common mode noise. if this is the case, its not easy to fix.. though there are a couple of things you can try which are easy but if that don't work.. likely u have to scrap the design. i hope it doesn't come to that but this is the whole reason behind learning at lower power levels how to do things properly because at higher power levels you have less margin to success. anyway, we'll see.

its hard to say if any of that common mode noise is contributing to your desat tripping off when it should not. noise can do funny things though.

Ok the last 4 pictures are the lower IGBT I will see if the diff probe shows the same. BTW I solved the desat problem a long time ago. It's a timing issue. But solved.

 

[Arlo1]

Ok this again is the low side. 3 pictures 1 of the gate then 1 of the common mode noise with both positive and negative of the probe on gate pin then 1 of the TP- To the Phase connection. I have 1 ohm off resistors and I bet the third picture will look much better with 2 ohm resistors for 6 in total.
All of this is with the Diff probe.

Its my understanding if you hook both the + and - of the same probe to the same spot and you still get noise the noise is not in the circuit but the probe is picking it up externally?

 

[HighHopes]

1st pic, G/E with diff probe. you serious have that huge spike on turn OFF command? that's brutal. there's nothing in the gate circuit that would be causing that.

2nd picture, you see two big spikes.. these are probably two turn OFF events (i say this because of 1st pic) and also the space between is about right for two turn OFF events. this is pic and also the 1st is good evidence that might be problem outside of gate driver.

then 1 of the TP-

not sure what TP- is.. hopefully not "test point". its not good to have test points in gate driver circuit.. way too much noise these things act like antennas. i've only done that once with a gate driver about your power level and it was hard. i put series resistor with each test point so that i could de-populate resistor and let test points float so they out of circuit normal operation.

 

[Arlo1]

Traction pack - (DC- )
Emmiter on the low igbt.
You never answered the question on common mode noise? If both probe positive and negative are hooked to the same spot and the noise is still there its not in the circuit but rather out side noise the probe is receiving right?

 

[HighHopes]

nooooooooo.... i typed so much and then i lost it. damn.

ok, here is the rub.. we still want to determine if noise is origininating from within gate driver or not. here are two more tests

1) solder a multi stranded wire, relatively large one like gauge 14 or 18, between gate driver power supply mid-point and IGBT emitter (these two points are already connected direct in the layout). you are strengthing your grounding here. make IGBT G/E measurement.. notice any improvement? if not go to next step

2) with 1) already in place, remove external G/E cap. this will speed up IGBT switching AND change your gate driver's PI network paramater. so we change the excitation frequency AND change the resonant frequency so now we hope to be well outside any situation that might lead to a resonance. make IGBT G/E measurement .. notice any improvement?

if still the same.. begin to susepect noise is coming outside and getting in. that noise is outside, ringing, is not such a big deal. there' salways ringing in the switching and the components are tolerant to it. that it gets IN to your gate driver is bad. or, that externally it is so large that it forces its way inside is bad.. but we have already seen screen shots of your IGBT switching, blue curves, it didn't look too bad. so i would say you more have a problem that your noise is getting inside your gate driver

one way for common mode noise to get in is to have a bad ground or bad ground bounce. so we will investigate ground bounce next by looking and understanding these 3 pics. well, 4 pics, but the first two are the same.

 

[zombiess]

I would also like to have you perform a quick experiment for troubleshooting. It's pretty easy to do too. The goal is to measure the amount of inductance in your bus layout. You can perform the test at a lower bus voltage if you prefer.

1. measure the ringing freq you have right now, measure it as accurately as you can on your scope using the traces.
2. add some capacitance across the C-E on the IGBT you are testing. It does not need to be a large amount, 100nF should be fine.
3. measure the ringing frequency you see now, it should be lower than in step one.

If you can post this, we can work through the math to see how much inductance is in your layout.

 

[Arlo1]

Thanks for the Reply high hopes. I will read it and follow it in the order you describe. I was at the shop testing before you said anything. Working on the low side I switched out the off resistors to 6 ohms as well. That didn't seem to help so I said WTF lets try more voltage as in the past the reverse recovery worked better at higher voltage. I scoped the gate signal to begin with and eased up on the current until I maxed out at 681 amps at 319 volts.
The gate signal looked better for the most part but the spike at shut off is big. Like 380v overshoot big.
I have tried a jumper wire to connect the common to the emitter and I do remember it making a small difference I will try it again but when I removed the cap that's parallel to the gate it got worse but I will try again.

Here is two shots from 681a at 319v The 380v overshoot is still within the rating of the igbt at 1200v. When things work perfectly what kind of a spike can I expect with a IGBT at these kinds of voltages? Probe is 200-1

Blue is the gate in the first picture and the DC- to Phase in the second pic the yellow is a very zoomed in measurement of the current. Its 1.76mv/a

 

[zombiess]

Can you post some detailed pics of your physical layout again? I can't find them in this monster thread, spent 20 mins looking too. I'm looking for the bus bar design and how much overlap of the supply bus you have.

Large overshoot at turn off often means large layout inductance. I noticed in the above shots that your ringing freq is 2.5MHz. I have only seen ringing frequencies in the +20MHz area in my limited experience. This ringing freq is determined by layout inductance and the devices output capacitance. Being that your ringing is 2.5MHz and the large overshoot at turn off, I suspect you have a large inductance present in the layout.

 

[HighHopes]

I am suspecting the same. But how it makes G/E ring like that. This is why I posted those three drawings last night

If you have already done those experiments and know the results you do not need to repeat its ok.

 

[Arlo1]

I am suspecting the same. But how it makes G/E ring like that. This is why I posted those three drawings last night

If you have already done those experiments and know the results you do not need to repeat its ok.

I will try them again because the info is convoluted with different tests. I actually made things better and the only thing I did was remove snubber caps you wanted me to put on

Anyways as the title says... new pics on page 78.... idk if I was looking to find some pictures of how something was built and it said that on the title that might be a place to start hu Zombies??

I realize it would be nice to have the collector of the igbt closer to the emitter of the high side but its not possible with this housing and I really wanted to use the heat sink. But maybe I can use some polish and silver conductive grease to make the aluminum conduct better and help this.

High Hopes do you think I can make it survive as long as I keep the spikes inside the rating of the IGBTs?

 

[Arlo1]

Large overshoot at turn off often means large layout inductance. I noticed in the above shots that your ringing freq is 2.5MHz. I have only seen ringing frequencies in the +20MHz area in my limited experience. This ringing freq is determined by layout inductance and the devices output capacitance. Being that your ringing is 2.5MHz and the large overshoot at turn off, I suspect you have a large inductance present in the layout.

I think the 2.5 MHz ring left when I removed the snubbers. (what HH calls snubbers NOT the other guys)

 

[Njay]

Arlin, do you have this trace but for the turn on part, in as much detail (zoom on time and voltage) of the switching on as possible (with info on time and voltage scales)?

 

[Arlo1]

Arlin, do you have this trace but for the turn on part, in as much detail (zoom on time and voltage) of the switching on as possible (with info on time and voltage scales)?

The on looks fine.
I am working on the off events its what needs attention if I make the off better I might then re-visit the on and see If I can speed it up. But for now its fine.

Here is a picture. The Blue trace Is E-C (DC- to Phase) on the low side. Don't worry about he yellow its very zoomed in of the current measurement.
As you can see the on is very clean. The Diff probe is 200:1

 

[Njay]

The on looks fine.
I am working on the off events its what needs attention if I make the off better I might then re-visit the on and see If I can speed it up. But for now its fine.

yes I know, this is just data to try to estimate your layout's inductance.

Here is a picture. The Blue trace Is E-C (DC- to Phase) on the low side. Don't worry about he yellow its very zoomed in of the current measurement.
As you can see the on is very clean. The Diff probe is 200:1

Yellow is probe x10 and 1.76mV/A, right?

Edit: Your "FET" swicth-on time is really 2us?

 

[Arlo1]

Not sure why the two slopes. But my goal is first make it clean then speed it up!

 

[Njay]

Which 2 slopes?

 

[zombiess]

I am suspecting the same. But how it makes G/E ring like that.

It's captive coupling inside the power device.

 

[Arlo1]

Which 2 slopes?

On the turn on it starts with a slow slope then a second faster slope. Blue trace as you asked for on the left side where is goes down. When it goes up its a off event that I need to address.

 

[zombiess]

Is the below picture fairly close on how your layout is setup now? I highlighted each bus according to polarity and I see very little overlap of the bus bars.