Continue to Site

Welcome to EDAboard.com

Welcome to our site! EDAboard.com is an international Electronics Discussion Forum focused on EDA software, circuits, schematics, books, theory, papers, asic, pld, 8051, DSP, Network, RF, Analog Design, PCB, Service Manuals... and a whole lot more! To participate you need to register. Registration is free. Click here to register now.

[SOLVED] IGBT Overshoot and Weird Slope

Status
Not open for further replies.

Antor

Member level 2
Joined
Mar 1, 2019
Messages
51
Helped
0
Reputation
0
Reaction score
0
Trophy points
6
Activity points
501
Hello,

I'm trying to build a 3phase-3phase direct converter (Explains my schematics). I'm testing IGBT stage with DC source. Circuit is now just triggering IGBT named SAAF.
Cutting to the chase, I'm triggering only 1 IGBT and chopping a DC for now. I'm sure there will be negative undershoot if I trigger SAAR negative path IGBT too.

There are 3 problems;
1- Overshoot
2- Very bad transient
3- A weird slope after turn-on.

Here is main schematic and its's chield
MainSchematic.JPGMainSchematic2.JPG

These are the PCB with top and bottom layers
PCB2.JPGPCB.JPG

Signal triggering the IGBT driver (HCPL3140) is generated by signal generator. 5V %50 duty cycle @1KHz
IMG_20190914_203538.jpg

Load is a rheostat (around 10ohms, may varry)

Yellow : Output of the IGBT, Blue : Signal triggering IGBT (HCPL3140 output)
DC source of the IGBT 1A/24.2V, Output have an overshoot and a weird slope during on cycle. PWM signal is not in a good condition also.
003- 1A 24.2V.jpg

DC source of the IGBT 1.24A/30V, Output have an overshoot and a weird slope during on cycle. PWM signal slightly better.
005- 1.24A 30V.jpg

DC source of the IGBT 1.33A/31.7V, Output have an overshoot and the weird slope is gone.
001- 1.33A 31.7V.jpg

Close shots of overshoot and transient state of turn off
IMG_20190914_203607.jpgIMG_20190914_203713.jpg

Workspace
IMG-20190914-WA0049.jpg

How can I condition the output ?
 

Hi,

There are electronic devices called capacitors.
Do you know them?

Klaus

Added:
Your schematic violates general rules how to use gate drivers as well as IGBTs.
The same applies for the PCB layout.

Every IGBT manufacturer provides datasheets, application notes and design notes how to use IGBTs.
Every driver IC manufacturer provides datasheets, application notes and design notes how to use IGBTs.
All the infirmations are free to use and are very urgent for proper operation.
I strongly recommend to read through them and follow the informations.
 

Hi,

I will only point out things that I think may be direct contributors, so the list is not exhaustive but just a starting point.

=Your IGBTs are not properly connected.
=See the "Applications Information" section of the driver datasheet and follow the recommendations there. Your PCB layout is poor. Also take note of Figure 19.
=You haven't mentioned what the gate driver supply voltage is. Look at that again.
=Your test points and signal supply points: The wires that you are connecting your signal generator and oscilloscope probes to are way too long. Are they even soldered? Pogo pins are preferrable.
That GW INSTEK power supply is a switched-mode type power supply (do not doubt this, I have it) with a lot of ripples. You need to properly decouple them.
=I do not understand why the triangular waveform of your signal generator is active instead of the square wave. That does not correlate with the near square wave that is being displayed on the scope though but you may need to consider checking that out. I also see a type of output labelled "GATE", maybe you should check what that is. Have you properly tuned you oscilloscope and other BNC probes that you are using?
=The peforated metallic stuff lying on the bench to the left of your setup, what's that? Hope it's not a source of EMI.
 
Last edited:

...
=You haven't mentioned what the gate driver supply voltage is. Look at that again.
...
Let me elaborate more on this very point. Each of your HS drivers need a bootstrap circuit. Look at your schematic again. You need to rearrange the whole schematic.
 

you appear to have no caps across the DC bus ...
There's no DC bus in a "direct converter", which constitutes already some of the circuit problems.

It's not clear where you measure the "IGBT output". Why we see a negative pulse although the positive switch SAAF is operated?

According to oscilloscope waveform, overshoot occurs on turn off. Overshoot can be expected with unclamped switching of the inductive rheostat load. Don't know what's the intended final converter application. Some means to clamp inductive overvoltage may be required to protect the IGBTs, e.g. active clamping.

You seem to supply the gate drivers by an external power supply, may be even a lab supply. Did you realize that the gate driver supply must have low output capacitance and sufficient isolation for mains voltage? A dedicated gate driver DC/DC converter should be used.
 

Thanks for reply mate. Going down the list one-by-one. Sorry if sometimes I look rude, it is unintentional and caused by my lack of english skills.

=Your IGBTs are not properly connected.
It is not an DC-AC inverter, it is an AC-AC converter. IGBT's connected the way how a bidirectional switch operates. The PCB is (3Pin 1Pout) only 1 phase power module of a 3Pin3Pout system.
1-s2.0-S1569190X10001954-gr1.jpg

I preferred common emitter bi directional switch, because common collector topology needs complex connections between all power modules. Drawback here is I need 9 isolated DC sources (for 3x3) instead 6. But testing layers seperately is avaible with common Emitter IGBT driver source is 0-15V.
bidir.JPG

=See the "Applications Information" section of the driver datasheet and follow the recommendations there. Your PCB layout is poor. Also take note of Figure 19.
According to datasheet and my need (peak Ig current 0.6A) Rg > (15-5)/0.6 = 16,67, so Rg choosen as 22ohm to reduce a bit emi.
If we are talking about same Figure 19 below, please consider that I'm not using the topology on figure 19.
fig19.JPG

PCB layout is poor if you say so, I have no experience on this. Could you point out my PCB mistakes so I can redesign and order again ? I couldn't find a good source to read about PCB design, so I tried to avoid drawing coils as much as I can.

=You haven't mentioned what the gate driver supply voltage is. Look at that again.
It's 15V, since I'm using low power IGBT I don't need negative voltage to turn-off. And yes, my isolated 15V sources schematic is below. There is bulk and hf cap at Vout, ready for PWM (I think).
3x15v.JPG

=Your test points and signal supply points: The wires that you are connecting your signal generator and oscilloscope probes to are way too long. Are they even soldered? Pogo pins are preferrable.
That GW INSTEK power supply is a switched-mode type power supply (do not doubt this, I have it) with a lot of ripples. You need to properly decouple them.

The wire I'm using for signals are shielded cables (6x0.22). Circuit and cables are earthed. Earth cable yes, it's long and consist of few bad condition cables.

What is to be decoupled ? If I decouple signal generator output doesn't it delay the signal ? Decoupling the output of HCPL3140, I guess it causes delay too. I guess I need a bit help in here.

Do I need shorter oscilloscope probes ? I was guessing their length is important so they are fixed. GW Instek is all I have.

I do not understand why the triangular waveform of your signal generator is active instead of the square wave. That does not correlate with the near square wave that is being displayed on the scope though but you may need to consider checking that out. I also see a type of output labelled "GATE", maybe you should check what that is. Have you properly tuned you oscilloscope and other BNC probes that you are using?

I'm using TTL/CMOS output, it's not important which button is active (sin, tri, square), it produces 50% duty cycle square wave at 5V fixed (unless I pop duty cycle pot).
How probes can be tuned ? I need help in here.

The peforated metallic stuff lying on the bench to the left of your setup, what's that? Hope it's not a source of EMI.
It is rheostat. Yes it's a good source of EMI. Also it screams with high frequencies (seriously). It's surface is earthed. I can try moving it further.

Let me elaborate more on this very point. Each of your HS drivers need a bootstrap circuit. Look at your schematic again. You need to rearrange the whole schematic.

Bootstrap constrained by the need of bootstrap capacitor refresh. It makes direct converter more complex.

It's not clear where you measure the "IGBT output".
Measurement taken at the SAAR collector and Neutral (Vout point in schematics)

Why we see a negative pulse although the positive switch SAAF is operated?
n channel igbt and common emitter causes 180degree phase shift.

According to oscilloscope waveform, overshoot occurs on turn off. Overshoot can be expected with unclamped switching of the inductive rheostat load. Don't know what's the intended final converter application. Some means to clamp inductive overvoltage may be required to protect the IGBTs, e.g. active clamping.
I was planning to build clamp circuit later but as you point out the bleeding wound I guess I need it quicker. Thank you. I'll be on this ne

You seem to supply the gate drivers by an external power supply, may be even a lab supply. Did you realize that the gate driver supply must have low output capacitance and sufficient isolation for mains voltage? A dedicated gate driver DC/DC converter should be used.
This is the gate driver supply schematic. Do you advice me to lower the capacitor values ?
3x15v.JPG
Do you advice me to use one of these ?
https://www.mornsun-power.com/uploads/pdf/QAUxx.pdf
 
Last edited:

This is the gate driver supply schematic. Do you advice me to lower the capacitor values ?
No. I referred to the isolation capacitance, e.g. the transformer interwinding capacitance. The linear voltage regulator circuit is O.K. But it's not so good to have the power supply and transformer 1 m apart from the gate driver circuit. You create large antennas radiating switching noise.

Measurement taken at the SAAR collector and Neutral (Vout point in schematics)
Do you mind to mark the connection of both probes in your circuit? I still can't relate your description to the shown waveforms.
 

No. I referred to the isolation capacitance, e.g. the transformer interwinding capacitance. The linear voltage regulator circuit is O.K. But it's not so good to have the power supply and transformer 1 m apart from the gate driver circuit. You create large antennas radiating switching noise.
15V supply enclosures are metal and well earthed. I have to use long cables for first prototype. Do you advice me to use twisted pair cables ? Shielded cables ? Ferrite Choker ?

Do you mind to mark the connection of both probes in your circuit? I still can't relate your description to the shown waveforms.
MainSchematic2.jpg

Oscilloscope earthing is isolated. Another terms, oscilloscope has no earthing for some reasons.

This slope exists if DC Supply current is under 1A and 30V
003- 1A 24.2V_LI.jpg

Disappears if DC supply current is over 1A and 30V (1.33A and 31.7V at this measurement)
001- 1.33A 31.7V.jpg

I'm using DC source for converter for testing at the moment. Triggering only SAAF, It should act as DC chopper for now.
 

How can you measure with different ground connection for blue and yellow trace simultaneously? The real circuit must be somehow different.

- - - Updated - - -

Oscilloscope earthing is isolated. Another terms, oscilloscope has no earthing for some reasons.
Disconnecting the oscilloscope's protective earth wire in mains connection doesn't provide good isolation. Besides safety issues, there's still a considerable capacitive load between probe ground and earth affecting the measurement signals.
 

How can you measure with different ground connection for blue and yellow trace simultaneously? The real circuit must be somehow different.

Yes they are different grounds, It is just ment to observe overall. Trigger is on CH1 and it's on the load which I want to observe best. Disconnecting blue signal doesn't effect yellow. Still same.
IMG_20190914_203542.jpg

Disconnecting the oscilloscope's protective earth wire in mains connection doesn't provide good isolation. Besides safety issues, there's still a considerable capacitive load between probe ground and earth affecting the measurement signals.

What is your suggestion about this issue ?

I'm more worried about the slope between rise and fall. Why there is a slope ? Why it decreases while current increasing ?
 

Besides unsuitable probe connection it may be dropping DC supply. I won't care about it much. Overshoot is much more a problem. It probably kills the IGBTs if you connect high input voltage.

To measure voltages with arbitrary ground reference, a differential probe is the preferred solution.
 

That rheostat must be a very large one. It must be having significant inductance and switching its current abruptly like that might be causing the spike across the SAAF IGBT. Do you mind comnecting a snubber across it?

- - - Updated - - -

How probes can be tuned?

You just connect the oscilloscope probe BNC end to a channel on your scope and then connect clip end to the trigger output (two metal connection points) of your scope as if to measure the trigger signal. When you are done with the connection, you adjust the pot on the scope probe until you get a nice square waveform on the scope display.

- - - Updated - - -

...
It's 15V, since I'm using low power IGBT I don't need negative voltage to turn-off. And yes, my isolated 15V sources schematic is below. There is bulk and hf cap at Vout, ready for PWM (I think).
View attachment 155525



The wire I'm using for signals are shielded cables (6x0.22). Circuit and cables are earthed. Earth cable yes, it's long and consist of few bad condition cables.
...

Why powering the driver with a supply not isolated from the mains? You should consider using channel 4 of your bench power supply to power it. That way you would be free from mains supply and multiple earthing connections and possible ground loop.

By bad condition cables, what do you mean?

- - - Updated - - -

What is GNDGRID referring to in your schematic of Post #1?

- - - Updated - - -

NOTE!!!

All the things I have suggested are for the DC testing only. You would have to analyze the risks before you carry out the AC testing.

- - - Updated - - -

WARNING!!!

Do not attempt these measurements if you are not sure what you are doing. Oscilloscope measurements on mains powered circuits are a major safety concern.
 

That rheostat must be a very large one. It must be having significant inductance and switching its current abruptly like that might be causing the spike across the SAAF IGBT. Do you mind comnecting a snubber across it?
Rheostat is a simple one, 100ohm total, 250Watt. Snubber is not an option at this point. If those overshoots are not a measurement error, I'll try varistor to oppress/suppress (I couldn't decide which one is suitable) or try a clamp like below.
Three-phase-matrix-converter-with-diode-clamp-circuit.png


How probes can be tuned?
You just connect the oscilloscope probe BNC end to a channel on your scope and then connect clip end to the trigger output (two metal connection points) of your scope as if to measure the trigger signal. When you are done with the connection, you adjust the pot on the scope probe until you get a nice square waveform on the scope display.

I really don't believe that our cheap probes have a pot but I'll check tomorrow. I'm considering to buy differential probes as FvM suggested

Why powering the driver with a supply not isolated from the mains?
Schematics doesn't show but it's isolated with a transformer. AC IN 151 means terminals of seconder coil 1 (independent from other 2.3...).


You should consider using channel 4 of your bench power supply to power it. That way you would be free from mains supply and multiple earthing connections and possible ground loop.
If you are telling me to use gw instek power supply unit, I don't believe in that power supply channel's are totally isolated. (Channels are in independent mode - series or parallel modes inactive).
I have only 1 earthing connection and it's grid's earth (building's earth). I couldn't understand "ground loop".

By bad condition cables, what do you mean?
Earthing cable, it's not a good cable but It does the job. It consist of multiple cables added to each other. That's what I ment with bad condition.

What is GNDGRID referring to in your schematic of Post #1?
When inputs are AC 3Phase, GNDGRID means grid neutral. For testing it's DC supply 0/neutral.

All the things I have suggested are for the DC testing only. You would have to analyze the risks before you carry out the AC testing.
I'll consider this advice and use a variac/autotransformer and test AC with 1/4 of grid voltage.

Do not attempt these measurements if you are not sure what you are doing. Oscilloscope measurements on mains powered circuits are a major safety concern.
I'm behind a few fuses, circuit breakers and leakage current protection relays (I'm not sure if the last one is translated correctly). Also I'm keeping a 1kg small fire extinguisher.

Thanks for all.
 

You really need a scope with properly isolated inputs - to measure things in your circuit - having too many ground connections at the same or different points can make it very hard to measure true currents - draw out you circuit very carefully to make sure what you build corresponds exactly with what you want ...
 

The earthing cable should be a solid copper conductor of a certain cross-sectional area. You know it already. You're just ignoring it.

- - - Updated - - -

20190916_074727.jpg

Even when you have a differential probe, you still have be able to tune your conventional oscilloscope probes to display correct waveforms when you are using them. You see that round stainless steel on the scope probe that looks like the head of a screw, that's where you adjust for the nice square wave I talked about earlier on.
 
Last edited:

The earthing cable should be a solid copper conductor of a certain cross-sectional area. You know it already. You're just ignoring it.

- - - Updated - - -

View attachment 155539

Even when you have a differential probe, you still have be able to tune your conventional oscilloscope probes to display correct waveforms when you are using them. You see that round stainless steel on the scope probe that looks like the head of a screw, that's where you adjust for the nice square wave I talked about earlier on.

I have been suspecting why there is a little screwdriver in probe packs. Now it's clear =) I'm heading to my workshop. Thanks a lot.

Update : There is screwdriver but no screw =( Rest of the probe is a whole piece that I can't apart.
IMG_20190916_114612.jpgIMG_20190916_114600.jpg
 
Last edited:

Adjustment is either at the probe or the probe connector. Why don't you read the manual?
 

Adjustment is either at the probe or the probe connector. Why don't you read the manual?

Probe connector is a simple BNC connector. There is no screw. I borrowed probes from a friend and yes they had a screw to calibrate. I calibrated the proble with test signal on oscilloscope. Nice sharp squares.

=Probe calibration check
=Using shielded short cables for gate driver source check
=Measured from the pins of the PCB check (no cables around)
=Load metalic surfice earthed
=Circuit board earthed

Problem is overshoot stil exists.

I didn't isolated heat sinks from IGBT, I used thermal paste for better cooling. Would it be overshoot problem ?
 

Thermal paste conducts heat. They do not conduct electricity. In fact, it acts like a dielectric with the IGBT and the heatsink acting like the two parallel plates of a capacitor, the thermal paste sandwiched between them. If the heatsink is connected to earth, then the setup produces common-mode noise when switched rapidly.

Single-ended oscilloscopes like the ones we find everywhere can pick up this noise.

- - - Updated - - -

If your scope was picking up common mode noise, I believe the noise would be at a much higher frequency.
 

Status
Not open for further replies.

Similar threads

Part and Inventory Search

Welcome to EDABoard.com

Sponsor

Back
Top