EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: ghoetic on June 08, 2020, 12:28:32 pm
-
I have designed a 3 phase motor driver, which currently when trying to start with no load, trips for overcurrent. waveform for topside IGBT's seem wierd.
gatedriver: IRS2334 with local bypass 10uF electrolytic, 1 uF & 0.1 uF MLCC @ 15V
bootstrap capacitance 2.2uF.
integrated deadtime 420ns.
gate resistance: 20ohm
IGBT:STGP10M65DF2
Qg: ~28nC according to datasheet
low side gate waveforms.
rise: https://ibb.co/6nxb6XD
fall: https://ibb.co/NsHWTbw
topside gate waveforms.
rise: https://ibb.co/WDGPKrm
fall: https://ibb.co/k3GC7PJ
assembled board: https://i.ibb.co/3BF29CG/20200601-135204.jpg
i have swapped one of the bridges low and high side input signal to driver, i will fix that in firmware. but it should not be an issue right? its just inverted pulse train.
any ideas for the waveform?
Any ideas?
-
Still same issue, when i increased the current trip point i could just about see faint smoke coming from the irs2334 before it tripped again.
-
The linked images are broken or inaccessible.
Always upload to the forum directly, if possible.
-
Hi,
can you show the schematic of the IGBTs and the IRS2334 in pdf format?
What are the part number of the boostrap diodes D3 -D5?
Jay_Diddy_B
(Use this feature of the forum to store the images on the forum:
[attachimg=1]
-
low side rise
-
low side fall
-
high side rise
-
high side fall.
posting from phone, will fix schematic and part nr soon.
-
So what are you not liking there? You are looking at 50ns per division, unloaded output with a probably not the greatest probing setup on a likely not ideal PCB layout.
The high side waveform is clearly showing a deadtime of about a 1us (likely, where the mid-voltage is located at).
What voltage you have at the DC bus? What is driving the inputs? Try attaching at least some small load to it.
Can you show us the PCB layout? Just to check, if there are any places of concern
-
Here are pcb and schematic.
What it think is wrong? i am trying to find out haha. yes i have very bad probing setup, but something tells me it must be shoot through cause i am tripping for overcurrent.
i scoped the shunt resistor right at the legs and measured spikes up to 2.7v the shunt is pararell 200mOhm =~ 100mOhm.
video with wiff of smoke haha: https://www.youtube.com/watch?v=rlYBoqvyziw (https://www.youtube.com/watch?v=rlYBoqvyziw)
notes:
1: the diode for the bypass relay and current shunt are not the same as the gate driver, they are of another kind. can get that part number tomorrow, but they are fast and 1000v atleast.
the pulldown of the bypass relay i added afterwards, but i managed to solder one via the pads on this board.
-
i did tests at both 125 and 230VAC. should a load really be neccesary?
-
I build both brush and brushless motor drives, using IR gate drivers. One of the things about them is that they cannot stand the midpoint (U, V, W)
or VS1, 2, 3 going much below zero. I had to put ultrafast diodes across the low-side FET to prevent that. Now, in my case, it only was a problem with an inductive load. So, assume you have the high-side transistor on, sourcing 10 + A current out to the load, and then shut off the high-side transistor.
The load inductance will try to continue accepting current from the drive, and the midpoint of the two transistors drops below ground.
It turns out the body diodes in these fets will allow forward voltages of 10-12 V for well over a microsecond before they start to conduct. That's why an unltrafast diode is needed. I use an ESD3 for this.
What happens is that as the midpoint drops to around -7 V currents start leaking all over the gate driver chip, and eventually, it turns on BOTH transistors at the same time. Using a digital scope and creeping up on the current where this started to happen, I was able to capture the event without catastrophic failure.
Jon
-
I build both brush and brushless motor drives, using IR gate drivers. One of the things about them is that they cannot stand the midpoint (U, V, W)
or VS1, 2, 3 going much below zero. I had to put ultrafast diodes across the low-side FET to prevent that. Now, in my case, it only was a problem with an inductive load. So, assume you have the high-side transistor on, sourcing 10 + A current out to the load, and then shut off the high-side transistor.
The load inductance will try to continue accepting current from the drive, and the midpoint of the two transistors drops below ground.
It turns out the body diodes in these fets will allow forward voltages of 10-12 V for well over a microsecond before they start to conduct. That's why an unltrafast diode is needed. I use an ESD3 for this.
What happens is that as the midpoint drops to around -7 V currents start leaking all over the gate driver chip, and eventually, it turns on BOTH transistors at the same time. Using a digital scope and creeping up on the current where this started to happen, I was able to capture the event without catastrophic failure.
Jon
Thanks for your input. This i will def keep in mind.
But i aint even there yet with load :(
-
Ghoetic,
Further to what jmelson points out, page 23 of the datasheet has another solution to Vs going negative by adding a series resistor Rvs and clamp diode Dvs. link: https://www.infineon.com/dgdl/Infineon-IRS2334-DataSheet-v01_00-EN.pdf?fileId=5546d462533600a40153567aa9fe280b (https://www.infineon.com/dgdl/Infineon-IRS2334-DataSheet-v01_00-EN.pdf?fileId=5546d462533600a40153567aa9fe280b)
Questions for JMelson:
1.) any thoughts on why the IGBT's diode is so slow to turn on? I note that the data sheet does not show the forward recovery time and the diode's Vf graph shows a higher than expected voltage.
2.) have you observed this effect in other IGBTs?
-
Hi,
You may benefit form the resistor highlighted here:
[attachimg=1]
There will be one resistor for each of the diodes D3, D4 and D5.
If U,V or W goes negative this voltage is added to the Vcc voltage in the boostrap circuit. There are internal clamps at 25V.
The series resistor will help prevent the bootstrap circuit peak detecting the negative spike.
They may simply damp the resonance between the bootstrap capacitor, and the wiring inductance. This would also cause over-voltage of the high side drivers.
Regards,
Jay_Diddy_B
-
There are some better mitigation techniques for the negative voltage clamping, apart from improving the PCB layout: Place the gate resistor in the source lead connection and then add a clamp diode from DCbus- to the Vs terminal of the driver IC. But it also has some other disadvantages.
-
It works without load now https://youtu.be/NztFkpEPASs.
in video probing highside gate. showing transition from low to high frequency including overmodulation.
what i did:
replaced irs2334.
replaced bs caps to 1uF
replace 10uF bulk to 47uF
changed gate resistor lowest i had, 150ohm
tested it with just replacing the resistors but it tripped. so in my haste iforgot to check for what it tripped. which after changing resistor i noticed it was undervoltage... maby works with 20 ohms resistor and 2.2uF bs caps.maby the driver was bad?
going for a motor load soon. :-+
-
Hi,
For clarification:
[attachimg=8]
You changed
C10 to 47uF
C11,C12 and C13 to 1uF
R22-R27 to 150 \$\Omega\$
and you replaced U4
Other Observations
Brake IGBT
You may need to add another IGBT. This IGBT connects a high power resistor across the DC bus if the DC bus too high. The DC bus will rise if you decelerate rapidly. Energy from the mechanical rotation will be transferred back to the bus.
Opto-couplers in the linear mode
[attachimg=1]
The two analog channels for setting speed control and trip level is bad design practice. The transfer function is strongly dependent on the current transfer ratio, CTR, of the opto-coupler.
I suggest that you encode the signals with PWM.
Something like the LTC6992:
[attachimg=2]
Read the datasheet to lower the frequency to around 20kHz.
Filter the signal on the isolated side to obtain the analog voltage.
Isolated supply
[attachimg=3]
These TMR 1211 modules are not really designed to provide safety isolation from the mains.
The datasheet shows:
[attachimg=4]
[attachimg=5]
The insulation system is Functional
The 1600V dc is for 60s.
Compare to Traco TIM series:
[attachimg=6]
[attachimg=7]
This is safety rated.
EMC
You will need to add some EMC countermeasures.
You will need a filter including a CM choke on the input
Probably need RC circuits on the U,V,W outputs
The switching edges from the PWM waveform will be capacitively coupled in the motor to ground.
Regards,
Jay_Diddy_B
-
I build both brush and brushless motor drives, using IR gate drivers. One of the things about them is that they cannot stand the midpoint (U, V, W)
or VS1, 2, 3 going much below zero. I had to put ultrafast diodes across the low-side FET to prevent that. Now, in my case, it only was a problem with an inductive load. So, assume you have the high-side transistor on, sourcing 10 + A current out to the load, and then shut off the high-side transistor.
The load inductance will try to continue accepting current from the drive, and the midpoint of the two transistors drops below ground.
It turns out the body diodes in these fets will allow forward voltages of 10-12 V for well over a microsecond before they start to conduct. That's why an unltrafast diode is needed. I use an ESD3 for this.
What happens is that as the midpoint drops to around -7 V currents start leaking all over the gate driver chip, and eventually, it turns on BOTH transistors at the same time. Using a digital scope and creeping up on the current where this started to happen, I was able to capture the event without catastrophic failure.
Jon
^this. these IR gate drivers are notorious for latching "ON", even at far lower levels than jon is mentioning. IIRC, there's an old IR app note about the IR2113 floating around, explaining this behavior and ways to mitigate it. and no, testing without load doesn't cut it. the issue is mostly related to di/dt at the halfbridge's output, so test your contraption at full load _and_ overload. everything else is a testing placebo and will fly into your face further down your project timeline.
good luck!
-
Hi,
For clarification:
(Attachment Link)
You changed
C10 to 47uF
C11,C12 and C13 to 1uF
R22-R27 to 150 \$\Omega\$
and you replaced U4
Other Observations
Brake IGBT
You may need to add another IGBT. This IGBT connects a high power resistor across the DC bus if the DC bus too high. The DC bus will rise if you decelerate rapidly. Energy from the mechanical rotation will be transferred back to the bus.
Opto-couplers in the linear mode
(Attachment Link)
The two analog channels for setting speed control and trip level is bad design practice. The transfer function is strongly dependent on the current transfer ratio, CTR, of the opto-coupler.
I suggest that you encode the signals with PWM.
Something like the LTC6992:
(Attachment Link)
Read the datasheet to lower the frequency to around 20kHz.
Filter the signal on the isolated side to obtain the analog voltage.
Isolated supply
(Attachment Link)
These TMR 1211 modules are not really designed to provide safety isolation from the mains.
The datasheet shows:
(Attachment Link)
(Attachment Link)
The insulation system is Functional
The 1600V dc is for 60s.
Compare to Traco TIM series:
(Attachment Link)
(Attachment Link)
This is safety rated.
EMC
You will need to add some EMC countermeasures.
You will need a filter including a CM choke on the input
Probably need RC circuits on the U,V,W outputs
The switching edges from the PWM waveform will be capacitively coupled in the motor to ground.
Regards,
Jay_Diddy_B
Clarifications right on!
super helpful tips! :-+ :-+ :-+
you are right about the functional part. i totally missed it. My design is based on AN1660 from microchip but with some modifications.
Spinning and injection brakeing a no load motor:
https://youtu.be/SNKqk37RwU8
-
Questions for JMelson:
1.) any thoughts on why the IGBT's diode is so slow to turn on? I note that the data sheet does not show the forward recovery time and the diode's Vf graph shows a higher than expected voltage.
2.) have you observed this effect in other IGBTs?
I can't say for IGBT's. Since my servo amps are relatively low voltage, I only use big MOSFETs in them. The "body diode" is, in most cases, a side effect of the MOS structure, and not a designed-for-purpose diode. You CAN buy MOSFETs that have a separate diode that has much better charateristics.
But, i THINK that many IGBTs have a similar body diode that is an artifact of the transistor structure, and may have similar (poor) characteristics.
I have seen in the literature that experienced designers have said "never let the body diode conduct."
So, my observations are only certain for FETs, not IGBTs. And, the measurement I reported required a fairly large inductive load sinking current from the half-bridge.
The problem with the added resistor on VS1 is that it goes in series with the gate charging circuit of the high-side transistor. To be large enough to limit current, it probably will be too large for your desired gate charge/discharge time.
Jon
Jon
-
I've yet to see a diode with any significant forward recovery. As far as I know, it does not exist. Turn on time is dominated by parasitic inductance. Or not? I stay being corrected.
-
But, i THINK that many IGBTs have a similar body diode that is an artifact of the transistor structure, and may have similar (poor) characteristics.
I have seen in the literature that experienced designers have said "never let the body diode conduct."
That thinking is incorrect. Substrate diode in a mosfet is an unintentional structure, whose parameters such as the Qrr/trr are a result of a compromise in the design of the mosfet. You can't simply change parameters of one, without affecting the other.
IGBTs do not have any intrinsic diode in its structure. Hence why you can get IGBTs with and without integrated diode. The integrated diode is usually a separate die, that can be made using a different process with significantly better Qrr/trr than a similar mosfet substrate diode would achieve.
That kind of rule of thumb of "never let the body diode conduct." sure has some truth in it, but mostly with old nasty mosfets. There is plenty of choice these days with reasonably fast substrate diodes. Also, in a motor control application, Qrr/trr of mosfet substrate diode is mostly non-issue. Switching frequencies are rather low.
-
Ghoetic, have you considered the change that Yansi and I have suggested? (page 23 of the IRS2334 data sheet I referred to above.)
Infineon/IRF suggests Rvs not be greater than 5 ohms, I suppose to not affect gate drive.
Also, when I was looking for something else I found this on parasitic turning on: https://www.controleng.com/articles/tutorial-mitigating-parasitic-turn-on-effect-in-igbt-output-drives-to-improve-drive-performance/ (https://www.controleng.com/articles/tutorial-mitigating-parasitic-turn-on-effect-in-igbt-output-drives-to-improve-drive-performance/)
Jon, I also worked with power FETs, but it was years ago. I developed a 3 phase MOSFET servo amp in the mid to late 80s with a switching rate of 100 kHz. MOSFETs at that time had bog slow body diodes so I had to isolate it with a Schottky rectifier and then add a fast recovery diode to handle the freewheeling current. Fortunately Unitrode made a 50 ns, 50 A part but it was only available in a stud mount package. I recall that the forward recovery time was doggone fast (<< 100 ns) and more of a function of the loop area and inductance. If I remember my device physics correctly, an increased forward voltage simply causes the diode's depletion zone to close and conduct quicker. Within a few years, similar parts were available in TO-247 style packages and a good bit of the series inductance went away.
I haven't designed with IGBTs. I understand that they do not have body diodes as do MOSFETs, however they have other parasitic devices like a thyristor. I wonder if the IGBTs that have anti-parallel diodes add another die to the package or do some sort of trick with a diode connected transistor hence the higher forward voltage drop? Maybe it's just a smaller die that is optimized for speed and has a higher VF.
BTW, I tore down a 20 year old Mitsubishi VFD that had popped its IPM. I wanted to see if I could convert it to generate a variable frequency PWM pulse train to drive an external H bridge for a single phase lab test source for the lab. The potting compound was clear and I could clearly see the damaged IGBTs and that it had separate freewheeling diode dice.
Cheers,
-
I've yet to see a diode with any significant forward recovery. As far as I know, it does not exist. Turn on time is dominated by parasitic inductance. Or not? I stay being corrected.
Not entirely sure if we can draw conclusions about the forward recovery, but I remember seeing almost 2us turn off response latency difference in using a 1N4148 versus a BAT85/54 as an anti-saturation diode for a small signal BJT. Of course, a us may or may not be significant depending on your application...
-
Ghoetic, have you considered the change that Yansi and I have suggested? (page 23 of the IRS2334 data sheet I referred to above.)
Infineon/IRF suggests Rvs not be greater than 5 ohms, I suppose to not affect gate drive.
Also, when I was looking for something else I found this on parasitic turning on: https://www.controleng.com/articles/tutorial-mitigating-parasitic-turn-on-effect-in-igbt-output-drives-to-improve-drive-performance/ (https://www.controleng.com/articles/tutorial-mitigating-parasitic-turn-on-effect-in-igbt-output-drives-to-improve-drive-performance/)
Jon, I also worked with power FETs, but it was years ago. I developed a 3 phase MOSFET servo amp in the mid to late 80s with a switching rate of 100 kHz. MOSFETs at that time had bog slow body diodes so I had to isolate it with a Schottky rectifier and then add a fast recovery diode to handle the freewheeling current. Fortunately Unitrode made a 50 ns, 50 A part but it was only available in a stud mount package. I recall that the forward recovery time was doggone fast (<< 100 ns) and more of a function of the loop area and inductance. If I remember my device physics correctly, an increased forward voltage simply causes the diode's depletion zone to close and conduct quicker. Within a few years, similar parts were available in TO-247 style packages and a good bit of the series inductance went away.
I haven't designed with IGBTs. I understand that they do not have body diodes as do MOSFETs, however they have other parasitic devices like a thyristor. I wonder if the IGBTs that have anti-parallel diodes add another die to the package or do some sort of trick with a diode connected transistor hence the higher forward voltage drop? Maybe it's just a smaller die that is optimized for speed and has a higher VF.
BTW, I tore down a 20 year old Mitsubishi VFD that had popped its IPM. I wanted to see if I could convert it to generate a variable frequency PWM pulse train to drive an external H bridge for a single phase lab test source for the lab. The potting compound was clear and I could clearly see the damaged IGBTs and that it had separate freewheeling diode dice.
Cheers,
Would need a greater load test before i make more charges.
As i posted above, i am up and running now with a motor https://youtu.be/SNKqk37RwU8 (https://youtu.be/SNKqk37RwU8)
👍😊
-
Locking the motor's rotor is a great load test. Your controller should survive that.
The "double pulse" test is an easy (usually non-destructive) test you can make to evaluate the "quality" of your power stage.
Especially with the shorter switching times you really need a good test setup (ground clips directly over the power devices pins, beware of where the GNDs of the several equipments are connected) to look at the waveforms around the power devices.