821 - 840 of 2284 Posts

#### PStechPaul

·
##### Registered
Joined
·
3,141 Posts
It looks like you have a rather large footprint for your DC-DC converters, indicating that they may be 2 watts or greater. I think the actual power required may be much less, as I explained in a post a couple weeks ago:

I was wondering about the formula for determining the power needed to drive an IGBT gate:

P=Ug²/Rg*(ton+toff)*fPWM

For a 30V gate voltage swing, with 10 ohms gate resistance, 100 nSec ton and toff, and 20 kHz PWM, I get:

(900 / 10 ) * 0.2uSec * 0.02MHz = 0.36W

Is this correct? Otherwise the ton and toff might be better calculated as the TC of the gate resistor and gate capacitance, which typically about 6nF for a TC of 60 nSec, which is even less than my estimate above. And the current through the resistor would need to be integrated over the charge and discharge times, making power even less.

There will also be whatever power is consumed by the gate driver, and that may be greater than that in the gate resistor. The A3120 appears to have a power dissipation less than 300 mW. So I think a 1W DC-DC should be adequate for under 20 kHz. Its current capacity is not an issue because the current for the charge and discharge pulse is taken from the output filter capacitors which are many times larger than the gate capacitance.
I also ran a simulation which showed that it might be about 1 watt, depending on various parameters. Can anyone confirm or refute my findings, and determine an actual power requirement for the DC-DC? The IGBT-specific DC-DC supplies I found are 2W, and are reasonably enough priced not to be an issue, but I'd like to know what's really needed and how to determine the power requirement.

#### jackbauer

·
##### Registered
Joined
·
2,229 Posts
Success! It works. Something very interesting happened. I built one side of one board as a test as can be seen in the pictures. Connected it to my "blown" Cm600dy module that I removed from the inverter figuring it would do for a test. The board would not drive the gate and kept tripping the "Fault" output. Now this transistor can be switched on and off just fine on the bench but would not run in the inverter in the E31. Changed to a know good CM600dy and it burst into life. So it would seem the ACPL-337J could detect the damage in the "blown" module. Desat detection works fine on the bench also.

Paul , I am indeed using the Murata gate drive part that you recommended. It is physically bigger than the 2W 1kv isolated parts I had used.

#### Attachments

• 927.7 KB Views: 132
• 914.9 KB Views: 142
• 979.9 KB Views: 136

#### Tomdb

·
##### Registered
Joined
·
895 Posts
Nice to see it coming along so fast. Keep up the good work, but dont forget to take the breaks you deserve.

#### jhuebner

·
##### Registered
Joined
·
971 Posts
Discussion Starter · ·
Can anyone confirm or refute my findings, and determine an actual power requirement for the DC-DC? The IGBT-specific DC-DC supplies I found are 2W, and are reasonably enough priced not to be an issue, but I'd like to know what's really needed and how to determine the power requirement.
I studied an 8A concept driver a while ago and they are also using a 1W model. So I think you do have a point.

Good work with the drivers!

I did a first shut of an assembly video. I noticed that I mumble terribly. Comments apart from that welcome

#### Tomdb

·
##### Registered
Joined
·
895 Posts
Have to see quite clear and understandable. Even without a personal Mic.

However when showing computer things a screen capture is easier with voice over.

#### jhuebner

·
##### Registered
Joined
·
971 Posts
Discussion Starter · ·
Have to see quite clear and understandable. Even without a personal Mic.

However when showing computer things a screen capture is easier with voice over.
Yes, good one. Should be easy to reproduce.

#### jackbauer

·
##### Registered
Joined
·
2,229 Posts
I watched the video while eating dinner. Very well done and informative. I have had some good results with the car. Changed boost from 1700 to 750 , idlekp from 1 to 0.7 and deadtime from 28 to 50. Starting surge on 500rpm idle is now an easy 20A. Changed brake param from 30% to 10% and throttle now feels much better. That Siemens motor is a beast. even at only 280V and with the gearbox running in emergency mode (4th gear) , it moves that big 840CI as well as any V8

I am very happy with the drivers. Going to do a few more tests over the weekend and run one all day at 8Khz to make sure no heating problems. When the new boards arrive I will build a complete set.

#### jackbauer

·
##### Registered
Joined
·
2,229 Posts
So as it looks like my IGBT driver is at least working on the bench , here are the design files as promised. Format is DesignSpark PCB 7 but you don't need this as BOM and gerbers are included.

#### Attachments

• 85.1 KB Views: 168

#### jackbauer

·
##### Registered
Joined
·
2,229 Posts
Attached PDFs of schematic and pcb. Forgot to include in the zip.

#### Attachments

• 66.6 KB Views: 165
• 51.1 KB Views: 171

#### Gurgel

·
##### Registered
Joined
·
3 Posts
to : J. Huebner . This video is very important, very good at good resolution, the audio should be amplified in the next post (this is common in the youtube videos), after my kit arriving, I immediately I'll get the powerful IGBTs

#### arber333

·
##### Registered
Joined
·
975 Posts
Hi Johannes

As i said before, i have motor wound for 75VRMS! I was using it at 120VDC from battery and it ran fine albeit with huge RMS current demand.
Well now i tried your inverter at 300V and it works in a car! Yay! But... this was when i was running with wheels up.
Yesterday i went and set up for drive testing. It went as follows:
1. I connected battery and phase cables.
2. I gave inverter 12V and i got communication with linux
3. I set up parameters relevant for 300V. Min slip i put at 3.2Hz as i calculated, Max Slip=6Hz, Fweak=180Hz and Fmax=240Hz. Min slip Freq=1Hz I had it set at 3Hz but it caused more oscillation.
Boost worked best at 2400.

4. I tried to start and inverter worked. But when i pushed the pedal, car jumped forward and then stopped and the oscillation continued...
No matter what parameter i changed it either got worse or was the same as this.

I fear that my motor has less phase resistance than controller can manage. Probably i should get someone to rewind it to 180VRMS.
In the end i was trying to control RPM with clutch and i lost one phase, DCDC brick, driver and its caps! I will have to repair that...

What do you think?

A

#### jackbauer

·
##### Registered
Joined
·
2,229 Posts
That is almost the exact failure mode i encountered. You have probably killed the IGBT also. In my case the inverter required a HUGE surge current on startup to get to 500rpm idle. My hope is with the new driver boards we can detect this and just shutdown without damaging hardware.

#### arber333

·
##### Registered
Joined
·
975 Posts
That is almost the exact failure mode i encountered. You have probably killed the IGBT also. In my case the inverter required a HUGE surge current on startup to get to 500rpm idle. My hope is with the new driver boards we can detect this and just shutdown without damaging hardware.
Well that is true, for surge protection you have better/faster detection. However this doesnt change anything, since my car would still be non-drivable because my motor requires too much current to spin.
Now that i started to think about it again, i figure motor wants to spin to higher RPM than i want. I guess it is natural with having low resistance coils with 300VDC...

A

#### PStechPaul

·
##### Registered
Joined
·
3,141 Posts
I have often heard about problems with low resistance and low inductance motor windings, but I don't understand why they cannot be dealt with. The peak current of each PWM pulse before saturation depends on the winding inductance, applied voltage, and time, so I can see where a low voltage motor driven from a high voltage source could quickly see current rise to the point of saturation, at which point only winding resistance (and the rest of the circuit) limits the current.

The construction and operating conditions of the motor may have some effect on the inductance. If the gap between the rotor and the stator is wide, the inductance will be lower, so for a rewound motor it may be important to choose one that has a tight fit. I would think such close spacing would also provide higher torque, but it may also contribute to friction at higher speeds due to the "windage" or resistance of air. Then again, it may act as an "air bearing".

The effective inductance is also influenced by the coupling of the stator to the rotor, and the inductance and resistance of the rotor squirrel cage will be reflected as a complex impedance at the motor's windings.

In any case, however, there should be a practical lower limit to the possible inductance and resistance that can be presented to the controller by a motor and its connections, and the controller also should be able to withstand a short circuit. For that to be possible, there may need to be internal resistance and inductance that limit the di/dt of the current as well as its maximum value so that internal sensors have time to shut down the system before damage is done.

Adding inductance to the motor lead connections, along with some capacitance, may be able to filter out the PWM carrier frequency to some extent and improve efficiency as well as reduce radiated noise and losses in the cables, motor windings, and bearings. I don't think such inductors would need to be very large or costly. Probably some ferrite, powdered iron, or even tape steel toroids, with a few turns of wire, would be sufficient.

Another idea I had was to make an adjustable voltage bus so that, at low speeds, a higher PWM value could be used, and a much lower dV/dt would be imposed on the motor windings. This could be accomplished with a simple buck converter front end that could reduce the bus voltage from, say, 350 VDC, to 180 VDC or even 90 VDC. The converter need not handle the maximum power requirement, as that could be accomplished by bypassing it entirely with a power relay. Another alternative would be a relay matrix that could connect two or four battery packs in combinations of series and parallel depending on need. Transition may be difficult, but there may be ways to smooth out the doubling or halving of the bus voltage.

This might be a topic for a separate discussion, if anyone thinks it may have merit. I think I have seen articles on something similar, and I may have even proposed it elsewhere some time ago. It may or may not be practical or even feasible, but it may be interesting to talk about.

#### Tony Bogs

·
##### Registered
Joined
·
746 Posts
For an IGBT to survive this kind of short circuit (probably saturation at low PWM freq, no back EMF) slow turn-off is necessary.
Otherwise the resulting voltage surge at the collector (-L dI/dt, L= bus bar inductance, dI/dt extremely high) will almost certainly exceed the allowed max Vce.
Unlike mosfets, that have high energy avalanche capability, IGBTs are fried by overvoltage.
ACPL-33xJ family gate drivers have slow turn-off built in, so Damien's new gate driver should prevent IGBT failure if the bus bar inductance and other parts of the HV layout are within the limits for the power rating.
For more information on HV layout requirements: Semikron IGBT module application notes (chapter 5 of the Semikron power electronics manual).

A dV/dt filter introduces additional inductance in the phase lines. For more information and design procedure: application note AN-1095 irf.com.

Forgot: if the overcurrent (sensor board) kicks in before the IGBT desaturates, the 33xJ desat detection and slow shut-off are inactive and the result could be the potentially deadly hard turn-off at high amps.

#### jhuebner

·
##### Registered
Joined
·
971 Posts
Discussion Starter · ·

I can't see any problem with the parameters, except perhaps boost. 2400 (=19V) seems a bit much for a 75V (@50Hz?) motor. What other values did you try?

Seems like I might have to admit that over current detection reaches it's limits here. That said, with the boards that Damien has designed the current sensors could be made redundant.

#### jackbauer

·
##### Registered
Joined
·
2,229 Posts
Got the full set of pcbs today. I think I'd like to keep the current sensors even if it's just to monitor current.

#### arber333

·
##### Registered
Joined
·
975 Posts

I can't see any problem with the parameters, except perhaps boost. 2400 (=19V) seems a bit much for a 75V (@50Hz?) motor. What other values did you try?

Seems like I might have to admit that over current detection reaches it's limits here. That said, with the boards that Damien has designed the current sensors could be made redundant.
Yes but i tought i could make better start since slip is 3.2Hz.

I also tried boost with 700, 1400, 2400, 3000, 4000!
At 700 motor just threw OClimit, when pedal reached 30%, it didnt even turn. At 1400 it was turning but with too much force. With larger boost things just went worse...
I also varied with slip, max slip from 4Hz to 6Hz and i tried even low slip from 1.5Hz to 4Hz.

It just seems my motor is reacting to high voltage with acceleration that requires lots of amps. I can spin the motor in neutral but when loaded motor jumps dangerously.

A

#### Tony Bogs

·
##### Registered
Joined
·
746 Posts
Contrary to what one might expect, inverter power devices are subject to high thermal stress peaks at low speed.
Source : Semikron power devices application manual.

Fig 5.2.14 in chapter 5 (IGBT module application) shows the peaks in the junction temperature at startup (low phase frequency).

Fig 5.2.13 shows the huge steady state differences in junction temperature at very low freq (close to 0 Hz), 5 Hz and 50 Hz. Simulation result.

So it's better to think twice before running a setup at low phase freqs without checking the cooling.

821 - 840 of 2284 Posts