Paul and Sabrina's Cheap 3 Phase Inverter (AC Controller) with Field Oriented Control

[MPaulHolmes]

One possible guess is with the 165degree out of phase, it could get a DC bias going in the line filter, and maybe the line filter saturates, which would drastically reduce the inductance and makes the current coming from the driver too high. I moved to the ixdn604si that is surface mount, and has a metal tab on the underside for removing heat. I was noticing that the "no heat tab" version would get hot the touch (but not freakishly hot), but still worked. Hearing your results makes me think that there is too little headroom for that driver chip. But with the metal tab, they stay cool even switching 1200v 600amp IGBTs (I tested it over the last couple weeks on a grid tie). I'll be doing new driver boards that are all on the top side. They also use cheaper 10uF 1206 caps rather than the big fat 1210 caps.

Another issue I discovered about a week ago is, the FOD8316 pin 8 is sometimes NOT connected internally to ground, even though the datasheet says it is. So, I had to add a connection to ground from pin 8 (it allows turning on the LED inside) on the 3 phase grid tie board I recently did. I'm going to get new driver boards made. I'm very sorry it's causing you so much trouble.

[e*clipse]

Thanks a bunch Paul - no worries about my troubles; this is the world of R&D

Yes, I was concerned about the out of phase (not 165 degrees out of phase, 165degrees instead of 180degrees) so, the problem was about 30 or 35 degrees. Still that could be the problem, and I couldn't figure out how to fix it; I tried adding some delays to the code or reversing the order of the switching, but that didn't help. One of the main issues may be the high frequency, and that any delay ends up being too much.

Also - regarding that DC bias in the driver & filter - can you see where hooking a 'scope across the output of the IXDN604 or possibly the "output" of the transformer could cause enough DC bias to fry things?

One possible solution might be to use one pin from the DSPIC to drive a pair of logic gates. They might put out little or no delay if the right combo of buffers and inverters were used. One thing I like about using the heartbeat circuit or some pin from the pic to drive the power supply for the driver circuit is that power will shut off if the the DSPIC latches up for some reason.

I also noticed what seemed like too much heat from the IXDN604's and the 5V regulator. During the successful part of the testing, I wasn't driving any IGBT's, so there shouldn't have been any significant current flow through the IXDN604's or the transformers.

That issue about the FOD8316 is very interesting. Was this causing any latch up or similar issues with the inverter that used it? When I connected and started switching the PWM circuits (at about 5kHz) For the tests, one part was always off, while I adjusted the pulse width of the second. I wasn't able to see any output of the circuit using my 'scope. There was nothing except the scope probes connected across the outputs of the driver circuit.

I'm also very interested in your new driver design; especially that they are all on the top. The current design is a bit "thick" considering the transformers on top and bottom. Also anything to build in some robustness would be good; Somehow it has to be robust enough to survive my experiments but also eventually run an EV.

Thanks again for your help - I'll look into the timing & FOD8316 pins.

-e*clipse

Originally Posted by MPaulHolmes

One possible guess is with the 165degree out of phase, it could get a DC bias going in the line filter, and maybe the line filter saturates, which would drastically reduce the inductance and makes the current coming from the driver too high. I moved to the ixdn604si that is surface mount, and has a metal tab on the underside for removing heat. I was noticing that the "no heat tab" version would get hot the touch (but not freakishly hot), but still worked. Hearing your results makes me think that there is too little headroom for that driver chip. But with the metal tab, they stay cool even switching 1200v 600amp IGBTs (I tested it over the last couple weeks on a grid tie). I'll be doing new driver boards that are all on the top side. They also use cheaper 10uF 1206 caps rather than the big fat 1210 caps.

Another issue I discovered about a week ago is, the FOD8316 pin 8 is sometimes NOT connected internally to ground, even though the datasheet says it is. So, I had to add a connection to ground from pin 8 (it allows turning on the LED inside) on the 3 phase grid tie board I recently did. I'm going to get new driver boards made. I'm very sorry it's causing you so much trouble.

[wehzeh]

I'm still waiting to get my hands on all the updated code/schematics! The repo needs commits!

[Astro]

Originally Posted by MPaulHolmes

...
I improved the PI loop tuning test so it's successful with a larger range of motors now. The problem was, with some really big motors, in the PI loop test, if you command a current step, properly chosen P and I constants will give convergence in close to 1millisec. So, I had been forcing successful PI loop tuning tests to converge in 2mS for all motors. But yesterday I hooked up a 1.5HP permanent magnet motor, and it required 15ms for convergence. So the test always failed with that motor. Now, I have it run through the possible PI values, allowing 2mS, and if it all fails, then all PI values allowing 4mS, and then 6mS, ... It also halts tests sooner if any oscillation is observed, since oscillation indicates the PI values are too big anyway.
...

Will the tuning test work with the motor in the vehicle?
On the bench it would be just the motor shaft that needed to move but in the car there may be a motor to flywheel adaptor, flywheel, clutch housing, clutch plate and gearbox input shaft. Would this additional additional weight (inertia) throw off the tuning test?
Just wanted to work out whether i need to pre-test the motor before putting it in the vehicle.

[MPaulHolmes]

Oh no, you can just put the motor in the car. It doesnt have to be tested on a bench. The PI test is actually a locked rotor test, but it only lasts like 1/20 of a sec so yhe post doesnt have time to spin anyway.

[e*clipse]

I really like this driver design; What do you think of this idea:

replace the ATTiny with a triple inverter: ( i really don't have anything against them; I'd just like to make things a bit simpler by reducing the number of microcontrollers...)

The main "heartbeat" signal from the DSPIC would still exist as one 50% duty cycle 62.5kHz signal. This would go to the driver boards where an inverter would be in the place of the ATTiny. The inverter I found has the same pin count and the same pin locations for VCC and GND. The inverter would act like a buffer for one of the drivers and invert the second. The inverter I found only has between 1.7 and 3 nS propogation delay, which is like 1/16 of a degree phase angle.

The inverter is a NXP Semiconductors 74LVC3GU04DP

Thoughts??

- Eclipse

Originally Posted by MPaulHolmes View Post
One possible guess is with the 165degree out of phase, it could get a DC bias going in the line filter, and maybe the line filter saturates, which would drastically reduce the inductance and makes the current coming from the driver too high.

[MPaulHolmes]

The only reason I like the attiny is, the very high impedance 62.5kHz signal is very short and never leaves the PCB, which has a ground plane running right under the high frequency signals. I've never had a problem getting the isolated supplies working using it. The dspic has a few interrupts running. Could it be that the 62.5kHz timer sometimes gets put on hold due to other interrupts? With the ATTiny It's an identical number of 1MHz cycles on as off at 1MHz. So, any variation would only be in the difference between turnoff and turnon time of ports. The inverter chip would work I bet, but I would just be worried that you would really have a perfectly consistent 62.5KHz signal due to interrupt issues. I know you could make the 62.5kHz timer the highest priority. maybe that could eliminate jitter.

[e*clipse]

Ahh, I see the point of the ATTiny as a custom "make switching signal" IC. Shoot - it might even be a good way to make a base signal for my resolver-driver.

I do also like the point of keeping the high-frequency signals isolated.

In this case, I'm using much different code. The application is actually a dual H-bridge DC drive, ( mini electric snow cat ) and I only have one interrupt to provide timing for this signal and a blinky LED. The driver signal was, from what I could see on the 'scope, very stable, it just had that lag problem.

You've encouraged me to try some new stuff. I'm going to get some inverters and see things work out with the ATTiny's as well.

Thanks for your help,
E*clipse

[chimchichclick]

hello im a newer. im know that the motor parameter is very very important in this FOC.we need it to estimate angle roto flux. im used to use it in my project with TI chip (f28069 ) but i have so many experiment to find some parameter (when i at experiment lab in school ) . about Tr = Lr/Rr = Rotor time constant. and if we need sensorless we need more parameter to estimate speed. but now my motor dont have them.its just have volt/apm/speed... .and i cant do experiment because i dont have enough equipment like when i at school. So. if u have a solution about this. pls help me. we will disccuss about this. thank u. and sorry for the Bad english skill lollll )

[chimchichclick]

And Attiny 25. i dont see sourgce code in wiki ! sorry because i not good to read english. so read 280 page to find it is teriable.where is it guys ??? thanks you very much!