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

[MPaulHolmes]

AWESOME VIDEOS!!!!! YOU ARE A NATIONAL TREASURE! Thank you so much for taking the time to do those videos. You are so thorough. I couldn't have asked for a better beta tester.

The data that I can stream is:
High voltage battery pack amps (found indirectly)
Id
Iq
Iphase1
Iphase2
Iphase3
Vd
Vq
dutyPhase1
dutyPhase2
dutyPhase3
rotorFluxAngle
rotorFluxSpeed
RPM (both electrical and mechanical)
base plate temperature
IdRef
IqRef
um.... I'm sure there are others. And I can put it into the RTD format no problem. Which variables do you want to see? Maybe we could have a variable where you can choose which set of variables will stream. Maybe a 16 bit word where it's 1 if you want that one, and it's 0 if you don't want it? What is the RTD format that you have now? You name it, and BOOM it's yours. haha.

For me, I think it would be interesting to see Id, Iq, IdRef, IqRef, magnitude of <Vd,Vq> (to let us know how much of the voltage is available to us still for a given RPM), electrical RPM, ...
Just a repost of the 24v test, showing up big so people are more tempted to watch! haha

https://www.youtube.com/watch?v=pehC...ature=youtu.be

[thingstodo]

Originally Posted by MPaulHolmes

The data that I can stream is:
High voltage battery pack amps (found indirectly)
Id
Iq
Iphase1
Iphase2
Iphase3
Vd
Vq
dutyPhase1
dutyPhase2
dutyPhase3
rotorFluxAngle
rotorFluxSpeed
RPM (both electrical and mechanical)
base plate temperature
IdRef
IqRef
um.... I'm sure there are others. And I can put it into the RTD format no problem. Which variables do you want to see? Maybe we could have a variable where you can choose which set of variables will stream. Maybe a 16 bit word where it's 1 if you want that one, and it's 0 if you don't want it? What is the RTD format that you have now? You name it, and BOOM it's yours. haha.

For me, I think it would be interesting to see Id, Iq, IdRef, IqRef, magnitude of <Vd,Vq> (to let us know how much of the voltage is available to us still for a given RPM), electrical RPM, ...

I like being able to toggle on and off data from the stream. I don't know if that will mess up RTD explorer, but it sounds like a GREAT feature.

I have a beta of RTD explorer that does not require XP SP3. I - sadly - have not been able to test it since my Cougar is down for the count. I think the Cougar is at 1.11b firmware.

What I was trying to do on the AC side was pretty simple:
- High voltage DC pack Volts and Amps
- AC controller output Volts and Amps
- Temperatures for everything you can. I need to make sure I don't cook something during a test!

I will still need to spot check temperatures on a bunch of stuff manually once in a while during tests.

I see pack amps, indirect. Is that accurate enough for doing an efficiency calculation? Or should I have a shunt and rig up a display of some sort?

Vd and Vq - do we need to do a vector sum or is it just the real power that is being measured? I think just the current doing work is measured ... it's been a while so I need to look up some stuff!

It'll be the same for Id and Iq.

Can you do electrical RPM and mechanical rpm? That would show slip under various loads (and the slip DOES change). When you reach peak torque, the slip reaches a critical point, then the motor loses synchronization and stalls (don't ask me how I know). This would be good for the racers to determine how far they can push the motor, and what their absolute maximum torque out is!

DC Volts and amps on the DC motor side would help - V and I in, V and I out. I'm struggling with the Cougar to determine how many MOSFETs failed and which ones.

The next video should be better (not as much background noise). I think I figured out (with the help of many tutorial videos) how to use Google Hangouts live to have 2 cameras and use a noise cancelling bluetooth headset so I can talk above the howl of the motors. So far, the main camera will be like the present videos. The second camera will be pointed at meters and displays. It should be synchronized so I don't need to spend time merging things together after taking the video. All I should need to do is remove the extra-boring stuff, or the parts when I have to look for tools, chase the cat away, etc

I may run out of multi-meters So - Paul - which information is most useful to you?

[MPaulHolmes]

Originally Posted by thingstodo

I see pack amps, indirect. Is that accurate enough for doing an efficiency calculation? Or should I have a shunt and rig up a display of some sort?

It was accurate when I used a 100mV = 100amp shunt connected to the DC bus, so I wouldn't bother with doing that.

Vd and Vq - do we need to do a vector sum or is it just the real power that is being measured? I think just the current doing work is measured ... it's been a while so I need to look up some stuff!

Ya, I get used to using the terms, and I forget what the heck they are. haaha. My understanding is, Vd is the voltage responsible for the stator field generation, so I think Vd*Id would be the instantaneous reactive power, and Vq*Iq + Vd*Id would be the true (or apparent) power being used by the system, so I guess the power factor would be Vq*Iq/(Vq*Iq + Vd*Id). Hey, I bet power factor would be a nice variable to see too.

Can you do electrical RPM and mechanical rpm? That would show slip under various loads (and the slip DOES change). When you reach peak torque, the slip reaches a critical point, then the motor loses synchronization and stalls (don't ask me how I know). This would be good for the racers to determine how far they can push the motor, and what their absolute maximum torque out is!

This is one of the many reasons why you are straight out of Fort Awesome. You bet we can do electrical and mechanical RPM. I just didn't know it would be so useful!

DC Volts and amps on the DC motor side would help - V and I in, V and I out. I'm struggling with the Cougar to determine how many MOSFETs failed and which ones.

Unfortunately I don't have DC voltage as a possible measurement We can look at the 3 phases of V and the 3 phases of I on the output side. We could then do an instantaneous Vphase1*Iphase1 + Vphase2*Iphase2 + Vphase3*Iphase3 to get the instantaneous power out?

The next video should be better (not as much background noise). I think I figured out (with the help of many tutorial videos) how to use Google Hangouts live to have 2 cameras and use a noise cancelling bluetooth headset so I can talk above the howl of the motors. So far, the main camera will be like the present videos. The second camera will be pointed at meters and displays. It should be synchronized so I don't need to spend time merging things together after taking the video. All I should need to do is remove the extra-boring stuff, or the parts when I have to look for tools, chase the cat away, etc

That is awesome. I would very much like to know how to do it too when you figure it out.

I may run out of multi-meters So - Paul - which information is most useful to you?

Gosh, I think there are a bunch of interesting tests I would like to see. First I think it would be good to see that Id really is following IdReference, and Iq is following IqReference on your motor. Then, because we used the "autotune" PI feature, I would feel good about it. I'm not sure of what the new RTD format is, but this data can be streamed right to a text file, and then pasted into Excel. That would allow updates at maybe 1KHz rather than just 10Hz, not that 1KHz would be required. You wouldn't see it live as it happens, but could go back and see what the last minute looked like on a graph.

[thingstodo]

Originally Posted by MPaulHolmes

It was accurate when I used a 100mV = 100amp shunt connected to the DC bus, so I wouldn't bother with doing that.

Excellent - Pack amps measurement not required!

Ya, I get used to using the terms, and I forget what the heck they are. haaha. My understanding is, Vd is the voltage responsible for the stator field generation, so I think Vd*Id would be the instantaneous reactive power, and Vq*Iq + Vd*Id would be the true (or apparent) power being used by the system, so I guess the power factor would be Vq*Iq/(Vq*Iq + Vd*Id). Hey, I bet power factor would be a nice variable to see too.

I don't believe I have a method to measure reactive power. Vq and Iq would be great to log. Vd and Id are optional for me as is the power factor. I don't think you can really do anything to change power factor unless you want to add some weight and put an LRC filter in parallel with the motor. It would be more efficient ... at moving the heavier vehicle ...

Unfortunately I don't have DC voltage as a possible measurement We can look at the 3 phases of V and the 3 phases of I on the output side. We could then do an instantaneous Vphase1*Iphase1 + Vphase2*Iphase2 + Vphase3*Iphase3 to get the instantaneous power out?

I can put a meter on it for now. If you want to calculate it and log it we can compare and see how accurate the calculation is (likely more accurate than my measurement!) But it would be good to prove it one way or the other.

That is awesome. I would very much like to know how to do it too when you figure it out.

I will post a link to the tutorial .. perhaps maybe 2 .. that were the most help if it works as described. Multiple video sources mixed to one video is kinda exciting for me!

First I think it would be good to see that Id really is following IdReference, and Iq is following IqReference on your motor.

Hmm.. As long as those are logged, that should be pretty easy. No specific test is required for this, right?

Then, because we used the "autotune" PI feature, I would feel good about it. I'm not sure of what the new RTD format is, but this data can be streamed right to a text file, and then pasted into Excel. That would allow updates at maybe 1KHz rather than just 10Hz, not that 1KHz would be required. You wouldn't see it live as it happens, but could go back and see what the last minute looked like on a graph.

I have no source code. I will power up the Cougar and see what it streams, then paste that into a message here. I am pretty confident that Adam's code will deal with the information in a different order .. maybe I should test that!

At 115,200 baud (I think) .. 10 bits per byte with start bit and stop bit = 11,520 characters per second. At 1 Khz or 1000 samples per second, there is only 11 characters, including CR and LF, per sample. Since we are talking about maybe 80 characters, plus a bit of headroom .. how about 100 Hz sampling instead?

If you are going ahead with a 16 bit word of flags for which data to log, can you add a logging rate as well? That way we can log only one or two values at super-high rate, or log everything at a slower rate ...

[thingstodo]

Originally Posted by thingstodo

I have no source code. I will power up the Cougar and see what it streams, then paste that into a message here. I am pretty confident that Adam's code will deal with the information in a different order .. maybe I should test that!

Sorry Paul. I powered up the Cougar, figured out (again) that I need a throttle connected to do much of anything, got a 36V pack connected, got the serial connected ... and it shows Cougar 1.11b just as it has many times, but the screen does not echo characters.

I tried rtd-period 1000 just in case - but no result.

I checked the serial cable .. it looked a bit rough so I made another one. Same result. I appear to have a part-time connection between the serial port and the processor that I need to track down.

I ran out of time tonight.

[thingstodo]

BTW - the Cougar pre-charged and all seemed fine at 36V with no motor connected. Do I need a motor (or at least a resistor in place of the motor) to continue troubleshooting which MOSFETs have quit?

Sorry - should be in the Cougar list

[MPaulHolmes]

Don't feel any pressure on my end. I'm getting almost nothing done for the next few days due to family crises and work.

Wait, so you can charge the caps, and they stay charged up?

[thingstodo]

Originally Posted by MPaulHolmes

Wait, so you can charge the caps, and they stay charged up?

Yes. I wan't really paying attention when I was trying to power up the Cougar. I tried with 12V controller power only and got the flashing yellow. Hooked up the throttle and cleared that. But the Green LED was off? So I brought up the small pack (3 sla from old UPS), put on the pre-charge resistor, brought it up to 32V, then bypassed the precharge. It held. Green LED (I didn't think I needed the high voltage pack to light the Green?)

So I connected the laptop, got the startup message .. but no echo.

Then I realized that there was a PACK CONNECTED! I checked all of the resistors driving the MOSFETS - no heat? With a failed MOSFET the B- and M- are shorted ... so I need a motor connected B+ to M- for current to flow, right? Is that when the resistors will heat up?

Sorry for the off-topic. I'll cross post to the Cougar list.

[e*clipse]

thingstodo - I'd just like to add a big thank you for this amazing testing you're doing.

On my end, I've been working on the driver circuits for a TO-247 switch equipped controller.

This is awesome from an R&D perspective, because you can actually purchase all leading technologies ( IGBT, Mosfet, and SiC mosfets ) in the TO-247 package.

To get a feel for this, I got a sampling data sheets for 3 different versions from different manufacturers of the three different technologies. The common point with all of them is they needed to be able to carry about 100A. The standard mosfets could handle a maximum of 600V, while the IGBT's and SiC mosfets could handle 1200V.

The question I've been hoping to answer: Is it possible to make one driver topology that can drive all three types of switches. Answer - yes

Would the one in the Paul's most recent design (the one thingtodo is testing) work for all three? - yes, with very small mods. It looks like you only need to change the zener diode that determines the high/low voltage outputs. For lower voltage mosfets, it may be necessary to reduce the supply voltage. This can probably be done my changing the duty cycle for the power isolator circuit.

Anything new yet? no, honestly - Paul covered this a couple of posts back. However, there are a couple of interesting points I learned from some Powerex driver app notes:

The power required to drive a switch - whether a "bruiser" IGBT or a SiC switch is determined simply by multiplying the gate charge by the frequency. This is similar to the power required to charge/discharge a capacitor. A large TO-247 IGBT has a total gate charge of 410nC (nano Coulombs) and the biggest TO-247 SiC Mosfet I can find has a gate charge of 161nC. Because various tail effects and so on, you probably don't want to run the IGBT over 10kHz. However, the SiC can be run comfortably at 30kHz.
IGBT: 410nC*10kHz = 4.1mW
SiC: 161nC*30kHz = 4.8mW

This is NOT addressing all the issues, such as whether you'd want to drive the switch that fast - just the power required to drive one switch. Most importantly, it shows that the transformer used to supply isolated power is very safely spec'd as it can handle well over 3A - - - 72W.

The other issue is the current that must be instantaneously supplied to turn on a switch. This is simply E=IR, assuming the parasitic stuff like capacitor ESR, etc is small relative to the gate resistor value. This is where the SiC switches are more demanding, because they can handle a faster gate turn-on time. For example, a SiC switch with a 19V supply and 2Ohm gate resistor would require 9.5A PER SWITCH. A system that uses three switches, like the one I'm designing, will require a control that can handle 3*9.5A = 28.5A! Ouch! This would require a very capable control switch for the driver circuit. I've found ones that can handle peaks of 30A; they are actually physically smaller than the current ones - this may be a realistic limitation:
http://www.nxp.com/documents/data_sheet/PHPT60415NY.pdf

This then begs the question about all the connections for the switch gate pins and the emitter connections. At first I thought the Radsok connectors capable of 30A were silly overkill. The connectors I chose for the gate pins can handle 36A, but I chose them primarily for the contact reliability. At any rate, the standard "spring finger" type connector won't cut it. I'm looking at this technology for a reasonably small connector:
http://www.empf.org/empfasis/2008/no...asis_11-08.pdf
So far I've sourced the spring and I think the housing would be pretty easy to make.

Of course, if you're driving an IGBT @ 5kHz using a more normal gate resistance of 10Ohms or more, this isn't a problem.

- E*clipse

[thingstodo]

A bit of an update for Nov 11

I've had to clean the garage a bit in preparation for winter. The attached pictures show:
- the High Voltage battery pack in the corner by the overhead door.
- the 24V pack is to the left of the High Voltage pack.
- the AC motor and DC motor coupled, AC Motor in front of the High Voltage pack
- the aluminum base of the Renault/Better Place pack in front of the AC motor
- the AC controller, contactors, power supply, etc on top of the base
- the base on top of 3 roughneck storage containers to raise the base to a useful height

The High Voltage pack is not connected to the AC controller. I'm trying to determine where to put the shunt resistor so that the voltage and current on the pack can be videoed together
The AC controller is not connected to the AC motor as yet. It just needs to be done.
The Low voltage pack is not connected to the DC motor. If I put a shunt resistor on the DC motor, perhaps I can voltage and current on the motor to be on the same camera as the high voltage pack
The voltage and current out of the AC controller will require a different camera since the meters will be on top of the cabling from the controller to the motor.

The other pictures are closer up on the high voltage pack, showing the acrylic touch-safe shield; the 24V pack which is now 6 batteries, 2s3p, or 2 batteries in a series string, and three of those strings in parallel; AC controller on the aluminum battery base that I'm hoping to use as a heat sink.

With 5 deep cycle batteries and an AGM, I hope to put 300+ amps into the DC motor for several minutes at a time during a test. The shroud on the DC motor has been removed so I can check the temperature of the brushes and armature. I still have not located the reference to the temperature that it should not go above. Perhaps it was Tom Brusa's EVCON presentation? Or a discussion of Michael Breem and Matt Hauber's BMW that did the Pike's Peak run with dual Warp11's? If anyone knows what temperature I should stop my testing at - please let me know.