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

[e*clipse]

Originally Posted by Hersbird

Soooo, is there a condensed version of this thread with details of what the title says? I just can't do 230 pages when just knocking around the idea of electrifying my Forester. I know the "kits" out there are way to much for me, but the couple posts above mine hardly give the impression of DIY.

That's a pretty open ended question. As you can see, the project is in development. Once the issues involving all this stuff are worked out, then - depending on your skills - there will various levels of DIY AC controllers available.

Perhaps a good way to start is to look at the open revolt DC controller.
Open ReVolt: open source DC motor controller - Fuel Economy, Hypermiling, EcoModding News and Forum - EcoModder.com

Glad to see you're interested in EV's!

- E*clipse

[MPaulHolmes]

Hersbird: Thanks to thingstodo, I'm pretty confident that with a couple minor additions, the software is basically ready to go if you are using an encoder with the AC board that plugs right into 3 IGBT half bridges. I'm probably going to include a datastream feature today.
Here's the instructable for how to put one together:

http://instructables.com/id/200kW-AC...-Electric-Car/

Don't get too excited though. I entered it into a contest on instructables, and got like last place. haha. I'm not sure what that means.

I basically am perpetually broke, and won't be making completed controllers ever. haha. So, you would have to put one together yourself if you wanted to go this route. It's not too complicated though. Almost all through-hole components, and very large 1210 package surface mount capacitors. The software is to a point where you can basically connect the motor to the controller, type a couple commands through the serial port, and it can figure out the characteristics of the motor and save them to EEProm. Then, you drive away. hip hip hurray! And it works with permanent magnet or induction motors.

The lower power version I made (well, the boards only exist on the computer) would be mechanically easier to assemble. You could easily just use a hand drill. But it would have an absolute peak power of maybe 50kW, probably less. It also has the feature where you could just have, say, a 144v battery pack, and boost it to 300v, allowing the motor to run at a much higher RPM under load. It is set up to be all through hole components, or you could use 1206 package components (I made 1206 pads with holes through them so you can mount through hole components instead if you like). The total cost would be around $350 in quantity 1, so you could probably get it down to $250 if you went into business buying enough for 10 controllers at a time. But that would require someone who is not in abject poverty. LOL. But I may be broke, but my job is at home, where I get to see my kids and wife throughout the day. So, I think it's a good trade.

That's about the state of things on my end. 2 controller types. One uses IGBT half bridge modules, and the other uses TO-247 package IGBTs.

Well, I'm reordering some of the type that plug right into the half bridge modules. I'll have 7 of them, since 3 are already bought. And as i was ordering it, there was an option for a metal screen thing that puts the solder paste on the pads. So, I could order some of the surface mount parts that go on the boards, and solder those, and sell the boards partially assembled maybe?

[MPaulHolmes]

Originally Posted by e*clipse

On the subject of saving some space, what's the reason for the two banks of four 22uF capacitors - the ones tied to +15V , IGBTemitter, -8.2V? Are these buffer capacitors or filter capacitors? Why not just have one capacitor each at about 48uF? Or if these are filtering capacitors, one 22uF and one 66uF capacitor each?

The 22uF caps are rated for 3 amps of ripple current each. So, I was trying to design for 12amp of turnon current and 12amp of turnoff current. But that is only because I was intending to drive a 1200v 600amp monstrocity in the worst case scenario. A typical 66uF electrolytic cap might have 1A of ripple current for its rating. Higher uF MLCC caps usually have worse ripple ratings. I am fairly certain that 40uF would be enough, and that 10uF 1206 caps would work fine. And those are like $0.05 if you buy a bunch of them, but I was using my philosophy of "stuff as much as you can into the available space, and if it still doesn't work, go home and cry, because there's nothing else you can do" TM/CIRCLE R. What powerex does is use 1000uF eletrolytic caps instead. But those are rated for 3000 hours, and the MLCC caps are rated for like 300 years at 125 degC. haha. But I know for a fact that you can use the 1000uF caps for more than 10,000 hours, since they are running my house right now. I made an inverter/maximum power point tracker charger thing that uses the vla500-01 driver with the 1000uF caps, and it's been running 24-7 for over 2 years. Well, the maximum power point tracker thing now has an issue but is working better now that it's cooling down, and I bet it's related to the electrolytic caps drying out. But it's 150 degrees in my shop during the summer, so what should I expect.

If you are using SiC mosfets, you could probably dump the whole current buffer stage, and just use an optocoupler + gate driver. I bet one 4.7uF cap would be OK. Well, maybe one from +20v to "emitter", and one from "emitter" to -5v, and maybe one from +20v to -5v.

This brings up the question about the gate drive's power consumption. I checked into the driver that push-pulls the transformer, and it's rated at 4 amps. So, if everything were matched perfectly 24V * 4A = 96watts - - - Close enough?
This would bring some opportunities to downsize certain components a bit - for example, it sounds like the gate drive transistors could be smaller.
and I've found a transformer that is much smaller, but a bit more expensive. It's made by Pulse and can take 5A, with something like a 1500V Hi-Pot test. This actually is a deal because many of the transformers I looked at could only handle < 100V. There's also a similar sized one my Eaton that can handle 4A and is "rated" for 300V for about $1 more.

The gate driver is actually rated for 4amp peak I think. I'm not sure about that, but I've always interpreted it to be "no peak gate current over 4 amps". I might be wrong though. The line filters I've been using I tested at my job to over 7000 volts from primary to secondary on a high (on) pot test. It was oregon/washington after all. What better place for a high pot test. haha. So that is probably overkill, but it has very low coupling capacitance. You would have to make sure that the transformers you found also have very low coupling capacitance. I think as long as secondary isn't on top of primary, it might be good.

About needing desat detection at all - VERY interesting point. Since mosfets are breaking into the high voltage realm ( with SiC Mosfets ) maybe just don't bother with IGBT's? The current is limited to what the legs of the TO-247 can handle, which is covered by Mosfets.

So desat really relies on a big RdsON to work? Very interesting - because I'm trying to get away from as many losses as possible with this design. Maybe I should just commit to Mosfets... Wow, that's got my head spinning.

If the stupid fod8316 or the like would allow you to change the 6-7v desat trigger point to maybe 3-4v, it would be probably better. But they hide that inside the chip.

Thanks again, Paul - I'd like to add that you are an excellent teacher.

Well, I did go to school to be a teacher! I still have nightmares about teaching at Timberline High School. "Why do you even come to school if you are just going to mess up the class the whole time?" "My P.O. says I have to." umm... I had to go home and look up P.O. haha

[MPaulHolmes]

Finding the magnitude of <Vd,Vq> was ALMOST identical to finding the electrical RPM, so today if I have enough time I'm going to try to make a correction array, so that it will fit exactly. Then, I'm going to gradually decrease the encoder counts that I pay attention to until it is down to maybe 1 tick per revolution that is used for computing the sensored RPM, and see if this new way of doing sensorless can effectively fill the gaps.

The proper thing would be to go back to all that crap I was doing for sensorless and do a similar reduction in encoder ticks to see if that could fill the gaps effectively. I bet it could, but this new approach is so spectactularly simple that I want to try it first. You don't need to know anything about rotor inductance, stator inductance, stator resistance, etc...

[Hersbird]

Originally Posted by e*clipse

That's a pretty open ended question. As you can see, the project is in development. Once the issues involving all this stuff are worked out, then - depending on your skills - there will various levels of DIY AC controllers available.

Perhaps a good way to start is to look at the open revolt DC controller.
Open ReVolt: open source DC motor controller - Fuel Economy, Hypermiling, EcoModding News and Forum - EcoModder.com

Glad to see you're interested in EV's!

- E*clipse

Ok, I get it now. I thought there was a AC one ready. I did see the other DC and the wiki and the webpage and that's what I was thinking was ready here.
Carry on and keep up the good work!

[e*clipse]

Originally Posted by MPaulHolmes

The 22uF caps are rated for 3 amps of ripple current each. So, I was trying to design for 12amp of turnon current and 12amp of turnoff current. But that is only because I was intending to drive a 1200v 600amp monstrocity in the worst case scenario. A typical 66uF electrolytic cap might have 1A of ripple current for its rating. Higher uF MLCC caps usually have worse ripple ratings. I am fairly certain that 40uF would be enough, and that 10uF 1206 caps would work fine. And those are like $0.05 if you buy a bunch of them, but I was using my philosophy of "stuff as much as you can into the available space, and if it still doesn't work, go home and cry, because there's nothing else you can do" TM/CIRCLE R.

Ok, I think I got it; I'll look into the ripple current issue. It's kind of like spec'ing the big one for the main power bus.

I did look into the transistor; the Dpak package is about as small as reasonable. The problem is I can't find the transistor you spec'd in a Dpak package - it only comes in a TO-220 package. (that's the DV45VH10) It's rated for 12A and the best I can find is rated for 8 or 10A VERY intermittantly. Also, I have no idea how to match the specifications for a BJT Time to drag out my "art of electronics" book and hopefully learn something.

What powerex does is use 1000uF eletrolytic caps instead. But those are rated for 3000 hours, and the MLCC caps are rated for like 300 years at 125 degC. haha. But I know for a fact that you can use the 1000uF caps for more than 10,000 hours, since they are running my house right now. I made an inverter/maximum power point tracker charger thing that uses the vla500-01 driver with the 1000uF caps, and it's been running 24-7 for over 2 years. Well, the maximum power point tracker thing now has an issue but is working better now that it's cooling down, and I bet it's related to the electrolytic caps drying out. But it's 150 degrees in my shop during the summer, so what should I expect.

On your solar inverter - the panels really don't like the heat. It's pretty ironic, but the solar panel's output drops dramatically as it heats up. My solar system's best output is on those strange days when there's full sun and it's cold. The middle of summer is not nearly as good. Perhaps you could sense the panels temperature and alter the MPPT accordingly?
Regarding capacitors - I really don't like electrolytic caps - I've seen too many things - from LED lights to inverters die because of the electrolytic caps. And that was easy stuff, compared to the automotive environment. I'm going to avoid them if possible.

If you are using SiC mosfets, you could probably dump the whole current buffer stage, and just use an optocoupler + gate driver. I bet one 4.7uF cap would be OK. Well, maybe one from +20v to "emitter", and one from "emitter" to -5v, and maybe one from +20v to -5v.

I'm trying to design the board to be as multi purpose as possible - so it's going to have some capability that won't be used in all cases. Believe it or not, I've got a double DC motor drive that needs a double H-bridge control and this controller could do that.

So, because SiC switches are pretty expensive right now, I'm leaving the option for other technology, like standard Mosfets or maybe IGBT's ....The place for the buffer stage will be there; maybe I can jumper around it as a simplicity/cost benefit of using SiC switches.

The gate driver is actually rated for 4amp peak I think. I'm not sure about that, but I've always interpreted it to be "no peak gate current over 4 amps". I might be wrong though. The line filters I've been using I tested at my job to over 7000 volts from primary to secondary on a high (on) pot test. It was oregon/washington after all. What better place for a high pot test. haha. So that is probably overkill, but it has very low coupling capacitance. You would have to make sure that the transformers you found also have very low coupling capacitance. I think as long as secondary isn't on top of primary, it might be good.

LOL! - my first experience with all that was in Oregon. Of course here in NorCal there's an awful lot of HiPot testing as well.
The transformer I found is a toroid with two windings; the windings are connected magnetically and don't overlap at all. They don't mention capacitance in the spec sheet. The inductance is 130uH and the DC resistance is 6.75mOhms. The AC impedance peaks at about 2Mhz @ 500 Ohms.

I thought the power circuit just had to supply a fairly constant power - say 3-4A at 24V. The gate drive would be drawing something like 10 or 12A at a very intermittant duty cycle. The poor capacitors would have to buffer this, but oh well - that's their job. However, if the power matches, the transformer and gate drive would just putt putt along happily at a much lower output. - Or am I missing something??

Oh - this brings up the question of power use for the whole inverter!! Do you have numbers for that? Maybe thingstodo could measure that??

The reason I'm asking is that I'd like to run the 24V inverter power in with all the other stuff in that big 35pin connector. The pins are rated at 16A under perfect conditions with gold plated contacts. It's probably much more realistic to think 10A or 12A. Waving my hands around, I'd say each gate drive runs at 4A - potentially 8 gate drives. Say the "brains" of the operation requires much less - 50W - 2A. That's about 800+ watts or 34A!! That would require 3 pins for just the power supply...

If the stupid fod8316 or the like would allow you to change the 6-7v desat trigger point to maybe 3-4v, it would be probably better. But they hide that inside the chip.

Well, I did go to school to be a teacher! I still have nightmares about teaching at Timberline High School. "Why do you even come to school if you are just going to mess up the class the whole time?" "My P.O. says I have to." umm... I had to go home and look up P.O. haha

LOL! High School is probably the worst from that perspective. I like teaching if the students are interested and want to learn - then it can be fun.

[MPaulHolmes]

It draws around 200ma at 24volts when switching the 600v 600amp IGBTs at 10khz and 30mips micro.

[e*clipse]

Originally Posted by MPaulHolmes

It draws around 200ma at 24volts when switching the 600v 600amp IGBTs at 10khz and 30mips micro.

Really?! Wow. That's really low - less than 5 watts!
Actually it's really good the numbers aren't like the ones I put up - that would be really inefficient.

Ok, so @ 3 phase and no boost stage: each phase leg is drawing less than 5W/6 >> about 0.8 watts for each gate drive...

So - OMG - we're only talking about 33 mA @ 24V MAX for each gate drive power supply (gate driver, transformer, and capacitors) The capacitors and BJT's would have to deal with 8A > 12A for a picosecond.

Sorry for the long reply... this is really awesome news.. . . . I'll limit the 24V power input to 4 pins max.

- E*clipse

[thingstodo]

Last part of testing for Nov 1 - spin test

Drive the DC motor with a deep cycle batteries (no controller - none of mine work right now). AFTER verifying that the battery turns the motors the same direction as the AC controller does

Put the AC controller and AC motor into regenerative braking. Take energy from the rotating motor and push it back into the high voltage battery pack.

I need to spend some time setting up some instruments - this was as painful to do as it is to watch (if you choose to watch it)

Video link https://youtu.be/zNtGYB99ZLM

Here is the information that is interesting

Motor rpm	AC motor current	AC Motor Voltage	DC Motor current
2500		42						35
2230		22						35
1950		9						38.5
1950		0.3						38.5
1970		-0.5						38.5
1970		-7.8			0 - 30			38.5
1949		-14						40
1880		-19.4						41

The high voltage pack shows 0.0 amps in, 0.0 amps out, so the power developed in regen must just be enough to power the electronics.

[thingstodo]

Nov 2 - 24V load test

Drive the DC motor with two deep cycle batteries this time around. Still no controller - none of mine work right now.

Put the AC controller and AC motor into regenerative braking. Take energy from the rotating motor and push it back into the high voltage battery pack.

AGAIN, I need to spend some time setting up some instruments - this was as painful to do as it is to watch (if you choose to watch it) Paul - what measurements can you stream from the AC controller? I won't bother coming up with a way to measure those!

Video link https://youtu.be/pehCWdH49KE

Just like Nov 1, here is a summary of the information that is interesting

Motor rpm	AC motor current	High Voltage Pack Amps	DC Motor current
3600		5.5			0.3				40
		0.0			0.0				40.5
3457		-30			-0.2				45
		-40			-0.7				50
		-60			-1.6				58
		-70			-2.0				64
2300		-80			-3.0				70

The DC motor pack was down to 22.7V at 64 amps, 22.4V at 70 amps
As the load was removed (AC amps were raised) I measured rpm at each current

Motor rpm	AC Motor Amps	
2520		-70
2750		-60
3000		-50
3100		-40
3550		-30
3800		-20
3950		-7.6
4000		-5.7
4060		0.0