Hey everyone,
I'm working on a 3 phase motor controller for a 100kw PMAC motor. After many weeks of internet reading and learning circuit/ pcb design I have a rough draft of a single phase I'd like to run by people who actually know what they're doing. My background is in Structural Engineering and this project is just oozing with things I don't know that I dont know... But that's part of the fun!

So my driving requirements (in my perceived order of importance):
- Packaging. Need it to fit in ~9" x 6.5" x 5" box (this is for a motorcycle, very little room in the frame Ive already started welding)
- ~84kw at 168V (500A from battery), plan on hitting up to 100kW in the future.
- Im using paralleled mosfets in the inverter so I need to be careful to match the impedance of each half phase's gate pins. (many reasons I'm using mosfets. not really worth discussing, but i really don't want to use IGBTs)
- Need a relatively High switching frequency. Induction of the motor is very low (10.3uH) and consequently people have a hard time driving it. The manufacturers recommend a minimum switching frequency of 16kHz, I'm shooting for 20kHz to be safe
- Everything is going to be very tightly packed... I'm worried about noise screwing with the drivers (inputs and outputs...). Need a smart PCB layout.
- in general want this to be super robust. going to buy high quality components and want any protections possible built into the circuits.

My questions/ concerns are:
A: What about this screams "I don't know what I'm doing"? I'm looking for industry standards, "good practices" I don't know I'm ignoring or any blatant schematic/pcb mistakes you can see. the more the merrier, I'm sure there are lots...
B: The PCB design is super preliminary, but I'd like to know if I'm on the right track or way off base. I only have 20mil traces right now, only one phase modeled but it just looks messy to me.... Its my first PCB so who knows maybe its not too bad... Also, in case its not clear, the mosfet gate resistors and diodes live on both sides of the PCB in the same place to try and make everything more uniform and compact.
C: What should be done about a ground plane and general interference protection on the PCB? I dont have anything modeled because I dont even know where to start. Any idea how worried I should be or know of any good references?
D: With regards to making paralleled mosfets work properly, any thoughts on how I have the power leads connected? Is the gate inductance the main thing to worry about? I have a little CNC mill so cutting sweet copper couplings is no big deal. I'm pretty attached to the Drain plate since it will be my interface with the heat exchangers, but idk if the amount of copper will mess with switching performance (haven't sized the plate thicknesses yet either, these are just estimates)

Thanks! I plan to do more iterations on this, but when i get it working I'll for sure share the plans. Its a fun project.

 

I have a few questions:

1. How do you plan to power this thing? You mention "168V (500A from battery)". Is this a series arrangement of 14 lead acid batteries or is it lithium 18650 cells or prismatic cells? Balancing all those batteries is certainly not trivial as I have found myself recently. How will you charge the bike?
2. How do you arrive at your power requirement for the bike? 100kW is one hell of a motor A few back-of-an-envelope calculations suggested to me that a 15kW PMAC machine would be about equivalent to the 28hp petrol engine in my quad bike.
3. Are you looking for bidirectional power flow ie regenerative braking?
4. Control - setting aside the power electronics hardware, an inverter of this size requires careful control. Will you be using a microcontrolller, DSP or FPGA?

Your copper fabricated conductor bars are impressive! Carrying those sort of currents in a small space isn't easy. Have you performed any thermal calculations yet based on your chosen MOSFETs, modulation scheme and switching frequency? If it's being retrofitted to a motorcycle, is it possible to keep the liquid cooling system in place?

With regards PCB design, when it comes to traces, short and thick is good. Don't be afraid to use filled areas on traces carrying large currents.
It is possible to use more than one ground plane filled area, one for power and one for signals. If need be they can be tied together at a "star point" with a 0 ohm resistor. This ensures all grounds are at the same potential but it also ensures that there are no large currents flowing in the ground plane near sensitive components.

As an aside, when you get a little bit further down the line and start thinking about control, Dave Wilson of Texas Instruments does an excellent series on motor control, focusing on PMAC machines and Field Oriented Control. It's entitled "Teaching Old Motors New Tricks" and is on YouTube.

 

Wow! That's a large project for a starter!
I second the comment from @qrb14143 about the current available from your batteries. What on earth are you using and how will they be kept balanced? Your battery charger is going to be an interesting project in itself. Maybe have individual monitored inputs for each cell?
A couple of years back I built a power controller with paralleled mosfets in, they work nicely because if the resistance of one is slightly lower than another (driven harder by accident), then it conducts more of the current which makes it heat up slightly, therefore increasing resistance again. The key issue I had was making enough current available fast enough. Basically any solid state switching device heats up most while it is in the process of switching, therefore to keep unwanted heat to a minimum you need to pump current into the mosfet gates as fast as possible. As they are effectively capacitors, I ended up with a logic level P-mosfet driving all the rest!

 

Looks pretty good. Why the caps in series? Also it is a good idea to have a local decoupling caps on the hv rail, at least a few uF.

 

I have made a 100kW units but never I succeed less than bankomat stan size. Just power caps will take some 30 liter at least, the power igbt with damping chain will take another 30 liter and water-cooling. If air cooling, then cubic meters, and not a few. The transformer itself will take about 30 liters and 50 kilograms, the ZVT choke will take some 20, anti EMP filter will take some 50 liters, and more, and more, and more. The hardest problem - its very problematic to get the high frequency at such kilowatts, but all with no except controller tablets are dedicated for NO-low frequency work. That is the basic reason why always all there is organized on arm processor, as software.
Second - DONT swirl Your head with mass parallelizing, but take a brick instead. The cheapest bricks has price so low as 65 Eur, however I hardly suggest to not buy a cheapest, but some 110-130 Eur are safe. I have suffered myself of buy the last cheap igbt, beat it out and then no way how to get the spare. No way at all.
Third, its not a hard job to kill a 1200 Amp 1200 Volt copatible igbt via simple 5 Amp fuze from 220Volt network. Its small sound like pop, and even fuse is still alive. But transistor is already at "Transistor Heaven". Its mainly because of improperly short phase shifts between gates. Take this in account, please.

 

About drivers: as You must to drive gates something about 0,1 microfarad or more, the all all and ultimately all ready made drivers are inappropriate, except those for 1000 Eur and more costy. Thus there are some cheap solutions, but seems Your destiny is to burn em before to start ask the right questions, sorry.

 

RE: ""when it comes to traces, short and thick is good""
Its not enough in such power, sad truth. The dV=L*di/dt, where if L=25cm*2directions*10nH/cm=0,5 microHenries, di=500A and t=20 kHz=25 microsec then dt=0.1 mksec, then dV=0.5E-6*500/0.1E-6 =2500 V what kills the igbt immediately. Therefore the bifiliary line is sth must to be as the least, however the two-platelet dissipated plane is far better than any best sandwich line. Thus this battle is about geometry and topology not a prost thickness of paches.

 

I suggest you build it in a strong metal case to keep the hot flying bits inside!
(I speak from experience)

 

I have a few questions:

1. How do you plan to power this thing? You mention "168V (500A from battery)". Is this a series arrangement of 14 lead acid batteries or is it lithium 18650 cells or prismatic cells? Balancing all those batteries is certainly not trivial as I have found myself recently. How will you charge the bike?
2. How do you arrive at your power requirement for the bike? 100kW is one hell of a motor A few back-of-an-envelope calculations suggested to me that a 15kW PMAC machine would be about equivalent to the 28hp petrol engine in my quad bike.
3. Are you looking for bidirectional power flow ie regenerative braking?
4. Control - setting aside the power electronics hardware, an inverter of this size requires careful control. Will you be using a microcontrolller, DSP or FPGA?

Your copper fabricated conductor bars are impressive! Carrying those sort of currents in a small space isn't easy. Have you performed any thermal calculations yet based on your chosen MOSFETs, modulation scheme and switching frequency? If it's being retrofitted to a motorcycle, is it possible to keep the liquid cooling system in place?

With regards PCB design, when it comes to traces, short and thick is good. Don't be afraid to use filled areas on traces carrying large currents.
It is possible to use more than one ground plane filled area, one for power and one for signals. If need be they can be tied together at a "star point" with a 0 ohm resistor. This ensures all grounds are at the same potential but it also ensures that there are no large currents flowing in the ground plane near sensitive components.

As an aside, when you get a little bit further down the line and start thinking about control, Dave Wilson of Texas Instruments does an excellent series on motor control, focusing on PMAC machines and Field Oriented Control. It's entitled "Teaching Old Motors New Tricks" and is on YouTube.

Thanks for the response!

regarding the additional info:

1 - Im working on assembling a li-ion battery from 18650 cells (LGCHEMMJ1's) they have a peak continuous rating of 10A and can handle short bursts of ~15 amps without many side effects assuming you keep the temp low. Ill have a single 16P40S pack with plans to add a second in parallel when I can rustle up the $$ (so its 168V max, not average.) Main problem with the battery im having is actually just assembly. welding cells together for 15 amps each is outside the realm of nickle strips and resistance welders... I have a 16S BMS design that can be daisy chained together (thanks TI) but haven't begun testing yet.

2 - initially it was "As powerful as i can" obviously lol but the more powerful motor is definitely worth it when you consider that its direct drive to the wheel and doesnt have the benefit of a transmission like gas motorcycles. I want to be able to comfortably ride on the highway and have the power to get out of tight spots when I'm already at 80mph. Also thats 100kW peak power, not continuous. continuous is ~50kW

3- I will be, but I havent started working on the control side yet, and as i understand it the inverter hardware wont change because of this.

4- My plan here changes often, but I recently discovered texas instruments piccolo chips and their (fairly plug and play it seems) motor control examples. Thats my next project, the actual control of this beast. trying to take this one step at a time and make the inverter as "standalone" as possible to attempt to skip a revision or two.

5- fairly comfortable with thermal analysis, that and structures is part of what I do for work. Once the schematic is done I will need to size those bus bars. I think they're pretty reasonably sized now but only did a first pass (did not include mosfets or stray conduction paths, just "amperage through bar"


The battery will be liquid cooled due to the huge thermal mass and the fact that its sealed, may be able to tie the inverter cooling into that, but for now planning on just a heatsink with forced air through it. Switching to liquid would just be a matter of swapping out the heat sink for a coldplate which may be necessary if this gets much fatter.


Thanks for the tip about Dave Wilson. Ill check it out. Texas instruments is my new favorite company. Idk if it’s actually their goal, but their “give working circuit examples out in order to sell chips” methodology is amazing. This wouldn’t be possible for me without good examples I could tweak.

 

Wow! That's a large project for a starter!
I second the comment from @qrb14143 about the current available from your batteries. What on earth are you using and how will they be kept balanced? Your battery charger is going to be an interesting project in itself. Maybe have individual monitored inputs for each cell?
A couple of years back I built a power controller with paralleled mosfets in, they work nicely because if the resistance of one is slightly lower than another (driven harder by accident), then it conducts more of the current which makes it heat up slightly, therefore increasing resistance again. The key issue I had was making enough current available fast enough. Basically any solid state switching device heats up most while it is in the process of switching, therefore to keep unwanted heat to a minimum you need to pump current into the mosfet gates as fast as possible. As they are effectively capacitors, I ended up with a logic level P-mosfet driving all the rest!

yes not easy to find things for this size a pack lol
The BMS was the first thing I tried to figure out. Its a li-ion pack with 40 cells in series so balancing is very necessary. Texas instruments has lots of great working examples of bms boards plus all of the build info and schematics. basically just needed to swap out the balance resistors to allow for more current (16 parallel cells per balance point) and i think im good to go.
Since this is for one specific purpose and because I will have very precise control over how much current my motor is drawing i opted to not include power protection in the bms (much like what tesla does in their packs). Ill probably have a contactor relay to switch the pack between "charge" and "discharge" and ill add a single outgoing fuse somehow as well (maybe a fuse on a normally open contactor i.e. when the pack hits a certain current, it fuses the relay input power?)

 

Looks pretty good. Why the caps in series? Also it is a good idea to have a local decoupling caps on the hv rail, at least a few uF.

Hi,
Thanks! The caps in series are because I'm trying to keep the board at a very low profile. 35v caps are .5" tall, 60+v caps are 3/4"+ tall. Somewhere people recommended a higher voltage cap at the DC/DC converter's input to protect against voltage spikes. After a bit of reading I couldn't decide what kind of cap to use so I just copied a working example I found. I would prefer to use ceramics since I could then fit the correct voltage capacitor in. Any idea if ceramic caps would be ok here? My lack of experience causes me to default to copying existing examples when I don't have enough info.

I will have a 500uF DC link capacitor at the HV rail (SBE 700D529). per a calc I found that should keep the ripple voltage at about 2v which I think will be ok... not impossible to add a second cap either.

 

About drivers: as You must to drive gates something about 0,1 microfarad or more, the all all and ultimately all ready made drivers are inappropriate, except those for 1000 Eur and more costy. Thus there are some cheap solutions, but seems Your destiny is to burn em before to start ask the right questions, sorry.

Hi thanks for the info!
Why do you say ready-made drivers cant handle high gate capacitance mosfets? As I understand it, these gate drivers are made specifically for high power fets and are pretty commonly used on paralleled high power fets? I've found a few examples that successfully use standard off the shelf gate drivers for paralleled power mosfets (granted they probably know what they're doing better than I do, but it at least tells me I'm not on a broken path)
Are you mainly worried about the feasibility of matching impedances? If so, does what I have look bad? This is exactly why I wanted to ask, I know its an issue but don't know how big of one or how to really judge my design.

 

I suggest you build it in a strong metal case to keep the hot flying bits inside!
(I speak from experience)

yes.....

My experience with mosfets thus far is that I kill them SO quickly they never even get the chance to heat up. So at least the flying bits wont be hot...

 

Thanks for the response!

regarding the additional info:

1 - Im working on assembling a li-ion battery from 18650 cells (LGCHEMMJ1's) they have a peak continuous rating of 10A and can handle short bursts of ~15 amps without many side effects assuming you keep the temp low. Ill have a single 16P40S pack with plans to add a second in parallel when I can rustle up the $$ (so its 168V max, not average.) Main problem with the battery im having is actually just assembly. welding cells together for 15 amps each is outside the realm of nickle strips and resistance welders... I have a 16S BMS design that can be daisy chained together (thanks TI) but haven't begun testing yet.

2 - initially it was "As powerful as i can" obviously lol but the more powerful motor is definitely worth it when you consider that its direct drive to the wheel and doesnt have the benefit of a transmission like gas motorcycles. I want to be able to comfortably ride on the highway and have the power to get out of tight spots when I'm already at 80mph. Also thats 100kW peak power, not continuous. continuous is ~50kW

3- I will be, but I havent started working on the control side yet, and as i understand it the inverter hardware wont change because of this.

4- My plan here changes often, but I recently discovered texas instruments piccolo chips and their (fairly plug and play it seems) motor control examples. Thats my next project, the actual control of this beast. trying to take this one step at a time and make the inverter as "standalone" as possible to attempt to skip a revision or two.

5- fairly comfortable with thermal analysis, that and structures is part of what I do for work. Once the schematic is done I will need to size those bus bars. I think they're pretty reasonably sized now but only did a first pass (did not include mosfets or stray conduction paths, just "amperage through bar"


The battery will be liquid cooled due to the huge thermal mass and the fact that its sealed, may be able to tie the inverter cooling into that, but for now planning on just a heatsink with forced air through it. Switching to liquid would just be a matter of swapping out the heat sink for a coldplate which may be necessary if this gets much fatter.


Thanks for the tip about Dave Wilson. Ill check it out. Texas instruments is my new favorite company. Idk if it’s actually their goal, but their “give working circuit examples out in order to sell chips” methodology is amazing. This wouldn’t be possible for me without good examples I could tweak.

1) Yes it was very nice of TI to give away that reference BMS design - I have used it myself. I'm a little curious where you're getting your LG batteries. I would like that exact part number too but despite manufacturers rambling on about buying the genuine product I'm struggling to find an authorised reseller for LG or Panasonic.

2) Think you'll have more than enough power to propel you to light speed never mind get you out of a tricky situation I have friends who are really into the fast bikes so I suppose I understand the desire for power.

3/4) Yes the Texas Instruments chips are nice. I have used a F28069 to drive induction machines before with great success. Be aware when designing your power electronics that you're going to want to incorporate at least current measurement on the three phases going to the motor to allow you to regulate current.

5) If you do thermal analysis in the course of your day job then you're already more clued up than me.

To clarify my previous post about PCB traces, I wasn't for a moment suggesting that these be used as the main current carrying paths in the converter. I was making a general observation that short and thick traces should be used where possible and current carrying paths kept as short as possible.

Have you considered using fewer MOSFET's in larger packages instead of the TO-220? These generally have MUCH larger thermal interface pads which simplifies cooling somewhat. They have screw terminals for ease of assembly and disassembly. They also often offer Kelvin connections which can help to alleviate some of the issues with stray inductance that previous posts have alluded too.

 

1) Yes it was very nice of TI to give away that reference BMS design - I have used it myself. I'm a little curious where you're getting your LG batteries. I would like that exact part number too but despite manufacturers rambling on about buying the genuine product I'm struggling to find an authorised reseller for LG or Panasonic.

2) Think you'll have more than enough power to propel you to light speed never mind get you out of a tricky situation I have friends who are really into the fast bikes so I suppose I understand the desire for power.

3/4) Yes the Texas Instruments chips are nice. I have used a F28069 to drive induction machines before with great success. Be aware when designing your power electronics that you're going to want to incorporate at least current measurement on the three phases going to the motor to allow you to regulate current.

5) If you do thermal analysis in the course of your day job then you're already more clued up than me.

To clarify my previous post about PCB traces, I wasn't for a moment suggesting that these be used as the main current carrying paths in the converter. I was making a general observation that short and thick traces should be used where possible and current carrying paths kept as short as possible.

Have you considered using fewer MOSFET's in larger packages instead of the TO-220? These generally have MUCH larger thermal interface pads which simplifies cooling somewhat. They have screw terminals for ease of assembly and disassembly. They also often offer Kelvin connections which can help to alleviate some of the issues with stray inductance that previous posts have alluded too.

I've bought cells from batterybro.com before, they were all legitimate and not unusually expensive either. I think the LG cells were 4.50$ at the 1000qty point.
The mosfets I'm using are to-247 packages and I was looking into to-264 packages but they tip the scales just past the point of me being able to package them. And yes, I plan on carrying all the current through those copper plates and the mosfet connections to the PCB are just for control.
On the subject of current sensors, any recommendations for compact or "in-line" hall sensors? I was planning on adding 2 lem has 600 sensors but have never worked with them so not sure they'd be a good match yet.

 

I have had a quote from battery bro but was reluctant to go ahead until I heard a testimonial or two. I find it strange that LG and Panasonic make no effort to advertise their authorised agents. Most OEMs have a "find your nearest genuine stockist" type feature.

I have used this exact current sensor in the past with satisfactory results. Our power electronics group who have a wealth of experience in this area also use this sensor so they must like it too. If memory serves me you will need a +/-15V supply for them which will require an isolated switch mode converter to generate. Not a great inconvenience really but yet more precious space on your PCB. We have used products by Traco Power for this before - they're not cheap but they have proved very robust. It is entirely possible to do this sort of thing using non-isolated means which are typically more compact, the trade-off being that care must be taken not to expose yourself or your control circuitry to dangerous voltages. A low value, high power shunt resistor can be used in series with the power carrying path to obtain current directly using ohm's law.

Check out this article for current measurement shunt examples http://www.analog.com/en/analog-dialogue/articles/optimize-high-current-sensing-accuracy.html

 

I have had a quote from battery bro but was reluctant to go ahead until I heard a testimonial or two. I find it strange that LG and Panasonic make no effort to advertise their authorised agents. Most OEMs have a "find your nearest genuine stockist" type feature.

I have used this exact current sensor in the past with satisfactory results. Our power electronics group who have a wealth of experience in this area also use this sensor so they must like it too. If memory serves me you will need a +/-15V supply for them which will require an isolated switch mode converter to generate. Not a great inconvenience really but yet more precious space on your PCB. We have used products by Traco Power for this before - they're not cheap but they have proved very robust. It is entirely possible to do this sort of thing using non-isolated means which are typically more compact, the trade-off being that care must be taken not to expose yourself or your control circuitry to dangerous voltages. A low value, high power shunt resistor can be used in series with the power carrying path to obtain current directly using ohm's law.

Check out this article for current measurement shunt examples http://www.analog.com/en/analog-dialogue/articles/optimize-high-current-sensing-accuracy.html

Yeah I've contacted LG chem multiple times about this and never gotten any response. I was happy with the battery bro order though. Arrived fast and were well packaged. The measured capacity was right where it should have been and they pull 10A continuous with a very reasonable amount of heat generation. Super happy with these cells. Heard of some new LG cells which may be even slightly more power dense (m50s I think? They're 21700 form factor)