BESC - beefed up VESC motor controller for high power ebikes

galp said:
I declare this project open source.

Outstanding!

I hope the 2 parallel shunts end up being a good alternative to the hall sensor. Those 2x 10W shunts in parallel per phase should be beefy enough to handle large current. Can't wait to see what you come up with!
 
Hi,

What about this driver https://vesc-project.com/comment/2760#comment-2760 ? You finally don't want to use it ? I can't wait to see the final result and some tests videos ! :)
 
galp said:
I have finalized plans for future work. It goes something like this:

  • 24 TO-220 FETS
  • Three low side shunts (phase current monitoring (hall) is expensive @400A)
  • Two configurations of HW:
    • 100V/350A peak phase (20S) with IRF135B203 mosfets (1.8€ per piece)
    • 150V/250A peak phase(30S)
  • Cost effective (BOM under 100€ for 20S version)
  • Standard extruded aluminum housing
  • I'll design everything in KiCad because what's the point of opensource project if you need 8k licence just to open files. :mrgreen: :mrgreen:

Maybe you could consider a third "low voltage" version with 16/18S version based on irf100s201 (100V 3.5mohms 136A@100°C and cheaper than IRF135B203, not by much) for 65/75V 500 or a baby version with only 12 fets and 250/300A ? Would also be a little and cheap beast :). I'm really curious to see how fast switching you could achieve to make these TO-220 mosfets !
 
@maxxximatoze

You may want to check out my controllers for lower voltage/power needs. Might be hard to pump 300A through a 12 FET controller though.

https://endless-sphere.com/forums/viewtopic.php?f=30&t=99152
 
galp said:
I have finalized plans for future work. It goes something like this:

  • 24 TO-220 FETS
  • Three low side shunts (phase current monitoring (hall) is expensive @400A)
  • Two configurations of HW:
    • 100V/350A peak phase (20S) with IRF135B203 mosfets (1.8€ per piece)
    • 150V/250A peak phase(30S)
  • Cost effective (BOM under 100€ for 20S version)
  • Standard extruded aluminum housing
  • I'll design everything in KiCad because what's the point of opensource project if you need 8k licence just to open files. :mrgreen: :mrgreen:

I'm a little confused by your numbers.
20S LION or LIPO at full charge is 82v
30S LION or LIPO fully charged is 123v

150v/4.1v is basicly 36S.

However, if your 150v MAX controller can do 32S or 131.2v reliably at 250 amps, well that would serve me really well!
This is a design requirement for my next EV build. I've been looking for FOC controllers that can do 150v MAX for a while now.
Read what John in CR said...I'll be building with a hub monster and running at 32S.
 
If @galp is using 150V MOSFETs, then I would think max operating voltage would be about 120V. That's where galp is possibly getting the 30S number from (maybe). It leaves room between max operating voltage and absolute max voltage that the components can handle.
 
John in CR said:
I've never found a $79 controller that will handle 16kw peak. I've also never run a HubMonster at 40kw. 30kw is the max I've used and never more the 17kw peak input without cooling mods. If you run much more than 100A per controller battery side, then you need to watch the heat. With my ventilated cooling approach I run at 150A/controller without any temperature worries, but I do have mountains to contend with and if I want to run hard to show up motos I don't want to have to worry about heat. Getting to max power is all about getting to a high enough voltage, because there's no getting around current limits, but rpm limits are still unexplored by anyone.

Get the motor out of the wheel, so you can gear it down, along with a pair of truly high voltage controllers, and then 40-50kw is perfectly reasonable with a ventilated motor that has centrifugal blades on the exhaust side. It would take over 200V for iron core losses to get to 1000W. While I wouldn't want to cruise continuously at that rpm and heat generation with my simple cooling, out of the wheel with some ducting and a filter on the intake, it's reasonable, especially since Miles' spreadsheet predicts 96% efficiency while making 127Nm of torque at 200V at 244A. HubMonster will actually do a bit better than his spreadsheet predicts, as I erred on the conservative side with my measurements, which others have proven with their own slightly better measurements. That was with cheapie controllers, and I'm seeing improved range with my PV controllers pushing one of my HubMonsters, so sine wave means true limits are higher do to improved efficiency.

When you find that miracle $79 controller that does 16kw, let me know. In between time, I think that's a fantasy!

We had talked about running the hubmonster at 40kw....that's why I want to shoot for that and of course run at 131v. AND I want to do it on FOC controllers. I've had PV 150v controllers for a long time now, but they are sinusoidal and not FOC. It's probably not a problem with the hubmonster since even at 131v the eRPM won't exceed 50,000 (PV controller limit). With other motors, I've exceeded this several times on these controllers.

Taking the motor out of the wheel...yeah why not? I was planning on running it in the back wheel and gear via tire diameter, but I honestly prefer mid-drives anyway! This would also let me open it up a lot...maybe make it into an actual outrunner.
 
shaman said:
If @galp is using 150V MOSFETs, then I would think max operating voltage would be about 120V. That's where galp is possibly getting the 30S number from (maybe). It leaves room between max operating voltage and absolute max voltage that the components can handle.

Yes, I agree about having a safety margin I'd never run 100v mosfets at 24S for this reason. I want about a 20v "slush factor" safety margin. Same for 150v mosfets. 131 volts ought to be OK and still maintain that 20v safety margin.

On that subject, I have 2 150v MAX sinusoidal controllers and I've run both of them at 36S (147.6v) on the bench, but I'd never do that under load. There's just no margin for error!
 
ElectricGod said:
John in CR said:
I've never found a $79 controller that will handle 16kw peak. I've also never run a HubMonster at 40kw. 30kw is the max I've used and never more the 17kw peak input without cooling mods. If you run much more than 100A per controller battery side, then you need to watch the heat. With my ventilated cooling approach I run at 150A/controller without any temperature worries, but I do have mountains to contend with and if I want to run hard to show up motos I don't want to have to worry about heat. Getting to max power is all about getting to a high enough voltage, because there's no getting around current limits, but rpm limits are still unexplored by anyone.

Get the motor out of the wheel, so you can gear it down, along with a pair of truly high voltage controllers, and then 40-50kw is perfectly reasonable with a ventilated motor that has centrifugal blades on the exhaust side. It would take over 200V for iron core losses to get to 1000W. While I wouldn't want to cruise continuously at that rpm and heat generation with my simple cooling, out of the wheel with some ducting and a filter on the intake, it's reasonable, especially since Miles' spreadsheet predicts 96% efficiency while making 127Nm of torque at 200V at 244A. HubMonster will actually do a bit better than his spreadsheet predicts, as I erred on the conservative side with my measurements, which others have proven with their own slightly better measurements. That was with cheapie controllers, and I'm seeing improved range with my PV controllers pushing one of my HubMonsters, so sine wave means true limits are higher do to improved efficiency.

When you find that miracle $79 controller that does 16kw, let me know. In between time, I think that's a fantasy!

We had talked about running the hubmonster at 40kw....that's why I want to shoot for that and of course run at 131v. AND I want to do it on FOC controllers. I've had PV 150v controllers for a long time now, but they are sinusoidal and not FOC. It's probably not a problem with the hubmonster since even at 131v the eRPM won't exceed 50,000 (PV controller limit). With other motors, I've exceeded this several times on these controllers.

Taking the motor out of the wheel...yeah why not? I was planning on running it in the back wheel and gear via tire diameter, but I honestly prefer mid-drives anyway! This would also let me open it up a lot...maybe make it into an actual outrunner.

131V fresh off the charger isn't going to cut it if you really want to get to 40kw unless these magical controllers have the ability to limit max current at will. At 131v you're looking at a max of 120V or less under full load. If you run more than about 250A of peak rms current, you're going to run into heat issues with a daily rider. Even to the extent you improve cooling, you still end up wasting energy on copper losses. The route to getting the maximum out of these motors is high voltage. 100A per controller at 200V is a far better path to 40kw than 167A per controller at 120V. It also opens up the potential for 50kw or even 60kw.

I'm not sure why you keep bringing up something I never said, but for my first 9 months of daily HubMonster use back in 2012, I did use dual $79 controllers at over 8kw peak from each controller.
 
I agree that higher voltage and lower current is more efficient. The problem may be practically acquiring motors, particularly not hub motors, that can accommodate 200V. Running something like a ME1616 on 200V would probably run it over it's max RPM. I've spent a bit of time looking for 30kW+ motors that are available to the average Joe but choices are limited that aren't way over priced.
 
shaman said:
I agree that higher voltage and lower current is more efficient. The problem may be practically acquiring motors, particularly not hub motors, that can accommodate 200V. Running something like a ME1616 on 200V would probably run it over it's max RPM. I've spent a bit of time looking for 30kW+ motors that are available to the average Joe but choices are limited that aren't way over priced.

The 3 variations of HubMonster motors are the only ones I know of that are so under-utilized in their designed use that they could be so drastically powered beyond spec. Right an ME1616 can't be pushed so far. Higher voltage and lower current isn't necessarily more efficient. I was just warning EG, that the all too common route around here of stuffing too much current into a motor is not the route to extreme power with a HubMonster. It cracks me up when people set their current limits, so high that they can't keep the front wheel down on their short wheelbase vehicle with a CG so high and rearward. Then they wonder why they have heat issues. The implications of heat in the copper increasing by the square of current is sobering once it sinks in. Unfortunately for most it never does.
 
shaman said:
@maxxximatoze

You may want to check out my controllers for lower voltage/power needs. Might be hard to pump 300A through a 12 FET controller though.

https://endless-sphere.com/forums/viewtopic.php?f=30&t=99152

Thx shaman, but I already follow, with a lot of attention, this thread https://www.electric-skateboard.builders/t/some-new-focers-84v-vesc-6-based-controllers/87227/93 ;) But thx, I'm also going to read this one !
 
John in CR said:
131V fresh off the charger isn't going to cut it if you really want to get to 40kw unless these magical controllers have the ability to limit max current at will. At 131v you're looking at a max of 120V or less under full load. If you run more than about 250A of peak rms current, you're going to run into heat issues with a daily rider. Even to the extent you improve cooling, you still end up wasting energy on copper losses. The route to getting the maximum out of these motors is high voltage. 100A per controller at 200V is a far better path to 40kw than 167A per controller at 120V. It also opens up the potential for 50kw or even 60kw.

I'm not sure why you keep bringing up something I never said, but for my first 9 months of daily HubMonster use back in 2012, I did use dual $79 controllers at over 8kw peak from each controller.


I think we should stop hijacking this thread...it's about the BESC controller, not hubmonsters.
 
You need some voltage margin for high frequency oscillation between capacitor bank and motor / power stage inductance.

Draft layout of power stage:
besc%20g2%20top.png

besc%20g2%20bottom.png

besc%20g2%20power%20stage%20layout.JPG


Here's few things I figured out:
- Two shunts in parallel would be overkill. It's rated for 15W which means 270 A constant, even more with a lot of copper surface on PCB. Everything else will probably melt before this shunt. :lol:
- Transistors are grouped in pairs to ensure current balance. Layout is also done symmetrically to further ensure current balance between transistors. They are also bent 90 deg for easy assembly and to be mounted on a heatsink.
- I'll be using 6x 22 uF film caps and 11x 470 uF low esr aluminum caps to decouple dc rail
- I'll be using two 4x3 mm bus bars to route phase and input current. You can see them on second picture. They will be re-flown directly to PCB. Bus bars will be pressed against heatsink to cool them and shunts with thermally conductive material that doesn't conduct electricity.
- To save some cost motor and input wires will be directly soldered to PCB. I hope this turns out okay and nothing melts. :banana:

For anyone still wondering, this thing is open source: https://github.com/galpavlin/BESC-G2
 
maxxximatoze said:
Hi,

What about this driver https://vesc-project.com/comment/2760#comment-2760 ? You finally don't want to use it ? I can't wait to see the final result and some tests videos ! :)

I didn't use it at the end because it made more sense to use isolated ones in combination with hall current sense. I think I'll go back to them for BESC g2 where cost is a priority. I'm also using low side shunts which means it won't be possible to have isolated current sense.
 
besc%20g2%20bottom.png


Looking at this rendering...

You want the wide portion of the mosfet leg to bottom out at the through holes. This is the portion of the TO-220 package that is rated for 75 amps. The thinner portion of the leg is rated for more like 20-30 amps. You will get considerable heating from the mosfet legs if you don't.

Though holes for mounting the phase and battery wires is a fine solution. It's used widely by cheap and expensive controllers. Just make those through holes large enough for 8 awg wire.

Ceramic insulators under the mosfets is pretty ideal. Graphite insulator is better, but costs a lot more.

https://www.ebay.com/itm/10pcs-TO-220-Alumina-Ceramic-Transistor-Triac-Thyristor-Insulator-Protection/263689683162?hash=item3d652140da:m:mTdFANsbpHqNco4boa3NGIA

Thermal gap filler on the copper and shunts will take care of any heating there.

https://www.ebay.com/itm/ARCTIC-COOLING-ACTPD00001A-Thermal-Pad-the-high-Performance-Gap-Filler-2x2x0-0/292003799708?hash=item43fcc8629c:g:wxYAAOSw~05Z0Ltz

I think your controller will sit on an aluminum carrier like Sabvoton controllers do. This lets you mount everything and then slide it into an inexpensive extruded aluminum shell.

MQCON%20heat%20spreader.jpg


MQCON%20heat%20spreader%202.jpg


MQCON%20board%20bottom.jpg
 
galp said:
maxxximatoze said:
Hi,

What about this driver https://vesc-project.com/comment/2760#comment-2760 ? You finally don't want to use it ? I can't wait to see the final result and some tests videos ! :)

I didn't use it at the end because it made more sense to use isolated ones in combination with hall current sense. I think I'll go back to them for BESC g2 where cost is a priority. I'm also using low side shunts which means it won't be possible to have isolated current sense.

For the mosfet used, why to use 24xTO-220 instead of 12xTO-247 (like the very good and not too expensive infineon IRF100P219, 1.4milliohms but only 100V) ? the board would be more compact and simpler with very good current capability, no ? What are the TO-220 ref you will use on your design, I can't find it anywhere ? Thx :)
 
It's IRF135B203. I think 15V margin is too low for such high currents. Fella from another forum has issues with 18S/300A and 100V mosfets.

F611DY226K100ZLH0J https://www.farnell.com/datasheets/2373766.pdf

A bit expensive but they have very low inductance (6 nH).
 
@galp

Do you know if there is a significant difference between polyester caps and polypropylene caps? Most documents I've found points toward metallized polypropylene for DC link in motor drives but I have no experience to base my opinion off of.

Also the package of the referenced capacitor is referred to as Pizza Box in the datasheet. I find this amusing.
 
galp said:
It's IRF135B203. I think 15V margin is too low for such high currents. Fella from another forum has issues with 18S/300A and 100V mosfets.

F611DY226K100ZLH0J https://www.farnell.com/datasheets/2373766.pdf

A bit expensive but they have very low inductance (6 nH).

The IRF135B203 is a 135v mosfet so 100v is not applicable.
The F611DY226K100ZLH0J is a 100v capacitor.

At 100v...
18S at 4.1v per cell is 73.8v or 26.2v margin.
20S at 4.1v per cell is 82v or 18v margin.

Where did this 15v margin come from?
 
Fully charged 20S lithium is 84V. So 16V margin. You have to understand that difference between saying 15V and 18V is insignificant for this case. For example one motor could produce 10V spikes and a different one 20V spikes at same phase current. It's really hard to know how much is enough but 2V difference for sure won't save you at 84V input voltage.

PP would be better but I couldn't find any. I'll research this more.

Mosfets are Vds rated for absolute maximum. If you exceed Vds voltage of a mosfet it will break almost instantly most likely with audiovisual effects. :mrgreen: Capacitors are rated for operating voltage. They usually survive up to twice that voltage but operating them over their rated voltage reduced their lifespan. So if some spikes go a little bit over their rated voltage nothing should go wrong.

This is also how planned obsolescence works. Say you have a computer monitor and you want to decouple a 15V rail in it. You could use a 16V or 25V cap but you can also use a 10V one and calculate its reduced life span according to your needs.
 
galp said:
Mosfets are Vds rated for absolute maximum. If you exceed Vds voltage of a mosfet it will break almost instantly most likely with audiovisual effects. :mrgreen: Capacitors are rated for operating voltage. They usually survive up to twice that voltage but operating them over their rated voltage reduced their lifespan. So if some spikes go a little bit over their rated voltage nothing should go wrong.

A while back I did a "test" on 36v infineon mosfets. This is a voltage range I don't really have much use for so blowing up a few didn't bother me at all. I forget what part they are now. Turned off, they had no problem at 48 volts. However as soon as I turned them on they immediately fried. They worked OK at 40v...which is still 4v OVER their specs. I'm not saying that anybody should rely on a "bit of extra" becasue who knows if it will be true on every mosfet used. However, it's clear that better brands have a larger "margin of error" than cheaper parts.

For example:
Real CREE LED's can be over watted considerably. I tend to run real CREE LED's at something like 20-30% OVER their ratings without issues. However Chinese CREE clones, won't hold up to any amount of over watting without quickly degrading or burning out. I ran a 6v XHP70.2 at 12v for 10 minutes before it burned out. It was insanely bright...like a welding arc. The thing that killed it was the build up of heat. The copper star it was soldered to could not pull heat away fast enough to the aluminum heat sink and the solder melted. I bet if I had soldered the copper star to a larger hunk of copper, the LED could have held up much longer and maybe even years. Running this LED at specs...6v and 4.8 amps...is easy. I tend to run at 7v and 9 amps. All that's needed is a bit better cooling and the LED runs for years and is 20% brighter than factory specs. There's plenty of documented cases of people running legit CREE LEDs well over factory specs. I do this for any real CREE LED.

I bet if I did the over voltage test again on legit AOT290's or IRFP4110's or whatever good quality mosfet, that I would see similar results. The mosfets while rated for X volts would actually keep working reliably at 5-10v more. Exactly how much more and would it be predictable for all mosfets of a specific model...who knows...but I bet it's pretty consistent like it is for CREE LED's.

AKA good components hold up to over watting and abuse better than cheap parts.
 
ElectricGod said:
AKA good components hold up to over watting and abuse better than cheap parts.

There are still some "gotchas" behind this. Yes reputable semiconductor companies often derate their devices to ensure reliability over full temp range or other environment conditions. Yes you may be able to exceed Vmax by X volts at 25C ambient but that may not be the case at 100C. You also may very well decrease the life span of the device by exceeding it's max parameters.

However we can often get away with exceeding the max parameters of devices for short durations. It's you're call as an engineer/DIYer/user on how much you want to push that limit.
 
Back
Top