Choosing a BLDC voltage

[robzy]

So I am currently working on an electronic scooter, and I'm trying to decide which BLDC hub motor to buy. (I.e. those funky wheels that have a built in motor).

I am looking at a 350W hub motor from Aliexpress. You can buy them as 24V, 36V, and 48V - and I'm trying to decide which.

My current thinking is this, and I am wondering if it makes sense:-

I will be using an FSESC controller. For power I am planning to use 2 battery packs in series, each containing 6 LiPo 18650 cells - for a total of 12 cells in series. 6 cells in series seems to be the maximum that most associated tech (e.g. chargers) supports.

This gives a minimum/nominal/maximum voltage of 36/44/51 respectively.

I was thinking that I should buy a 36V motor. Am I correct to assume that it will be okay to use this with higher voltage batteries with the appropriate PWM duty cycle?

[Siwastaja]

Battery pack voltage can be higher than the nominal motor voltage, just don't go ridiculously overboard as that would make duty cycle very small. Also if you use a high-voltage controller, it will have MOSFETs dimensioned for that high voltage, making it a compromise for current handling (given same price and size of semiconductors), yet with low-voltage motor you need a lot of current.

So trying to match the motor and supply/controller voltage ratings is generally a good idea, although you may want to give a small bias towards having a bit of extra voltage available if you want to be sure you can extract all the torque also at higher speeds.

With BLDC motor ratings, the "nominal" voltage tends to mean the input DC voltage to the controller which allows most of the motor performance to be extracted. If you have higher voltage available, then you'll be able to run with the combination of higher motor current with high RPM. This meaning more power available, also means the motor may not be designed to handle such power dissipation continuously, but you don't have to run the motor outside the specs. Everything depends on the configurability of your controller.

Having some extra voltage means your DC link is running at lower current given the same motor current. (The combination of controller + motor is equivalent to a synchronous buck converter.) Going too far for mismatch, and the DC link capacitors have to see the high current peaks from the high motor current.

With the cheap Chinese motors, nobody knows what the ratings actually are or what the given ratings are supposed to mean. Chances are, the motors are identical. Or, maybe the windings are really configured differently so that a lower voltage rating motor requires higher current to create the same torque; basically fewer turns (current * turns defining torque), larger number of strands in parallel.

In motor design, the sweet spot tends to be large. In a 350W range, you can definitely design a well-performing, efficient motor just as easily to work with 24V, 48V or even maybe 12V or 250V nominal. They all have similar windings, the difference is between the configuration of strands being in series or parallel. Each strand contribute to the magnetic field according to the current flowing through it.

So there is no easy answer which is best, but if you already have a controller in mind, that obviously sets the maximum pack voltage in stone. Seeing 12s pack specified, minimum pack voltage at 3V/cell would be at 36.0V, so a 36V motor would sound like a sane choice, you get the rated performance with an empty pack, so during normal use you have a bit of excess voltage, but not too much for a mismatched design.

(Looking at the product images of the Chinese VESC clone, it's kind of cute how they are cooling the MOSFETs through the package plastic yet they would have the copper pours with shitload of thermal vias right next, but they decided not to extend the heatsink coupling to that area )

[Hiemal]

I can't add anything on the motor stuff, but I will say, if you can, avoid the "4.2" version of the FSESC; it's not an official design from the original creator of the board, but rather Flipasky's redo of it with cheaper fets (instead of one for each leg of the H bridges, they use two), and a modified layout. Heard they actually used suggestions from people on their forum boards, due to simple design mistakes they made on some of their earliest revisions.

The 4.12 is what I would personally go for and trust more.

[robzy]

Thank you Siwastaja, that's a really fantastic set of information. It confirms some of my unspoken assumptions about my setup.

Hiemal, thanks for the notes on the FSESC. I've already ordered the 4.2 (I was drawn in by the heatsink) but I'll keep that in mind.

[uer166]

Curious, do any scooter or other small controllers do field weakening? That would allow lower voltage packs at high speeds with back-emf that would be higher than pack voltage otherwise.

[Pawelr98]

Generally speaking, extra voltage is also needed because of the battery discharge curve.

An average cell under medium load (say bit less than half the rated continuous current) will have around 3.5V when discharged at 50%.
So if a 36V motor is used, then a 10S pack at 50% discharge you will already be running below the rated voltage which may affect the bike performance.

Myself I have a 36V motor paired with an 11S pack, which I charge as 6S+5S and use without BMS, just fused (two in series just in case) and charged with a charger that also does balancing.
So extra cells are quite helpful, because you have all the performance, even when the battery is getting close to flat.
Or you may just use a 5S and 6S BMS wired in series if you want to.

11S like in my case will be flat at around 33V (2.75-3V/cell is around 10% capacity left, no need to push that far, less deep discharge prolongs the battery life) while fully charged at 46.2V, which can give the 36V motor quite some speed "boost" if the controller allows for faster switching.
That's what I would call a decent compromise for a 36V motor.

[amyk]

Curious, do any scooter or other small controllers do field weakening? That would allow lower voltage packs at high speeds with back-emf that would be higher than pack voltage otherwise.

I believe these are permanent magnet motors, not wound-field ones, so field weakening isn't applicable.

[uer166]

Curious, do any scooter or other small controllers do field weakening? That would allow lower voltage packs at high speeds with back-emf that would be higher than pack voltage otherwise.

I believe these are permanent magnet motors, not wound-field ones, so field weakening isn't applicable.

Field weakening exists precisely for permanent magnet BLDC motors. Most EVs use that.

[robzy]

Thank you too, Pawelr98.

I just had a thought, though. Should I get a 24V BLDC motor and run it from 36-51V at an appropriate PWM cycle?

I wouldn't mind having the option to over-volt the motor during initial acceleration, even with low batteries. I am sure if you ran it even semi-permanently at 36V it would overheat and die, but I don't care too much about top speed, and keeping to spec should be fine after initial acceleration.

[amyk]

Curious, do any scooter or other small controllers do field weakening? That would allow lower voltage packs at high speeds with back-emf that would be higher than pack voltage otherwise.

I believe these are permanent magnet motors, not wound-field ones, so field weakening isn't applicable.

Field weakening exists precisely for permanent magnet BLDC motors. Most EVs use that.

https://en.wikipedia.org/wiki/Brushed_DC_electric_motor#Field_weakening

Also see attached, page of book from 1914 when permanent magnet motors were basically nonexistent.

[uer166]

Of course field weakening existed before BLDCs did. My point is it's applicable and very relevant to BLDCs. You can "artificially" increase the top speed/RPM of a BLDC motor by adding field component that is co-axial to the rotor field (this is trivial with Field Oriented Control). Essentially "fighting" the rotor field by applying one of opposite direction.

What that allows to do is, spin the motor at a speed where the back-EMF would normally be substantially higher than the battery source voltage. This allows lower battery voltage yet high top speed. The disadvantages are: less efficient, this field needs real current/power yet does no real work. And it's dangerous: you need to actively keep control, if you lose control at high speed in deep field weakening region, the BEMF will rise much beyond the pack voltage, destroying the inverter and/or pack thru the body diodes.

[Siwastaja]

Curious, do any scooter or other small controllers do field weakening? That would allow lower voltage packs at high speeds with back-emf that would be higher than pack voltage otherwise.

Even if they did, the problem with electric traction, even with e-bikes, is you really need more power for more speed, because the power requirement due to air drag goes up quadratically. Outside field-weakening region, available power is only rising linearly per speed, even this isn't enough to match the physical requirements, so the motor needs to be sized for the power required at high speeds.

In field-weakening, you are increasing speed while keeping power constant which is definitely not what you want to do, unless you have way oversized motor.

That's why field-weakening isn't super important for a e-bike or EV controller. All the complexity and risks you mention, and the feature wouldn't be practically ever used.

You might use it if you have a strong, oversized motor which can do more torque than you need but you have way too little voltage available, then you would be able to trade off some excess low-speed power for high-speed power you need. But it's easier and less hacky to just to use high enough pack voltage (or a motor wound for lower voltage).


This "quadratic power requirement" is based on expectation of getting constant acceleration out of it at all relevant speed, of course in real world we accept our vehicles lose ability to accelerate quickly at high speeds, hence we are somewhat satisfied with linear power. But with only constant power available, we quickly totally hit the wall so basically the maximum speed will be right after the boundary where we switch to field weakening.

[opampsmoker]

Looking at the product images of the Chinese VESC clone, it's kind of cute how they are cooling the MOSFETs through the package plastic yet they would have the copper pours with shitload of thermal vias right next, but they decided not to extend the heatsink coupling to that area

Agreed...this will be your problem more than the motor problems.
Heatsinking FETs through plastic. And even the heatsink itself is pathetically small for your power useage.
They say they use 3oz copper foil.....but the heatsink in no place even contacts to that copper foil...as pointed out by Siwastaja, the thermal vias go nowhere.

Hope you dont get voltage spikes on those 60V FETs when you're running them at 51V battery voltage.

[Siwastaja]

The phase wires and the board itself will act as the heatsink. The aluminum heatsink is helping just a tiny bit (way less the designer likely expects). Whether it's enough or not, who knows?

They rate "voltage spikes up to 60V". Seeing they use 60V rated MOSFETs, this means they have exactly 0 volts headroom to the MOSFET max Vds rating. This is just beyond laughable when you consider this is a hard switching loop which always rings, more or less. Yes, the FET is avalanche rated but this is still just nuts. Layout parasitics need to be just about perfect.

My layouts tend to peak somewhere around Vin + 20% under typical conditions, if I would be designing a 12*4.2 = 50.4V max input inverter, I'd rate max input at maybe 55V then verify my layout adds at most some 20% peak, then add another 10-20% of leeway so pick MOSFETs with Vds_max = at least some 75V.

This product is very marginal on a 12s pack, don't be too surprised if it just blows up. Obviously, this isn't surprising with cheap hobby modules anyway.

[james_s]

That looks very typical of the ESCs used in RC aircraft, they also usually have a piece of aluminum over the top of the mosfets. It isn't ideal but it does usually work, a good quality ESC does not produce a great deal of heat, so the plate over the top of the packages is normally adequate.

[robzy]

My layouts tend to peak somewhere around Vin + 20% under typical conditions, if I would be designing a 12*4.2 = 50.4V max input inverter, I'd rate max input at maybe 55V then verify my layout adds at most some 20% peak, then add another 10-20% of leeway so pick MOSFETs with Vds_max = at least some 75V.

This product is very marginal on a 12s pack, don't be too surprised if it just blows up. Obviously, this isn't surprising with cheap hobby modules anyway.

That's a fair point. They seem to claim that it supports up to a 12s battery pack, but I can see how that might be cutting it a bit close, especially at the [relatively] high powers that I'd like to play with.

I'm thinking that I'll move down to 2x5s battery packs instead.

I wish I could find an affordable(-ish) 18650 battery kit, so that I could DIY batteries and reconfigure as needed, but no such luck.

[Dave]

Even if they did, the problem with electric traction, even with e-bikes, is you really need more power for more speed, because the power requirement due to air drag goes up quadratically. Outside field-weakening region, available power is only rising linearly per speed, even this isn't enough to match the physical requirements, so the motor needs to be sized for the power required at high speeds.

...

This "quadratic power requirement" is based on expectation of getting constant acceleration out of it at all relevant speed, of course in real world we accept our vehicles lose ability to accelerate quickly at high speeds, hence we are somewhat satisfied with linear power. But with only constant power available, we quickly totally hit the wall so basically the maximum speed will be right after the boundary where we switch to field weakening.

It's even worse than that, I'm afraid.
The force of air drag is quadratically proportional to speed. Because power is speed times force, the power requirement is actually increasing cubically with speed.