mounting hall sensors to R/C brushless motors

Beautiful!

Looks fantastic Jeremy!

You make mine look bad :( But I did set mine up on a slotted base plate, so it can be rotated easily to adjust timing to dial things in. But mine are just ABS plastic :( I'm excited to see how it runs for you!
 
I agree, it looks great Jeremy. I'm still trying to understand the concept here. Are you saying you can have the sensors outside the motor case, and they will still work? Wow, that will definitely simplify things. :)

-- Gary
 
Are you saying you can have the sensors outside the motor case, and they will still work? Wow, that will definitely simplify things.

Absolutely. I tested it last night with a hall sensor hooked up to the 'scope and the motor driven from a sensorless ESC. The sensor switched cleanly at over 1/4" from the can of the motor, so running them in an alloy ring just 1/16" or less from the can should be very reliable. The sensors I'm using are Honeywell SS411A latching hall switches.

In parallel with this mod, I'm also fitting halls internally to my other Towerpro motor. It was a bit traumatic putting the motor stator in the milling machine and milling out the slots a tiny bit to take the sensors, but wasn't too bad in the end. I reckon the job could be done with a Dremel if you've got a steady hand. The hard part is wiring the things up, as even the thinnest heatshrink sleeving I've got is way too big. I've ended up wrapping the sensor leads in kapton tape and embedding them in epoxy resin. Seems to be OK, but I'll have to wait another hour or so before I can test them.

Jeremy

edited to add this picture of a hall sensor fitted to the motor stator:

3429671488_97f20d9b3d.jpg
 
It works!

I'm frankly just amazed at how well it works, too!

The set-up

- A Towerpro 5330-10t motor, fitted with three Hall sensors spaced at 120 degrees around the stator, in slightly machined-out stator slots. All three sensors are fitted with the "spot" face outwards (see photo in previous post that shows how they are fitted). I haven't got around to testing the external sensor idea yet, and may not now bother with it.

- A baby 6 FET 48V, 15A Infineon controller, with stock current limit and no firmware changes.

- A 48V bench power supply, current limited to 2A.

- a 5k pot wired as a temporary throttle control.

Here's a picture of the test set-up:

3431645880_705b56e2fa.jpg


The results

- The motor starts very, very smoothly indeed, there's absolutely none of the jerkiness that the big 120A Swang Hobby ESC gives at all.

- The slow speed torque is simply incredible. I cannot stall this motor by grabbing the shaft, even when it's only running at a few rpm. "Idle" speed seems to be right down around 50rpm or so, with an idle current of just 91mA off load at this speed! This is a direct contrast to the performance with the sensorless controller, where low speed torque was non-existent by comparison.

- At full speed, no load, the motor draws much less current than it did with the sensorless controller, even allowing for the higher operating voltage, implying a bit better efficiency perhaps. The RC ESC drew over 200mA even with the motor stationary at zero throttle, the Infineon is drawing just 59mA in this condition.

- The only downside is that at very low speeds with a lot of torque loading on the motor the controller FETs started to get very slightly warm (maybe 30deg C), even though the supply current was pegged at just 2A. I was running it without the case, though, plus the FETs in this controller are still the high Rdson ones it came with. Fitting the IRFB3077s should reduce the controller losses by a factor of four, so leading to less heating, although I don't think it's really a significant issue.


I managed to get the Hall sensor and phase wire orientation right first go, by the simple expedient of driving the motor with the RC ESC and putting the 'scope on each Hall sensor in turn, comparing it's switching point with the zero crossing of the phase voltage.

I very strongly recommend this way forward for driving these big RC motors. The big, 18 FET Infineon controller should be OK for 80 to 100A at maybe 72V, enough power for the biggest budget RC motors. Even a Kelly controller might well work well with this mod, opening up even more powerful motor possibilities.

The smoothness of this controller/motor has to be seen - it really is super smooth compared to the sensorless controller, plus the motor runs much more quietly at all speeds. The massively increased low down torque will also make gearing easier, as there shouldn't be any need to consider using two speed systems I think - just gear for max speed and let the increased motor torque take care of the low speed end.

I'm dead chuffed with the outcome of this experiment, it's turned out better than I would have hoped.

Jeremy
 
Very interesting results, indeed. :) It would still be interesting to see if your external approach would be just as smooth. That would make it easier to implementwith existing motors. I Think this is something David said he can do with his new big motor, and I'm also going to talk to Bob Boucher at AstroFlight about adding hall sensors to special 3210s and maybe a 3215/3220, but for existing motors, your external idea might be easy enough to do for other motors.

At first I was against the notion of using existing ebike controllers with RC motors, but the fact that I can't use 90% of the RC controllers with a "real" 48V/16s LiFePO4 setup has really soured me on RC controllers. I have only one controller that is rated high enough, voltage-wise, which is the Kontronik Power Jazz, but it is "uber"-expensive at well over $400, and I still don't know if will last in an ebike environment (Matt blew one up...). I will have the caps added, which I hope will keep it together, but I still don't know. I'm not going to buy another one of these in any case, so for my other setup, which is the Hacker A60-18L on my smaller folder, I don't know what to do. I'm definitely not going to rewire my 16s5p a123 packs to turn them into 15s configurations, so the HV110 is not an option now. This sensored approach could be, however.

-- Gary
 
Gary,

I think the real advantage (apart from price) of using existing ebike controllers with these motors has to be the starting torque. The Hall sensors allow the controller to maintain very high torque levels at low rpm, which not only makes it potentially more driveable, but also means that the controller may prove to be more reliable. I have a sneaking suspicion that it's the difficulty that sensorless controllers have at maintaining sync at low speed, high torque load, that might cause some of the reliability issues (they don't usually have any form of current limiting).

Add in the much higher voltage capability of ebike controllers and it becomes a bit of a no-brainer to go down this route.

I may try the external sensor option, but now think it's probably a lot harder to implement. The problem is accurate sensor placement - it's tricky to align the outer ring with the stator. Fitting internal sensors is a bit simpler. These motors come apart very easily indeed. The Towerpro has a retaining collar held on to the shaft by two grub screws. Undo this and the rotor just pulls off. Cutting the slots a bit wider is a bit nerve-racking, but no harder than replacing the Hall sensors on a Crystalyte motor. The only tricky bit is working out which sensor goes with which phase wire.

Having got this motor to run well with the Infineon, there's no way I'd contemplate using a sensorless controller now - it really is a massive improvement in my view.

Jeremy
 
Jeremy,
Cool result!
now we have two conversion system working :D
I managed to get the Hall sensor and phase wire orientation right first go, by the simple expedient of driving the motor with the RC ESC and putting the 'scope on each Hall sensor in turn, comparing it's switching point with the zero crossing of the phase voltage.
this is what was i trying to do, but i could not resist the simplicity of blindly trying all combinations :mrgreen:

i am building a dyno to test the motors but unfortunately the coupling b/t the motor and the load seems to introduce too much noise into the torque readout (using simple kitchen scale)...still improving it....
You tried to dyno the motor before?


-George
 
I like that there's experimentation in this regard(Potentially cheaper and more appropriate for ebikes), but I really haven't noticed a lack "low RPM" torque with my Castle Creations controller. At start-up from absolute 0 mph(Not even a slight .1 mph push), it might buzz for upto half a second but it inevitably finds its way and starts applying current in huge quantities quickly so there doesn't seem to be an issue with a lack of "Low RPM" torque with CC controllers.

I also wonder... is higher voltage truly better for everything? It increases the no-load RPM of the motor which may lead to greater heating of bearings, but if you're doing some form of "speed limiting"(Even if you're manually keeping yourself under 40 mph!), that shouldn't matter. What other potentially negative affects would higher voltage have? I noticed some motors were rated for a given voltage(like 36 volts), so would going above that with the controller matter to the motor beyond the RPM consideration above? What about the power dissipated in the motor and the motor's efficiency? This is something that it seems a few graphs could help elucidate so I may want to check a few graphs using my simulator.

To avoid over-volting and its currently unknown unintended consequences, one could just go with a controller at a lower voltage that has a higher current limit. But, seriously, are there any cheap brushless 100+A e-bike controllers out there that'll take 36-48 volt? I've also noticed that knuckles doesn't offer a warranty on the Infineon controllers and one possible way to ruin them seems to be to go up a huge hill, so it seems they're not infallible from a heat view-point, but they might be considerably better in that regard than current R/C controllers.
 
I've not yet tried the dyno approach, George, my "measurements" are purely subjective. I have thought about making a drum dyno with a band brake and spring balance. I remember using one of these many, many years ago at college. It always seemed a pretty simple thing to build. At the end of the day, though, I'm just after something that works and I'm pretty sure that's what I've now got will do what I want.

swbluto,

The starting torque at very low motor rpm really does have to be seen to be believed. It's really, really smooth, the motor just picks up from zero rpm to just a few rpm in an instant, with lot's of torque, even when just slowly idling. Maybe the expensive Castle Creations controllers are much better than the cheap Swang Hobby ESC I've been using, but as the Infineon controllers are so cheap, there's no real contest in the "bang for your buck" stakes.

The main advantage of high voltage systems is the much reduced battery current demand. Doubling the voltage reduces the resistive power losses for a given power by 75% (as they are proportional to the square of the current). 50 amps is pretty easy for an Infineon, by all accounts, as is 72 volts. This means that more than 3.5kW is pretty straightforward and reliably achieved, which is more than enough for most people, I suspect. As an 18 FET Infineon fitted with 4110 FETs (capable of around 3.5 to 5kW) is only around $100, it compares pretty well with the highest power RC ESC units that are around two or three times this price for the same power.

The other advantage that ebike controllers like the Infineon have is repairability. If you do overload it, then it's odds on that you can just change a few FETs and get it going again. This isn't an easy option with an RC ESC.

Jeremy
 
Oh yes, definite bonuses, especially the reparable aspect. I'm pretty convinced that the infineon w/ 4110s installed is a pretty worthy contender, and I'll definitely consider going that way if I decide to continue with e-biking and eventually get a geared motor setup.

I have more questions about the infineon controllers, but those are probably better reserved for knuckles.
 
I'll take high voltage over high current all day long. Heat losses go up as a square function of the current, so the higher the current, the more heat you have to deal with. Even ignoring the fact that for most of the motors we've all been playing with can come in lower kV versions, we aren't really at any where near the rpm limit for any of these, so they could be run at a higher voltage and higher rpm, if it is "convenient" to the gearing options The 3210-10t has a kV of 135 and at 50V, they will run at about 6500 rpm, but these could easily handle 8-10k. These things get their max efficiencies at about 7500 rpm, so you really need to run the 10-turn at 60V to come closer. It can certainly handle the 4-5V more that a 16s a123 setup will put out over a 12s LiPo configuration, and get the rpm up closer to the 7500 rpm value. The problem is the RC controllers just can't hack the extra volts.

For a 72V setup, I would use the 12-turn version, which drops the kV down to about 113: http://www.astroflight.com/pdfs/3210.pdf. which would get very close to the max efficiency point.

-- Gary
 
Great work Jeremy, sounds like you're onto a winner.

The litmus test would be if this setup could handle the sort of punishment handed out by liveforphysics. :D
 
We will soon find out if it can handle abuse from me :) I finished modding a pair of 12fet infinion controllers with irf4110s, I made adjustable brackets to mount the Hall effect sensors. I also have a pair of the 18fet infinions comming, which will also get irf4110s.


But, I have a question. I don't know much about working with the hall sensors, but the brackets I made only wrap 180deg around the motors. I'm have 41s honeywell hall effect sensors. George said he was able to mount them all on one side by flipping the one in the middle over. Now, I'm confused. If these just latch high or low when in a strong field, how does flipping them over help? I'm also still a little puzzled at the exact degrees spacing I need between the sensors. My mounts can adjust the full width of a stator pole, so I should be able to dial things in if I can get them mounted with the right spacing, but I'm still unclear on this.

If I do two of them facing towards the magnets on a 120deg spread, and then one centered between them fliped away from the magnets, is this the correct spacing? Then line up the outside two fets with a gap in the poles, which will make the fliped sensors sit in the middle of a pole (I think). Does that sound right?

-Luke
 
If you fit the outer two sensors with 120 deg between them (assuming your motor is a 12 slot, 14 magnet one) then the third sensor can be fitted dead centre between the outer two, but flipped the other way up. This will fool the controller into thinking that the third sensor is really on the other side of the motor, in effect.

There are other spacing options for them with the Infineon, as it will work with 60 deg spacing I believe, but the above method is pretty much the easiest.

I fitted my internal sensors all around the motor, as it was just as easy to do as any other placement arrangement. The idea is to aim for the sensor pulse edges to align with the FET switching points (at the zero crossing point). When I checked my motor on the 'scope with the sensorless ESC driving the motor and the Halls separately powered, I was pleased to find that the centre of the Hall rising/falling edges was exactly aligned with the centre of the rising/falling edges of the trapezoidal phase drive voltages. As I'd just fitted the sensors where they'd best fit, this was a pleasant surprise!

Jeremy
 
This is great news! Anybody wanna buy a HV110? :wink: So is there any way the common man without access to a scope can try this? Is that what you are doing Luke? Can you just turn your sensor mounts to fine tune the timing, kinda like on the distributor of a car? There has been so many groundbreaking discoveries this year, it's just amazing!
 
Thank you Jeremy, I will mount them just like that when I get a little time off work again.


Here are some pictures of how the mounts adjust. Much like a car distributer, slotted holes and screws to clamp them in place. The hall bracket is made of ABS. Crude, but it should work ok for testing and seeing how I like it before I step-up to something more permanent.

p4110323.jpg


p4110324.jpg


p4110325.jpg


p4110326.jpg


Best Wishes,
-Luke
 
Looks fine to me, Luke, I can't see any obvious reason why it wouldn't work.

I've spent the morning fitting IRFB3077 FETs into my baby Infineon, complete with mica insulators and a lapped-flat spreader bar. One thing worth looking out for on these controllers is the burrs inside the case where the spreader bar bolts on. On my case these were pretty big and stopped the spreader bar from fitting tight to the case. Easily sorted, but something equally easy to miss when putting the thing together.

The rest of my time has been spent making a programming interface for the Infineon, following the excellent post by philf here: http://endless-sphere.com/forums/viewtopic.php?f=2&t=8317&hilit=regen&start=30#p126939. It works a treat and I've already played around with some of the settings.

Jeremy
 
Nice work Jeremy,

Another approach might be to sneak the hall sensors in on the end of the stator, something like this:Hall sensor in RC motor 2.jpg

On the end, it may work fine to have the sensors mounted with the face aimed toward the front of the motor, rather than toward the side. This would avoid the need for machining the slots and it looks like there may be adequate space inside the housing. There might be enough room to slide one into the end of the slot on top of the windings, but below the stator teeth.

The axis of flux measurment is perpendicular to the flat face on the sensor, so if you rotated them so the leads were coming out in the direction of the gap between the stator and the end plate, it should work the same. This might make fitting them easier in some motors.

The external approach seems like it should work fine too, as long as there is adquate flux outside the ring. If you can strongly attract a paperclip or piece of steel to the outside of the ring, it should be good. I like the adjustable timing feature, but I think you'll find that neutrally timed works the best.
 
I will start with neutrally timed, then measure acceleration and current draw. Then advance 5deg, measure acceleration and current draw, then retard 5deg and repeat.

I've never had any experience with a motor with this many poles, but I remember with low pole count RC motors, timing advance can sometimes make a big power improvement. Neutral may be the best for these motors, but the tinkerer in me won't let me be content until I give various advance and retard positions a whirl :)
 
Fantasic brake through! Jeremy, will a non latching hall device work also? (I have some on hand). I think there is excess plastic on the sides of the device, file to fit?
 
George said he was able to mount them all on one side by flipping the one in the middle over. Now, I'm confused. If these just latch high or low when in a strong field, how does flipping them over help? I'm also still a little puzzled at the exact degrees spacing I need between the sensors. My mounts can adjust the full width of a stator pole, so I should be able to dial things in if I can get them mounted with the right spacing, but I'm still unclear on this.

flipping the sensors is equivalent (almost) to adding an inverter to the output of the sensor, so it is like making 1 to 0 and 0 to 1. your turnigy motor looks identical to mine, so i guess you could simply use my setting of mounting the sensors at the middle of two slots.

for example:
slot1 xxxxxxxx(sensor1, facing out)xxxxxxx slot2 xxx slot3 xxx(sensor2, facing in) xxxxx slot4 xxxxxx slot5 xxxx (sensor3, facing out) xxxxxx slot6
---------------------------->|<--------------2 slots------------------------------>|<------------------2 slots------------------------------->|

hope this helps to clarify.

-george
 
Thank you George. That does help.

You have not even been here a month, and your work has all ready inspired and helped to bring new ideas to solving the RC controller problems. Jeremy seems to be very happy with the control this gives his motor, and I'm excited to be happy too! I can't wait to see what else you will bring the forum in the future.

Best Wishes,
-Luke
 
Fechter, I'm pretty sure that putting the sensors at the end can be made to work, the only problem I found when playing about with a sensor as a probe (with the motor driven by the RC ESC) was that there was a fair bit of jitter on the sensor signal near the edges. This may have been induced by vibration though, as I was just holding the sensor against the stator through one of the end slots. Certainly fitting them in the centre has made the signal rock-solid, but I would be surprised if this is totally associated with that position. I chose to do it that way just to support the sensor leads a bit better, but some epoxy dribbled around them when fitted at the end would probably do as good a job.

JEB, I'd go for latching sensors, as they are cheap and reliable. Having the sensor latch tends to make it less vulnerable to noise, too. Honeywell SS411A sensors are readily available from Digikey for around £2 or so, other places sell them for less if you shop around. I guess you could try filing the edges of the sensor down, but the leads are pretty close to the edge of the package, so the chances are than the lead header is only "just" covered. Filling it down might expose the lead header and risk an inadvertent short to the stator.

George, The Infineon controllers can be purchased directly from Keywin Ge (e-crazyman(AT)gmail.com). The price for two 18 FET boards, with cases, all components except the FETs and all the mounting hardware, was $130 including shipping to the UK (I think shipping to the US is about the same). Keywin also sells on eBay, just look for his user name, e-crazyman. He currently has the 6 FET 48V, 15A controller on eBay for $22 plus $23 shipping. He also has the much bigger 12 FET, 72V, 45A controller on eBay for $75 plus $35 shipping. All told, if you want really high power, then buying bare 18 FET boards and fitting your own FETs is probably the best bet.

Jeremy
 
Jeremy,

I am also very excited about this! I have been having good success with my modded (caps added) HV110s. However, I am still not 100% confident they are the best option considering your success here. At this point, assuming adding halls is easy enough for most motors, the only real benefit to using HV110s is the programability and very small size.

Also, I am not an electronics guy. I know just enough to be dangerous. I can do anything mechanical, though. So, mounting the halls is the easy part for me. Getting them wired up properly is another story, though.

Now I just want to see how easy it would be to mount halls to my Astro inrunner motors! :wink:

Matt
 
Back
Top