Powerchair Direct Drive Brushless Motor (High-Torque)

[bigmoose]

Thanks for opening her up Amber. Looks like 2 linear halls, and they have a multipole ring magnet spinning by them. Might be interesting to figure out how many poles on that ring magnet, you might be able to place digital hall sensors and use that ring magnet to trip them.

[amberwolf]

That's a thought...but given where that ring is inside, I'd have to "know" where to place the sensors, because putting them in there means reopening the motor for every adjustment to their position.

I'd much rather play with positioning them on the outside, if possible. Failing that, just physically putting them in a 120 degree physical arrangement on the stator itself.


If I did want to determine the spacing and polarity of the poles on the ring, what would I do? Wire up a single hall to resistor and LED, and rotate the ring past it to watch it toggle on and off?

[rehab2021]

Your idea is pretty much what people used on cars with mechanical distributors: a timing light.

[Harold in CR]

Is there a name or identifying model numbers on that motor ? Looks like a good candidate for the hills here.

[rehab2021]

The motor is for Invacare powerchairs: part #1152665 or 1152666. The only difference between the two is the orientation of the brake release lever.
Invacare's parts search: http://www.invacare.com/cgi-bin/imhqprd/inv_catalog/partsPHII_home.jsp?s=0&partsHome=partsHome&newSearch=true&area=Main&WT.svl=topNavLink3

[Harold in CR]

Thanks for that info. I was all over Invacare site, BG searches, and the info for making these things run. Just could not find a picture to be sure of what I was reading. There was another named company that controls the info, and, makes it difficult to
obtain. Can't remember the 3rd company name.
Anyhow, They sense off of 2 sensors, and, the controller makes up the third sensor info, or not, depending on the application.

[amberwolf]

I found an old floppy drive that has halls for motor position sensing. Gotta get it apart enough to get them out without damaging them, and see if they'll work for this.

[amberwolf]

Haven't had time to get the drive apart yet, but Icecube57 had lowered a pre-wired/pigtailed hall set to $6 (from $12),
viewtopic.php?f=31&t=33602
so in the interest of time saving plus having a known-good set of halls that I know should work for this purpose, I bought them, and they will probably be here in a little while. LIkely before I will have time to get back to this project, anyway.


I also changed the thread title back to "Powerchair Direct Drive Brushless Motor (High-Torque)" from the "Use SIN/COS position sensor output to recreate 3hall signals", since it's unlikely I will be able to accomplish the latter anytime soon. (though I would still like to find a way to do that, just because it would make it easy to use off-the-shelf motors that sense using SIN/COS on our ebike controllers, with no modification of the motor at all).

[amberwolf]

I finally remembered to measure things:
Stator: 50mm width, 170mm diameter
Magnets: 53mm width, 171mm diameter (0.5mm airgap at most)

DSC05357.JPG (51.95 KiB) Viewed 949 times

[amberwolf]

Got a donated Lyen 12FET sensorless controller that probably has a bad drive transistor. If I can fix it, I can test this motor with it.
http://endless-sphere.com/forums/viewto ... 99#p493399

[amberwolf]

The hall sensors from Icecube57 arrived:

DSC05452.JPG (28.25 KiB) Viewed 772 times

now to find time to install them in the motor (probably at 120 physical degrees for ease of figuring out positioning, at least at first), and then fix one of the several controllers I have around here to test it, or maybe use the Lyen 6FET that's on DGA.

[amberwolf]

Amber before you go too far, put 5V and Gnd on the pins and scope the Sin and Cos simultaneously while spinning the motor and make sure you get the functions out in analog. What is neat if this is true Sin and Cos is that one could build and advance function pretty easily... easily being a pejorative of course! If you get Sin and Cos the next question is to figure out how they are doing it so cheaply and reliably.

DSC05463.JPG (45.95 KiB) Viewed 677 times

[liveforphysics]

You totally rock Amberwolf!


And this motor looks like a beast! That should be capable of a ton of power, and looks like a pretty quality design and build.

[amberwolf]

And these motors are all over the place on all those newer powerchairs--the ones with two large center wheels and four outer "caster" wheels. Not all of those have these motors, but the larger ones and higher end ones should.

As I understand it, the places that service these may not be able to get replacement parts for the brake assembly, so they may just toss out the WHOLE MOTOR ASSEMBLY when the brake wears enough to start slipping. After all, medical insurance / medicare pays for the repairs of the equipment, so the end-user doesn't have to foot the expensive bill for that....

So...go hunt down your local repair places and see what kind of deals you could make.

[amberwolf]

After I get the chance to open it up and find the screw that came out of the handle pivot, I'll use this latest donation from a local friend to try to capture both signals at the same time:

DSC05468.JPG (42.77 KiB) Viewed 643 times

It's a Hitachi V209 portable oscilloscope, although I don't know if the battery still works or not (easily fixed if it doesn't ). Since it can be powered by external 11-14V, too, I could power it off my lighting pack on CrazyBike2, if I needed to take it somewhere for off-grid troubleshooting or repairs or something (like at a racetrack, though hopefully nothing would require a scope at a track!).

DSC05470.JPG (49.09 KiB) Viewed 643 times

It came with two presumably Hitachi probes with 1x/10x switches, as well as a Tektronix standard probe and a Tek current probe (so now I have two Tek current probes...that could come in handy someday, like for comparing phase currents?).

DSC05471.JPG (69.44 KiB) Viewed 643 times

There was also an ammo can and some other assorted electronics, including a bunch of TO-3 style 7812 voltage regulators (and a few TO220 7812s).

DSC05467.JPG (55.24 KiB) Viewed 643 times

Some TO-3 style 2N6285s, which are PNP Darlington comp-pair transistors, rated at 10A 60VDC. Some 2N6282 same family NPN style.

Some TO3 style 2N5875s, PNP 60V 10A power transistors.

Some triacs, TO202 style, L6004F51, 600V 5A.

Some triac drivers, 6-pin DIP, MOC30411 which are zero-crossing detecting and optoisolated.

Guessing these were all for PSU repair, or control system repair. I'm sure I'll figure out something to use them for.

[amberwolf]

Easy enough to find the loose part; was just a brass disc for the handle screws to thread into. reinstalled n handle and fixed that.

Noted this scope has seen some action over the years, but it does still work at least on AC.

DSC05472.JPG (66.41 KiB) Viewed 1313 times

Internal NiCd (12V 1.8Ah) isn't charged up enough to run it, and I don't have a cord to hook up to the scope's 12V input. It's kind of like a laptop two-pin AC cord, but square on the sides instead of rounded, and is quite a bit deeper than any of the cords I have. So I'll have to make one up for it.

DSC05473.JPG (58.43 KiB) Viewed 1313 times

According to the specs, it does only take about 800mA on DC to run the scope, so it's pretty low power, at only about 10W power usage typical. It only charges the NiCd with trickle charging, at 180mA, and says it takes 12-13 hours for a full charge. On discharge, it cuts off at 10.2V. At 800mA, with 1.8Ah, I guess it has something less than two hours of runtime on battery, depending also on self-discharge of cells, etc. and how long it has been since they were last charged.

DSC05474.JPG (47.35 KiB) Viewed 1313 times

Anyway, it powers on and both traces respond to input, so once I have a bit more time I can see about using another motor to spin up that power chair motor (maybe friction drive against the front wheel of CrazyBike2 or DayGlo Avenger) while videoing the scope monitoring both sin and cos hall outputs.


Then I need to wire up the new 3 halls for power, ground, and signal, and mount them up at a physical 120 degree spacing (for simplicity). I'm not sure exactly where they need to go relative to the stator, so I will mount them on a ring that I can rotate and then glue into position once I figure out where they should be. (unless someone here has a good starting point idea, I'll start with them between stator teeth, as opposed to centered on them).

Then figure out the phase/hall combo on one of the controllers, and see if I can make it work on ebike controllers.

[amberwolf]

Got the video of the sin/cos sensors working with the motor spinning (powered by friction drive from CrazyBike2's front wheel), and also of the motor's phase wire waveform the same way.

DSC05481.JPG (73.74 KiB) Viewed 1278 times

Test setup was very simple, just the lab supply for 5V on the halls, then Ch1 of the scope on one hall and Ch2 on the other. Signal ground provided by a lead from scope's Trigger input ground to the lab supply ground. Both channels are on 2V/cm. Ch1 zero is center of screen, Ch2 zero is bottom of screen.



Phase waveform test is just ground of Ch1 on one phase, signal of Ch1 on another, Ch1 on AC input setting and 5V/cm (max this scope goes to) and zero is center of screen.


Unfortunately until after I had already put everything away, I didn't think of measuring the actual voltage output from it vs a specific wheel speed, which I could have used to calculate what kind of (unloaded) voltages I could generate from the motor from BEMF at various speeds. Maybe I'll measure that later, when I have it set up to test the three-phase halls once they're installed.

[amberwolf]

Been doing some searching of ES regarding trapezoidal vs sine BEMF, and I am not quite sure I understand it all. But if I do understand it right, this motor would be much better driven by a sine controller since it has sine BEMF, rather than a trapezoidal controller like most ebike controllers (including all the ones I have here, AFAIK).

Also been trying to figure out an easy way to determine the winding resistance and inductance. Since I have that big Sorenson with adjustable voltage and current limit, I should be able to use that and a voltmeter across a pair of phases, plus Ohm's law, and calculate the resistance for a pair of windings and divide by two.

But I don't have anything that will measure inductance at all, so what I was pondering was using a signal generator (which I have around here somewhere, if it works) to pass a specific amplitude and frequency thru a pair of windings, and see if it is delayed by an amount I can measure on my scope, and maybe calculate it from there? I have to look up the procedure for doing this, because I am probably forgetting an important step or component, like having a known capacitance in there, too. I remember doing it in the lab at school, as one of the lessons, but that was 1988 and I don't think I have done it since then.

[amberwolf]

I took teh hour or so today to wire up and test the hall sensor idea, but there is no way to read the magnets from the place I wanted to put them (for ease of figuring out positioning without opening up the motor repeatedly). That would've been on the lip of the mounting box, where the red arrows point.

file.jpg (52.58 KiB) Viewed 1242 times

I should have tested for that before doing all the wiring for all three sensors and rigging up the scope and multimeter for monitoring, and digging out this 36V 12FET to run the motor with:
viewtopic.php?f=4&t=21830
but naturally I didn't think of that, and just started setting up and playing around, then finding there was no hall signal changes happening.


I can't get readings from outside the rotor (behind or either side of the flux ring), as there's not enough flux leakage to trigger the halls from any position I could find.

If I could get them up inside the rotor and still move them around from outside, I could get them up about 3-4mm from the ends of the magnets and it'd still work, but the gap is too narrow for the wire insulation; it'd rub thru it, especially since I wouldn't have it glued down during the experiments till I figured out positioning.

I could machine off enough of the inner lip of the rotor housing (see red marking on rotor in above picture) to give me the gap I need, which would also lighten up the motor some, but I'd rather leave the motor unaltered for first experiments, especially since I can't undo machining metal off, and I can't replace the motor if I screw it up.

So it looks like for the moment, I'll have to fix the halls in place, install the rotor, test it, and if it doesn't behave as desired, remove the rotor, move the halls, install the rotor, and repeat as necessary. That's gonna take a while if I don't get it right the first time.


In the interests of saving some of that time, then, some more of the motor info: It has 45 teeth and 40 magnets; assuming that it is wired one phase per tooth it should then be 15 poles, right?

Assuming that, and assuming I just want neutral timing for now, would the best place to put each hall (in a 120-degree physical separation) be at the center of a stator tooth? Or at one edge or the other?

[rhitee05]

But I don't have anything that will measure inductance at all, so what I was pondering was using a signal generator (which I have around here somewhere, if it works) to pass a specific amplitude and frequency thru a pair of windings, and see if it is delayed by an amount I can measure on my scope, and maybe calculate it from there? I have to look up the procedure for doing this, because I am probably forgetting an important step or component, like having a known capacitance in there, too. I remember doing it in the lab at school, as one of the lessons, but that was 1988 and I don't think I have done it since then.

You can easily do this by supplying a square wave input to the motor hooked in series with a moderately-sized resistor, say 1k or something. Measure the voltage across either the motor or resistor and you can calculate the time constant from the waveform, which will be L/R (R = the external resistor). This is a pretty easy trick that I teach my students in intro EE lab. You might need to choose a different value for R, depending on how good the resolution on your scope is and roughly what you expect L to be. If L is 100 uH, for R=1k the time constant would be 100 ns. The period of the square wave should be quite a bit longer, so 1 MHz or less probably.

[amberwolf]

You can easily do this by supplying a square wave input to the motor hooked in series with a moderately-sized resistor, say 1k or something. Measure the voltage across either the motor or resistor and you can calculate the time constant from the waveform, which will be L/R (R = the external resistor). This is a pretty easy trick that I teach my students in intro EE lab. You might need to choose a different value for R, depending on how good the resolution on your scope is and roughly what you expect L to be. If L is 100 uH, for R=1k the time constant would be 100 ns. The period of the square wave should be quite a bit longer, so 1 MHz or less probably.

Thanks! That actually sounds familiar, so it might be how we did it in lab way back then. If I had the signal generator out I'd try it right now, but it'll have to wait until I have time to find it. The scope is a 20MHz bandwidth, but I don't remember what the best div on screen is (for either the giant Tek 531 or the mini Hitachi).


Earlier, I finally decided to just get the halls in there to test this thing, because I wanna see it spin on i'ts own. So I unbolted the stator from the base plate,

DSC05513.JPG (46.33 KiB) Viewed 1230 times

and mounted the halls from underneath it (so the wiring will all be protected from the rotor, and so that the halls will be closer to the air circulation from the baseplate/rotor gap--right next to it, in fact).

To keep them clear of the rotor, but without doing anything to the stator lams, I notched the filler sticks between the teeth instead, and superglued the halls to those (I didn't want to use epoxy yet, as I didn't even know if they would work there, plus I have no epoxy that would hold up better than superglue anyway).

DSC05516.JPG (76.94 KiB) Viewed 1230 times

Naturally, it can't all go right. I managed to break a leg off the yellow hall while holding it and it's wires down while the glue set.

DSC05515.JPG (43.35 KiB) Viewed 1230 times

I did manage to fix it, after over an hour's dinking around with magnifying glasses, several flashlights and other lights to help me see, scraping plastic off the case away from the remainder of the leg (not even a quarter of a millimeter!), and several tries with the tiniest soldering tip I have. Eventually I got the leg glued down first and then soldered with a glob of solder to the hall. It's the output pin, so I'll know pretty well when it comes off again later. Probalby when I have forgotten it was even broken.

All the duct tape came off after I got the wires all secured down (dots of superglue here and there to the stator endplates; they won't stay glued but will hold long enough for assembly) and halls glued in place. That was at least an hour and a half to hold wires while glue set (tape ripped them loose when removed, so had to hold by hand utnil set).


I had to take it apart twice because of pinched wires, as I didn't move them all out of the way of the baseplate's riser edge, apparently, and ended up with continuity from power and ground to the baseplate. There went another hour or so.


This is the EVAssemble "36V" 12FET I used to test it with, and it works well enough, though the red LED on the PCB flashes all the time it's powered on (can see it thru the end of the case for the wires, which I don't have screwed down). I bolted it tot he motor with the bottom motor baseplate screw, thru the end plate of the controller. Can't quite do what I planned and bolt it diagonally across the baseplate, because the phase wires are a cm or so too short.

DSC05517.JPG (56.16 KiB) Viewed 1230 times

The phases are bullet-style, female, so I just removed the thick plastic housings that wouldnt' fit in teh D-style motor connector, wrapped the bullets in electrical tape to prevent accidental contact, and plugged the bullets onto the large round phase connector pins in the D-connector.

The motor hall wires I just tinned the ends of and stuck them into the female contacts of teh controller's white hall connector.


The whole test setup, using the Sorenson DCS 55-55 at 40V, 10A current limit, to simulate my 36V NiMH pack that isnt' currently charged up enough to use for the test.

DSC05518.JPG (48.72 KiB) Viewed 1230 times

Scope is setup with a Phase (blue) on top at 10x / 2V/cm, and a hall (blue) on the bottom at 1x 2V/cm. I forget what the time setting was at. WattsUp from AussieJester used to monitor power usage.

DSC05521.JPG (37.56 KiB) Viewed 1230 times

It took one full set of hall combinations (6) and then one phase swap and then 3 more hall combinations to find a smooth low-current no-load spin condition, at about 1.6A no-load current full throttle at 40V.

DSC05519.JPG (40.93 KiB) Viewed 1230 times

Ended up with Y-G-B on the phases on the motor connector, and G-B-Y on the halls. (in case I forget later)

Video of operation at 40V:



Later I took it up to nearly 56V, which is as high as the Sorenson will go, and also as high as I think I should go on the 12FET.


But it does work, on a cheapy type of ebike controller, at least at no load. I'll have to bolt the baseplate to something and try loading down the motor, probably via friction to CrazyBike2's front motor wheel in regen mode to a depleted battery.

If it still works ok, maybe I can mount it on CrazyBIke2 in the center frame, bolt a chainring to the face of the rotor, and connect it up to the chainline, and see what breaks first.

[kfong]

Looking good AW, sounds like you now have a very powerful motor for your bike. Looking forward to the final build.

[amberwolf]

"Final build" That could be a while! Eventually this thing will go in the new bike, at least as a test motor, running the wheel via the NuVinci hub with autoshifter. Hopefully, at least.

A few minutes digging htru the chainring pile finds only two that can easily be made to fit: Shimano I24 and I34. Dunno if I even ever had the 44.

DSC05524.JPG (114.61 KiB) Viewed 1219 times

Most likely I'd use the smallest ring, and up the voltage, but I don't actually know what speed the motor goes at a particular voltage (kV) yet. I'll probably setup a speedo on it and find out soonish.

Either way, you can see that hte bolt holes don't quite line up, but I shoulc be albe to drill into them if I can borrow a friend's drill press. Either that or use my round file and lots of elbow grease, to lengthen each hole in the outward radial direction about half again their current size.

DSC05525.JPG (68.91 KiB) Viewed 1219 times

But I will need new bolts, as the existing ones are a bit short:

DSC05527.JPG (32.35 KiB) Viewed 1219 times

I might have some that are long enough but I am not sure they're the same thread pitch. Have to find them and see.

I also found a bracket that coincidentally has teh same bolt hole width as one side of the rear mounting plate:

DSC05528.JPG (37.68 KiB) Viewed 1219 times

It's that computer-chair bracket (went between back of chair and base of seat) that I was planning on cutting up for the swingarm pivot plates on the bike this motor may go on.

DSC05529.JPG (43.72 KiB) Viewed 1219 times

Too bad it's not exactly 90 degrees, or it'd be easy-peasy to mount up. I'll have to shim it up at the bend portion, or something, probably, if I use it (which I probably will because it does fit and is very thick and strong).

[amberwolf]

Some thoughts on RPM measurements:

A friend has a tachometer he can bring over to actually measure it easily, but I am still pondering other ways to measure or calculate it.

I was thinking I could measure it's RPM by friction driving the front wheel of CB2 and calculating what it would have to be to drive CB2's wheel at whatever speed shows up on the CycleAnalyst. Or by putting the speedo sensor on the motor directly with a magnet stuck to the hub's outer diameter (which I plan to do tonight anyway), and figuring out what the math would be to turn that into RPM.

I was also thinking that I should be able to calculate something close to it's actual RPM by knowing the speed at which it would have driven a powerchair at 24V. I presume it would be either 4MPH or 8MPH; I think I was told once but I forgot.

Another thought on RPM measurement is to just put a frequency counter on one of the hall signals, and divide by the number of magnets (40) or maybe that would be by the number of pole-changes (20) or maybe I am just lost.

[Harold in CR]

Do you happen to have a Dremel tool ?? Drifting those sprocket holes would be MUCH easier with a Dremel. Drilling is going to be tough and probably end up snapping the bit. I have a cross feed vise I use on my drill press, and, HSS or Carbide die makers bits I can use for harder stuff. I made both sprockets from saw blade material using the Dewalt side grinder and Dremel for finishing work.

Good to see that motor spinning. You Da MAN.