MAC-BMC Motor "600 watt", Star Connection, Efficiency Study

[mrbill]

This motor was used on the Currie, USProDrive, and Synergy kits (and probably others). Ecospeed also uses or used a controller-less version of this motor, otherwise known as the PowerPack motor. This is not a hub motor.

I recently acquired another instance of the larger "600-watt" version of these motors with a non-working controller. As usual I removed the internal controller or enough of it to make room for dressing new cables connecting the power leads and installing new Hall sensors. With this motor I decided to study its behavior when its phase leads are connected in Wye or Star rather than in Delta, as is done on the stock motor. In theory, torque should increase by sqrt(3) and RPM should decrease by sqrt(3).

I had tried using a Star connection on another instance of these motors I had in my possession last year, but I abandoned the idea after I observed relatively poor efficiency (by about 5-6%) and rough rotation. I suspect there was a problem with one of the Hall sensors or the rotor magnet on that motor as I could find no other difference between it and another motor I had that ran smoothly.

This time everything seemed to go well. The motor ran smoothly the first time I hooked it up. (1 out of 6 chance of that.) I tuned the position of the Halls to give minimum average current draw at wide open throttle, no load, and with minimal variance. That meant aligning the Hall devices with the edge of every third stator tooth as seen in the photos linked below. I saw about 1 Amp at 25 volts and 2 Amps at 50 volts as a minimum.

The peak efficiency of the Star connected motor was similar to that of the Delta connected motor with an important difference. Performance at high power was poorer than with the Delta connected motor at the same voltage. I suppose this should be no surprise since resistive losses are higher with a Star connection as current must flow through twice the length of wire in each phase.

In the back of my head I had thought I might be able to use a Star connected motor at 48 volts and enjoy a similar performance envelope to that of a Delta connected motor at 24 volts, but the efficiency curves show that this is not quite so. Efficiency in the mid-band of the Star motor at 48 volts is roughly equivalent to that of the Delta motor at 24 volts, but at the low power end, the 24-volt Delta motor is more efficient, and at the high power end, the 48-volt Star motor is only slightly more efficient than the 24-volt Delta motor, but not enough to compensate (in my mind) for significantly reduced efficiency at low power. The best compromise may be to use 36 volts with the Star connection, if a Star connection must be used.

Perhaps the Star connected motor doesn't really come to life until it's fed 60 volts or more. My conclusion is to stick with the Delta connection on these motors. The Delta connection will put 1/sqrt(3) less wear on the gearbox, too.

Measured Kv(Star) = 70
Measured Kv(Delta) = 125

Photos of motor w/Hall sensor positions for Headline and Infineon controllers: http://bit.ly/q8HDJ5

Efficiency chart of Star connected motor: http://bit.ly/oFGBbK

Efficiency chart of Delta connected motor: http://bit.ly/opMQbg

 

[neptronix]

Good data, thanks!

 

[ptd]

awesome data, as always. do you maybe have this in spreadsheet form, with voltage and amperage readingss as well ?

 

[mrbill]

awesome data, as always. do you maybe have this in spreadsheet form, with voltage and amperage readings as well ?

Thanks. I do have the data in spreadsheet form, but I didn't record voltage or current for this test. You could get a pretty good estimate for voltage and current by assuming the following: open circuit voltage starts at 26.5 volts and falls to 23 volts at the highest power reading for the 24 volt nominal test. Since I use the same batteries for the 36-volt nominal test, you can scale the starting and ending voltage figures by 1.5. For 48 volts I used smaller capacity batteries wired in series that sagged more at high load, starting at 53 volts and dropping to 44 volts at full power. Interpolate voltage for each input power level and calculate (input power) / (interpolated voltage) to get current at each power level.

Let me know if you want me to send you the original worksheet as a starting point.

 

[ptd]

that would be fantastic, much easier then cutting and pasting pdfs, lol. by interpolate, do you mean just figure the formula for a straight line starting at 26.5 for 33 watts (or would that be for 0 watts ?) and 23 for 894 watts ? i get volts = 26.63 - .004065*watts , which yields 1.25 and 38.88 amps, respectively. sounds about right. thanks

the reason i ask, is i'm trying to see if there's a similar correlation between, peak efficiency, and a specific ratio of volts to amps, like there was in your mars axial flux data. i was wondering if there was a way to "force feed" it a specific voltage and amperage using pwm, and offset the rpm by a shift in gears (maybe nuvinci optimally).

was also wondering about the efficiency of the drivetrain. iirc, it was more like 84% in that other test. is this a different setup ? and speaking of nuvinci, have you ever tested the efficiency of one ? inquiring minds want to know, and fallbrook ain't giving it up, lol. thanks again.

 

[mrbill]

that would be fantastic, much easier then cutting and pasting pdfs, lol. by interpolate, do you mean just figure the formula for a straight line starting at 26.5 for 33 watts (or would that be for 0 watts ?) and 23 for 894 watts ? i get volts = 26.63 - .004065*watts , which yields 1.25 and 38.88 amps, respectively. sounds about right. thanks

Yes. Linear interpolation works well enough for this. You might even get close enough by using the midpoint voltage if you don't want to bother with the interpolation. 33 watts or 0 watts for the start, it will make little difference in the estimated current. 33 watts is probably closer. Heck, I'm even going on memory about the voltage endpoints. And, don't forget that the CycleAnalyst accuracy is rated +-3% and PowerTap hub is +-1.5%, so there's not much to be gained by splitting hairs.

the reason i ask, is i'm trying to see if there's a similar correlation between, peak efficiency, and a specific ratio of volts to amps, like there was in your mars axial flux data. i was wondering if there was a way to "force feed" it a specific voltage and amperage using pwm, and offset the rpm by a shift in gears (maybe nuvinci optimally).

There is an efficiency penalty when using a duty cycle less than 100%. As you can see from the difference in "full-throttle" and "half-throttle" curves, the latter are at roughly 50% duty cycle. What you gain in feeding the motor what it wants to see ideally may be lost in the electronics that accomplishes the same. Lunches aren't free.

was also wondering about the efficiency of the drivetrain. iirc, it was more like 84% in that other test. is this a different setup ? and speaking of nuvinci, have you ever tested the efficiency of one ? inquiring minds want to know, and fallbrook ain't giving it up, lol. thanks again.

Power out is read at the rear hub. Between the Cycle Analyst (power in) and the rear hub (power out) we have wiring into the controller (a short length (<1ft of 10ga wire and an Anderson 45A PowerPole connection), the controller itself (Infineon 12-FET, using IRFB4110s), the power phase leads going into the motor (about 1ft of 12ga wire and an Anderson 45A PowerPole connection), the motor, a ANSI #25 chain connecting an 11t sprocket to a 90t sprocket on a mid-drive, and standard 3/32" width 1/4" pitch bicycle chain connecting a 15t sprocket to a 20t - 34t sprocket at the hub. The wheel is held off the ground and a load is introduced at the rim by a rim brake, so rolling resistance from rubber (tire) loss doesn't enter into the model.

The main losses are in the motor, followed by the two-stage chain & sprocket drive, then the controller. The controller losses are probably somewhat greater than the drive train losses when the former is run at <100% duty cycle, but I'm only guessing here. Based on others' research I estimate the drive train losses at about 6.5% (93.5% efficiency), perhaps less at high power when the chain is under more tension, perhaps more at low power. Wiring and connector losses are probably less than 2%. So, when you see total efficiency at 77% that puts the motor/controller/wiring losses at about 0.77/0.935 = 82.5%. If we estimate 96% efficiency for the controller at 100% duty cycle and 98% efficiency of the wiring and connectors, we get 0.825/0.96/0.98 = 0.88 = 88% efficiency for the motor alone.

This is the same setup I've used to test all of the motors I've tested except for the MAC-BMC w/Currie or Lashout planetary gearbox and the Headline motors w/gearbox in CCW rotation that I tested on a different bike.

I'd like to try riding a bike with a NuVinci. People seem to like how it works but not how much power it soaks up. I haven't seen any third-party efficiency test on this hub, and I wish I could test it to see how much energy is lost. One could probably get a very rough idea by feeling the hub after it has been ridden for some distance. If it's uncomfortably warm, then it's probably wasting significant energy.

My "dynamometer" is built into one of my wheels, so I have no way to test the NuVinci (or a hub motor, for that matter) unless I adapt a resistance trainer with accurate power measurement or build my own rig. I doubt very much that it is as efficient as a chain & sprocket drive-train, either one or two stages. As I see it battery energy is still too valuable to throw away as heat in exchange for the luxury of continuous gear ratios.

Here's a link to the MS Excel workbook I've been using to prepare the pretty charts and tables.
http://bit.ly/ocJYqf

 

[ptd]

wow. seriously. WOW, lol. was just expecting one motor's worth. this is the whole kit and kabootle. fantastic. thank you. i've got data for days, weeks maybe, lol. the helpfullness of this forum and most of it's members never ceases to amaze me.

i really love how you do you the half throttle as well. might be able to interpolate that too. i imagined that there'd be a cost to pwm, didn't think it was going to be that high, i guess. i've been talking with a guy about transformers, and i thought they were pretty inefficient, like 70-80%. i actually made the same observation (that there'd be no free lunch transformer). But i did a little research, and the buck or boost that exhibit those numbers, have a relative, that uses both, the split-pi, and it's supposed to operate around 96%, with the ability to vary voltage up and down (i think by up to 40% iirc). sounds like it might still be effective. your data's also getting me back to thinking that a dual drive might be in order, say a 250w in the front, and 750w in the rear. run one for acceleration, and one for cruise. or even both for more acceleration or billygoatin 2wd.

thanks for the detailed description of the drivetrain. i had pictured an at the wheel dyno of some kind. wondering if it'd be simple enough to just add an inline set of dropouts up above for the ability to test hub motors. then maybe if you had a nuvinci with a disk brake mount, you could just throw a second sprocket on instead. i think i might have heard someone say 90% efficient at 1:1. guessing more like 75-80% for highest or lowest ratios. i get the feeling the sponge analogy is spot on.

well, many thanks again for the data, paul