AT12070 motor (similar to many 12070, 12090... motors)

mxlemming

100 kW
Joined
Jul 17, 2020
Messages
1,117
I bought this from ant innovation technology. Similar ones are available from freerchobby etc but Arvin at ant was very forthcoming and useful so got the business.

Quick write up as requested. Not yet on my bike, that'll come later.

IMG_20211012_012926119_HDR.jpg
Copper fill is not huge, but the resistance is low and this does allow air flow so maybe not so terrible at it looks. The primary losses in this seem to be iron losses. Hall sensors installed and seem to be fairly regular in preliminary investigation.
IMG_20211012_012913689_HDR.jpg
It's big compared to an 8080 motor
IMG_20211012_012757494_HDR.jpg
But only twice the weight at 2375g (3d print weighs nearly 100). 8080 motor weighs 1156g. The other 8080i have is about 1250g.
IMG_20211012_003702317.jpg
VESC finds 10mohm and 7uH inductance... My MESC finds slightly lower but basically comparable. Ant measured 19mohm phase to phase which is meant to be double the single phase measurement vesc does.
Edit: new 5.03 VESC firmware finds nearer 20uH. Substantial difference. New VESC firmware far far more reliable and generally better so I'll have to investigate this.
IMG-20211012-WA0005.jpeg
The back emf waveform has quite a lot of 5th (edit 11th?) distortion, which makes it a bit hard to drive but VESC does ok in unloaded bench tests. MESC also runs it ok but struggles to control the d axis current.
IMG_20211011_171024864.jpg

Windings are fairly tidy and the shaft is a custom I drew and Arvin had made for me. It looks to have been hardened. No significant quality issues.

Free running power is 63W at 1300rpm (24V)and about 180W at 2500rpm (45V).The 8080 motor is about 1/3 of this. Strangely, the heat seems to be being generated in the rotor, which seems to get hotter than the windings and core. This is very provisional, I've not run it for long. This might be an artifact of it being quite tricky to control and inducing high frequency components in the rotor.

The biggest downside is the noise. It's got a fairly loud whine that's present in BLDC and FOC mode. Not sure what's causing this yet, the cogging torque is low.

14PP (28 magnets) 118mm diameter.

I requested the 8 turn 62kV option. The kV has come out nearer 57... Close enough.

The fixing screws are annoyingly m4 not m5 like every other RC motor in the world uses.

Overall looks promising. Waiting for the mounting bracket.
 
Today I put a drill on it and used the FFT funciton of my oscilloscope.

Seems like the predominant harmonic is 11th (11x fundamental frequency) which is weird.
IMG_20211016_182550908_HDR.jpg
The implication of this I think is that there will be high frequencies generated, which are likely to result in the Eddie currents in the rotor. The fundamental frequency will travel with the rotor synchronously, whereas higher harmonics will not.

I wonder if there is a big win to be had by reconfiguring this to star topology...

The iron losses should be roughly... say 0.8kg stator at 2500rpm (46V 55kV as measured) is about 590Hz
Core loss of 12070 motor guestimate.PNG
So should only be looking at about 50W from core loss. I know this is the wrong steel, they say they use Kawasaki 0.2mm, but I can;t imagine it's orders of magnitude off.

Loss to IR should be virtually nothing since it is 10mohm phase... it spins to max speed with VESC saying 5A so that's I^2*R = 0.25W. Practically nothing. There is clearly something going awry here. I mean... It is perfectly useable already, once running at several kW a loss of 200W or so is meh, but perhaps it could be a lot better for very little effort.

I shall ask Arvin if
1) It is easy enough to unwind the wires and make it star and
2) The extent to which I void any warranty I may or may not have...
 
Measured the laminations today.

No amount of my memory that they said 0.2mm will change the reality that they are 0.35mm.

IMG-20211019-WA0011.jpg

13.5 laminations in 4.98mm. so 0.368... I guess with some photo aberration error that's probably 0.35.. lams.

Will update this with a comparative snapshot of the losses expected between 0.35 and 0.2... it's not disastrous but it is a bit disappointing.

A comparative loss diagram for 0.35mm steel (recognising that this might not be the steel used - they say Kawasaki, this is from BaoSteel but I can't find a datasheet for Kawasaki)
0.35mm lamination loss.PNG

Assuming 1.5T field (guess) and 1000Hz (=4285rpm) the loss per kg is ~150W instead of ~100W. Hardly a disaster; not going to stop my ebike running and does not account for the 180W free running consumption at 2500RPM, but for this grade and price of motor I think 0.2mm laminations would have been appropriate.

I can't reconcile these datasheet numbers with the often cited in literature rule that losses are proportional to t^2, which would give (0.35/0.2)^2 = 3.06x the losses.
 
Hi Mxlemming,
Thanks for the pictures and write-up.

Looks really nice in quite a lot of ways, and then not so nice in others.

The whine.... maybe because of the fifth harmonic distortion in the back-emf? That's a bit of a nasty surprise. I mean, I expected it to be somewhat noisy, but I assume it's much worse than the 8080 you're comparing it to (and by extension the 80100 i'm considering)

The free running power is a bit big. Meaning, will not be efficient at low power (as in street-legal mode). Which is a b bit of a bummer. But then at 19mOhm phase to phase for a 60kV motor, it's definitely not designed for low power :D
The neumotor 8053 has bigger phase to phase resistance for a 75kV, and I considered that one "omg".

Btw, the number I got quoted for this motor by Freerchobby at 10V free running is 4.6A, so 46W at ~800 RPM. So more or less in line with what you got.

Ow and interesting tidbit: Freerchobby also told me that their customers usually estimate / measure a lower kV than they do. For example they measure the kV of their 80kV 80100 at 92! But they sell it as 80, because that's what the clients "measure". Kinda matches your experience with the expected vs measured kV.

But what surprised me the most:
mxlemming said:
Strangely, the heat seems to be being generated in the rotor, which seems to get hotter than the windings and core.
This. Haven't heard the rotor as a major source of losses yet :).

Thanks again for the detailed writeup.
Would like some comments on how it compares with an 80100 if you have one around (no load losses, noise, phase to phase resistance and inductance, etc)
 
badgineer said:
Hi Mxlemming,
Thanks for the pictures and write-up.

Looks really nice in quite a lot of ways, and then not so nice in others.

The whine.... maybe because of the fifth harmonic distortion in the back-emf? That's a bit of a nasty surprise. I mean, I expected it to be somewhat noisy, but I assume it's much worse than the 8080 you're comparing it to (and by extension the 80100 i'm considering)

The free running power is a bit big. Meaning, will not be efficient at low power (as in street-legal mode). Which is a b bit of a bummer. But then at 19mOhm phase to phase for a 60kV motor, it's definitely not designed for low power :D
The neumotor 8053 has bigger phase to phase resistance for a 75kV, and I considered that one "omg".

Btw, the number I got quoted for this motor by Freerchobby at 10V free running is 4.6A, so 46W at ~800 RPM. So more or less in line with what you got.

Ow and interesting tidbit: Freerchobby also told me that their customers usually estimate / measure a lower kV than they do. For example they measure the kV of their 80kV 80100 at 92! But they sell it as 80, because that's what the clients "measure". Kinda matches your experience with the expected vs measured kV.

But what surprised me the most:
mxlemming said:
Strangely, the heat seems to be being generated in the rotor, which seems to get hotter than the windings and core.
This. Haven't heard the rotor as a major source of losses yet :).

Thanks again for the detailed writeup.
Would like some comments on how it compares with an 80100 if you have one around (no load losses, noise, phase to phase resistance and inductance, etc)

Added some more data above, including the harmonic orders.

Worth considering when comparing that you could run the 12070 at half the speed and get the same torque as the 80100 geared down by 2; the torque scales with the square of air gap diameter which is 106mm vs 70mm so 106^2/70^2 = 2.29x as much torque. That would go a long way towards making up the difference, and would give you a massive top speed and overall ability boost.

From memory, at 2500rpm:
8080 is about 60W loss
Only really mechanical noise
70kV one is 45uH, 45mohm ish edit: 130uH Ld
50kV one is 62uH 60mohm ish edit: 220uH Ld

AT12070 is 10mohm phase, 6uH phase inductance.
Edit: this was VESC detection figures the real figures are 10mohm phase (double that phase to phase) and 13.5uHLd 24uHLq. The flux linkage on my 62kV version is 6.4mWb
Electrical screaming noise
about 190W loss (kinda hard to get an exact reading, I'm just reading off my power supply display)

Still waiting for my machinist to supply me a mounting bracket; going to hassle him a bit ;)

Overall, I anticipate great things from this motor, but it is going to take quite some taming.
 
Hi Mxlemming.

Thanks for the info. Quite some valuable nice details

mxlemming said:
Worth considering when comparing that you could run the 12070 at half the speed and get the same torque as the 80100 geared down by 2; the torque scales with the square of air gap diameter which is 106mm vs 70mm so 106^2/70^2 = 2.29x as much torque. That would go a long way towards making up the difference, and would give you a massive top speed and overall ability boost.

Don't understand where you're going with this. I mean, At least I think of it this way: a say 100kV motor, be it the 80100 or the 12070 will give me the same 0.1 Nm for each Amp I feed it.
The difference will be the phase resistance thus the copper losses. A 80100 will melt before the 12070 even got warm.
So I can gear the 12070 2x slower, and thus halving eddy current losses but then...
1) I could do it as you said with the same kV, get the much higher top speed, but then I need to feed it twice the amps for the same wheel torque, which has implications on controller.
2) Or I could double the kV for the same top speed, same torque at same amps, and probably still get significantly lower phase resistance.

I'd have to run some simulations get a feeling of what the difference will be, (the grin motor simulator with its custom motor feature is an endless joy for me to play with). But I have already a feeling that the 12070 will be better in the end for anything except efficiency at low power.

But then I have some doubts that the bell to shaft physical mounting interface can handle a high torque for a long time. I mean, it seems to be designed to turn a propeller bolted to the bell, not the shaft. Hope I'm wrong though.

I really like the form factor of this. Also love the look, with the visible copper... And the low resistance is wow!
But it's slowly dawning on me that it's probably a not very good motor for an ebike that will be mostly driven in street-legal mode.
The noise you're describing seems a recipe for unwanted law enforcement attention, (also probably just plainly annoying), and high core losses will make it inefficient at street-legal power levels ... :(

mxlemming said:
they say they use Kawasaki 0.2mm, but I can;t imagine it's orders of magnitude off.
Hmm. Freerchobby told me their 12070 has .35 lams. Does this AT12070 really have .20 lams? That would mitigate some of the core losses I'm complaining about. But it would be kind of odd, since I expect all these companies just buy the same parts from the same big suppliers, and get more or less the same motor at the end.

mxlemming said:
The implication of this I think is that there will be high frequencies generated, which are likely to result in the Eddie currents in the rotor. The fundamental frequency will travel with the rotor synchronously, whereas higher harmonics will not.
I wonder if there is a big win to be had by reconfiguring this to star topology...
Do you mean that the harmonics will just circulate in a loop through the delta? and in star there's no closed loop for them to recirculate? Interesting, would be nice if you could shut the whine up this way. (but yeah, 1.732 lower kV).
Wait, you said rotor! not stator! Eddy currents in the rotor?
The basic shape of the back EMF is not dependent on the winding, but on the stator teeth and magnets shape, isn't it? In this case you'd still get the same harmonics in the back EMF regardless of winding delta/star/kV etc.... I think. And thus the same noise in the rotor... Just the recirculating currents would be gone with delta.

Something else what could happen with a delta reconfiguration. Just the 1.732 lower kV and greater inductance will probably make the thing easier for the controller to drive. That might improve things by itself.
 
badgineer said:
Hi Mxlemming.

Thanks for the info. Quite some valuable nice details
.....

Take what I'm saying at this stage as rather preliminary. I'll tidy it up as I get more experience with it.a lot of what I'm saying is speculation, hypothesis... Maybe one day I'll find out for sure, maybe not

Re. The torque, gearing etc. I found that beyond about 60A for a 70kV 8080 motor you really can't generate more torque. The iron saturates and it just gets hot really fast with no appreciable difference in force generated through the wheel.

With the bigger motor, with over twice the mass of magnet and stator, that 60A will become 150A or so. You can figure the torque difference. Controller amps are a bit expensive but not that expensive. You can easily push 150A for short periods with a Spintend Ubox for example and even that little flipsky is claiming 120A for 100$. You then get the same pick up at low speed from a lesser gear ratio, lower motor rpm and the losses become more equal for the same low speeds.

E.g. 8080 would ideally have about 8:1, but with 12070 you could run 4:1 and half the rpm. Or meet half way (since losses proportional to square or rpm iirc) and use a 6:1.i intend to run it at 15:93 or 24:148 (6.2 ratio). At 15mph, 26 inch wheel then gives 1250 rpm and the losses are quite acceptable at about 60W. But you've easily got the oomph to get 45mph or so at 3750rpm if you derestrict.

45mph isn't really feasible with an 8080. Not for long anyway. I've seen about 35mph on mine.

You have a much higher top speed, assuming you're gear ratio limited on the 12070.

At 1250rpm, you won't attract much attention. It's not that loud at lower speeds. I'm thinking it might even be the fan making the noise now, TBC.

Shaft:
I have no concerns about the torque capability of the shaft and bell.

Harmonics and rotor:
Yes, rotor. Higher harmonics will be present in the rotor as well as the stator since the higher harmonics are not synchronous with the magnetic field (the fundamental is).

In Delta they can circulate around the D unseen. In star, they are seen and can be quenched by the controller. These currents still generate magnetic fields that are seen by the rotor.
 
Hi Mxlemming.

mxlemming said:
I found that beyond about 60A for a 70kV 8080 motor you really can't generate more torque. The iron saturates and it just gets hot really fast with no appreciable difference in force generated through the wheel.

Oooh OK, understand now. This is where my "100kV is 100kV" falls apart. You get the same torque per amp.... until the core saturates, and then you don't. Haven't thought of that. The 80100 I'm eyeing will do some more torque until it saturates than the 8080, but yeah, 12070 will be again one big step ahead. Thanks for making me less ignorant, one small increments at a time :)

1:8 was the gearing I had in mind for the 80100.... just with a 20" wheel :D. And about the gearing. What drive chain / belt / etc are you planning to use?

About the lams though - I looked at the Ant Innovation website. They state the .2 lam thickness everywhere (for example they would have the only 80100 I know of with .2 lams). Everywhere except the 12070. So I'm really curious if the lams are really .2. If you ever open the motor to do the delta - wye thing, please check :)

Br,
 
badgineer said:
Hi Mxlemming.

mxlemming said:
I found that beyond about 60A for a 70kV 8080 motor you really can't generate more torque. The iron saturates and it just gets hot really fast with no appreciable difference in force generated through the wheel.

Oooh OK, understand now. This is where my "100kV is 100kV" falls apart. You get the same torque per amp.... until the core saturates, and then you don't. Haven't thought of that. The 80100 I'm eyeing will do some more torque until it saturates than the 8080, but yeah, 12070 will be again one big step ahead. Thanks for making me less ignorant, one small increments at a time :)

1:8 was the gearing I had in mind for the 80100.... just with a 20" wheel :D. And about the gearing. What drive chain / belt / etc are you planning to use?

About the lams though - I looked at the Ant Innovation website. They state the .2 lam thickness everywhere (for example they would have the only 80100 I know of with .2 lams). Everywhere except the 12070. So I'm really curious if the lams are really .2. If you ever open the motor to do the delta - wye thing, please check :)

Br,
Checked. See photo above. Disappointed slightly.

My setup (see Giant trance thread) uses HTD5M 20 tooth to 148 tooth at the moment. I have a pile of different belt brands, and they basically all behave similarly - the big problem is under breaking they do not like step up gear ratios.

I also have a 15:93 HTD8M 20mm wide belt, which is completely immune to skipping, but is louder and the belt is not rated for 15 tooth pulleys so I anticipate a reduced life.

For this motor, I am going to try a 24:148 ratio HTD5M with an Optibelt Omega 5MHP belt. If that is not capable, I have a 28 tooth pulley and I could always revert back to the HTD8M. The higher noise with this motor is probably not that significant.
 
mxlemming said:
Checked. See photo above. Disappointed slightly.

Oh boy. I expected it, given the no load power readings, but still, disappointing. The difference is negligible at a few kW, but at 25kmh (~220W) its the difference between 80% and 60% efficiency. The only way to lower the core losses then it is to gear the motor to a very low RPM, but then you lose most of the high power potential.

I'm curious if it's the same for the ant 80100, which the website explicitly states to have .2s. Pretty sure it's .35 also. I'll ask out of curiosity, but not getting my hopes up.

I looked at your giant trance thread. Love a few of your solutions, especially the pulley mounted to the brake disk. The side mounting of the motor does have some disadvantages, but the clear plus is the short belt! At that length the extra 5 teeth on the small pulley might make a difference in skipping teeth.

Thanks again for all the Info :)
 
Update re. Lamination thickness...

Arvin tells me there is a serious supply issue and the factory changed the steel without him knowing... Apparently Kawasaki stopped exporting 0.2 to China so it sounds like no one is getting a 0.2mm stator from any of these companies for the foreseeable future.

He's promised me a new stator for free as soon as 0.2 material is available again. TBC

I'll continue this thread with results from the motor as is, it's clearly useable...
 
badgineer said:
But what surprised me the most:
mxlemming said:
Strangely, the heat seems to be being generated in the rotor, which seems to get hotter than the windings and core.
This. Haven't heard the rotor as a major source of losses yet :)

I’ve had the same thing in revolt motors, whining noise and a rotor that gets hot after 30sec of noload use so it happens. I assume it’s eddy losses driven by the harmonics and a poor balance in magnet thickness/back iron/field strength. I planned to switch to segmented magnets at one time, but never got to it as i was scammed by the magnet seller (poor dimensions)
 
larsb said:
One thing to mention is also the revolt motors leak a lot of flux as the rotor backiron is too thin. I guess it’s the same with this motor as they seem dedicated to flight and weight savings?

Yeah, there is a lot of flux leakage; a screwdriver sticks strongly to the outside of the can. Substantially more strongly than the Alien 8080.

I would imagine that this doesn't result in too much extra heating (if any?), but still not optimal.
 
Stray flux = steel is saturated and it makes flux link outside of the proper magnetic circuit, it lowers efficiency a bit.

I guess more flux might also go circumferentially through the magnets so heats them more than when it goes straight to the iron..
 
The DIY ebike motor customer base may be too small for them to make any production changes, such as a heavier back-iron ring.

That being said, its impossible to hide the lamination thickness, so I believe them when they say the China factory changed it without notice. Its nice that they offered a new stator when they come back in stock with the 0.20mm lams.

I like the size and form factor for ebikes. Thanks for posting pics and details.
 
I’m thinking about installing this motor on my classic vespa for EV conversion.

This motor will rotate the wheel shaft directly without any gears.

Would it be feasible/powerful enough?
 
rikoari said:
I’m thinking about installing this motor on my classic vespa for EV conversion.

This motor will rotate the wheel shaft directly without any gears.

Would it be feasible/powerful enough?

how fast do you want your wheel to spin? check out the rpm of the motor...it's probably a tad high for a direct connection. very likely you need either a different motor (like a hubmotor) or some form of gearing between the motor and the wheel, even if it is just different sprocket sizes at each end of the connecting belt or chain.
 
amberwolf said:
rikoari said:
I’m thinking about installing this motor on my classic vespa for EV conversion.

This motor will rotate the wheel shaft directly without any gears.

Would it be feasible/powerful enough?

how fast do you want your wheel to spin? check out the rpm of the motor...it's probably a tad high for a direct connection. very likely you need either a different motor (like a hubmotor) or some form of gearing between the motor and the wheel, even if it is just different sprocket sizes at each end of the connecting belt or chain.

Amberwolf is exactly right here. It's definitely not the right motor for your application.
 
So with summer on the doorstep I've actually been using this motor in anger.

It's better than I'd expected from initial testing.

1) under load it actually seems to generate less heat than on the bench which is weird. Even running 100A and going WOT up a hill it barely gets warm.
2) the noise is the biggest irritation but again under load seems less bad. It's still bad though.
3) running higher frequency pwm is a must. It greatly reduces heating, running 15kHz (VESC 30, the VESC standard frequency) it gets quite warm but with 35kHz (equivalent of 70kHz VESC) it's definitely better.
4) it responds very nicely to HFI (the version I wrote).
5) another irritation that's getting me now is the screws that hold it. M4. I'm not even sure that's really a screw, real screws are at least M6. Technically they have a huge breaking force but the threads into the motor shell don't feel all that great because, well they're M4.
 
As soon as it’s that annoying it’s worth it to tap it to what you want :wink:

Not going into too many badgineer details but m4 could be a relevant sizing for proper tensioning of the bolts vs fatigue stress. I bet there’s a bolt tension calculator online to check it.
 
So conclusion after a few hundred miles of testing... This motor is pretty good. I've been running it at 140 phase amps, 75 battery amps at 20s. 62kV version. Off road. It pulls hard and smooth for a bike. With a 6:1 ratio to a 26"wheel I can comfortably ride up a 30° slope. At those settings it starts to get warm but not uncomfortably so.

The noise is still annoying but I'm getting used to it.

Responds well to FOC, I've dropped the pwm frequency from 35khz (equivalent to 70k VESC) to 20khz without issue.

Overall, it's good. I'm not getting toolman2 levels of performance from it, but it's absolutely a completely different beast to the 8080 motors.

The price tag is still a bit painful considering its qs138 money but what can you do...
 
larsb said:
I run the 80100 motors i’ve got at 250A phase for 120kV, isn’t 140A a bit too conservative :wink:

After all it’s a motor about twice the size really?

140 is fairly conservative. It seems to be the higher speed that causes the heating more than the current. I could try going higher... But I'll have to change so the connectors for 8mm bullets.

I guess your 80100 motor must be water cooled on a surf board or something. The ones i have would definitely melt with that much current, even scaled for kV.
 
Back
Top