Definitive Tests on the Heating and Cooling of Hub Motors

Best science based thread in ages.
 
My face in Grinning, and my mind is spinning. If anyone asked me right now what bicycle is the one that could replace a car, number one on my top-ten list would be the Edgerunner with stokemonkey. The 20-inch wheel is stronger than a 26, and it lets the weight of actual cargo sit lower. Dave Kaufmann has a Yuba that needs to tackle extra-steep hills, and he kept moving up in "mid mount" motor size to keep from overheating until he got to a de-spoked Cromotor. If we add ventilation holes and a 20-inch rear wheel, I'm sure a 406 would have been more than adequate.

Hats off to Justin and the team for not only doing this, but for publishing it.
 
justin_le said:
jk1 said:
awesome thread, interesting if you could compare the data with ferro fluid as well.

View attachment 6
:D :D Very interesting data coming up next!


We requested the lowest viscosity ferro-fluid that they could could offer which would still handle the temperatures and conditions in a hub motor, getting samples in the 60-70 cps range, which is like a really runny oil, since my main concern with fluids was the additional drag on the hub. This stuff is nowhere near as exciting as the youtube videos of ferrofluids, it's just loosely drawn to powerful magnets forming smooth clump without any neat spiking effects.

If we assume a 0.5mm air gap, then the estimated the volume requirement to fill the entire space was ~0.5mm*45mm*198mm*Pi = 14cc. So we milled a little "access window" to one of the MXUS side covers and squirted in the full 20mL sample jar. It definitely seemed to just wick right into the airgap space as hoped, and at the end when you turned the motor you could see a solid oil sheen everywhere bridging the magnets and stator.
Inserting Ferrofluid to MXUS.jpg

Our biggest concern was the effect the drag torque of the wheel, and when we ran the no load testing this confirmed that the presence of the full fill of ferrofluid caused a significant increase, in the case of this MXUS hub it more than doubled the drag forces on the wheel at speed:
FerroFluid Drag Effects.jpg

That result wasn't unexpected, but was a bit of a bummer that it was so pronounced. At most bicycle cruising speeds it's an additional ~0.5-1Nm of resistance. Under power this is of minimal consequence, and actually the net drag of this combo (MXUS V2 with 0.35mm lams + ferrofluid) is still less than that of other similar sized motors with 0.5mm lams, like the Crysatlyte Crown, with no oil.

Anyways we went ahead and ran the full thermal testing in the wind tunnel, and the results are shown here. Total thermal resistance at low speeds is better than the case of the drilled side covers, but at high speeds it's not as good:
20cc Ferrofluid Thermal Resistance.jpg
Specifically, at 10kph the double drilled side cover was 0.38 oC/watt while the ferrofluid is about 15% better at 0.33 oC/watt, however at 50 kph the drilled side plates trump with 0.15 oC/watt versus almost exactly 0.2 oC/watt for the ferrofluid.

Here is the full dataset for the ferrofluid test, showing the heat conductivity graphs (as opposed to thermal resistance)
Ferrofluid Thermal Test Results.jpg

What you'll notice is that the heat conductivity from the stator to the shell is almost perfectly constant with only minimal dependence on the motor RPM. It stays pretty well pegged at about 6.5 watts/degree over the course of the experiment, while in all the other tests there was a definite increase in conductive coefficient as a function of the motor RPM as we'd expect from internal air convection.

I'd hypothesize that the thermal conductivity of the ferrofluid from stator to shell is nearly perfect, and that this constant 6.5 watts/degree is simply the result of heat flow from the copper windings (where the thermistor is located) to the steel stator. There is no metal to metal contact between the windings and the iron, a paper liner covers the insides of each stator slot and the ends of the stator have a fiberglass plate on them, and so even though the coils and iron are nominally touching they aren't thermally coupled all that well, which fundamentally limits how well you can cool the motor from the stator directly. At the 40-50 kph speeds you can see that about 75% of the barrier to heat flow comes from this constant term.

If this is the case, then when we do the same test using sufficient oil fill not just to fill the air gap but also to wet and splash the copper windings, then I'd expect we'll see a much better performance in the copper->shell resistance. Similarly, it would suggest that if the stator was potted with a thermally conductive resin to link the coils to the steel, then we could get similar benefits without the mess. Suppose potting the windings decreases the copper->iron resistance by a factor of 3, then at 50kph the ferrofluid fill would have a net resistance of around 0.1 oC/watt. You could dump 500 watts of pure heat into the copper and only experience a 50oC increase in core temperatures.
 
Love the new trip simulator and the ferrofluid data thanks! What ferrofluid type did you end up using?

Here is a simple comparison with all setting default (30km/h)

10.039km trip with linear 1000m rise

temps at end of trip

motor_type core_temp(C) shell_temp(C)
4503 stock 70.5 34.3
4503 ferrofluid 62.4 37.3
4503 1 side drilled 63.0 31.8
4503 2 side drilled 57.3 31.5

Core temp is all we really care about assuming we are not overtemping mag

Let's see what happens with a slow climb @ 15 km/h

motor_type core_temp(C) shell_temp(C)
4503 stock 59.6 34.5
4503 ferrofluid 49.9 36.4

The slower the speed and the more you are lugging the bigger the advantage ferrofluid has over stock. Ferrofluid with a finned cover is optimal for everyday ebikes. Drilled covers are optimal for semi-hi performance apps. Drilled covers and fans for 2nd place. Liquid cooling for the win.
 
And here's where things got unexpectedly interesting. During the course of the wind tunnel testing with the ferrofluid (which involves about 10 hours of motor run time), a lot of the fluid managed to leak out of the motor. We put a nice acrylic window seal over the pour hole that we filled things from, so the leaking was actually from the side covers themselves which weren't sealed, just screwed down to the normal tightness that we'd screw down a side cover bolt. Here's the splatter in the test chamber when we were done, you can see the two rows of dots from leakage out the left of the right side plates:
Ferrofluid Splatter.jpg

I was a little concerned that we might have lost so much oil as to compromise the test results, but looking at the data it was hands down different than the unmodded motor right up to the last 50kph run. So I thought we might be able to estimate how much oil was lost by repeating the no-load drag torque. I was expecting that we'd see slightly less resistance than what we measured right after filling it with the fluid, but instead the resistance was almost identical to the drag test with no oil at all!
Ferrofluid Drag after being depleted.jpg

Judging from that, I would have assumed that there was no oil left bridging the stator and magnets in the air gap, yet the thermal resistance tests show that there was clearly still enough oil present to transfer the heat, just not so much as to cause any perceptible drag increase! If this is the case then it's excellent news on 2 fronts, since it means
a) You can use much less ferrofluid than required to fill the entire gap, which given the crazy price of the stuff (~$1.50 / mL from ferrotech) makes it way more economic, and
b) The small amount of fluid needed for good thermal transfer does NOT in the end increase rolling drag of the motor. If you've increased the drag, you've put in too much.

When we looked in the viewing window after the test, you couldn't see any fluid left right on the edges of the magnets and stator like before, but if you shone a flashlight in you could clearly see that there was still glissening sheen of oil fill further down in the gap.
Fluid from Window.jpg

I was still somewhat skeptical that the loss of this much fluid didn't effect the results, so we repeated the entire wind tunnel testing program a 2nd time. We also put down a paper liner in the chamber to see if there was any new leakage from the 2nd run and this time around only collected a few drops. Net result is pretty much just like the first ferrofluid test.
Limited Ferrofluid Thermal Resistance.jpg

Next up I'll want to do the reverse, rather than overfilling the air gap and letting it leak we'll keep adding say 1-2mL at a time and then measuring the thermal transfer coefficient, and find out at what point you get diminishing or no returns. We'll also do a test with conventional ATF oil fill which splashes on the copper windings. And finally we'll try to pot the stator in a thermally conductive resin and see how well that improves the ferrofluid approach by better coupling the windings to the iron stator. Expect all this to take a couple months.

In light of these results, the previous direction that I'd been excited about with nested concentric heat sink fins between the shell and stator support is now on the backburner, since it appears that ferrofluid can achieve even better ends at lower cost and weight and with no retooling of side cover or stator core castings required. In both cases of ferrofluid and concentric nested fins, the limiting bottleneck I feel is going to be the heat path from copper winds to stator iron, and I'm really hoping that filling all the winding air voids with a resin can result in much better results.
 
this is pretty much the same with loudspeaker drivers. you only add enough just the fill the gap. adding more will just reduce the efficiency (more drag on voice coil) while providing almost no additional benefit thermally.

also note adding too little will make it more likely to leak

and also make sure any glue/paint/whatever is ferrofluid compatible or it can cause fluid separation long term

this would mean the carrier fluid would leak out the motor while the particles remain in the gap. the ferrofluid likes to travel along crevices. u may want to add grooves to make sure it cant cross the road
 
Do yo think a thick stator fluid would have less leakage and running.. assuming at the cost of added drag. Or do you think that if the remaining amount is probably the perfect amount that would remain intact by magnetic forces without it slinging it every where. Any benefit with a vented setup.
 
icecube57 said:
And would it to be better or faster to keep rolling at a lower speed and load or just simply stop to aid in motor cooling.

Justin answered this about 10 pages back, but I wanted my own proof.

https://endless-sphere.com/forums/viewtopic.php?f=2&t=48753&hilit=john+bozi&start=350#p954675

Of course, keep moving will cool faster. But using any electric power will slow down the cooling. Just stopping and waiting will allow the motor to cool quicker than using 200w and pedalling.

(unless maybe you can pedal your 40kg bicycle to high speeds of course).
 
also note the viscosity lowers as the fluid warms up. there is another "loosening" effect that is not temp related (less particles clumping) after being "worked"

make sure to do the drag test on a "cold" motor as-found after a fresh fill and then repeat the test after the fluid has been worked/warmed
 
Excellent work :) The options you're exploring for sealed motors are particularly interesting because open-sided motors just aren't practical for many and almost certainly unsuitable for a commercial product.

FWIW, I don't think it's at all unreasonable to install the sidecovers with a bead of sealant to prevent leaks. It's a lot to ask the ferrofluid to resist the considerable centrifugal force in a spinning motor, it'd be nice if just refrained from escaping via the bearings or phase wires :)
 
flathill said:
this is pretty much the same with loudspeaker drivers. you only add enough just the fill the gap. adding more will just reduce the efficiency (more drag on voice coil) while providing almost no additional benefit thermally.

For sure, but what surprised me is that the fluid which remains in the motor is quite LOT less than required to fill the gap, and even with this fractional fill it still has the same thermal benefits as the full coverage. I wish I could quantify it easily but from peering in with a flashlight I'd estimate maybe like 20-30% of the gap is bridged by the fluid. In the case of a rotating motor I'm guessing that this gets distributed and pushed around enough to keep the thermal transfer coupling well over all areas.

also note adding too little will make it more likely to leak

Curious, what's the presumed reason for that? Just that there wouldn't be enough to bridge the air gap and be held in place by the complete flux path?

this would mean the carrier fluid would leak out the motor while the particles remain in the gap. the ferrofluid likes to travel along crevices. u may want to add grooves to make sure it cant cross the road
Interesting thought. Wish I had a high speed camera located inside the motor while it was spinning to see just what was going on with the stuff. We thought that the irregularity of the magnets and stator teeth could have resulting in a "pumping" action that would squirt the liquid out sideways, where the centriptal force encouraged it to seep out between the cracks in the side cover plate.

icecube57 said:
Do yo think a thick stator fluid would have less leakage and running.. assuming at the cost of added drag. Or do you think that if the remaining amount is probably the perfect amount that would remain intact by magnetic forces without it slinging it every where.

The latter is what I think. What we could do now is install a side cover plate with large holes around the perimeter to encourage every drop that breaks free from the air gap to get flung out, and then see what's left after spinning for a few days. My hunch is that not a whole lot more would come out than already did, but I could be wrong.



Any benefit with a vented setup.

It wouldn't hurt I suppose, but the benefits to be gained are a lot less than when you are starting with a sealed unvented motor. Once you have fresh ambient air flowing directly over the stator and motor windings, there isn't much need to move the heat from the stator over to the rotor shell and then to ambient.


Punx0r said:
Excellent work :) The options you're exploring for sealed motors are particularly interesting because open-sided motors just aren't practical for many and almost certainly unsuitable for a commercial product.
FWIW, I don't think it's at all unreasonable to install the sidecovers with a bead of sealant to prevent leaks. It's a lot to ask the ferrofluid to resist the considerable centrifugal force in a spinning motor, it'd be nice if just refrained from escaping via the bearings or phase wires :)

Yeah yeah, nobody should read the fact that oil escaped the side plates here as a downside. I deliberately didn't seal them because I was curious if the magnetic forces alone would keep it in check, but in any kind of production run they already put a sealant on the side plate covers to keep water out so this would all be par for the course.

In in regards to the bearings and cable exits, if this WAS a concern or issue (I doubt it will be), then it would be quite easy to install small magnetic traps that would keep any fluid down in this non rotating axle area from getting drawn outside. That's how the ferrofluid is used as shaft seals in hard drives and such.
 
Too little may cause the fluid to become aerosolized which may not be a problem if it "lands" back in place.

Maybe paint the inside white and look for signs of misting

Also filling the gaps between the magnet and teeth may help with long term fluid stability (less shearing) and reduce drag
 
flathill said:
also note the viscosity lowers as the fluid warms up. there is another "loosening" effect that is not temp related (less particles clumping) after being "worked"
make sure to do the drag test on a "cold" motor as-found after a fresh fill and then repeat the test after the fluid has been worked/warmed

Yeah, this is what we had planned to do. The first drag test after 20cc's was added was performed with a cold motor. We were going to repeat the test when it was hot, but as you know most of the fluid escaped so it wouldn't gave been a fair comparison. And then when we repeated the drag test with the fluid that remained, it was indistinguishable from having no fluid, and so not much point in doing that again at a higher temp.

Also filling the gaps between the magnet and teeth may help with long term fluid stability (less shearing)
That would be one nice consequence of potting the stator too, we would present a smooth circular OD for the fluid to move around over without all the stator teeth edges and gaps and such. The magnets in this MXUS case are already butted up and make a fairly smooth ID, but in my other motor project I've got a gap between the magnets and would certainly want that filled to be smooth.

and also make sure any glue/paint/whatever is ferrofluid compatible or it can cause fluid separation long term

Indeed it will a fair amount of burn in "field testing" to see if there aren't other consequences here. Are you speaking in this case from firsthand experience of having had the ferrofluids react with chemicals in this kind of environment or just precautionary? We wouldn't want it to react in any way with the magnet wire enamel or the adhesive holding magnets in place, but given how the intended application for the product is around magnets and wire I'd be surprised if they used compounds that were incompatible with coatings typically used on those.

Hyena said:
Those are some whopping big holes in the size covers! They are larger than I'd be comfortable running from a strength/reliability and foreign object entry point of view. I know this was largely to test the best case scenario for the experiment but do you have any thoughts on the biggest holes people should run ? Personally, and without any quantitative science to back it up, I think around 20mm is as large as I'd want to go.

I can't say that I do. You're right that I was aiming for "best case" in this experiment for air flow, which means having a smaller number of large holes, but don't have any firsthand experience running a drilled hub with this little metal remaining in an offroad situation. There's no motor torque transfer along the side plates so I think (similar to a spoked wheel) that the remaining metal could be pretty thin and still strong enough.
 
First hand experience with fluid separation causing a huge problems. Don't rush this to market without beta testing in the field. As you probably know ferrofluid is messy and stains. Still this is the optimal solution for passive cooling of a semi-sealed outrunner direct drive wheel motor since it runs at relatively low rpm.

I put some in an ultra motor (like on a2b) but it is press fit (no side plates) so no chance of leaking on the perimeter. I also added a sintered metal vent plug for filling and to equalize pressure. I still ended up burning up the motor so I don't know how well it holds up long term stability wise in a rotary motor application.

In stop and go travel, a ferrofliud motor might have an advantage over a vented motor (without a fan).

"In devices, ferrofluids come in contact with a wide variety of materials. It is necessary to ensure that ferrofluids are chemically compatible with these materials. The fluids may be exposed to hostile gases, such as in the semiconductor and laser industries; to liquid sprays in machine tool and aircraft industries; to lubricant vapors in the computer industry; and to various adhesives in the speaker industry. Furthermore, ferrofluids may be in contact with various types of plastics and plating materials. The surface morphology can also affect the behavior of the fluid. The selection of ferrofluid is carefully engineered to meet application requirements."
 
Love your work Justin!

Hey why don't you try making a side cover out of plexiglass or clear acrylic? Obviously not practical for riding on, but would be awesome for being able to see (and video) what's happening inside the hub motor's when spinning.

You could then try mixing some UV reactive substance with the ferofluid or ATF and with a UV light you would really be able to see what was going on. :)

Cheers
 
spinningmagnets said:
Dave Kaufmann has a Yuba that needs to tackle extra-steep hills, and he kept moving up in "mid mount" motor size to keep from overheating until he got to a de-spoked Cromotor. If we add ventilation holes and a 20-inch rear wheel, I'm sure a 406 would have been more than adequate.

Ha, that's a bit of a stretch, but if you're needing to use a cromotor as a mid-drive then you certainly have some serious hill issues! One of the things that I'll detail later and which will come up for sure with anyone playing with the updated trip-simulator is that the heat capacity of the motor still plays a huge role in short term hill climbs. If you run a really steep ascent of realistic length and compare the unmodified motor to one that is vented or with ferrofluid, you'll often see that the peak temperatures don't differ all that much. Often it's only like 5-10oC, even though the modded motors have less than half the thermal resistance. But those that are modded for better cooling will cool down much faster on the downhill or flat section following the hill, and will be better prepared for the next climb.

Hats off to Justin and the team for not only doing this, but for publishing it.

Moreover I should thank all our loyal customers who support us and the Grin shop which is really what makes this whole shared R&D thing possible. Without them there wouldn't be any funds for any of this.
 
icecube57 said:
Will the 4T Mxus be updated in the simulator?

Done! All the 4 and 6 turn MXUS variants are updated now too. Refresh your browser:
http://www.ebikes.ca/trip-simulator-test

Also I should point out that the geared eZee motors are listed too but the model is only considering copper and core motor losses at the moment, not any of the gearing transmissions losses, so it's showing higher than actual efficiency #'s.
 
Great work Justin!

Regarding the ferrofluid, you mention the drag, but did you see any decrease in Kv? That's what sparked my interest in ferrofluid added to a hubbie, to increase Kt without increasing copper resistance. If there's that benefit too with the very slight amount that you showed has little effect on drag torque it gets even more attractive.
 
John in CR said:
Great work Justin!

Regarding the ferrofluid, you mention the drag, but did you see any decrease in Kv?

It's shown right there in the subtitle of the graph data which summarizes the numerical results. 6.67 rpm/v without the ferrofluid, and 6.62 rpm/v with it. I suppose technically that's a decrease in Kv of 0.8%, but I can't really say if it's within the margin of measurement error or not. Pretty insignificant in any case. That said this particular fluid has very low magnetic attraction, the jar of fluid itself can just barely pick up and support a rare earth magnet. It's like its overall magnetic interaction is pretty low, just enough to keep it in the air gap, and I doubt the relative permeability is far from 1. I mentioned to the supplier that it would be in an environment with strong ~1T magnetic fields and wanted the least amount of drag friction in those fields so they produced or spec'd the compound we got accordingly.

That's what sparked my interest in ferrofluid added to a hubbie, to increase Kt without increasing copper resistance. If there's that benefit too with the very slight amount that you showed has little effect on drag torque it gets even more attractive.

I think to see this effect we'd need to spec a ferrofluid that has a high relative permeability and isn't saturated at the already quite strong air-gap fields in the hub motor, and I'm not sure if there is such a product that would help out much when the air gap field is already as strong as it is. IIRC in the research paper talking about the use of ferrofluids to decrease motor Kv's, they mentioned fields in the 50-100mT range with large air gaps.
 
Cowardlyduck said:
Not sure if you missed it, but what did you think about my idea for a clear side cover and UV additive above Justin?

Oh we've talked in the shop about machining an all clear plastic side cover for a while now, or easier installing some relatively large clear windows on a stock metal side cover. It would be hard to see much of anything going on when the motor is spinning without a high speed camera though, and for getting the thermal analysis the plastic side plates would skew the data. So until we have a suitable ~>1000 fps video recorder to see the fluid motion there isn't too much point in making the clear side plate. Another option is to use a stroboscope with a more conventional camera, but it's not an endeavor I could justify the time/money input yet.
 
Great work Justin. Now we know the stuff is worth looking at more deeply for certain.

With several different blends/ different volumes of liquid and long test times make this a perfect candidate for a design of experiment (DOE) to get to the pay dirt in much shorter order and have better feel for what is driving the results. You could also add in several cooling fin and venting options into the experiment at the same time. It will blow your mind! Some reading. https://controls.engin.umich.edu/wiki/index.php/Design_of_experiments_via_taguchi_methods:_orthogonal_arrays
http://www.isixsigma.com/tools-templates/design-of-experiments-doe/most-practical-doe-explained-free-template/

cheers!
 
speedmd said:
Great work Justin. Now we know the stuff is worth looking at more deeply for certain.

With several different blends/ different volumes of liquid and long test times make this a perfect candidate for a design of experiment (DOE) to get to the pay dirt in much shorter order and have better feel for what is driving the results. You could also add in several cooling fin and venting options into the experiment at the same time. It will blow your mind!

cheers!

1+
 
As I suspected and as Justin alluded to in the real world with up and down and stop and go is where the ferrofluid really shines. It even beats 2 sides drilled.

A fan blown on a sealed motor with ferrofluid should drop core temps dramatically.

Here are the results of going up and down the same hill 3 times

x,y (km traveled, elevation)
0,0
2,400
4,0
6,400
8,0
10,400
12,0

at 30 km/h (all other settings default)
core temp at end of travel (12km point)
4503 stock=134C
4503 2 holes=103C
4503 ferrofliud=92C (42C improvement over stock!)

at 15 km/h (all other settings default)
core temp at end of travel (12km point)
4503 stock=111C
4503 2 holes=88C
4503 ferrofliud=74C

at 15 km/h (all other settings default)
core temp at last hill crest (10km point)
4503 stock=142C
4503 2 holes=126C
4503 ferrofliud=103C

very exciting!
 
Back
Top