Definitive Tests on the Heating and Cooling of Hub Motors

[fechter]

Great test!
With my bad hearing and the background noise in the video, I wasn't able to make out most of the narrative (bad hearing sucks).

Any idea how much you improved the continuous power rating?
What was being measured on the DVM?

After adding the FF, it appeared like there was a lot more heat coming from the shell right where the magnets are. I guess that's how it's supposed to work. Adding some fins right there would help.

Do you know what 'flavor' of FF you were using? There seems to be quite a variety out there. Guys using ATF for oil cooling haven't seemed to have much problem with the glue being affected. I wouldn't worry much about the varnish on the windings but the magnet glue could be suseptible. Most magnets are stuck with some kind of epoxy, so depending on the oil base used in the FF, there should be some literature indicating compatibility.

 

[nicobie]

This thread keeps getting better and better.

 

[liveforphysics]

Im using FF I got on amazon for ~$25 shipped IIRC.

I don't know its base. I don't know that it won't slowly harden or disolve the magnet glue or who knows.

~1min to achieve a internal thermistor reading of 1.3kOhm before FF.

>5min (more like 10min) to achieve 1.6kOhm (indicating lower temp) internal thermistor reading after adding FF.

That's pretty impressive effect for a few ml of anything.

 

[justin_le]

OK, great news is that we got our full bulk supply of FF's at long last, and the detailed experiments will resume.

Earlier I had posted about the potted motor core which we'd made intending to improve the copper->stator heat conductivity since we saw this as a potential bottleneck. However, one thing I had neglected to do was actually characterize the difference between the copper winding temperature and the steel stator temperature, and see how this temperature difference is affected by the ferrofluid cooling. To conclude that potting is effective we'd need to first know a) that there is indeed a large temperature gradient between the copper windings and the stator metal, and b) that this temperature gradient is greatly reduced in the potted core.

So we went back to the original MXUS 4505 motor that all previous tests were done on and reworked it to have additional temperature sensors. Here is the core with the original sensor, which was glued to the windings but also touching the aluminum part of the stator support:
View attachment 1

This was left in place for reference, but then we also added the following, with one sensor clearly embedded in a hole inside the cast aluminum stator support away from any copper, another wedged right in the thick of the copper windings, and a 3rd located under the loop of one of the end turns, touching both copper and also the fiberglass end plate that they put over the final stator lamination.
View attachment 2

This meant now repeating the the same tests in the windtunnel all over again which is a bit of a drag. So in order to make things more useful and interesting we also installed a mock-up rim and tire in front of the motor test in order to have a more realistic measure of how much the tire creates a wind shadow.


I've completed all the runs without ferrofluid yesterday, and am getting close to finished runs with the 5mL FF today. Expectation is that the FF run will should show a much greater copper->stator temperature difference if the heat path is now predominantly through conduction via the stator.

 

[flathill]

I don't expect much difference in the stator temp and average copper temp with or without FF

Another way to measure the total average winding temperature is to setup a relay box on one phase winding, and switch in a constant current source when you want to sample temperature. You can use the temp coefficient of copper but the more accurate way is to use a water bath to do a two point calibration. Just put the stator in the bath and measure the voltage drop across one phase with 1A of current running thru it at room temp, and then again at elevated temp temp water bath. Downside of this technique is you cannot use it while running but it is extremely accurate. You may not need this accuracy depending on the temp gradient in the windings

I have been trying to find a way to use a similar technique to sample winding temp with controller with no temp sensors. That way we can run our motors with no hall sensors and no temp sensors in the future. Saves cost but more importantly it is ultra robust. True sensorless controller

 

[fechter]

This is like Mythbusters for bike motors. :wink:

 

[spinningmagnets]

From Fechter:

After adding the FF, it appeared like there was a lot more heat coming from the shell right where the magnets are. I guess that's how it's supposed to work. Adding some fins right there would help

BINGO!!

Fast forward to 9:57 on the video to see the FLIR image of the magnets being hotter than the sideplates.

Here's an idea...a finned heat sink inbetween two sections of aluminum angle. The "chimney" would create flow, and the airflow through these would remove heat faster. Several dozen of these epoxied to the steel backing of the magnets, maybe with some type of banding to add more holding to them?



An aluminum or 3D printed plastic disc connected to one side (or both) with radial fins to create flow, like a ventilated disc brake from a car...

 

[sendler2112]

FerroFluid wiki
.
https://en.wikipedia.org/wiki/Ferrofluid
.
Ferrofluid has been used for decades to cool loudspeaker coils.

 

[justin_le]

I don't expect much difference in the stator temp and average copper temp with or without FF

And this is why we do experiments! Here's a quick summary of some of the data from yesterday's and today's collection.



The temperature sensors are the motor as shown in this this image which I posted earlier, the "copper loop" is the thermistor on the right (touching both copper and core), while the "copper junction" is the one in the middle.


So without any FF in the core we had a 4.5-5 oC temperature differential between the stator and copper wire, and with the FF this increased slightly to a 5.5-6°C difference. It's not nearly as much as I was expecting. I suppose this would be a larger difference at say 300 watts of heat dissipation, like 10°C. But these results clearly put an upper limit to how much of an effect potting the stator core is going to have in the grand scheme of things. It looks like it will be at best a small incremental improvement rather than another large boost.

As for having a tire in front of the wind flow versus running just the bare hub into the air stream, well that had almost zero effect on the results. Here you can see the original test data that I posted some time ago in blue diamonds, and the green squares are the 3 tests I just completed with the tire in front of the motor. So that's good news, it means that all the previous model parameters are still valid and the rim and tire don't cause a measurable wind shadow effect downstream on the hub.
View attachment 1

 

[liveforphysics]

You rock brother, thank you for such well presented precise work.

 

[justin_le]

Here's an idea...a finned heat sink inbetween two sections of aluminum angle. The "chimney" would create flow, and the airflow through these would remove heat faster. Several dozen of these epoxied to the steel backing of the magnets, maybe with some type of banding to add more holding to them?

An aluminum or 3D printed plastic disc connected to one side (or both) with radial fins to create flow, like a ventilated disc brake from a car...

So one thing that I think almost everyone on ES needs to get out of their heads is the idea that harnessing the rotational motion of hub motors to generate air flow has ANY merit in the case of hub motors. This kind of thinking makes sense when you have a high speed motor running in still air, which is why we're familiar with seeing it and I think that's why it's permeated the thinking of so many of these improvised venting/cooling ideas. In a hub drive on a bike, you have the exact opposite, a slowly rotating motor in a very fast air stream. The speed of the air blowing over the hub as a result of your bike moving forwards is way greater than the relative rotational speed on the hub, and that's what you want to be harnessing. What you're proposing in that complicated design would have way less airflow over the finned heatsinks than just having the simple finned heatinks on the hub pointing out to the air, no angle plates or chimney structure or anything.

To put some numbers in the mix. A typical DD hub motor laced in a 26" wheel (33cm radius), spinning at 300rpm. Radius of shell just inside the stator where fins etc. can fit ~7.5 cm
Relative speed of the fins = 300rpm *2Pi/60 * 0.075 = 2.35 m/s = 8.5 kph.
vs.
Relative air speed from forwards motion = 300rpm *2Pi/60 * 0.33 = 10.4 m/s = 37.4 kph

So you have almost 5 times more air velocity at your disposal from the bike's forward motion, which would completely drown out any slow moving rotational air effects. My suggestion for people trying to maximize the air cooling of their vented motors is to pretend that the motor isn't even spinning at all and think about how you would design something that captures the forwards air velocity and directs that to pass over the windings and through the motor.

Justin, for high performance looks like venting is still the way to go.
Edit: Just saw that you already tested most of if not all hole placements for intake/out take.

This is why I think the whole idea of intake/outake is completely misconstrued. What you see spinning a motor at high RPM's in still air is largely irrelevant to what matters when you have the motor spinning at modest RPM's in very fast moving air, which is what we have in practice with an ebike. When I get a smoke machine setup to show the air stream patterns in the wind tunnel then I'll likely run a few illustrative experiments on this front.

 

[justin_le]

From Fechter:

After adding the FF, it appeared like there was a lot more heat coming from the shell right where the magnets are. I guess that's how it's supposed to work. Adding some fins right there would help

Fast forward to 9:57 on the video to see the FLIR image of the magnets being hotter than the sideplates.

To put some numbers into the mix, when I ran the recent tests on the MXUS hub without FF, the 3 IR temperature sensors on the side plates and right over the steel rotor were all within about 1 degree of each other, with the outside of the side plate being hottest, the steel rotor 2nd, and towards the middle of the side plate coolest, but as I said all within 1 degree C, consistent with what I saw with the thermal camera. Pretty much uniform temp.

WITH the FF in the mix, the steel rotor area was consistently about 4 deg. C hotter than the side plates. This is with ~180 watts of power. So it's a noticeable difference for sure, although the bright white autoscaling of the FLIR camera in luke's video definitely exaggerates the magnitude of it. It is enough to make me want to test using thermal grease rather than a conventional silicone sealant under the side plates before screwing them on.

 

[flathill]

awesome

definitive data makes all talk just talk

you can apply science/engineering to many aspects of your life

now we know what to concentrate on

you are a real G

 

[macribs]

My suggestion for people trying to maximize the air cooling of their vented motors is to pretend that the motor isn't even spinning at all and think about how you would design something that captures the forwards air velocity and directs that to pass over the windings and through the motor.

I think Zero MC did this a few year back. They had a hose pointing to the motor, IIRC the hose was somewhat restricted to create a "venturi effect", where the air flowing out was faster then intake. Take this with a grain of salt, I am free basing this from the top of an old brain.

Kawasaki tried the same thing with their air flow through the frame creating "blower" effect at high speed, to try to increase the amount of air entering the intake.

Maybe a scoop like what we see on some radiators would work, air from scoop going into restricted hose then blow directly and at high speed across the motor?

 

[litespeed]

Justin,

Have you ever checked or tried the fans in the stator sucking in cool air and blowing them across or through the air gap? Holes only on one side cover drawing fresh air in towards the axle and blowing in out at the magnet gap. Kind of a "C" air flow. I wonder if this is quantifiable in your test box. I have noticed a significant amount of cooling as has CowardlyDuck.

Tom

 

[John in CR]

So one thing that I think almost everyone on ES needs to get out of their heads is the idea that harnessing the rotational motion of hub motors to generate air flow has ANY merit in the case of hub motors. This kind of thinking makes sense when you have a high speed motor running in still air, which is why we're familiar with seeing it and I think that's why it's permeated the thinking of so many of these improvised venting/cooling ideas. In a hub drive on a bike, you have the exact opposite, a slowly rotating motor in a very fast air stream. The speed of the air blowing over the hub as a result of your bike moving forwards is way greater than the relative rotational speed on the hub, and that's what you want to be harnessing. What you're proposing in that complicated design would have way less airflow over the finned heatsinks than just having the simple finned heatinks on the hub pointing out to the air, no angle plates or chimney structure or anything.

To put some numbers in the mix. A typical DD hub motor laced in a 26" wheel (33cm radius), spinning at 300rpm. Radius of shell just inside the stator where fins etc. can fit ~7.5 cm
Relative speed of the fins = 300rpm *2Pi/60 * 0.075 = 2.35 m/s = 8.5 kph.
vs.
Relative air speed from forwards motion = 300rpm *2Pi/60 * 0.33 = 10.4 m/s = 37.4 kph

So you have almost 5 times more air velocity at your disposal from the bike's forward motion, which would completely drown out any slow moving rotational air effects. My suggestion for people trying to maximize the air cooling of their vented motors is to pretend that the motor isn't even spinning at all and think about how you would design something that captures the forwards air velocity and directs that to pass over the windings and through the motor.

Justin, for high performance looks like venting is still the way to go.
Edit: Just saw that you already tested most of if not all hole placements for intake/out take.

This is why I think the whole idea of intake/outake is completely misconstrued. What you see spinning a motor at high RPM's in still air is largely irrelevant to what matters when you have the motor spinning at modest RPM's in very fast moving air, which is what we have in practice with an ebike. When I get a smoke machine setup to show the air stream patterns in the wind tunnel then I'll likely run a few illustrative experiments on this front.

Justin,

I agree that straight fins between the spoke flanges would almost certainly be better than the complexity of the chimney thing, and that interior structure is difficult to create much through-flow. You proved earlier in the thread that interior straight radial blades perpendicular to the side covers are ineffective at creating flow or turbulence at the hot parts.

Yes, the motor moves through the air differently than spinning while stationary, but it would be foolish to ignore how it rotates as the wheel rolls across the ground when trying to create flow, because the rotation of the covers around the axle could stuff up the flow you're trying to create. I've settled on a simple implementation with external blades that creates a centrifugal fan that keeps my stator temps low enough that heat is a non-issue even running at 27kw peak input. I take advantage of both the air flow from forward motion as well how the blades move through that flow, since during part of the rotation the blade tips see air speed greater than bike speed. The blades are on only one side and extend slightly beyond the motor's perimeter, so it draws air both through the motor, much of which flows through the air gap, as well as across the outside of the shell. The exterior flow was a surprise that showed up using a dense smoke machine.

I've been through a lot of iterations of ventilating hubbies, and while each had some benefit, this is the only one that showed significant merit without going to noisy active blowers. It may work better because it's motion, but like any centrifugal fan it relies on the rotational aspect to work at all, and the blades don't seem lose effectiveness moving in repeating arcs relative to the outside world.

John

 

[Offroader]

I'd like to hear more about what internal fan blades actually do in a sealed motor. The way I see it without blowing fresh air in and hot air out you are only blowing hot air around inside the motor. I bet the air inside the motor quickly becomes the same temp as inside the stator because of the time it takes for heat to be pushed outside the motor case, this would be the limiting factor.

I am sure it helps a little bit but nothing significant.

Having internal fan blades with properly placed vent holes would make a bigger difference.

 

[John in CR]

I'd like to hear more about what internal fan blades actually do in a sealed motor. The way I see it without blowing fresh air in and hot air out you are only blowing hot air around inside the motor. I bet the air inside the motor quickly becomes the same temp as inside the stator because of the time it takes for heat to be pushed outside the motor case, this would be the limiting factor.

I am sure it helps a little bit but nothing significant.

Having internal fan blades with properly placed vent holes would make a bigger difference.

With a sealed motor you're right, not much. With a vented motor it's not that much either. For that matter commonly used venting doesn't do much more than help get rid of maybe a few hundred watts as Justin's early tests showed...that is until you get a meaningful flow of fresh air through the motor, which I've only been able to accomplish using external blades. To get to my no heat concerns bliss required a high efficiency motor, conservative controller tuning, AND faster than typical speeds and rpm to make the fresh air flow great enough to move the heat. Air can't carry much heat with it, so real flow is required.

 

[spinningmagnets]

Thanks for doing the calcs, Justin. Now that I see the numbers, it's not as effective as I thought it might be.

 

[Cowardlyduck]

Fresh air is definitely the key here, however it gets in/through the motor.
The difference is, centrifugal blade attachments rely on the motor spinning, and spinning at speed. Internal fans continue cooling at slow and near stall speeds where most of the heat is generated.

Justin,

Have you ever checked or tried the fans in the stator sucking in cool air and blowing them across or through the air gap? Holes only on one side cover drawing fresh air in towards the axle and blowing in out at the magnet gap. Kind of a "C" air flow. I wonder if this is quantifiable in your test box. I have noticed a significant amount of cooling as has CowardlyDuck.

Tom

Indeed!
Although I'm not doing a "C" motion air flow, I'm sucking in one side at the perimeter and out the other perimeter so air passed right over the windings twice.
I'm about to take it to the next level also.

12 Fans this time in a Leaf '1500W' motor I hope to keep cool at 3-4KW.
The thing on the left is my Adaptto Mini-E with a heat-sink and 3 fans.

Cheers

 

[markz]

What kind of CFM or those fans?
I saw some fans in an electronics store and I thought about your mod. Are you gunna cut open the side plates? or just have the fans enclosed?

 

[justin_le]

I'd like to hear more about what internal fan blades actually do in a sealed motor. The way I see it without blowing fresh air in and hot air out you are only blowing hot air around inside the motor. I bet the air inside the motor quickly becomes the same temp as inside the stator because of the time it takes for heat to be pushed outside the motor case, this would be the limiting factor.

The idea of internal fan blades is to help move the internal air around so that there is more convection present for moving heat from the stator to the shell. However, as has been remarked a few times on this thread (see here and here and here) a likely much more effective means to to achieve this isn't to think of blades for stiring/moving the air, but as heat sink fins to increase the surface area for convective transfer. This would apply both on the side plates and also on the stator itself, something we rarely see.

Whether you have internal fins/blades or not, the air temperature inside the motor will be sitting pretty much mid-way between the stator and shell temperatures. Once the heat is brought to the shell of the motor, the passing air flow from the vehicle moving forwards does a very good job of removing it. Fins would help even further, but this is rarely the limiting factor.

Having internal fan blades with properly placed vent holes would make a bigger difference.

Of course, but that's a difference scenario that involves opening up the motor to the outside world, allowing water, dirt, salt, etc. into the hub as well as the cooling air, and I've been trying to keep these approaches somewhat separate. The question then to ask here is that if you have vent holes to the outside world, do you really need fan blades to move this air through the hub or does the much faster moving air outside already do that for you? A perfect analogy, when you're driving in a car and open the windows to cool off, do you also hold up a fan inside the car to help coerce the outside air to come in through the window? No, the air comes in pretty well on its own. Especially if you open up windows on both sides of the vehicle.

But if you DID want to really create a draft inside, then you'd put out a large scoop of sorts out the window to seriously re-direct the passing air into the vehicle. That is what I think is most at play when people make these "external fan" type of devices, like the one that pendragon8000 discusses here:
https://endless-sphere.com/forums/viewtopic.php?f=2&t=48753
It would probably work just as well if the motor wasn't spinning at all and the hot hub is just moving forwards through the air, with the pocket design creating local areas of higher pressure that helps push this outside air into the hole in the hub.

 

[justin_le]

Justin,
Have you ever checked or tried the fans in the stator sucking in cool air and blowing them across or through the air gap? Holes only on one side cover drawing fresh air in towards the axle and blowing in out at the magnet gap. Kind of a "C" air flow. I wonder if this is quantifiable in your test box. I have noticed a significant amount of cooling as has CowardlyDuck.
Tom

I have not tried or experimented with this yet, and given all the small muffin fan failures that I've had to deal with over the years from failed computer power supplies and battery chargers, I'm pretty disinclined to want a stash of them inside my hub motor. Those things are delicate and not meant for abuse. But I also have no doubt that independently powered fans like this will work better than anything else as a cooling strategy at very low speeds. So I think what you and cowardlyduck are doing is awesome and quite viable for a modder/hot-rodder, but I'd hate to go there on a commercial product.

There's gonna be a certain vehicle speed where the effects of external moving air flow would start to dominate and become more significant than what the inside fans are contributing, and it would be interesting to find at what point that is. 10kph? 20kph? 30kph? It wouldn't be too hard to test, but when/if I do pursue more experiments with vented side covers the focus will be on passive techniques of directing external passing air flow through the hub, rather than using an active approach. In the past when I have held up an anemometer behind those types of fans I've been pretty surprised at how low the actual air velocity is that they expel, on the order of 10kph or less.

 

[sn0wchyld]

Justin,
Have you ever checked or tried the fans in the stator sucking in cool air and blowing them across or through the air gap? Holes only on one side cover drawing fresh air in towards the axle and blowing in out at the magnet gap. Kind of a "C" air flow. I wonder if this is quantifiable in your test box. I have noticed a significant amount of cooling as has CowardlyDuck.
Tom

I have not tried or experimented with this yet, and given all the small muffin fan failures that I've had to deal with over the years from failed computer power supplies and battery chargers, I'm pretty disinclined to want a stash of them inside my hub motor. Those things are delicate and not meant for abuse. But I also have no doubt that independently powered fans like this will work better than anything else as a cooling strategy at very low speeds. So I think what you and cowardlyduck are doing is awesome and quite viable for a modder/hot-rodder, but I'd hate to go there on a commercial product.

There's gonna be a certain vehicle speed where the effects of external moving air flow would start to dominate and become more significant than what the inside fans are contributing, and it would be interesting to find at what point that is. 10kph? 20kph? 30kph? It wouldn't be too hard to test, but when/if I do pursue more experiments with vented side covers the focus will be on passive techniques of directing external passing air flow through the hub, rather than using an active approach. In the past when I have held up an anemometer behind those types of fans I've been pretty surprised at how low the actual air velocity is that they expel, on the order of 10kph or less.

it'd be interesting to see if adding active cooling to a FF filled motor to assist with low speed cooling could bridge that gap... a good size blower pointing at the covers of the motor, perhaps with a side cover with higher sufrace area too. use speed and temp so it only kicks in at low speed and high temps...

 

[John in CR]

Once the heat is brought to the shell of the motor, the passing air flow from the vehicle moving forwards does a very good job of removing it. Fins would help even further, but this is rarely the limiting factor.

This is why I avoid the electrical analogy of heat transfer using terms like "resistance". You've demonstrated that the exterior flow of air over the shell does fine when transferring a small amount of heat, but you'd probably need to be at near stall conditions on a very hot day for the less than 200W of heat like in the tests to cause a failure. If hubmotor shells as designed were very good at removing heat, then they would never be hot, something we all know not to be the case. That final step of heat transfer (from the shell surface to the world) is always a limitation, and adding surface area to the shell and/or directing a greater and more turbulent flow of air at it help move the same amount of heat at lower temperature, or more heat at the same temperature.

For several years others have demonstrated the limitation of the stock shell's ability to transfer heat after creating a better heat pathway from the stator to the shell using oil. It's great that FF eliminates the mess and the relatively small amount of drag losses, and no doubt FF alone is all that's necessary near legal power limits. That's not where the interest lies. People want to unleash the potential of their motors, which will enable extreme power of current sized motors as well as lead to smaller lighter motors at moderate power.

I see low power solutions as being a small market compared to the untapped potential of what lies between current electric assist bicycles and full blown motorcycles.