Definitive Tests on the Heating and Cooling of Hub Motors

[justin_le]

I liked the idea of having your copper fill be the heat pipes! That is Bruce Lee wisdom. (Become the water...)

Well put! Given that some of the motors are running just like 3 turns per pole, it's not totally out of context for that to be an engineered copper tube rather than like a dozen parallel strands of solid wire, but it is well out of the reach of any prototyping by hackers. That said copper tubing with water cooling is used all the time as a conductor in giant electromagnets, so that's some kind of precedent.

Has anyone try to quantify how much heat is removed via the phase wires? And how much heating of the outside phase wires do people see? Is the temp rise noticeable at all?

From what I've seen the phase wires are definitely a source of heat much more than a path of conducting heat away. If you look at a motor from the outside with an IR camera, the wires going into the motor will usually be the hottest thing that you see. And if you look at my earlier calculations on the hypothetical conduction of heat out of the motor through a solid copper axle, then you'll know without any calculations that it's a pretty hopeless path to use the much much thinner diameter copper wire as an avenue for heat removal.

[justin_le]

With a few tests in the wind tunnel ...

Oh man, that was not easy, but after a few weeks of fussing around and false starts we finally have the process dialed in with this. The tunnel control panel has two CA3's, one for controlling the fan and wind speed via the stokemonkey motor drive which also monitors the wind tunnel ambient temperature. Another CA is monitoring the power flow into the test motor along with the motor core temperature via an embedded thermistor in the windings.

WindTunnelMXUS.jpg (123.84 KiB) Viewed 9086 times

Then on the backside of the test chamber we have a small hole trilled in the wall from which an IR camera is able to look at the motor side plate, with spot points measuring the inside, middle, and outer plate temps as well as an average area.

WindTunnel IR Camera.jpg (128.1 KiB) Viewed 9086 times

The motors are spun up using a Field Oriented Controller (in this case a BAC500+ with a massive heatsink) using substantial field weakening current in order to dump substantial heat in the windings without needing any mechanical load. The data from both CA3's and the IR camera is all streamed to a laptop while the tests are run at constant wind speed and power until the motor core and shell temperatures reach steady state.

WindTunnel Full Test.jpg (141.84 KiB) Viewed 9086 times

What you see is an example run with an MXUS V2 5Tturn hub motor. We averaged around 200 watts of pure heat into the motor, and then ran the wind tunnel at ~5, 10, 20, 30, 40, and 50 kph. While doing this we also adjusted the motor RPM to 10x the wind speed. So the hub was spun at 200rpm for the 20kph run, 300rpm for the 30km wind and so forth.

The IR camera was set to measure both spot and average temperatures, and confirmed that the temperature gradients along the side plate are very minimal. The inner spot was always a tiny amount warmer than the outer edges, and I would presume that's because the outer edges would have better contact with moving air for convective heat flow.

Example IR Camera File

IR_TIMELAPSE42601.jpg (34.15 KiB) Viewed 9086 times

[justin_le]

... And the results. Here is the data summarized for the V2 MXUS hub, with the motor going from 75 to 500 rpm while the wind speeds increased from 5 to 50 kph.

MXUS Wind Tunnel Conductivity Results.jpg (52.74 KiB) Viewed 9082 times

And graphically, MXUS shell to ambient conductivity:

MXUS Shell to Ambient.jpg (38.76 KiB) Viewed 9082 times

And the much lower Motor Core to Shell conductivity (note that the graph X axis should read motor RPM, not wind speed):

MXUS Core to Shell.jpg (39.05 KiB) Viewed 9082 times

What you'll notice is that at higher wind speeds, the conductivity from the shell to ambient air increased dramatically, almost 4x improved when traveling at 50kph versus a walking speed of 5kph. At 14.5 watts / degree, that means that the side covers with no additional fins or anything can dissipate 300 watts of heat while being just 20 degrees warmer than ambient.

Meanwhile, the conductivity from the stator core to the shell didn't even increase by 50% over the same equivalent speed range. So while very low stall speeds, the barrier for heat flow from the stator to the shell and the shell to ambient are somewhat comparable, at fast cruising speeds, almost all the resistance to heat flow is from the stator to the shell. This would suggest that if you are going to mount heatsinks to your motor shell, you are far better off putting them inside the side plate with fins facing inwards, rather than sticking them on the outside.

And it really validates the idea that a much tighter thermal coupling between the stator and side plates as illustrated here


Could go a long way towards increasing the continuous power handling capability and keeping the windings cool on a conventional DD hub.

[Punx0r]

Excellent Like they say, one good test is worth a thousand opinions.

The model for the concentric fin design looks great. Looks promising.

The hollow winding idea is an interesting one. On the face of it, the decreased copper fill should make for an all-round worse motor (unless the frequency is very high and skin effect comes into play), but thinking on it, at some point a hot solid winding will have equal resistance to a hollow cool one, so there may be an advantage.

[speedmd]

Would think the concentric/ opposing fins heat transfer would benefit greatly by a bit of waviness as opposed to being smooth and parallel so they can generate some turbulence. Even if they were made in smaller sections and angled slightly from one another.

The simple addition of a tiny amount of ferofluid just to fill the gap and thermally connect the two most massive elements in the motor looks to be a must try.

[icecube57]

My Mxus 4T is vented but the data that you showed mirrors my experience quite a bit. While riding at lower speeds and watching temps on the v3 I get no real decrease in temps other than the predictable slow decrease after a hard run which is almost similar to the motor being stationary cooling off. But as the speed increases and I cruise to a higher speed I see about a 20-30C degree difference between my Delta T during acceleration vs cruising at my top speed of 40-45 mph. This pushes the temps down to about 70-80-90C. Below the level I have set for my thermal throttling. So Im able to sustain peak power almost indefinitely If im able to keep my speed up in the higher rpms.But lugging around at lower speeds gives me no real mentional benefit. Its actualy a compound snowball effect. The temp just rises and plateaus or equalizes to a point to where the air can remove any more heat than what you are generating. Accelerating there is a temp spike and then it decreases below the low speed cruising value.

[justin_le]

Excellent Like they say, one good test is worth a thousand opinions.

Here's the same data presented a slightly different way, showing the total thermal resistance (rather than thermal conductivity) split up between the heat flow path from the Core to Shell, and then Shell to Ambient. You see that by 20kph, nearly 75% of the resistance to heat flow is coming from inside the motor, and at 50kph this reaches over 80%.

ThermalResistances.jpg (53.88 KiB) Viewed 8916 times

My Mxus 4T is vented but the data that you showed mirrors my experience quite a bit. While riding at lower speeds and watching temps on the v3 I get no real decrease in temps other than the predictable slow decrease after a hard run which is almost similar to the motor being stationary cooling off. But as the speed increases and I cruise to a higher speed..

Yeah, I think with a vented side cover motor the effect would be quite pronounced. Ultimately what this shows is that the effect of your forwards wind speed on motor cooling is much more significant than the effect of the motor spinning, and if this forwards wind speed can make its way into the motor core directly (via the vent holes) then that would be a big boost to heat extraction. My previous tests with drilled side covers were done just on the lab bench and not in the wind tunnel, so I'm really keen to see how that data looks with actual significant air velocity in the mix.

Would think the concentric/ opposing fins heat transfer would benefit greatly by a bit of waviness as opposed to being smooth and parallel so they can generate some turbulence. Even if they were made in smaller sections and angled slightly from one another.

For sure I was totally thinking that too, but for making a quick illustrative CAD model the smooth rings were quicker and easier. Again if anyone on this board has access to a CFD package that can model the air turbulence and convective heat transfers of different intermeshing spinning heatink fins, that'd be really invaluable.

The simple addition of a tiny amount of ferofluid just to fill the gap and thermally connect the two most massive elements in the motor looks to be a must try.

I've sent out a few RFQ's to ferrofluid companies to see if they have anything in their product lines that they think may be suitable, so hopefully we'll be able to get our hands on some. Having oil cooling stay exactly where it needs to without any needs for oil sealing the bearings and wire exits etc. would be golden, assuming that the resulting viscous drag isn't too bad.

[speedmd]

The simple addition of a tiny amount of ferofluid just to fill the gap and thermally connect the two most massive elements in the motor looks to be a must try.


I've sent out a few RFQ's to ferrofluid companies to see if they have anything in their product lines that they think may be suitable, so hopefully we'll be able to get our hands on some. Having oil cooling stay exactly where it needs to without any needs for oil sealing the bearings and wire exits etc. would be golden, assuming that the resulting viscous drag isn't too bad.

I hear that! Golden for certain. Your cutaway shows it very clearly what a potential benefit the straight path across the gap would bring to thermal loss. Wondering if this magic liquid can be made to be smart? Certain feed frequencies having it attach/ shear differently or made to get out of the way slightly with secondary controller data feeds at higher rpms. Have to dream a bit here.

[icecube57]

Looking forward to the vented results with any motor.

[litespeed]

This is one of the coolest (pun intended!) threads I have read on this forum. Well done Justin and looking forward to all your data going forward.

Excellent work!

Tom

[made_in_the_alps_legacy]

For sure I was totally thinking that too, but for making a quick illustrative CAD model the smooth rings were quicker and easier. Again if anyone on this board has access to a CFD package that can model the air turbulence and convective heat transfers of different intermeshing spinning heatink fins, that'd be really invaluable.

there is hardly such thing as a CFD tool to model convective heat transfer - CFD model calcultaing heat transfer rely on semi-empirical equation to assess convective heat transfer. (they calculate Ma numbers, Re numbers, and from those calculated numbers, they evaluate Nu numbers based on correlation) and I would suggest you to get a book summarizing convective heat transfer correlations such as "Idelchik - Handbook of Hydraulic Resistance": you will find in that book far more than pins and ribs correlation, which can be assessed at the 1D level.

EDIT : I think I found a downloadable version here
EDIT 2 :
various spread sheet to evaluate convective heat transfer
some more about "pin bank" enhanced convection

My question to you, are you planning to make some transient IR shot while the hub is heating up ?
because at the moment you seem focussed on steady state thermal state and there is probably more to see than homogenous temperature during the heating phase of the hub

[justin_le]

"Idelchik - Handbook of Hydraulic Resistance": you will find in that book far more than pins and ribs correlation, which can be assessed at the 1D level.

EDIT : I think I found a downloadable version here
EDIT 2 :
various spread sheet to evaluate convective heat transfer
some more about "pin bank" enhanced convection

Hey and thanks for the references.

My question to you, are you planning to make some transient IR shot while the hub is heating up ?
because at the moment you seem focussed on steady state thermal state and there is probably more to see than homogenous temperature during the heating phase of the hub

We had the IR camera set up to take a still shot every 60 seconds so I've got the full transient response too. But so far it's not all that interesting, just a steady and uniform change in temperature, and the slight thermal gradient from the insides of the side plate to the outside edge being slightly (<1oC) cooler is there all along. Here's an example of it transitioning to steady state after an increase in the wind speed from 20 to 30 kph

Time lapse of side cover cooling during transition from 20kph to 30kph wind tunnel speed

IR Transient, MXUS at 30kph.jpg (127.33 KiB) Viewed 8878 times

And here's another from the first test at 6kph, when the hub started off at room temperature and was heating up.

IR Transient, MXUS at 6kph.jpg (130.48 KiB) Viewed 8878 times

I guess it is a bit interesting that in these ones point 2 (middle spot) is slightly colder than 1 and 3, while in the higher wind speed tests the outer spot (3) was always coolest even though in principle we'd think that would be the hottest as it is closets to the windings.
I think if we add thermal coupling pins/fins on the inside then we'd hopefully see something of a hot band on the side plate where these are located if they are doing their job.

[made_in_the_alps_legacy]

View post history

[speedmd]

This is great work Justin. Looking at the photos and noting the spoke holes and just wondering what quantity of heat the cool running in the breeze spokes pull out of the hub. Depending on the gauge I am thinking it may be significant. I missed it if you already addressed the topic.

[Sancho's Horse]

Love the work! I settled on bolt on. No cover modification, just sandwiched between what comes factory. No direct contact between the fluid and the windings, but far easier/less mess, and impressive results. I still used transmission fluid though, now whether this was a real gain (especially considering time spent), or imagined is a good question.

Best of Luck!

[flathill]

The simple addition of a tiny amount of ferofluid just to fill the gap and thermally connect the two most massive elements in the motor looks to be a must try.


I've sent out a few RFQ's to ferrofluid companies to see if they have anything in their product lines that they think may be suitable, so hopefully we'll be able to get our hands on some. Having oil cooling stay exactly where it needs to without any needs for oil sealing the bearings and wire exits etc. would be golden, assuming that the resulting viscous drag isn't too bad.

I hear that! Golden for certain. Your cutaway shows it very clearly what a potential benefit the straight path across the gap would bring to thermal loss. Wondering if this magic liquid can be made to be smart? Certain feed frequencies having it attach/ shear differently or made to get out of the way slightly with secondary controller data feeds at higher rpms. Have to dream a bit here.

I would recommend you try https://ferrofluid.ferrotec.com/product ... dio/apgCD/

The name brand fluids are really more stable in the long term. Other fluids evaporate and turn to sludge over time, or the particles separate from the carrier fluid.

Ferrofliud works great with regards to improve cooling of sealed outrunner motors, given they have high quality magnets. Just watch heatsoak into the magnets at a stop after a high power run. The drag is noticeable but how many people pedal with large direct drive hub motors anyways. I did not notice any torque improvement with the two fluids I tried, but that was seat of the pants. Can't wait to see your dyno tests to get hard data.

[speedmd]

Great flathill that you went for it. I also look forward to the data/testing.

For the heat soaking issue on the rotor ring, I could see a simple addition of a thin base standard cooling extrusion formed and wrapped around the circumference between the spoke flanges as a big help. All you would need to do is cut through the fins every 5mm or so with a thin blade on a radial arm saw and it will bend/ form nicely and allow you to bond it directly or use safety wire style clamp and thermal grease. Tons of exposed area there.

[jk1]

awesome thread, interesting if you could compare the data with ferro fluid as well.

[justin_le]

So I decided it would be a good idea to first repeat the earlier tests (see viewtopic.php?p=721949#p721949) that we had done using drilled vent holes in the side plates now that we have the wind tunnel to vary the wind speed, as I'd expect that the effects of vent holes with 50+ kph wind speed would be quite a bit more pronounced than when I just had an external fan blowing on the motor.

Drilled one side on stand.jpg (67.44 KiB) Viewed 8669 times

We drilled out some pretty large 1.375" (~35mm) holes around the very perimeter of the side plate, with about as much material removal as we'd be comfortable with before worrying about inadequate side plate strength. We repeated the tests first with just one side plate drilled out, then with both side covers drilled, and finally with both covers drilled and the entire insides of the motor painted black.

Drilled and Black.jpg (78.11 KiB) Viewed 8669 times

For reference, here was the test results we had with the MXUS motor and no modifications. The motor RPM's were adjusted to 10x the speed, so 200rpm at 20kph, 300rpm at 30kph etc:

ThermalResistance Unmodded.jpg (44.85 KiB) Viewed 8669 times

Total resistance is 0.55 oC/watt at 11kph and drops down to 0.35 oC/watt at 50 kph.

With a single side plate drilled out, the resistance went from 0.48 oC/watt all the way down to 0.22 oC/watt at 50kph. So at speeds, there was about a 60% improvement in the heat flow out of the stator.

ThermalResistance 1 side drilled.jpg (46 KiB) Viewed 8669 times

With both side covers drilled out, the results are quite a bit more striking. 0.38 oC/watt at 11kph (close to the unmodded motor at 50kph), and just 0.154 oC/watt at 50kph. That's a 130% improvement in heat conduction at fast speeds:

ThermalResistance 2 sides drilled.jpg (46.32 KiB) Viewed 8669 times

The addition of black paint all of the motor insides had a measureable albeit slight effect. Final resistance at speed was 0.147, compared to 0.154 oC/watt before the painting.

ThermalResistance 2 sides drilled + black.jpg (48.44 KiB) Viewed 8669 times

In all the above graphs, the data is still split as though the heat was flowing from stator to shell, and then shell to ambient, when this doesn't really reflect what's going on with the drilled out side plates when there is direct ambient airflow on the windings. I left it like this for modeling sake but what really matters in the end is the total sum.

The other thing we noticed was that it took WAY longer for the drilled motors to reach steady state temperature. For the unmodded motor, after ~1 hour in the wind tunnel things seemed to be in pretty good thermal equilibrium, while it typically took ~3hours to get each data point in the drilled side cover case since the side plate temperatures would drift quite slowly.

awesome thread, interesting if you could compare the data with ferro fluid as well.

FerroFluid Jars.jpg (25.4 KiB) Viewed 8669 times

Very interesting data coming up next!

[Hyena]

Great data Justin! This confirms what myself and a number of others have been finding in the real world for a number of years now.
I ran my old style 9C (28mm stamped steel stator) aircooled at 40-60km/hr with 5kw peaks and it never skipped a beat in nearly 2 years of daily commuting. This data nicely explains why/how.

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 know you're largely looking at thermal conductivity but if you already have the data in excel could you possibly plot the winding temp over speed with the sealed and drilled motor on the one graph ? That would nicely summarize the benefits of air cooling and how it avoids cooking the windings at higher power levels

[icecube57]

The sealed motor is creating its own micro climate. I found the sealed motor took a while to shake the heat off and would essentialy stabilize but still drop over an extremely long period of time. While even slight disturbances such as a draft affects the drilled covers alot more so when the drilled motor is stationary litte currents of air were still removing heat and cooling better than the sealed. Justin do you think its possible to possibly introduce some smoke or something on a heated motor to see if the motor is generating any convection air currents that will show up visibly or through a FLIR.... What I mean is heating the motor up and letting it sit stationary to see if the smoke gets sucked in through the bottom holes and possibly exiting the top.

[icecube57]

Also can you grab any data on how long it takes for a sealed motor to cool and stabilize to the point where it makes no sense to wait because it will be like watching paint dry vs a vented in the same scenario. Basically I guess Time from Critical to Ambient. 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.

[Hyena]

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.

It's definitely quicker to keep rolling to cool the motor, and the above graphs show the faster you pedal the quicker you'll get you e-mojo back.
Prior to the days of the CAV3 and thermal roll back I fitted simple thermostats to the windings in my motors. It was somewhat crude but a pretty foolproof solution to stop noobs cooking motors. It was a 110oC thermostat that didn't reset until 95 degrees. When it clicked open it was game over and you either had to get off and twiddle your thumbs until it cooled down or start pedalling. If I was on a hill I opted for the former, if it was fairly flat I'd just pedal for a few minutes and then it'd kick back in shortly later (typically 5 minutes, and on a heavy bike it'll be the longest 5 minutes of your life ) With thermal rollback you get a bit of warning that you're headed to shitsville and you can back off a bit before you hit the city limits. But even 200w will allow you to keep rolling along and take a little load off, and then pedalling as fast as you can will reduce the motor heat as quickly as possible. I know with my old thermostat set up as soon as it cooled enough to kick back in I would get the bike back up to speed in a sensible fashion and then the temps would rapidly get back in order. Again, pretty much exactly what Justin has logged above.
Again, a real world example of the above data was a trip to work one morning when I was running late so pinned it full throttle all the way there. It must have been pretty hot but because I was doing 60-70km/hr continously it was able to shed the heat. That was until I hit a red light, and the 2 minutes sitting with no airflow pushed it over the edge.

[justin_le]

The sealed motor is creating its own micro climate. I found the sealed motor took a while to shake the heat off and would essentialy stabilize but still drop over an extremely long period of time. While even slight disturbances such as a draft affects the drilled covers alot more so when the drilled motor is stationary litte currents of air were still removing heat and cooling better than the sealed. Justin do you think its possible to possibly introduce some smoke or something on a heated motor to see if the motor is generating any convection air currents that will show up visibly or through a FLIR.... What I mean is heating the motor up and letting it sit stationary to see if the smoke gets sucked in through the bottom holes and possibly exiting the top.

Yes, that's totally been on the radar and I've been keeping my eye out for a smoke machine to show up on craigslist so that we can modify it and make a grid of smoke traces to see just what the airflow pattern is. I get such a trip watching videos like this:
https://www.youtube.com/watch?v=q_eMQvDoDWk

Also can you grab any data on how long it takes for a sealed motor to cool and stabilize to the point where it makes no sense to wait because it will be like watching paint dry vs a vented in the same scenario. Basically I guess Time from Critical to Ambient. 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.

Much better than that, I've got all the results fully modeled with each of the thermal conductivities fit with a curve of the form K = Ko + K1(s)^n, where "s" is either linear speed in m/s in the case of the shell to ambient temperature, or the rotational speed in rad/sec in the case of the core->shell thermal resistance, and a new thermal model based around this which is on the development site here:
http://www.ebikes.ca/trip-simulator-test
So you can just create any initial conditions and load/speed etc/ to see how the core and shell temperatures evolve. And then switch between the motor model that is unmodified, with holes, with ferrofluid etc.

TripSimulatorDropdown.jpg (75.35 KiB) Viewed 8730 times

Once it's complete I'll have a much more comprehensive explanation on the use of this tool to get the kind of info that you are asking about, but for now I'm still trying to get all the raw data collected.

[made_in_the_alps_legacy]

Much better than that, I've got all the results fully modeled with each of the thermal conductivities fit with a curve of the form K = Ko + K1(s)^n, where "s" is either linear speed in m/s in the case of the shell to ambient temperature, or the rotational speed in rad/sec in the case of the core->shell thermal resistance, and a new thermal model based around this which is on the development site here:.

very nice, makes sens to me knowing that convective heat transfer coefficient corrrelation usually take a "Re^n" term as driver, Re being the Reynolds number, mostly depending on speed at our scale, well done