Definitive Tests on the Heating and Cooling of Hub Motors

[Arlo1]

BTY I got 3 year warranty with my 12v fans then I run 25v (fully charged) to them I just run 4 in series and hooked pack voltage to them

That's a good idea actually. I run 50v fully charged, so my 6 fans would run at about 8v each if I did that. That's a little slow, but still acceptable if I needed to.

Hook them into 3 sets of two so you have 3 in series then two strings in parallel. Then you get 18v each

 

[John Bozi]

Thanks cowardlyduck for answering all my questions,

it seems we have a lot in common. I do a variety of riding on my 4065 and on the very steep off road stuff is the main area that these hubs fail.

Bitumen is fine as I can hit the base of the hill hard and fast and conquer it and then let it cool, but the rough stuff with no run up is the killer challenge in many many many ways.

I just wonder whether it would be out of sync with all the other limitations, so not worth doing the mod unless finding solutions to all.

In the same vain of steep off road:

fan with air filter on controller as that heats up just as much as my hub does
traction and bike geometry to climb without front kicking up or sliding out

and even whether the fans could handle drops, jumps jolts etc. Not sure if they were designed for anything but sitting on a desktop at one particular angle with clean or relative dust in an interior. My hub only after a few months of riding up the mountain was caked in fine brown dust in almost every part possible. The hardest to blow out is between the magnets on the outer part of the motor. Must invest in a high pressure blower to do that after each ride....

I am interested to follow how tough those little fans are. The motor itself is a pain in the 234 to fix for halls or bearings etc but opening up for so many more moving parts.... mmm

Wonder if you could design your own motor from scratch designed around running fans..... because I have trouble just fitting my wiring through the axle as is...

 

[justin_le]

I was under an intuitive impression that light pedalling with a bit of power might increase the cooling of the motor more than just not using power at all. It actually takes 1 minute to cool from 90 to 80 degrees with no power used, or three times longer when using a few amps at pedalling speed.

Hi John, well nothing like collecting solid data to set things straight and either confirm or dispel ones conceptions. So thanks for taking the time to do this more quantitative experiment!

Going back to the original start of this thread, part of my goal was to establish and validate some baseline models for characterizing the heat generation inside a hub motor and properly assessing the effectiveness of different cooling strategies against this benchmark. Then stuff happened and it got put to the back burner for a while. However, the last few weeks we've had a bit of time to devote some resources to this topic again and have started the work to make an online motor temperature modeling software much like the ebike simulator.
View attachment 2
http://www.ebikes.ca/MotorTemp/index.html

On the top line, you can draw a profile of the motor phase current vs. time. Right click to create a node, drag it up and down to move it around, and double click to delete. Then on the graph below you'll see the resulting motor core temperature update in real time. What we are working on next is to have it so that rather than sketching the motor phase current on the top, you can actually draw or import a trip elevation profile, and then indicate your vehicle type and weight and then the program will compute the required motor torque (and hence phase amps) and then get its input that way, which I think would be more fun. You could draw a mountain and then see how hot different motors would get running to the top at various speeds.

But even in the present implementation there is a ton of neat stuff you can see. For instance, see what happens when you set a small wheel diameter and a fast speed (like 60 kph and a 16" wheel), the internal core losses become so high that even with NO power through the motor windings you can still get 80 degrees C in the motor core. The same 60kph speed in a 26" wheel and the core only reaches 50oC, since the iron losses are quite a bit less at this lower motor RPM ( 58 watts vs. 127 watts)



Right now it's just the nine continent 2807 motor that I have the parameters for since this is the hub that I did all the tests on earlier in this thread. The hope is that if the model parameters match closely to the physical units, then we wouldn't need to accurately characterize the thermal behavior of each motor type which would be a massive undertaking. Instead we could just compute the heat capacities from weighting the stator core and the motor shell and making assumptions about their material composition, and use some empirical rules of thumb related to the surface area for the thermal resistances from core to shell, and shell to ambient.

 

[GCinDC]

so glad you're back on this, justin!

looks like an awesome tool you're developing!

another huge incredible gift to the community!!

 

[John Bozi]

thanks Justin,

I tried your link but I don't seem to be able to get anything show up on the graph. Is there a submit button?

Also for newbs like me, can you maybe provide below the input and graph area, an explanation of how to measure the various motor parameters? Or what they exactly mean...? I doubt I probably even could.

In your example of 16" vs 26" at 60kmh, makes me glad my 20" custom job's average speed is 20kmh

I wonder how many people actually use their average speed as a gauge for what the most efficient motor/wheel diameter for their purposes would be, instead of just looking at the top speed they want to achieve occasionally...

 

[John in CR]

Great to see you back to this Justin. I've moved to exterior blades and one sided exhaust, and if you're inclined to give it a try let me know if you need any input. It can be added quite simply to any of your previous iterations of cover, and really turns the motor into a centrifugal fan. While higher rpm creates more flow, the smoke test I did should clearly that smoke was drawn into the intake side well below 200rpm. I run higher rpm but my hubbie has about the same surface area as a 9C, and I'm now pushing 27kw peak into it, and only the exterior blades cured the heat issues and they're nowhere near optimized.

Even though 127W of iron losses at less than 800rpm is abysmal with your 9C, it's still a fixed amount, and I'm pretty sure the reduced copper losses will be greater than the 70W increase in iron losses so don't write off that 16" quite so fast. That's before even considering the greater heat dissipation of higher rpm that you already proved in your tests way back when.

 

[Samd]

Great to see you back to this Justin. I've moved to exterior blades and one sided exhaust, and if you're inclined to give it a try let me know if you need any input. It can be added quite simply to any of your previous iterations of cover, and really turns the motor into a centrifugal fan. While higher rpm creates more flow, the smoke test I did should clearly that smoke was drawn into the intake side well below 200rpm. I run higher rpm but my hubbie has about the same surface area as a 9C, and I'm now pushing 27kw peak into it, and only the exterior blades cured the heat issues and they're nowhere near optimized.

Even though 127W of iron losses at less than 800rpm is abysmal with your 9C, it's still a fixed amount, and I'm pretty sure the reduced copper losses will be greater than the 70W increase in iron losses so don't write off that 16" quite so fast. That's before even considering the greater heat dissipation of higher rpm that you already proved in your tests way back when.

This is great John.

John Bozi, this is what I meant about not using small internal PC fans like Pendragon has, but letting the motor "be the fan" instead, internal fans would just be barriers to flow IMHO.

 

[justin_le]

thanks Justin,
I tried your link but I don't seem to be able to get anything show up on the graph. Is there a submit button?

Oops, I should have been a bit more clear. At first the graph will be blank, but then every time you right click somewhere in the current vs time plot on the top it will insert a graphing node, and so you can build up your motor input current profile that way, and once you have a node you can left click to drag it around to reshape the curve, or double click on it to delete. On my example curves showing the no-load plot, I had made a single line across the top graph at 0 amps with a node at the far right, so you can't really see it.



There is no "submit" button, the graph on the bottom showing the temperature plot of the motor core will always draw and update itself in realtime.

Also for newbs like me, can you maybe provide below the input and graph area, an explanation of how to measure the various motor parameters? Or what they exactly mean...? I doubt I probably even could.

Oh, this is a preview/development state, thorough documentation like you describe usually happens like 2 years afterwords. But in a pinch. K/W is the thermal resistance in degrees per watt, J/K is the heat capacity in joules per kelvin, Core loss is the combined eddy current and hysteresis inside the iron laminations at a given motor RPM (ie the cogging drag), and RWinding is the room temperature winding resistance of the motor phase. It's exactly the same as the motor model 2 in the very first post of this thread:

Where now:
R1 is K/W Core->Shell
R2 is K/W Shell->Ambient
Cstator is J/K Core
Cshell is J/K Shell

If you choose the motor from the drop-down list, then it will compute the core loss directly from the speed and wheel diameter, but if you enter a custom motor then you have to put this value in yourself for the motor RPM you are trying to model. Core loss at a give speed is super easy to measure by looking at your no-load power draw, running with the wheel off the ground.

I wonder how many people actually use their average speed as a gauge for what the most efficient motor/wheel diameter for their purposes would be, instead of just looking at the top speed they want to achieve occasionally...

That very much depends on the class of user. I used this particular example of a 16" wheel size because a lot of people are encouraged to run a smaller wheel diameter for better motor efficiency and higher motor RPMs, which is true in general, but not necessarily if you are doing high speeds at lower power levels where the additional core losses can become quite significant. On a normal ebike this wouldn't usually be the case, but if say someone was using a 9C hub as a mid-drive stokemonkey style motor and decided to have a nice gear reduction to the bike wheel, then their expected gains in efficiency might be mostly negated.

 

[justin_le]

Even though 127W of iron losses at less than 800rpm is abysmal with your 9C, it's still a fixed amount, and I'm pretty sure the reduced copper losses will be greater than the 70W increase in iron losses so don't write off that 16" quite so fast. That's before even considering the greater heat dissipation of higher rpm that you already proved in your tests way back when.

Yes this is all very true. On the current model, the thermal resistance parameters are all constants, although we hope to similarly make these parameters both functions of the RPM and wind speed after doing more systematic data collection, and that should take this factor into account. If the goal is to get 1000's of watts from the hub then for sure the gains or the higher RPM outweigh the losses, but in the example of someone wanting just like 500-800 watts or so then it's probably not worth it.

I'll let you know John when we start more actual dyno tests, it'll take a little while to setup the lab station for that but I'm looking forwards to populating a more complete list of motor models and provide options to simulate the different mod effects (holes, fan blades in and out, oil cooling etc.). The external blades seems like a neat idea.

-Justin

 

[John in CR]

Even though 127W of iron losses at less than 800rpm is abysmal with your 9C, it's still a fixed amount, and I'm pretty sure the reduced copper losses will be greater than the 70W increase in iron losses so don't write off that 16" quite so fast. That's before even considering the greater heat dissipation of higher rpm that you already proved in your tests way back when.

Yes this is all very true. On the current model, the thermal resistance parameters are all constants, although we hope to similarly make these parameters both functions of the RPM and wind speed after doing more systematic data collection, and that should take this factor into account. If the goal is to get 1000's of watts from the hub then for sure the gains or the higher RPM outweigh the losses, but in the example of someone wanting just like 500-800 watts or so then it's probably not worth it.

I'll let you know John when we start more actual dyno tests, it'll take a little while to setup the lab station for that but I'm looking forwards to populating a more complete list of motor models and provide options to simulate the different mod effects (holes, fan blades in and out, oil cooling etc.). The external blades seems like a neat idea.

-Justin

If someone is going to run their 9C in a 16" wheel at the 100V+ needed to get to that rpm, then 500-800W isn't even possible. The simulations of the 2 wheel sizes are interesting though. Take the 2807 and run it at the 104V needed with the 16" wheel, and then at the 64V needed with the 26" wheel, both at comfortable 20A, and it pretty much proves what I tell these guys all the time. That is how much better a hubmotor runs in a smaller wheel. Even with the extra iron losses you mentioned it still has better efficiency, and that's with higher power on tap too. Pedalists are so prejudice against small wheels, but if you want optimum performance with hubbies then small wheels rule, especially if you combine them with higher Kv windings that can handle more current and don't need impractical voltages to get speed. One of the best deal you ever had was the 9C2805 in 20" wheels selling on clearance at $110 when people didn't understand how much more powerful they were at our 100V max available at the time. Simply dodge the holes or stick to smooth roads and a 20" is a none issue other than it looks different. Off road I understand the need for larger wheels, but not on pavement. It can even look good...doesn't Miles bike run 20's?

 

[Cowardlyduck]

That's an awesome development Justin. It would be great if you could also model the internal fan mod like several of us have done now...although I realize it's a highly variable option.

John Bozi, this is what I meant about not using small internal PC fans like Pendragon has, but letting the motor "be the fan" instead, internal fans would just be barriers to flow IMHO.

Incorrect. http://endless-sphere.com/forums/viewtopic.php?f=30&t=56965&start=25#p946731

The fans are not barriers, but significantly cool in slow/stopped conditions. Adding blades to the motor can't do this as effectively, and only help where speeds are high enough.

Although one catch I've recently discovered is the dirt/dust build up over time is substantial, but this would also be the case with letting the motor 'be the fan'.

Cheers

 

[Samd]

Correct?
Incorrect comparison. I was not talking about it at stall. At run speed.

You can't seriously think I proposed the motor could have airflow through it if stopped.

 

[John Bozi]

Thanks again, I'll wait until more experienced people start posting screen shots of their parameters to get the data I'm not sure how to measure.

It will be interesting to see a cross over or overlay of the data from the various analyzers that Justin has created now. I am very interested in particular in the trip analyzer (one day I might feel geeky enough to purchase the logger and gps apps or whatever to run it.)

I see the temperature mapping there much more useful and real world solid data.

I know it would be a pain but having two people same weight riding the same bike with one parameter different would be the smoking gun in all discussions.

Hopefully this is not too much of a digression
For example, although not perfect:

recently I rode with a friend who had the same motor as me.

He had a stock 4065 26" unvented hookworms running bms 10ah 100v 80amp max
I had vented 4065 19" rim knobbies 10ah 88v 60amp max

By the end of the ride he had used almost twice the amount of battery and his motor and controller were almost twice the temperature.
Our maximum speeds were are 63kmh and average around 35kmh. I accelerated away much faster especially on hills.

I wish we both had the equipment to log our rides....

If we had enough trips populated and shared online these real world rides would give great information for people trying to decide on a set up.

ps. Justin, how the hell did you not kill your motor on the steep section portland to sherwood at 199C ? http://www.ebikes.ca/tools/trip-analyzer.html?trip=blog01

pps Another intuitive impression I have that defies "logic" is that many times using more power means less heat and possibly less used power to overcome obstacles.

When I have time I want to record this.

Start at a base of a mountain and WOT to top and then disc brake to base.
Start at the base of it and pedaling speed to top and then disc brake to base.

record power and temperature difference. And possibly allow a longer cool off period for the WOT since you did everything in a shorter period of time.

I am guessing that being in the right rpm, higher convection at speed, and the longer cooling off (waiting for your partner to catch up) will cool faster with less stress on the motor than staying at pedaling speeds.

 

[Cowardlyduck]

Incorrect comparison. I was not talking about it at stall. At run speed.

You can't seriously think I proposed the motor could have airflow through it if stopped.

Apologies. I didn't intend to offend.
The intended cooling for my usage is where slow, to near stall speeds are common. I've found fans to be one of the best options for this type of usage.

Cheers

 

[Emmett]

ATF.

Mineral oil is a bad choice for many reasons.

Would some low-vis transformer oil be the best thing to use? I assume it is certainly the best insulator. Or some vegetable oil which is real common like Peanut or Canola/Rapeseed Oil ? I would guess that some ATFs might contain additives which can conduct electricity, so they cannot be trusted.

When guys here write about "no load amp" increases, are they implying that the ATF is conducting electricity?

I suppose that controller damage could result if the fluid is too conductive and controller has no safety logic. Is that right?

Any insight most appreciated. Thanks in advance.

 

[izeman]

Would some low-vis transformer oil be the best thing to use?

also a good idea. but several test have shown that, as already stated several times, ATF is a very good fluid. it's very thin and is not conducting electricity. it can dissolve some plastics though. the heatshrinks used in my motor have been almost dissolved and useless. so take care!

When guys here write about "no load amp" increases, are they implying that the ATF is conducting electricity?

no. no load amps raises just because there is some resistance added as the windings have to move through the fluid. you can imagine yourself walking on land and trying to walk in a swimming pool

 

[Emmett]

Would some low-vis transformer oil be the best thing to use?

also a good idea. but several test have shown that, as already stated several times, ATF is a very good fluid. it's very thin and is not conducting electricity. it can dissolve some plastics though. the heatshrinks used in my motor have been almost dissolved and useless. so take care!

When guys here write about "no load amp" increases, are they implying that the ATF is conducting electricity?

no. no load amps raises just because there is some resistance added as the windings have to move through the fluid. you can imagine yourself walking on land and trying to walk in a swimming pool

Thanks for the quick reply.

In that case I can ignore no load amps increase. Since it's cooling (stator to outside) that I'm after, in low speed high torque riding conditions. Also the fluid drag is not a factor for me.

FYI, Just by subjectively playing with some ATF I have in the garage, versus some soy vege oil, the vege oil actually looks/pours/stirs as though it's lower vis. I know there are many types of ATF. In any case the vis difference of ATF vs vege (soy, sunflower or peanut) oil doesn't seem like much.

You've made me wonder if vege oil has multiple advantages over ATF:
- Vege oil is guaranteed to not have additives which might conduct, such as friction modifiers containing metals like zinc, tin etc.
- Vege oil is totally safe to the polymers and seals inside the hub
- Vege oil is biodegrable so safer for the env (and me and my tire) if it leaks out.
- Vege oil has no smell or a nice smell, but ATF simply stinks.

I don't see how I can trust ATF, when it's not a dielectric fluid which can safely go inside a transformer.

My third question is still out there ... if some fluid used in the hub (say 1/4 full) was slightly conductive, then what components would be damaged first? My guess is the controller, because more amps is drawn by a mild short circuit in the motor/fluid. Assuming the controller has no safety logic.

 

[Punx0r]

I don't see how I can trust ATF, when it's not a dielectric fluid which can safely go inside a transformer.

And vegetable oil is?

Veg oil is not lubricant, it is a foodstuff. Any lubricating or dielectric properties are coincidental and not controlled in the manufacturing process. I.e. various veg oil will be good to eat, but you may find mechanical properties are not consistent. You can get biodegradable hydraulic oils based on veg and canola oil and these will have controlled mechanical properties.

You thing you may want to do is put a pot of veg oil in the freezer. IIRC it turns into a stiff gel in cold weather. ATF does not.

The problem a conductive fluid in the motor would cause is electrolytic corrosion of exposed wiring. The conductance between phases might also cause the controller to enter a fault trip, if it's that smart. ATF is non-conductive and not hygroscopic. Verify the resistance with appropriate test equipment, if you wish.

You say veg is totally safe to polymers and seals. Do you know this, or are you guessing? ATF (and mineral oils generally) are very common in industry and used with various sealing polymers. A bearing seal, for instance, is much more likely to be rated as resistant to ATF than veg oil.

Use veg oil if you want and it fulfils your needs, but don't think it's mechanical, lubricating or electrical properties are superior to ATF.

 

[izeman]

water is conductive as well. even if this ATF is slightly conductive: it depends on voltage. and even 100v is not enough to be a problem.
vegetable oil has one major problem as well: it degrades. it is not meant to be outside a sealed container. it will smell bad quite soon - better or worse than ATF is up to your nose

 

[izeman]

... but don't think it's mechanical, lubricating or electrical properties are superior to ATF.

i guess we can neglect mechanical and lubricating properties here. he's not gonna lube his cam shaft or stuff under high pressure. it's only used to transfer heat. and as i said: everything but ATF and maybe transformer oil has been sorted out already as you say.

 

[Emmett]

And vegetable oil is? ...

I'm trying to learn here, and your responses help. As does you answer on the controller. So thanks.

Yes some veg oils are dielectric fluids. Search for "vegetable oils in transformers". Seems that canola, soy and sunflower are.

Castor oil is one of the best lubricants available for some motor racing engines. But for my hub motor, I don't need this fluid to lubricate anything. Just transfer heat, like is does in a transformer, and leave all wire insulation and seals alone.

Nothing ever freezes outside where I live and ride. Quite the opposite most of the time.

You might claim ATF is not conductive, but how many people here have tried the many different formulas of ATF in their hub motor? There are so many brews of ATF. Conductivity seems to be a "don't care" property in [edit: some] ATF spec sheets.

Actually, has anyone on ES used any type of ATF in a hub yet, or just speculated?

Something tells me that transformer oil, mineral or veg, has to be a better option than a randomly chosen ATF for getting heat out of an electric hub motor.

We know Inox M3 works. I can buy that in 4L bottles. It's about $20 per litre.

I suspect this oil cooling solution will leak at times, and cleaning up some sunflower oil doesn't seem too bad. Changing it once a month is not a problem either. It's harmless and about $2 per litre at my local shopping store.

 

[Arlo1]

I have 2 hub motors with ATF in it and I know others who have run ATF in other motors. There is others who are pretty big in the industry using it and I will let them post when they feel they want. It started with a company rewinding induction motors to use as traction motors there is a paper floating around from BGSU that might still be online you can read how they took a 10hp rated motor and made it do 100hp continuous and 225hp peak.
http://ivanbennett.com/forum/index.php?topic=66.msg388#msg388

 

[melodious]

Actually, has anyone on ES used any type of ATF in a hub yet, or just speculated?

ATF works and is proven (although, not as effective as flowing air around and through the hub stator itself). You just need to find ways to keep the fluid from leaking out, using stuff like gasket sealant. Otherwise it will play havok with your brakes.

 

[John in CR]

Actually, has anyone on ES used any type of ATF in a hub yet, or just speculated?

ATF works and is proven (although, not as effective as flowing air around and through the hub stator itself). You just need to find ways to keep the fluid from leaking out, using stuff like gasket sealant. Otherwise it will play havok with your brakes.

I favor ventilation and as opinionated as I am I still wouldn't make the statement in parenthesis. Oil cooling has the potential for far greater results than those using it obtain. Sure it transfers the heat from the stator to the shell much faster than air can, however, the benefit ends there. It needs to be transferred to the environment, and if the outside surface of the shell is unchanged, then so is the ultimate transfer of heat. ie If the outer shell gets to 100C, then it doesn't matter whether air or oil heated it up, it can only shed heat at the same rate. Sure oil heats the shell up faster and sooner, but there's little benefit in continuous operation.

GCinDC's testing with oil demonstrated this quite well showing us multiple graphs of temperature over time for the same longish route. The stator stayed cooler at first, because the oil transferred the heat more quickly to the shell, but late in the rides the temperature had climbed to toward the same temperature as without oil. Of course, his ventilated cooling didn't fare much better, but that's because he didn't make use of blades to get meaningful fresh air flow through the motor.

Why risk ruining the motor and have your ebike drip oil like a gasser if you don't significantly increase the power at which it can continuously run? The missing part of the oil cooling strategy that I've seen no one use is to increase the exterior surface area of the motor, so the same bottleneck of heat transfer remains unchanged. Double the surface area doubles the heat it can transfer to the environment at the same temperature. Add exterior blades and/or direct more air flow at the motor you can probably double that again fairly easily by creating a faster and more turbulent flow over the greater surface area that you've created.

Sorry but I'm not going to lead by example on this one, because I have no capacity to weld aluminum or do precision machine work. If I add oil to a motor it will be to build an all weather bike for more worry-free operation at the beach and in the rain, not to increase power.

 

[itchynackers]

Actually, has anyone on ES used any type of ATF in a hub yet, or just speculated?

http://endless-sphere.com/forums/viewtopic.php?f=3&t=41798&hilit=+pikes