Converting a hubmotor to a middrivemotor

[izeman]

i can't believe that this axle won't break. i know that there are some amazing compounds but if this can hold an axle togethet that transmit a lot of torque???

 

[crossbreak]

ok get it. you really turned around the shaft and elongated the threaded end into the motor. So this must have been a lot of work. Why not simply insert a key now that you have a thicker shaft? Will definitely hold up more torque than a press fit, even with lock tite. Press fits have quite challenging tolerances... hard to reproduce if you dont have very expensive tools.

i can't believe that this axle won't break.

The weak point now is in between the two bearings. So this is now at a place that takes less load (torque/twist load is none). makes sense to me. Should hold up now if the press fit does not fail. Anyway, this method seems to be more work/more expensive than building a new shaft from scratch

btw, my twisted shaft still works and is in service. it never broke. You just need a plain surfaces in between all these parts: bearing->spacer->sprockethub->spacer->freewheel-adapter. If all these surfaces are flat and smooth, the shaft almost only sees torque/twist load. It only twisted from abuse. With tha 100A phase current it wont twist. More than this is devious for this size of motor anyway.

 

[evb]

transmit a lot of torque???

There is phrase from datasheet

After 7 days @ 22 °C
Compressive Shear Strength
Steel pins and collars - 29 N*mm2 (4200 PSI)

it means when axle diameter is 15mm, on each mm of axle length torque is 10.25Nm, adapter wide is 24mm, total torque 246Nm.

Why not simply insert a key now that you have a thicker shaft?

it's more difficult, I can use milling machine to make a groove into axle, but i can't make groove into adapter, I need use special machine for do it.
Loctite 638 is cheap in China and users no need to divide adapter and axle.

 

[crossbreak]

you would need a machine like i showed in the video on the last page. You can also abuse your CNC mill for that

 

[TomHed50]

I have a mac 10 and thinkin of convert to middrive. Is it better to have a chain on each side? Or is a aapter like crossbreak best? Both chains on one side(should be thinner overall?) anyone selling?

Which battery should i use? I have a 48/15 and a 36/12, can i feed the little thing with 72v if i change cables?

 

[crossbreak]

sorry.. too little demand for that.. i wont make them anymore. the MAC has several drawbacks: Heatsink and hollow shaft are on the wrong side...no load loss is much higher than the bafangs too.

If you wanna convert the MAC, i propose you include the heatsink into the mount like spinningmagnets and iceman did. Also i would make a new shaft like both did as well. So you can make your own sprocket+freewheel adapter. Simply fasten it with a key and a bolt in the center. If you dont have machines for doing so, i propose you visit a machine shop of your choice

btw, i finally start builing downtube mount battery for one of my bikes: http://endless-sphere.com/forums/viewtopic.php?f=14&t=59940&p=895915#p895915

 

[TomHed50]

Ok can u plz come with a suggestion on how the new shaft would be machined? Length/thickness ,m12 or m16 bolt?

Any other upgrades to do? Aircooling?,change of bearings?
Which controller would work best with this motor if i go with a 48v15a batt?
Recommended gearing for a 29er?

 

[crossbreak]

Shaft dimensions:
The dia of the shaft should be trimmed to 16mm. Nut dia and thread pitch you can reared in the specs of your favorite motor

Controller+Battery:
I do 30A at 74V with the BPM2 ATM. It can take more but cooling gets in issue even if you install a heatsink on long hill if you go FT. I like Kelly controllers, since they have phase current feedback control. If you have a look at their homepage you may find one that fits your needs. For 48V/15A i would better go for a Bafang BBS02 instead of a converted hub..that is easier... chain will still last a while at only 15A even if you do not have a dual freewheel drive then. Just use a cassette that has 13T as the smallest sporcket and a large chainwheel like 50T.

Gearing:
For a 29" would be 18T..depending on what you aim for. Less then 18T leads to high chain tension which many frames to not stand well. Better go for 26" that rides better IMO and does less stress to the already overstressed chain.

Frame:
I would always use a 2-linker frame with large wheelbase and straight downtube.

Cant tell you what you want as long as you dont tell... do you go offroad of just steep roads?

 

[TomHed50]

Thank you. Probably a bbs2 would be the "shit" because my use of electric bikes have so far been 90% commuting and drive kids to preschool in the bikewagon
But since I got a mac motor i thought it would be fun to convert to middrive. I dont have much time to build and think though so if its to much of that i will probably sell the hub and buy a bbs2
I have two other bikes , commencal supreme dh 27,5 and scott ransom. A kit like l rods would be nice on one of those

 

[crossbreak]

sounds like you go for the bbs02 in your situation this should be better... you can improve a lot on your bike else where and be happy of what you get. 25A and 48V BBS02 is fun is you ride it on the street with the right gearing setup too

 

[cwah]

OK, got my BPM2, it's an 11T Too fast for Star->Delta conversion, too slow for a real high power setup :x The threaded side plate also cost me some of my nerves.

Tried to open it. One needs a special tool for that. I made one using a sheet. I drilled 3 holes into the sheet, one 35mm one for the BSA threads on the sideplate. And another 2 for the teeth that engage the grooves in the side plate. Just filed down two hex srews till they fit into the grooves and screwed them on my sheet. Used only two, but there are 4 grooves. A long lever on the sheet helps to input the right torque. A BSA threaded nut screws the self-made tool on the motor after the teeth engaged into the grooves (the nut has also 6-holes, that is a bit confusing, just had no other BSA nut at hand). This way, there is no way the tool can tilt and damage anything.

I was also able to open it without lacing it into a rim: I used the 6-hole interface for mounting the motor in my benchwise by drilling 7 holes into a 4mm alloy sheet: one 35mm hole in the center and 6 holes for the disc brake interface. After that I heated the motor (not the sideplate) and screwed of the sideplate anti-clockwise, I needed around 30Nm to do that, but this motor is unused. In a use motor this can be a lot more, since the chain drive further engages the threads of the sideplate when pedaling hard.

I need to open mine too, so I need to buy an aluminium plate? Maybe 5mm?

 

[crossbreak]

Yep. My tool has only two dogs. Quite simple To Make. Just be a bit creative.. good luck

 

[izeman]

The best would be to dril the holes before removing the axle so it is still perfectly centered. Problem is the long axle hits the drill chuck.

Pics of the hall sensors for Fechter below.

Thanks for the pic. Yes those hall sensors are NOT located in the slots, but offset slightly. I assume this will result in more neutral timing and better controller compatibility.

To drill out the holes without the chuck hitting, you could use a really long drill bit (aircraft drill).

how was the offset calculated? i'm struggeling with hall sensors in my MAC mid drive as well. maybe this solution could help?!

 

[fechter]

how was the offset calculated? i'm struggeling with hall sensors in my MAC mid drive as well. maybe this solution could help?!

Each motor 'pole' is actually a group of 3. This is a concentrated wind configuration. This reduces the amount of resistance in the wiring between poles and makes winding easier. When a winding is energized, the magnets will try to align with the stator poles with the middle stator pole of the 3 group being dead center. At this exact spot is where you want the hall sensor to switch states. Since there are 18 stator poles and 16 magnet poles, the slots don't exactly line up with the gap between the magnets. I suppose there's a way to calculate the exact offset. It's not much.

This slight offset causes the hall timing to be slightly advanced. At very high RPM, the advance is probably OK, but it sucks at low RPM unless the controller is specifically designed to compensate for this.

 

[izeman]

thanks. i never thought of different # of stator and rotor poles. but of course this would matter, and i can imagine this being an issue for high sophisticated foc controllers that rely on hall sensor data, especially with a high e-rpm motor.
i guess i can try making the hall sensor plate adjustable for bench testing and see if it helps moving the sensors back or forth. i hope the sine wave kelly controller can work with that. lebowski's controller could do it for sure but it don't have at home to test it right now.
the question now is: how would i calculate the correct position and distance between the hall sensors? and if it can't calculate it and have to do try&error - how will i know that i've found the correct position?

 

[izeman]

i did some testing now with the hall sensor board moving around. the distance between halls was unchanged (do i need to adjust that for the MAC as well??). i could see a significant difference between different positions.
the kelly sine wive controller has some weird rpm overshoot if you go WOT from stand still. so it goes up to eg 2.200 rpm and then settles at 1.900 rpm after a second. then if you let go the throttle it jumps up to 2.200 again an then goes down to 0 rpm. really strange.
this behaviour was gone with a offset of around 1 width of a hall sensor. so the halls where around 3mm moved. can't say if in turning direction or against it. have to check that.
but i can't say if this may cause something else strange ... don't know the kelly's internals.
what do you think? how can i be sure that there are no negative side effects?

 

[crossbreak]

you made it, attaboy!! the timing difference may correct errors that arise from too slow processing/ misinterpretations of the FOC algorithm. You have to find a position that works for both high and low rpm. You can calc hall sensor position by pole count.

http://mitrocketscience.blogspot.de/2011/08/hall-effect-sensor-placement-for.html

360/16p/3 = 7.5° deg mechanical or 120° deg electrical

an offset of ~3mm like you try is a lot. around 30°is 40° electrical is my guess . Into turning direction (positive timing advance). Normal hall placing is "neutral" which in fact means the signal is delayed by some degree, due hysteresis of the sensors. Another delay comes from motor inductance at high rpm, which FOC should compensate - in theory - if i got it right.

Plz drop us a picture of your hall board. I really like to try it too (if I find the time though ). And maybe a pic of the new design MAC axle, we talked about on the phone (if you have one). I can imagine how it must look like but it would be nice to have some reference, maybe someone comes up with a good idea how to convert them. Maybe also interesting in terms of a potential 2016 double wide MAC https://endless-sphere.com/forums/viewtopic.php?f=25&t=68784&start=50#p1098795

 

[crossbreak]

now that you made it... i still doubt my (and your old) no load measurements somewhat. they seem to be quite high for the MAC. if you find the time, it would be awesome if you could redo no load current measurement for two (or better three) different rpm with corrected timing. As you can read rpm real time from the software, it should not be a big deal for you? You can use the same battery voltage for all 3, just alter rpm by throttle (makes no difference IMOE). So you can measure all three in a single run, should take you just a minute once everything is hooked up

I hope there is a significant improvement, as high no load current may arise from bad timing. This would be a final proof that the MAC is superior to the Bafang, at least with corrected timing

 

[izeman]

thanks crossbreak
nothing special about the hall sensor board. it's the original MAC's board. i just heated the halls (the epoxy around it) and lifted the board, pulling the halls out of their gap between the teeth. than i moved the board around a bit, fixing it with duct tape only. the test was done with the motor alone, held by a vise.
i can do more testing on no load current. sure. no problem, as the controller and CA is attached all the time. i just need to write it down. i can do some testing with different controllers as well. but it will take some time as my wife starts complaining more and more about me sitting in the workshop every evening while she irons my shirts ... LOL
edit: the new axle is in the works. it's not done yet.

 

[crossbreak]

thanks iceman, the FOC measurements are all i'm after as we have a sufficient number of figures for other controllers

 

[madin88]

As the MAC has quite high ERPM, i think the response characteristic of the halls plays a big role. A hall that switches faster would have the same effect as if you shift the position mechanically, but you don't make it worse at lower RPM.

I think it would make the most sense to adjust the angle for lowest possible consumption at very high RPM as you cannot vary it across the range of speed.
Keep an eye on no-load RPM as well and try to keep it in the range of what kV x voltage would let. A too high timing would be similar to what field weakening or infineon 120% setting does and definitely lowers efficiency.
Additionally it may help to run a separate and shilded wire for the halls from the controller til inside the motor, because the phase wires can create noise in the single sensor lines which could disturb the controller.

looking forward to see your findings and hopefully you have success in getting this bitchy motor running well in the end

 

[izeman]

i have some hope now, and i guess it will be better with the new halls from you!

 

[izeman]

A too high timing would be similar to what field weakening or infineon 120% setting does and definitely lowers efficiency.

now that you say that it really looks as halls' original position gives a wrong timing. the motor runs with around 110% for a second and then goes down to 100% - maybe because some internal logarithm tells the controller that something is wrong and the controller tries to correct that.
but that of course is only an assumption as the kelly is closed source and we can't look inside.

 

[crossbreak]

Habe you tested the difference in max rpm for Infineon? If it differs more than a few %, it tells you that Kelly does something wrong

Good point on hall hysteresis maddin. Thanks for the support

 

[izeman]

fany from kelly said that the top rpm of their sine wave is around 15% lower than trapezoid controller due to the way sine wave works.