Rewinding a 15470 outrunner

[mxlemming]

by SafeDiscDancing » Jan 12 2022 10:31am

mxlemming wrote: ↑Jan 11 2022 7:28pm
This.
Or rewind to 50kV and run the same gear ratio, it amounts to the same thing.
I see the problem. The guy is asking questions about basic kV issues while at the same time very deeply into software coding.

As we know software is all about abstractions... the virtual realities.

So to anyone looking it's very obvious to see "oh, he simply needs the voltage increased" but his mind is focused on where he wants it to go even if simple honest sincere advice is dismissed.

This happens a lot.

In fact sometimes the deeper someone goes into the software side the more you start to ignore the beauty that is the real reality that is this dielectric / magnetic world we live in. (as in the Tesla wisdom)

As I see it this is a simple problem of needing to drive his 27kV motor at a voltage to equalize it with one at 50kV and at that point you finally are making really clear headed comparisons of motor performance.

Motors do have differences and often they can have odd unexpected behaviors due to their design but you always need that open mind to look at what you have.

Grammar, Logic, Rhetoric... the Grammar is the Evidence, the input, the "listening" part.

We all need more of it... my self included because I can fail to listen just as often.

I am not really sure what you are getting at here, or what exactly you are trying to say. But I have built a complete trials bike with the help from my brother, using motorbike rims and tyres, a 120100 outrunner, 1300W battery pack and the complete bike weighs under 48kg. This includes my own speed controller and BMS. The bike performs so well that it can hold its own against petrol machines. We have now gone on, with the investment from a UK company to build a second model with a 15470 outrunner and 1900W battery pack which is fully waterproof and weighs in at 51kg. It has just been a big set back having the torque issue, which with help from the community on here we can now address. Other than this, the bike is 99% complete. I will post pictures of the bikes in the coming months, however currently they have not been out in the public eye. To maximize the performance of the bike, I have to understand how every part of it works, the information I needed was that regardless of KV rating the maximum torque for a motor is the same.

Cheers
Danny

I find safe disc dancing's monologue more confusing than your problem tbh.

Well done on your progress, you've overcome a lot of difficult problems on software, electrical, mechanical fronts.

Look forward to seeing it.. do you take it to uk comps?

 

[SafeDiscDancing]

All I was saying is that kV and voltage are related.

The core premise was "why" don't the motors perform the same?

And the answer is not down some investigation into Delta and Wye (which was one misdirection) but simply to match a voltage to equalize the voltage to kV.

We might simplify:

Voltage * kV -----> Will make the motors mostly equal.

 

[trialspower2]

Once the new bike is completed I will be taking it to comps, all being well I want to ride it at a few nationals on the expert route this year.

While the old bike is really good, it was never well enclosed enough to take it out in wet conditions so its been by no means tested to its full potential. The problem is with making things yourself which im sure many people on here can relate to, you have to solve problems and design things using the processes you have available to you. Once the old bike had served its purpose and got an invester interested, we went back and completely redesigned the bike using all the additional manufacturing processes we then had available to us.

We have literally just got the first prototype running after 18 months of hard work, other than the body work everything has been designed by myself and my brother. Its a real shame we made a bad call on the motor KV otherwise we would have almost been over the line now.

We still of course have the issue of cooling, we put a fan on the motor shaft inside the stator, it blows air through the core and around the windings in a circle. Its inside a finned aluminium enclosure. We also tried to have good heat conduction from the stator to the side of the aluminium plate that makes up one side of the enclosure. Im not sure this alone will keep it cool, we might well have to suck cold external air in yet. Although riding trials, im hoping it will be ok, and just be an issue when blasting between sections etc.

I was thinking regarding the torque, ideally we need a motor KV which reaches saturation at round 180A which is what we are going to call maximum current. We could then gear it for the 40MPH ish top speed, this would give us the absolute maximum torque setup for the motor, of course at the cost of efficiency.

 

[SafeDiscDancing]

I was thinking regarding the torque, ideally we need a motor KV which reaches saturation at round 180A which is what we are going to call maximum current. We could then gear it for the 40MPH ish top speed, this would give us the absolute maximum torque setup for the motor, of course at the cost of efficiency.

Efficiency is a high rpm issue.

Yes... you need to figure out the motor resistance at the present at 27kV then count the turns in there now and figure out how many turns LESS will achieve the 180 amps. (assuming it's falling short now)

But do you see my point also?

If you can increase the voltage to the appropriate level you can get there without rewinding.

And think in terms of "Amp Turns".

Just as "Watts" relate to "Watt Hours" then "Turns" relate to "Amp Turns".

Saturation is a combination of phase Amps and the number of Turns... thus... "Amp Turns".

All the motor "talk" thinks like that. Copper Fill then determine efficiency and heat.

But saturation and hysteresis are different. Saturation is the limit of the iron to achieve peak magnetization but hysteresis is the sluggishness of reversing that magnetization.

Thank Steinmetz for that knowledge. (the guy that perfected Tesla's design)

And I might add... certain iron cores have extreme magnetic peaks, but suffer badly in hysteresis. Normally they are a bad idea but for trials it would be of benefit.

Obviously for now it's a "Get Er Done" situation, but down the road.

Moly Permalloy......... 8,700
MuMETAL®.............. 6,500
Supermendur...........24,000
Permendur...............24,500
Alloy 48....................16,000
Alloy 49....................16,000

 

[trialspower2]

Obviously for now it's a "Get Er Done" situation, but down the road.

Moly Permalloy......... 8,700
MuMETAL®.............. 6,500
Supermendur...........24,000
Permendur...............24,500
Alloy 48....................16,000
Alloy 49....................16,000

This is interesting, as I see a bespoke motor as a great possibility down the road. We have room for another 15mm width without to much of an issue, would take the magnet width up to 50mm.

Where would the material in the stator fall on this list now? Are the two best ones very expensive materials and what percentage would the additional losses be roughly?

The other consideration I had was maximum running temperature, while I have not looked into this, I presume we could use magnets which can withstand 150 degrees rather than 80 degrees. Having a temperature limit of 150 degrees rather than 80 would really help with the chance of keeping it cool enough when travelling between sections.

I realize the additional temperature would affect the performance of the motor, but its really a consideration for when not riding trials and this will make the most heat.

 

[SafeDiscDancing]

Where would the material in the stator fall on this list now? Are the two best ones very expensive materials and what percentage would the additional losses be roughly?

The other consideration I had was maximum running temperature, while I have not looked into this, I presume we could use magnets which can withstand 150 degrees rather than 80 degrees. Having a temperature limit of 150 degrees rather than 80 would really help with the chance of keeping it cool enough when travelling between sections.

To my knowledge no one builds the motors using Permendur laminations because the hysteresis is poor.

But for trials I can see this being the exception to the rule because you care most about maximum flux and not hysteresis.

As for temperature... Yes... some magnets are designed to handle much higher temperatures.

I recommend you search some sellers of magnets. Hold on... let me dig up a link:

https://www.kjmagnetics.com/

These guys are pretty good because they offer the scientific analysis to their magnets as information on their website.

And I bought from them about a decade ago.

 

[mxlemming]

Obviously for now it's a "Get Er Done" situation, but down the road.

Moly Permalloy......... 8,700
MuMETAL®.............. 6,500
Supermendur...........24,000
Permendur...............24,500
Alloy 48....................16,000
Alloy 49....................16,000

This is interesting, as I see a bespoke motor as a great possibility down the road. We have room for another 15mm width without to much of an issue, would take the magnet width up to 50mm.

Where would the material in the stator fall on this list now? Are the two best ones very expensive materials and what percentage would the additional losses be roughly?

The other consideration I had was maximum running temperature, while I have not looked into this, I presume we could use magnets which can withstand 150 degrees rather than 80 degrees. Having a temperature limit of 150 degrees rather than 80 would really help with the chance of keeping it cool enough when travelling between sections.

I realize the additional temperature would affect the performance of the motor, but its really a consideration for when not riding trials and this will make the most heat.

Taken from https://www.researchgate.net/publication/310599234_Soft_Magnetic_Materials
(1) FINEMET nanocrystalline alloys Fe 73.5 Cu 1 Nb 3 B 9 Si 13.5 ; (2) amorphous alloys Co 67 Fe 4 B 14.5 Si 14.5 ; (3) amorphous alloys Fe 78 B 13 Si 9 ; (4) Fe 15 Ni 80 Mo 5 (Mumetall); (5) grain-oriented Fe-(3 wt%) Si sheets; (6) Fe 49 Co 49 V 2 alloys (Permendur); (7) Nonoriented Fe-(3.5 wt%)Si laminations; (8) Low-C steels; (9) Mn-Zn soft ferrites; (10) Ni-Zn soft ferrites; (11) Fe-(6.5 wt%)Si; (12) Fe powder cores (Soft Magnetic Composites).

The kind of silicon electric steel commonly used in motors saturates at a bit under 2T, permendur at 2.4

I think motors commonly use magnets rated to 150 degrees rather than 80. My brief look into this showed you could easily use N35 magnets instead of N52 and it would be cheaper and more commonly available in high temperature version.

See also
https://www.femm.info/wiki/softmagneticmaterials
Which has a pile of nice graphs

 

[trialspower2]

This is interesting, would I be correct to presume that on the Y axis we have the level of magnetism? (tesla) and on the x axis we essentially have the current required to reach this level?

So some materials make the initially magnetism more efficiently? Where as permendur takes a little more to get going, but is then extremely linear all the way until it flattens off to saturation? Which line would be the closest to the typical silicon electric steel in common motors? 5?

 

[SafeDiscDancing]

Which line would be the closest to the typical silicon electric steel in common motors? 5?

I think you get it.

Yes... you pay more up front to get the higher peak and also pay more on the way out.

That's hysteresis.

This is why I'm saying for continuous power the permendur laminations would be horrible as far as losses at higher rpm but for trials who cares?

You get an automatic boost of something like 20% above a normal motor.

Things like transformers are way more concerned with hysteresis because they can just add more weight and not care. You want maximum flux at the lightest weight possible.

 

[larsb]

…

Maybe i missed it, i just skimmed the thread but was the complete system of the drives ever described?

Controller?
Controller setting?
Battery size, voltage and max current
Reduction?
Wheel sizes?

If they were, just ignore this post, otherwise please note that a question like ”why don’t the motor output xxx?” cannot be answered without knowing the full drive system details.

 

[trialspower2]

Hello,

Sorry, I thought I had posted enough details for my original question, but here is some further information;

Controller - Home made surface mount setup, 4 x 160A surface mount mosfets per phase, (24 in total). 80 x 22uf polymer capacitors, 2.2A ripple current, PWM frequency 25Khz, synchronous rectification.

24S battery pack, made up of 120 x 21700 molicell P42A batteries

gear reduction, 2 stage giving 5.4:1

motor 27KV 15470

Wheel size is a normal trials wheel, its almost exactly 2 meters circumference.

 

[larsb]

So you run the motors off the same battery and controller when comparing them?

What max battery current is possible/what do you measure?
what max phase current? What do you measure?

For each motor that is..

I am just trying to understand if both motors get equal scaled feed vs the respective kV values. If not then they cannot be compared. Also what rpm are you comparing them at?

Apart from that, as mxlemming said, if motors have roughly equal torque potential and one is higher kV and geared down more then that motor drive system wins on max torque as long as you feed it enough phase current. That would also mean that the comparison isn’t fair since it is fed more power.

 

[trialspower2]

So, I have finally got around to splitting the motor assembly and I have started to remove a winding. It looks like it is 10 turns, with 24 x 0.3mm strands in delta.

I would like to rewind for double the KV in star, but this of course doesnt quite add up so the closest would be 2 turns. I feel like two turns is not very much and I think it would be difficult to winds loads of wires of not many turns.

Currently I think each phase is effectively split into four groups, as in the image below. What would be the effect of winding the four groups in parallel rather than series? I cant just get my head around this, but I think the current is slit four ways, so to keep the same amp turn ratio, I would need 4 times as many turns? But then the voltage across each coil would be four times as high, so I dunno.....

 

[mxlemming]

Your logic is correct.

The only worry with this is that wiring in series forces balance between the coils, wiring in parallel means they can recirculate so you lose much of the advantage of star vs Delta.

Bear in mind also if you change from star to Delta your hall sensors phase will shift.


An aside,

What's the lamination thickness on that motor? And where did you get it?

 

[trialspower2]

I cant be sure what the lamination thickness of the motor is, I think ive seen somewhere its supposed to be 0.2mm. The motor is from alien power system, it has a custom core which I designed for this application. A fan sits in the recess at the front inside the motor to circulate air through the core. The core is thick to try and track heat into the side plate of the bike which it bolts onto.

I have now managed to remove all the wirings without damaging the isolating material. After removing a few more windings I confirmed that the motor had 9 turns, not 10.

I am not sure what wire thickness to attempt the rewind in, I understand that thicker wire wont bend as tight, but I think the tiny stands are really inefficient in actual copper content. If I used thicker wire, could I maybe try and bend it in more are a circular arc around the tight stator corner rather than having two sharp 90degree bends. There is plenty of running clearance in this area.I even thought about 3dprinting a semi circuit and gluing it on, I know the wire would be slightly longer, maybe it would affect the efficiency? just thinking I could pull the wire harder around this without going through the coating on a sharp corner.

I think the 24 x 0.3mm strands in the motor only gave 1.7mm cross section area, I have just found some solid wire at 1.2mm diameter (1.13mm2) and without any attempt to bend it tight, it fits easy. If the original design had been wound with two of these it would have been 2.26mm2 which surely would have been better?

Im changing from delta to star, but I can swap the phase wires around and the hall sensor signals if need be.

Would wiring in full parallel be better or worse than having 4 parallel groups of 3 in series?

Considering full parallel, (for star in series at approx 50KV I would need 2 turns) I have 12 teeth of the same phase, so that would be 24 turns with them all in full parallel? If I did that in say 1.2mm wire, that would give me 1.13m2 x 12, so 13.5mm2?

Image below shows the stator with 9 turns of 1.2mm diameter wire (just something I had lying around to wrap around it)

 

[mxlemming]

I don't think dividing by 12 will work. 4 clumps of 3 might but 12 I really think there will be an issue. There's no symmetry around the 3 teeth. They're not going to be changing flux quite in phase with each other.

I'd like larsb to chime in on this one before you spend hours winding tbh. He's wound far more motors than me.

I favour 0.7mm wire as a compromise between workability and fill and so on. I've wound... 2 stators ever.

Regarding hall sensors, I think it's a30 degree phase shift so you can't just swap them round. You'll need smarter logic. Again... Hope Lars can confirm.

Your board... Is it still running bldc? Have you smartened up the commutation and current control since your thread on the controller?

 

[trialspower2]

Regarding hall sensors, I think it's a30 degree phase shift so you can't just swap them round. You'll need smarter logic. Again... Hope Lars can confirm.

My hall sensors are adjustable, If I run out of adjustment, I can always laser cut a new bracket for them.

My controller is just running BLDC, its runs with synchronous rectification. I wrote a speed control equation which takes into it throttle position and motor RPM. The equation has coefficients which you can change, what it effectively does is it limits the output based on motor RPM. This naturally then limits the current. When considering full throttle, if you put more load on the motor, it slows it down, this reduction in RPM naturally reduces the output the motor, limiting the current. If you hit something hard in trials and it locks the wheel up, the controller reduces the output down to the maximum allowed for locked rotor. It also takes the battery voltage and adjusts to this as well.

It also controls regen in the same way, it actually works really well. It doesn't oscillate and when considering trials it makes the bike ride almost like it has a flywheel.

I hope larsb does pass on his experience, as like you say, a little advice can save hours of wasted time.

 

[mxlemming]

Why not put some current sensors on the motor phase wires and control the current directly?

You can do this quite easily, just find some inductive sensors.

Less than 200A use acs758
Over 200A you can use a linear hall sensor and a steel concentrator ring, then your current capability is virtually unlimited.

You'll probably find this is far far kinder to your whole system, and gives you true torque control with near instant response.

Below is one I bodged up in half an hour on a 12070 motor. Limited to 100A only by the air gap on the ring

 

[trialspower2]

You'll probably find this is far far kinder to your whole system, and gives you true torque control with near instant response.

I will be honest with you, I dont really understand what you mean by this. I am interested in the bike giving me the reponse I ask for from the throttle, the torque required to do this is down to the conditions at any moment in time. Without measuring the output i.e. rpm how can make speed control decisions.

My feeling with measuring current for control is that it will always be lagging and prone to causing oscillations.

 

[larsb]

So, I have finally got around to splitting the motor assembly and I have started to remove a winding. It looks like it is 10 turns, with 24 x 0.3mm strands in delta.

I would like to rewind for double the KV in star, but this of course doesnt quite add up so the closest would be 2 turns. I feel like two turns is not very much and I think it would be difficult to winds loads of wires of not many turns.

Currently I think each phase is effectively split into four groups, as in the image below. What would be the effect of winding the four groups in parallel rather than series? I cant just get my head around this, but I think the current is slit four ways, so to keep the same amp turn ratio, I would need 4 times as many turns? But then the voltage across each coil would be four times as high, so I dunno.....

4.jpg
3.jpg
2.jpg
1.jpg

Seems like a nice candidate for a rewind
Not so high copper fill there for sure!

I wouldn't change the winding scheme and keep to delta-connection, 10T/2/1,7 is about three turns, (not 2?) to reach your goal if using wye, it will work but each turn wire or bundle has to be massive. a five turn delta is slightly better but still not easy to get nicely wound. I'd start with drawing the slot in CAD or on a paper and start playing with the different wire options out there. When it comes to parallell / half-parallell windings i cannot say much. i've used half-parallel on my revolt winding job on a 12s14n motor but not sure same pattern can always be used. I've used RCgroups for questions like that

 

[mxlemming]

You'll probably find this is far far kinder to your whole system, and gives you true torque control with near instant response.

I will be honest with you, I dont really understand what you mean by this. I am interested in the bike giving me the reponse I ask for from the throttle, the torque required to do this is down to the conditions at any moment in time. Without measuring the output i.e. rpm how can make speed control decisions.

My feeling with measuring current for control is that it will always be lagging and prone to causing oscillations.

I presume you're not controlling speed, you're measuring speed with the hall sensors and using a bit of math/look up table to determine the voltage/pwm you need. From this you assume a current without measuring.

If you measure current, you can set a control loop in your software to control said current.

Given the motors you're using, it's entirely possible to have the current reach and be steady at it's set point within 100 micro seconds.

The current, if applied at the right angle which you read from the hall sensors is the factor that defines torque. Including a measure of speed in your control loop is not necessary. The hall sensors and speed only update 1 time per several hundred pwm cycles. Current measurements can update every single pwm cycle.

If you want speed control not torque control, which you probably don't, you probably want to actually have your throttle produce a speed request, which then gets converted to a current request via a (very aggressive) PID loop and handed to the current controller.

I say this because you're talking about large fans to cool large motors. The motors you're using aren't particularly efficient but the can produce great torque for their weight. Trials, while explosive doesn't seem to be an event which requires high continuous power, even at the highest level it uses a tiny fraction of the average power used in say Motocross so from what I've read that you've written so far I'm guessing you can probably reduce your heat burden.

I don't think you'll get better outright performance from this because your control mechanism probably produces enormous peak currents, but you can probably dramatically increase the performance power energy used and great created.

Or maybe not. Maybe you've got your speed based control loop completely dialed.

 

[mxlemming]

To put my point more simply, the performance of your bike is limited by the flux in the air gap of the rotor. The maximum drive your motor can possibly give is when your current starts to exceed the magnetic saturation of materials. For your motors this is probably in the 300A region.

Excess current beyond this only causes heat.

To get the absolute max out of your motor, you need to be accurately controlling the current as close to this limit as possible. Your motors have about 15mohm resistance so a 1V error causes something like 66A error.

If you're not measuring and controlling the current you're either missing out on performance or generating excess heat.

 

[trialspower2]

Seems like a nice candidate for a rewind
Not so high copper fill there for sure!

I wouldn't change the winding scheme and keep to delta-connection, 10T/2/1,7 is about three turns, (not 2?) to reach your goal if using wye, it will work but each turn wire or bundle has to be massive. a five turn delta is slightly better but still not easy to get nicely wound. I'd start with drawing the slot in CAD or on a paper and start playing with the different wire options out there. When it comes to parallell / half-parallell windings i cannot say much. i've used half-parallel on my revolt winding job on a 12s14n motor but not sure same pattern can always be used. I've used RCgroups for questions like that

oh yeah, slight calculation error there, I once I had removed more of the motor windings it was actually 9 turns. As you have pointed out this comes out as 2.5T. I am considering the 4 groups of 3 parallel option, this would give me 10 turns in star. I think the main practical issue with parallel is that I will have tails coming off in various locations which all need keeping the same length.

I will have a look at drawing something up on solidworks.

To get the absolute max out of your motor, you need to be accurately controlling the current as close to this limit as possible. Your motors have about 15mohm resistance so a 1V error causes something like 66A error.

While it is quite difficult to test without a rolling road, depending on which output level I choose the current sits quite static during full acceleration.

The equation I made is based on a straight line plot of unloaded motor RPM. The output limit then tracks this line running parallel to it. I do intend to put the bike on a rolling road once its running again, so will see how it comes out.

I didnt mention, but the motor also has an incremental encoder on the shaft for accurate speed feedback, really helps when going really slow. The calculations run every 10ms.

I thought to run a more advanced control you would have to drive the motor with a sine way.

 

[mxlemming]

An incremental encoder certainly will give better results than halls and allow more frequent updates like you say. It certainly can close the gap to what's achievable by current sensors.

How much current are you actually feeding these motors?

Sin and field oriented control will be better yes. Experience has shown me the difference isn't as huge as the expectation but it is better and much quieter. For trials though the bldc noise might actually be quite pleasant and provide some auditory feedback to the rider.

 

[trialspower2]

I was looking at up to 200A on 24S, although from what I now know about KV ratings and maximum torque I might consider moving down onto 15S at 300A as I feel this is a safer voltage. I know there will be more losses with the higher current, but I feel the safety factor might outweigh this.

Does sin and field oriented control make any difference to the stator losses at high RPM?

Like you say, I dont think the BLDC noise is an issue in trials.