Multiple motor & controller approach?

General BVS related area
User avatar
Jeremy
Posts: 472
Joined: Thu Jul 31, 2008 6:35 pm
Location: Salisbury

Multiple motor & controller approach?

Postby Jeremy » Tue Aug 19, 2008 5:28 pm

The thread on high voltage versus low voltage, plus Malcolm's timely reminder on the IoM thread about the power to weight ratio of brushless model motors, has got me thinking.

We're used to thinking in terms of a single big motor, or perhaps a set of wheel motors, as being the preferred approach for an EV. This used to the the preferred approach in other areas, like computing, but we've gradually come to realise that a distributed architecture can provide some benefits, like better performance, a measure of redundancy or lower overall cost.

We're also used to the idea that high voltage means lower losses and better efficiency. However, that comes at a price - any vehicle that has a power supply that's greater than 50V is supposed to comply with some pretty stringent safety requirements, to reduce the risk of death or injury from electric shock, both in use and in the event of an accident. Clearly there is some safety (and potentially regulatory compliance demonstration) benefit from opting for a low voltage (sub-50V) system, if only the efficiency issue could be adequately resolved.

One way to reduce the effect of losses is to look at a multiple motor/controller solution. What if an array of small, efficient, brushless motors were connected together mechanically, with each run by it's own controller, and used to provide a multi-motor that could be treated like a single, bigger motor?

The obvious questions revolve around technical feasibility, efficiency, cooling, weight and cost.

Technical feasibility

Brushless, sensorless motors will happily work together if mechanically locked and driven by individual controllers. This has been proven by modellers who have experimented with motors on a common shaft.
Motor starting is now good, with even budget controllers having good start algorithms. These motors can be simply reversed, either by the controller or by simply swapping two phase wires. Not all motors would need to be reversed to provide a reverse gear, as reverse wouldn't require full power. There are no good technical reasons as to why such a configuration shouldn't work well, provided that the mechanical installation is adequate.

Efficiency

Brushless motors in the 1 - 5kW range can be around 90% efficient, with fairly broad and flat efficiency curves. Controllers are around 98 to 99% efficient, even at fairly high power outputs and operating currents up around 80 to 100A. There would be some additional loss caused by the combining transmission, but this could, perhaps, be as low as 3% or so. Overall motor/controller system efficiency of a multiple motor system could be over 85%, with good design and careful component selection.
To reduce battery losses, and perhaps add redundancy, separate battery sub-packs for each motor would seem a good idea.

Cooling

This is an area that needs careful attention, because small motors have a low thermal mass, so will overheat quickly if overloaded. On the other hand, a multiple motor array has a high surface area to volume ratio, so should cool more effectively than a single large motor. Because of the mechanical configuration it's likely that some form of forced air cooling would be needed.

Weight

This is where the multiple motor/controller approach starts to look attractive. A typical brushless motor that will deliver a couple of kW continuously (maybe 3kW peak) will weigh less than 1kg, complete with it's controller. Even allowing for the mechanical components needed to both couple the motors together and reduce the output shaft speed to a more useful rpm will still result in a motor and controller combination that is much lighter than an equivalent power single motor and controller.

Cost

Another advantage is the significant cost benefit from the multiple motor/controller approach. I've compared the watts per pound for some typical brushless motor/controller combinations with my own Mars ME0709/Alltrax 4834 system. Brushless motor and controller combinations working at around 30V, delivering continuous power outputs of about 1500 to 1800 watts, deliver around 35 to 40 watts per £. The mechanical and transmission parts will add to the cost, but I would be surprised if the this reduced the above figure below about 30 watts per £. For comparison, the Mars/Alltrax combination delivers around 12 to 15 watts per £, so it looks as if the multiple motor/controller could be around half the price for any given power output.


Obviously there are some snags with this approach, the most significant being the added mechanical complexity and potential reliability issues this may bring. If the coupling transmission used toothed belts, then it could use standard, off-the-shelf, components, with good reliability and efficiency, plus the advantage of low noise levels. If CNC machined kits were made available, consisting of a pair of alloy sandwich plates to mount the motors, plus accurately machined spacers, bearing housings and an output shaft, then such a system would be relatively easy to assemble.

The sandwich plates could incorporate the cooling system, if the sides were enclosed and the space between fed by a high pressure fan, feeding air to each motor via it's perforated end bell, to exit via the spinning rotor end vents that would project out from the plates. The controllers could be mounted inside the space between the plates, to produce an integrated unit that only needs power supply and throttle connections.

To sum up, what I'm proposing is a multiple motor/controller unit that could be substituted for a single large motor and controller, but at a lower cost and with a lower weight. I would like to encourage a sensible and constructive debate on the pros and cons of this sort of approach, rather than start a contentious debate on what's intrinsically "right" or "wrong" - I fully appreciate that it's a novel and unusual approach that might offend the traditionalists!

Jeremy

PS: If anyone wants an idea of the sort of motor and controller that I'm considering in this (lengthy!) discussion starter on low voltage, multi motor, systems, then take a look here:

Motor: http://tiny.cc/xPlns

Controller: http://tiny.cc/eavKS

MalcolmB
Posts: 423
Joined: Sat Jun 23, 2007 8:07 pm

Postby MalcolmB » Wed Aug 20, 2008 8:14 am

Just thinking out loud:
I reckon it's important to identify first what benefits this approach would offer over a single large motor. These motors have been developed for the radio control market, and the special demands of this market have pushed the technology much faster than in any other area. Unfortunately for us the RC crowd don't use big motors, so we don't yet know what the possibilities are at the bigger end of the scale. I wonder if this is just a transition solution? Would multiple small motors on a common shaft have significant benefits over a single large motor of similar design? Or is that the wrong comparison - to take full advantage would you want the motors to run totally independently? This of course would make linking them together mechanically more of challenge...

hyve
Posts: 60
Joined: Sat Oct 06, 2007 1:35 pm
Location: Chesterfield, Derbys.

Postby hyve » Wed Aug 20, 2008 8:59 am

Stimulating idea, Jeremy and to my mind just what BVS should be doing.

Logic says to me that lots of motor ends instead of just 2 has to be heavier, but I'll take your word for it on the efficiency side.

I like the idea of making the sandwich plates form a plenum for cooling air, as well as an independent fan. This has for some time seemed to me a far better way of cooling, permitting control to suit demand. Isn't this how electric trains do it ?

Instead of belt drives, could I suggest arranging the motors in a circle with a small gear on each driving one large internally toothed ring gear around the lot ?
This deals with the very large step down ratio which will probably be needed for small high rpm motors, since using an internal gear gives more tooth contact. If this assembly is placed horizontal with a vertical shaft axis, the output can go straight down to a conventional spiral bevel final drive with differential and swinging half shafts. This sits in the centre and will probably project little lower than the motors, making a neat radial motor/transmission unit with a very low CoG.
Space will need to be left between a pair of motors each side for the half shafts, to maintain the compact, below-boot-floor layout, but otherwise a good number of motors could be employed this way. In fact for such a large ring gear it might be possible to use one of the plastics, since each motor is delivering fairly low power. This could help keep noise down and perhaps simplify lubrication/packaging issues.

I suspect so many interfaces, whether belt or gear driven, must make a deal of noise though - and belt drives are not entirely silent. Noise insulation will be needed.

Such a rear engine unit leaves little space for batteries. To put a large diameter assembly like this in the front of the car will be difficult, however because of intrusion into the footwell and probable interference with the steering gear. So the snag is that the batteries may be rather further from the motors than is ideal.
Batteries tend to be a large mass and to have both them and the motor/transmission all at one end of the car is not ideal for suspension/handling dynamics either. The inevitable designer's conflicting requirements scenario, to be resolved by compromises. Perhaps the batteries in two packs, with the less frequently used full power set at the front ?

I'll leave the electrical experts to debate that first premise: that a number of small motors can be more efficient than one large, but otherwise this sounds good to me.
Peter Ph

User avatar
Jeremy
Posts: 472
Joined: Thu Jul 31, 2008 6:35 pm
Location: Salisbury

Postby Jeremy » Wed Aug 20, 2008 2:48 pm

Thanks for the contributions, chaps. I think that a large brushless motor, using RC model motor technology, would be a better answer, but at the moment the more powerful (up to about 12kW) RC motors are very expensive, around £800 each.

The multiple small motor approach would have a small efficiency penalty, largely as a consequence of the additional losses from the transmission required to couple the motors together. I'm hopeful that this can be kept down to 3% or less though, in which case it probably isn't too important.

I like the idea of gears, although noise might be a problem, as might fabricating the prototype using off-the-shelf parts. HTD belts and pulleys have the advantage of being readily available, if a little expensive.

I've been doing some more work at lunchtime today to get an idea of cost and weight. This motor: http://tinyurl.com/6bko5k looks to be a reasonable candidate. It weighs 679g (about 1.5 lbs), will deliver 2800 watts maximum, and can accept up to 80 amps, although best efficiency is at 60 amps. It costs just $48.79 (about £24.50), so is very cheap indeed. It has a kV (rpm per volt) of 215.

Four of these small motors, coupled together with a 2:1 reduction would give give an output shaft speed of around 2580rpm if run at 24V. Maximum power from four of these motors at this voltage would be around 7.68kw, which is similar to something like the smaller Perm or Agni motors, or my Mars motor I'm using on my bike.

Total weight for four of these motors would be about 2.7kg, but to this would need to be added the weight of the mounting plates, pulleys, belts, output shaft etc. I think it should be possible to build a completed assembly that weighs around 4.5kg though, maybe less. This is less than half the weight of an equivalent power PM motor of the types mentioned.

Controllers present a challenge, but for an initial experiment I think I might opt for some relatively cheap, low voltage ones. If the concept works, then it would be fairly simple to upgrade these for higher voltage ones. The best value I can find are these: http://tinyurl.com/5ngdsy which will handle over 100 amps at 30V. These controllers cost just $29.99 (about £15.00) each, so can be considered expendable if they don't work out. They only weigh 115g each, so would only add about half a kg to the total assembly weight.

I'm seriously contemplating taking the risk and ordering four motors and controllers to build a proof-of-concept modular motor test model. The aim would be to produce a motor and controller combination that could meet the following specification:

- 7kW maximum power output

- 5kW continuous power output

- 24V operation (from four 24V battery packs capable of at least 80 amps each)

- <5kg total weight (including controllers)

- Shaft speed about 2500rpm maximum

- better than 80% efficiency at continuous power rating

- Total cost, including motors, transmission, mechanical parts and controllers < £250

Although I think that the most sensible layout would be to have the elegant circular gearbox that Peter suggested above, for the proof-of-concept model it will be easier to use two pairs of motors mounted at either end of a couple of rectangular plates, driving a common centre shaft. This can be engineered easily, as it only need two motor pulleys (two facing motors can couple into one wide pulley, used as a coupler) and a wide single central pulley driven by the motor pairs on either side (on pair being set to contra rotate).

Before ordering the motors I need to find a source for nice alloy HTD pulleys. RS do them, but they seem expensive. I think I may have to pay a visit to my local bearing/transmission shop and see what they have. I probably have enough other stuff, like 3/8" alloy plate etc, so it should then just be a matter of machining up the parts, bolting them up and looking at ways to build a dyno to test the motor. The latter should be easy enough, as I made a drum brake dyno a few years ago and probably still have the remnants around somewhere.

If this works as I hope it might, then it would be worth investing in some better parts for a reliable final version.

Jeremy

hyve
Posts: 60
Joined: Sat Oct 06, 2007 1:35 pm
Location: Chesterfield, Derbys.

Postby hyve » Thu Aug 21, 2008 8:07 am

There are two reasons I suggested gears instead of timing belts: one is, as you're discovering, that timing belts and their sprockets are very expensive; scarcely any cheaper than gears, in fact.
The other is that, even for just four motors you cannot get enough wrap on the driver to use only one belt around the lot. This means, as you suggest, a wider driven sprocket and two belts in this case.
You now have your motor pairs staggered, getting away from your nice simple flat sandwich plates or adding spacers and motors with different length output shafts.
For any serious power the complexity will be worse, and the possible reduction is never going to be as good as with gears. It will probably not only be heavier, but also require a lot more space.
Further, the belts call for either automatic tensioners or provision for regular adjustment. Plus they need to be nearly as precisely aligned as gears. To my mind belt drive for such a device is pretty much rejected completely by all these factors.

Have a look at http://www.hpcgears.com, an excellent firm near me who make off the shelf gears as well as sprockets/chains, and also stock various types of timing belts/sprockets. They have a sister coy called Ondrives, http://www.ondrives.com, which makes a vast range of gearboxes and motors plus all kinds of useful mechanical bits and pieces. Well worth a look through their catalogues, both.

Using some HPC off the shelf gears would leave only the internal ring gear to be made, as HPC's largest is only 180mm PCD. ( This might do for your experimental model)
This can be bolted to an alloy drive plate, which would bolt to the prop. coupling on a conventional independent rear differential unit, as mentioned.
For a motorcycle with fewer motors and therefore a much smaller overall diameter the circular unit stood vertically should fit very comfortably in the usual motor/tank space, but as always battery space in a motorcycle seems very difficult to find. The packaging challenges for a motorcycle are difficult at the best of times - I've built a few - but adding batteries turns it into a challenge I'm afraid I just wouldn't want. Good luck !
Peter Ph

User avatar
Jeremy
Posts: 472
Joined: Thu Jul 31, 2008 6:35 pm
Location: Salisbury

Postby Jeremy » Thu Aug 21, 2008 6:47 pm

I agree that gears would make a good final solution, but for a proof-of-concept model they would entail some extra work in machining up a gear case, sealing it to contain the oil and fitting seals to the motor input shafts and the output shaft. It's all perfectly possible to do, but would be wasted effort if I can't get the basic idea of multiple motors connected together to work.

The advantage of using belt drive for a proof-of-concept model is primarily that it's quick and easy to build. I can then do some bench testing and get a feel for the way that a group of motors and controllers interact and perform.

The unknowns at the moment are largely to do with the way that the controllers will work together. I know that RC modellers have used pairs of motors connected on a single shaft with some success, but I've also heard that the power output may not be double that of a single motor and that there may be some issues with harmonising the power contribution from each motor.

It's also been suggested that I may need to add sprag clutches to the motor shafts, in case of a controller problem. One possible scenario may be that a controller fails in such a way that it locks up one motor (I don't know how possible this is) in which case the other motors will suddenly see a massively increased load and may fail. Adding a sprag clutch removes this worry and is fairly easy and cheap to do with drawn cup roller clutches, but it does remove the possibility of reversing the drive direction.

I've been doing some work on motor and controller selection, trying to formulate a way to compare value for money and power to weight ratios. I've singled out a 100amp continuous (125A maximum) controller that seems to have good reviews, yet only costs about £30. When it comes to motors, it seems I have three main choices.

Tower Pro 5330 10T - 80A max, 60A best eff, 215kV, weight 679g, about £25

HXT 73-64 - 75A max, 60A best eff, 200kV, weight 790g, about £35

HXT 80-85A - 125A max, 110A best eff, 250kV, weight 1230g, about £75

There are some others I'm still looking at, but these give an idea of power, weight and cost.

The comparison figures I have, using the 100 amp controller at 24 volts and not exceeding the motor ratings, are:

Tower Pro 5330 10T - Continuous power = 1920W, total price = £55, watts/kg = 558, watts/£ = 26

HXT 73-64 - Continuous power = 1440W, total price = £65, watts/kg = 635, watts/£ = 22

HXT 80-85A - Continuous power = 2400W, total price = £100, watts/kg = 565, watts/£ = 24

From these it looks as if the Tower Pro 5330 comes out on top. Four of these would give about 7.68kW, weigh 3.216kg complete with controllers and cost about £220. To these figures would need to be added the weight and cost of the mounting and transmission, plus account would need to be taken of the slight power loss from the gear/belt drive arrangement.

Using a gear/belt total reduction ratio of 2:1, the motor cluster output shaft would spin at about 2500 rpm maximum.

7kW+ from a unit weighing less than 5kg at a cost of maybe £300 - £350 or so (complete with controllers) seems like a reasonable deal to me, if it works as well as I think it might.

If the idea works on the bench, then the next stage would be to switch to gear drive and get some cases CNC machined up. If I can use off-the-shelf standard parts, then it should be fairly straightforward to make self-assembly kits for anyone that wants to have a go with a similar set-up. The only slight problem might be with motor mounting holes, as I don't think there is a standard for RC motors. This might well create some interchangeability issues if people wanted to use other motor types.

Next step is to order some motors and controllers to see how well they work in this application.

Jeremy

hyve
Posts: 60
Joined: Sat Oct 06, 2007 1:35 pm
Location: Chesterfield, Derbys.

Postby hyve » Fri Aug 22, 2008 9:02 am

I will definitely be watching to see how this goes, Jeremy. I've been wracking my brain to find a reason why this is not a good idea and apart from complexity (quite a good reason, actually) it sounds very promising. The performance levels of these little motors are so far beyond the big old dinosaurs that we generally have to use and their price per kw so much better that it seems this idea just has to be tried. We should also not forget that most ICE's are similar in that they use a number of small pistons instead of one big one. This way of doing things has plenty of precedent.

My initial thought was that very high performance machines rarely suit conversion to longer term use. You can bet that all these Westfields with motorcycle engines in will have a pretty short life, for example.
But what is there in a brushless motor to have a short life other than the bearings ? Since these normally have to handle propellor axial thrust they must be reasonable.
You've already covered the cooling issues; reduced mass for heat absorption also means more rapid cooling - not a disadvantage in my book, especially if, as you suggest and I support, a seperate temperature regulated cooling fan is utilised.

Using a sprag clutch as a safety device also aids redundancy when running at low power, another of your justifications for the layout. Ondrives make a range of these too, but from memory they do not permit very great speed differences and this may prove a problem. Assuming you really think they should be included.

One small difficulty I see is drive from the plain motor shafts. I note in the advert on the website you linked that the motor comes with an adaptor to take the propellor, which is locked to the motor shaft by a few grub screws !!
OK for an airscrew, perhaps but I don't think this would do for road vehicle drive.
Again, as you point out it will do for a test model. If there were to be any number of such units built the quantities would justify asking the motor maker to roll a spline onto the shaft at manufacture. Since for a car I can see around 15 motors per assembly, the volume could soon justify this.
Incidentally with a motor diameter of only 63mm, 15 in a circle with space for a couple of halfshafts results in an overall diameter for the assembly of only 390mm approx., and only 100mm or so thick, including gear assembly. Not bad for 33kw continuous, including gearbox !

The potential problem of variable controller output: ought it not be possible to modulate the signal from the throttle pot ( or load cell, an idea suggested by Brian Gregson ) where it enters each controller, according to controller output ?
It sounds like a microprocessor job to me, taking it's cue from the throttle signal and ensuring every controller is delivering identical power. But I'm a total amateur here.

One way of testing whether the problem exists I suppose is an ampmeter across each individual motor supply. Could gear backlash or spring in the belts cause small loading fluctuations which might run into a harmonic in the power supply ? Creating only a noise problem, perhaps at such low power, but you do have to try to think of everything.

As I say, very best of luck with this Jeremy and go to it !!
Peter Ph

User avatar
Jeremy
Posts: 472
Joined: Thu Jul 31, 2008 6:35 pm
Location: Salisbury

Postby Jeremy » Fri Aug 22, 2008 10:46 am

All good points, Peter, many thanks for the useful contribution.

The shaft problem can probably be overcome fairly readily, I hope, by using one of the proprietary clamp-type fixings, although the motor bell itself it held on to the shaft by grub screws. I've discovered that the shaft is just a plain, hardened, steel one that goes right through the motor. It can be slid from one end to the other by loosening the screws.

Although a potential area for unreliability, it seems that this fixing method works well in practice. This is most probably because each motor isn't producing very much torque, compared with the sort of figures that we are used to, plus these motors are so light that the loads on the shaft from rapid changes in rotor speed are quite modest. I've no doubt that reliability might be improved by providing small locating dimples in the shaft and perhaps loctiting the screws.

The discovery that the shaft is just a plain steel one opens up another possibility, that of building an axial array, rather than a radial one. There are limits on the torque that can be safely transmitted by a small steel shaft, but having just done some quick calculations this limit is very high. In practice, the shaft diameter doesn't seem to be as much as an issue as the allowable torsional deflection. With three of these cheap motors connected to a common shaft, all running at maximum torque, the torsional deflection at the output end would still be less than 0.02 deg; the torsional stress in the shaft is so low as to not be worth worrying about.

Although the axial array isn't as neat in terms of packaging, in that it doesn't lend itself to a short, pancake style configuration, it does mean that the coupling losses between the motors would be reduced. In effect, the unit would become one long, thin, motor, with an output speed of around 5160rpm maximum from 24V.

Cooling an axial array might be simpler, as it should be possible to simply duct air down through the motors, perhaps from a central feed point, exhausting at either end. I'm not sure if the temperature difference this might create from one end to the other would be a problem or not. My guess is that it could be managed by careful design.

Switching to an axial array makes it worthwhile to investigate some of the bigger motors again, as there will be a saving in transmission component cost and weight. It might be worth using a smaller number of larger motors in this sort of configuration, particularly as the larger motors tend to have 10mm or 12mm diameter shafts.

It looks as if I will have to do a bit more research into what's available before I commit to spending some money. An early find are these motors, either of which may be possible candidates:

http://www.allelectricrc.co.uk/a705006-841-p.asp (perhaps a bit expensive, but of known good performance)

http://tinyurl.com/6bmwd2 (much cheaper and possibly a better bet)

Both of these motors seem to have a remarkable performance for their size, weight and cost.

Jeremy

MalcolmB
Posts: 423
Joined: Sat Jun 23, 2007 8:07 pm

Postby MalcolmB » Fri Aug 22, 2008 12:28 pm

Being able to mount several motors on a common shaft sounds like a big advantage to me, just from the point of view of simplicity and reduced weight. You could even combine motors in a multiple axial/radial array if you wanted to restore some complexity :)

You might need some intermediate bearings to support the shaft between motors, but I guess you've already thought of that.

I've had an eye on those HXT motors for a while. I've been trying to find some justification to buy one.

I look forward to seeing how this turns out.

User avatar
EVguru
Posts: 305
Joined: Fri May 25, 2007 5:17 pm
Location: Luton
Contact:

Postby EVguru » Fri Aug 22, 2008 1:04 pm

I wonder what the life expectancy of these motors is.

500 hours would be a lot of weekend flying, but not all that much driving every day.

Persuading a University to put one through a driving cycle simulation on a dyno would be a big help.
Paul

http://www.compton.vispa.com/scirocco/
http://www.morini-mania.co.uk
http://www.compton.vispa.com/the_named


Return to “General”

Who is online

Users browsing this forum: No registered users and 33 guests