by JoeC » 21 Feb 2012, 20:26
There are two major sources (that I can think of) why the powerchair motors have poor efficiency when lightly loaded; that's to say that they require a lot of power to spin freely and not drive the chair.
The first source is the gearbox. Powerchair manufacturers need to provide a cheap, reasonably robust way of stepping down by ~20:1, and they often require a 90 degree turn in power transmission. Most of the chairs I've seen in the US have a helical/worm gear making the first stage of reduction, coupled with straight spur gears for the second reduction. I believe that the first stage of gearing has a lot of sliding friction, and is not very efficient at any level of loading. The UK sunrise F55 uses a single stage worm reduction that is apparently somewhat more efficient, at least at some levels of loading. Certain US Quickie chairs use other methods- the Quickie P200 used a two stage inline spur gear reduction, and in my experience was significantly more efficient than most chairs with right-angle gearboxes. The P222 used a single stage bevel reduction (Cylkro, as built by ASI Drives), and this chair seemed to work extremely well for having a 65 amp brushless drive.
The second source, I think, is in the motor itself. The condition of moving along lightly loaded isn't as common for powerchairs; much of the time they're starting or stopping, turning, or compensating for obstacles in their path. Efficiency near the top speed just isn't something that the manufacturers both to optimize. I'm sure there are plenty of eddy currents in the armature, as well as resistive losses. Powerchair motors often have resistances significantly higher than 0.1 ohms, and many are above 0.2 ohms. A much better motor of similar size and speed at a given voltage could have less than half the resistance.
Another thing that you may not have taken into account is the parking brake. These are always energize when the chair is driving, and they can draw from 0.5 to ~1 amps each, and the chair has one on each motor. These are built as a solenoid holding a sprung plate off of a friction disc. If something with (for example) a dog clutch were used, or something else with more complex mechanics, a much less power hungry solenoid could be used. I don't think it's worth pursuing, but if you want to have a good look at no-load current draw, it must be kept in mind.
Sam, if you're a retired software engineer, you really might have a lot to offer in terms of turning a Roboteq controller into something a little more "civilized" for wheelchair use. As far as we can tell, almost all of the useful safety and convenience features of mobility controllers can be emulated by writing some of their "micro basic" scripting and wiring a few external switches and buttons. If you have any interest, and your software experience matches that, I'm sure that a few people around the board would appreciate your input.