Squirrel-cage induction motors are obsolete except in environments where weight and size is of no concern; BLDC motors are now superior in every way, often by two orders of magnitude.
I have a ¼-horsepower 1430-rpm electric motor here, adequate for running a grinding wheel or something through the V-pulley it’s got stuck on it. I think it’s originally out of a washing machine; I bought it used. It’s got a big capacitor wart on the side, because it’s a single-phase motor, and that’s a much less inefficient way to get a single-phase motors to start up than the hack in a shaded-pole motor. I haven’t opened it up, but I’m pretty sure it’s a squirrel-cage induction type, because it says “1430 rpm”, and does seem to run at a pretty fixed speed. It’s got cooling slots in the ends so you can blow dust into it.
It’s rated for 2.4 amps at 220 volts, which works out to 530 W. ¼ hp is only 186 W, so presumably 2.4 A is the current it draws if the spindle stalls; these motors aren’t that inefficient. I haven’t measured either its power input or its power output.
I don’t have a scale handy, but I feel like it weighs about 10 kg, and it’s 150 mm in diameter and about 200 mm long.
The thing I want to point out here is that this is not a very good power density, on the order of 20 W/kg and 50 W/ℓ.
I was just watching a YouTube video by Stuart de Haro entitled “Milling machine anatomy”, which is largely about the Bridgeport Model J head for Bridgeport milling machines, the most popular milling head for hobbyist metal milling in the US. We’re talking about a floor-mounted mill with an X–Y table a couple of meters long, capable of precision on the order of 25 or 50 microns. (In this case it’s equipped with a DRO that reads in microns, or 2.5-micron units in medieval-unit mode.) It’s got this big honking motor on the back, hooked up to a variable-frequency drive, bigger than the motor I’ve got here, and my man Stuart explains that the Model R ½-horsepower milling head is “relatively light duty” (and for that reason he’s never actually seen one, though he teaches machining at a college).
So we can deduce that, for Stuart and other Bridgeport machinists, one horsepower (746 watts in modern units) is adequate for the meter-scale workpieces they like to chew on, but half a horsepower (373 W) is “light duty” and would slow them down. And that, I suppose, is why they’re willing to deal with these heavy honking motors on their milling machines.
Modern electric quadcopter drones result from truly astounding progress in battery and motor technology. On MercadoLibre here in Argentina, I find a Turnigy BC2836-8 motor designed for quadcopters for AR$2090 today (at AR$44.50/US$, that’s US$47.) This motor weighs 70 g, measures 28×28×36 mm, wants to be driven with a 30-amp or 40-amp “ESC” (the kind of VFD you use for a BLDC motor), is rated at 1100 “KV” (rpm per volt), and is rated at 336 W, intended to be driven from 2–4 LiPo batteries. (A Singaporean vendor lists the same motor at US$13.)
That is, this motor is almost twice the power of the big honking 10-kg beast I have here in my living room. But it weighs a bit under 1% of what the big motor does. I’m guessing it costs about 10%, too.
Four 3.7-volt LiPo batteries in series would be 14.8 volts, and 336 watts at that voltage would be 22.7 amps, so the volts and amps pretty much check out. At 1100 “KV”, its maximum speed should be a bit over 16000 rpm, which is pretty plausible.
So this works out to 4800 W/kg and 15000 W/kg, about 200–300 times better than this big motor.
In part this is made possible just by running the thing eleven times faster, which is made possible by having much better bearings and designing the thing to depend on an ESC. At a given torque, running the motor eleven times faster is going to give you eleven times the power. Permanent-magnet brushless “DC” motors like this one (“BLDC motors”) also use rare-earth magnets, typically NdFeB, which gives them higher field strength and thus higher torque — although in theory it’s possible to achieve similarly-high flux densities in induction motors. (The “electrical steel” used in squirrel-cage cores saturates at 1.6–2.2 T, while NdFeB’s remanence is “only” 1–1.3 T.) Another significant factor may be cooling: the drone motor is of course designed to operate in the propeller downwash, which is a wonderful level of air-cooling. Finally, the drone motors are presumably designed for an MTBF of tens of hours, while the ¼-hp motor is designed for an MTBF of tens of thousands of hours.
BLDC motors can also maintain the optimal phase relationship between the rotating magnetic field applied to the stator and the magnetic field of the rotating rotor, enabling them to maintain the same torque at any speed; induction motors, by contrast, have a fixed torque–speed relationship which reaches zero torque at their natural or unloaded speed.
Permanent magnets are much smaller than field windings, which exist in the induction motor in the form of the copper “squirrel cage” within the rotor. This may account for a substantial fraction of the mass and volume of the motor, though far from 99%.
Neodymium magnets also have about an order of magnitude greater resistivity than iron does (1.1–1.7 μΩm, compared to iron’s 0.1, though electrical steel’s resistivity is higher than that), which might diminish eddy-current losses, and of course there are no hysteresis losses; although these are more properly efficiency concerns rather than power-density concerns, they do have some effect in that more power losses result in more necessary cooling.
Sometimes the low Curie temperature is cited as a disadvantage of neodymium magnet motors — it’s only 310–400°, so the magnets will be destroyed if the motor ever overheats that much. However, I think the solder joints and winding insulation will fail at a lower temperature than that.
So how could you build a milling machine or engine lathe with these little motors? Well, you probably need several of them, such as ten of them. (You don’t want to be running the motors at their maximum power if you want them to last; as I said above, they aren’t designed to last.) You’d need to gear them down. Our homeboy de Haro tells us that the spindle speeds in his classroom range from 50 rpm to 5000 rpm; trying to run that off this wimpy ¼-horsepower squirrel-cage motor I have here could require gearing it up by a factor of up to 3.5, or down by a factor of almost 30, although in actual fact you would probably use a VFD.
By contrast, you’d probably want to run the drone motors faster and gear them down. Maybe you could run them at 2000–6000 rpm, for example, although that difference in speed is less than I expected: you’d be gearing them down by factors of only 1.17 to 40, probably. You could imagine, though, that you’d want to take advantage of the drone motors’ higher speeds to support higher milling speeds for small cutters.
Dremel-style small-cutter milling is potentially really interesting, especially coupled with the larger number of axes of control that smaller, cheaper, higher-power motors and control systems make possible. The conventional CNC way to mill weird shapes — shapes that are far from the shape of the stock they’re milled from — is to convert most of the stock into chips, a slow and expensive process that becomes somewhat faster if you apply a large roughing milling cutter and a high-powered motor to the task. Smart machinists, and all but the most stubborn manual metalworkers, often use slitting saws and bandsaws to quickly, but imprecisely, remove much of the material ahead of time, thus avoiding much or all of the milling. With a six-axis machine, though, you could do this step under CNC control, using not only slitting saws but also narrow endmills or even drill bits or wire saws. Narrow endmills demand potentially much higher spindle speeds to reach the same surface speed, which is necessary for optimal tooth life.
Narrow endmills suffer from extremely low rigidity, resulting in chatter, imprecision, and potential breakage inside the part. Tapered endmills improve this situation dramatically; they already exist and are occasionally used in CNC milling even on three-axis machines. ConicalEndMills.com suggests using them for “draft angle & chamfer machining in all materials,” for example. But maybe they could be much more widely applicable with high-speed motors and five- or six-axis control.
Wikipedia tellls me that the obstacle to even higher power densities in small BLDC motors is commonly heat dissipation: the motor must be small to spin fast, but this limits its available cooling surface area. Permanent-magnet BLDC motors put the windings, which are the primary heat-generating part of the motor, on the stator, which has a solid connection to the outside world — this facilitates getting the heat generated within them out, as well as getting the prodigious amount of current they use in. However, most of them still rely on air cooling at this point, especially in the realm of quadcopters, where unobstructed high-speed airflow is not only guaranteed by construction but kind of the whole point.
Heat pipes are one possible way to improve the situation: you can run heat pipes through the stator to get heat out quicker. Heat pipes, unlike thermal conduction, can transfer heat at a rate that does not diminish with distance.
Another alternative, though, is forced-fluid cooling, in which a coolant fluid is pumped through channels in the stator to transfer heat out. Air is one common coolant (and, as I said above, in common use for these motors) with many advantages but also some drawbacks — its heat capacity is orders of magnitude lower than other viable fluids. (See Coolants for a survey.)
One of air's key advantages as a coolant is its low viscosity, which enables it to cool even long, thin channels effectively. But, by adopting a fractal geometry for the cooling channels similar to that of vertebrate circulatory systems, we can enable the rapid, efficient circulation of even fairly viscous coolants. See Heat exchangers modeled on retia mirabilia might reach 4 TW/m³ for more details and further applications.