I've been watching other build threads on DIY brushless attempts.
Instead of tacking onto others' threads, I'll do my own here.

I would like to achieve a wide-range, high speed spindle, up to 30K RPM sounds good.

I'm playing with ScorpionCalc: Scorpion Calculator (SC) - RC Groups
Which is supposed to be pretty good. We don't care about prop selection, so I have to tweak the prop diameter and pitch manually to achieve the desired watt loading to read the estimates of motor performance.

I think any reduction gear IS a huge mistake, and defeats the purpose. And it's unnecessary- these motors run just fine over a wide speed range!

The DIY brushless motor could probably compete well with the $300 2.2KW Chinese spindle + $250 VFD. That's a lot of cash. And the Chinese spindle doesn't have consistently good bearings- in fact, for the size they use, they're just not rated for high RPM. Even if you buy your own replacement precision bearings, I don't think they'll be capable of all that much more speed. WE can build a spindle with a cheap, blank ER8/ER16/ER20 precision shank where the shank diameter is lower. The smaller the bearing's diameter, the higher RPM it handles. I don't think a wood router, or even aluminum-cutting, spindle will have problems with a 12mm spindle- just my guess. We can get high rpm steel, even ceramic, bearings at a decent price in this diameter. Plus, the 2.2KW spindle is a 220V motor, which generally requires a 220VAC VFD, but AFAIK a router won't even draw more than a 110VAC outlet can do anyways. Not everyone has 220VAC available.

There are a number of design issues here of major note:
System voltage:
Voltage drop in the supply wires. If it's a router, this is pretty major because a 15 ft run is quite possible. If it's a low voltage system, like 12v, then the drops are problematic. Most RC systems use silicone wire because it won't burn up- but a 15 ft run of 14ga wire is 0.076 ohms. At 1/2hp, the current is 31 amps @ 12v, and 2.35v is lost in the wiring, which is FAR too much. If we go with a 48V system, 1/2hp is only 10.3amps, 0.783v dropped, which is only 1.63% of the source voltage.

There's several switching power supplies on eBay for 400W-500W in the 24V-60V range. I'm just not sure if that's enough. A switching supply won't tolerate overcurrent for even a second, that'll cause a foldback which could stall the motor entirely. The CNC software has no capacity to slow down the cutting if it's drawing too much power. But Taig's motor was only rated for 1/4hp- 186.5W.

Most of the RC controllers have a Vmax of 30V at most (~24v systems in practice). It'd be hard to find a 32v/48v/60v sort of controller, but they're actually not wildly complicated to build. In fact, due to their open-loop nature and other reasons, I'm pretty much "off" the idea of using an RC controller, period. I'd DIY a 24v controller if I had to.

Motor speed KV rating:
The KV rating of the motor is the RPM per volt when unloaded at 100% duty cycle. Yes, the units are totally nonsense ("K" means 1000, and it should be "/V", it should be "RPM/V"). 30K rpm on a 48V system would be 625KV, which is a purchasable item. You'd want to go a bit higher to ensure it can reach that speed under load, supply wire drops, controller drops, and winding resistance reduce the effective voltage by a bit under full load.

No-load current:
This looks like the dominant cause of motor heating, which has nothing to do with cutter loading.
This one is throwing me for a loop as far as understanding and getting data on. As far as I can tell, this is the current is supposed to be mostly constant over the voltage and RPM range- that is, if it's got a 1A no-load current on a 700KV motor at 14.3v=10K rpm, it requires 1A from the supply, which is 14.3W wasted as heat in the motor itself (pretty significant). At 21.4v=15K rpm, it's still wasting 1A, but that's 21.4W now.

Except that approximation is flawed- At higher RPM, the no-load current is higher. Unfortunately, almost all eBay suppliers who are so kind as to list a no-load current don't list the voltage it's supposed to be from, and the voltage the motor's designed for is a wide range.

Now here's something I'm looking at. On eBay, the powerful EMP3548 comes in multiple KV ratings. Same case design, more or less "the same motor" with different KV for comparison.
The 790KV is an Io of 2.1A. That's huge, and problematic for motor heating. That's 26.6W of heat to maintain 10K rpm unloaded.
There's a 1100KV, that would require less voltage to the wattage lost should be lower. But the Io is HIGHER, 3.5A. So at 10K rpm there's 31.8W of loss. again I don't know where these no-load Io numbers were measured so it's not conclusive.

What that's telling me is whether you go for a low-voltage, high-KV rating or high-voltage, low-KV system, the motor heating is similar.

Running the motor at lower speeds and actually using a STEP-UP pulley to get a higher spindle speed would actually make the spindle idle much cooler. But, this does exponentially increase resistive losses. The 790KV motor at 10K rpm direct-drive at a 1/4HP load would need 12.66V 14.7335A. The winding resistance is 0.048ohm, for 10.4W of resistive heating on top of the 26.6W Io loss. Actually much less than the Io loss. 37W total loss. So it's not actually gonna get all that much hotter when cutting as just spinning there.

If it were on a 2:1 step-up, it would halve the Io loss to 13.3W, but double the current and quadruple the resistive loss to 41.7W, 55W total.

If we used a 1:2 step-down, Io loss alone is 53.2W, resistive heating is only 2.6W, but the total's still 55.8W, quite high.

There's this glorious huge, low-speed beast for contrast:
http://cgi.ebay.com/C80100-KV130-EMP...#ht_3555wt_927
130KV, but a 6.5KW motor. That'd require 77v to get even 10K rpm, and I'd be afraid to put 230V on it needed to reach 30K- I doubt the winding insulation could even handle that. It's got an Io of 2A, which doesn't sound that different, but that's deceptive- 2@77v no-load would be 154W to idle at 10K rpm. It's a big motor and can dissipate more heat to stay cool, but that's a LOT.

Seems to me like the no-load current loss is easily a dominant form of heat generation. Changing the motor's KV doesn't seem to change the situation, but rather, the idle heating seems to be directly related to the maximum capacity of the motor. Bigger isn't always better. The 6.5KW monster above takes a buttload of power to keep turning, even if not used, and its low resistance windings don't even matter there. More weight in iron core, more iron core losses.

Heating at idle is not entirely a problem, as long as the motor's temp is within range. It seems like the bottom line is they're gonna take a lot of power to idle in a powerful motor, it's gonna get warm, and no gearing can fix that. But again, this just isn't a deal-killer as long as the losses are calculated and within spec. I don't give a rat's ass how much wall electricity is wasted as long as it works well and works reliably.

Well I'm a bit confused here. More technical notes to follow as I work it out.

I'd REALLY like to know how many watts a router motor actually cuts with on a router hogging out oak or whatever, and at what RPM. This would give me a great deal of insight to calculate from. From what I can tell, the 2.2KW Chinese spindle never actually uses anywhere near that, and I may be wasting my time trying to calculate and figure out the power supply for even a 1KW spindle. Like I say, Taig is only a 186.5W motor- a 500W drive motor sounds like quite a hogging beast, relative to that.