ball screws or lead screws

I've been planning a second printer, and am actually torn between belts and core-xy (which maximises speed) or running screws.

* A 20t GT2 belt will be 40mm/rev.
* TR8x8 (4-start 2mm pitch) leadscrews used by prusas for the z axis are available, and are 8mm/rev.
* 1204 ballscrews can be found on aliexpress for not much money - these are 4mm/rev but have the advantage of low friction and negligible backlash
* Dodgy M8 thread from the hardware store is 1.25mm/rev

If you want to run your machine at 100mm/sec (which is probably as fast as I'd really like to print, but maybe you might disagree) then:
* belts = 150RPM
* TR8x8 = 750RPM
* ballscrew = 1500RPM
* M8 thread = 4000 RPM

Now carl is right in that it's common practice to assume that typical printer stepper motors pack it in at 1000RPM - but this assumes you're running at 12V. If you up the voltage to 24V, or better still 30 or even 36V, the torque curve at high speeds gets a whole lot better, and speeds of up to 2000RPM should be achievable in a reasonably low-friction system (e.g. ballscrews... see where this is headed... ).

The regular threaded rod would be pretty useless.

What you will probably lose is acceleration: the limitation becomes the intertia of the motor rather than the weight of the carriage.

Steprates (limited by controllers, particularly the 8-bit arduino variety) can be kept in check by reducing microstepping on the drivers. This would bring both steprates and resolution back to comparable with the belt (e.g. 2x ustep on 4mm/rev would give a resolution of 0.01mm and steprate at 100mm/s of 10kHz, while 16x ustep on 40mm/rev (typical of belt setups) would be 0.0125mm and 8kHz). Using a faster controller (e.g. smoothieboard, azteeg x5, etc. with their 32-bit 72MHz+ LPC's) is now always an option.

Another thing preventing wider uptake is probably the cost, and also the design and assembly/alignment/maintenance required to get them to work well - a badly done leadscrew arrangement will be barely functional (think solidoodle z axis... urghhhh). We're not all toolmakers/engineers (and neither are half the 3d printer manufacturers).

tl;dr
I agree they're slower, but "too slow" - I'm not so sure about that.

It depends on your board. My Azteeg X3 Pro for example has seperate power inputs for the logic, the steppers, and the bed all rated at 12 to 30 volts independent of each other. Other boards like RAMPS have two inputs but they are just parallel on the board and not seperate.

To view the schematic for the Azteeg you need Diptrace installed which is actually a very nice PCB design suite for free. Roy for some reason made the schematic in a format only that program will read. Maybe I will do some screen caps of the three pages.

That's more or less the idea right there.

* 100uF cap should ideally be per driver, and close to the pin as you can get it. Voltage rating should be well above the VMOT voltage (35V and 50V caps are common - if you're doing 24V, either is fine).

* Ground should be shared, and shared well!! Tie the -ve terminals of the power supplies together ideally, and leave all the pololu ground connections as they are (ground path will then be back through the RAMPS and to the main PSU, as per normal). You only need to modify and hook up the VMOT pin to the 24V +ve. I think the pololu modules share ground anyway, so trying to do anything sneaky is most likely going to make more noise.

* Do it to any motor that might need more torque at speed - X and Y certainly, but you might want to do Z if you have a fine Z thread and want to drive it faster, and E if you want slightly faster retraction on the extruder. Basically it might be helpful if the motor ever moves over 500RPM or so, under that there's not much point!

Not a bad idea - a bit less efficient than a second PSU potentially, but not that we really notice.

I'd guess that for each motor:
* The peak power is the motor driver current (maximally say, 2A) at the supply voltage (24V in this case): so P = V * I = 48W.
* The stationary power is the driver current through the winding resistance (e.g. 1.3ohm for this motor), so P = I^2 * R = 5.2W.
* The average working power will be somewhere in between those two. This could be estimated by measuring the supply current while printing I suppose - of course that would involve building it first haha.

The item you picked, at what looks like from the description to be 90W continuous and 150W peak, looks good for at least 2-3 motors IMO, with some careful experimentation you might be able to get away with more or a smaller converter. Especially if you stick to doing just the X/Y to start with can't see what would be wrong with it.

EDIT: above is probably wrong - see below

This is all very theoretical, I'm kind of relying on you to try all this out and see how it goes!!