Are these servo/steppers appropriate?(Fastech EZI)

1:1 wrote:

www.fastech.co.kr/English/Ezi-SERVO.html
Particularly the NEMA34 '86' series steppers - specifications below ... They can go up to 32,000P/R res which can be divided down to 500

Bear in mind that 32,000 ppr at 1000 rpm is 500kHz, you won't get that out of software stepping, but can with external hardware.

- They also claim some impressive advantages over standard stepper drive system, or so they have led me to believe - maybe as a beginner I have been bamboozled by jargon

They seem to be using a stepper motor with current control and encoder feedback. People have tried a similar approach with EMC2 (but without the active current control) with mixed results (it is possible to configure a step generator in "velocity" mode and run it closed-loop as a servo if you fit an encoder. The main difference in that case is that all you can do if you miss a step is speed up, and that is no help with a stepper. With active current control you can increase the current at the same time, which is more likely to work.

I'd use them firstly in a mill/drill CNC conversion, where they may be over specced (?) - and over priced (?)

Not especially over-specced, no idea about the price.

I imagine a breakout board would be required, which might be cheap enough, but what about 40~70v power supply per axis for 4 axes ?

You can probably make a BoB, or do it all in the cabling. If the PSU is big enough then all the drives can share it. (try eBay)

They'd run open loop with respect to EMC ? but closed loop internal to themselves ?

Yes, though the drives have encoder outputs too, so you could close the loop in emc2 too. It would not give you any more control, but would allow for accurate following-error detection.

1:1 wrote:

The big NEMA34 motors and drivers giving 6 N.m of torque up to 1500RPM are about $500 US each

That is expensive. You can get conventional servos for about that price
Compare to www.dmm-tech.com/Products.html
Those will take step-dir or 0-5V input too, approx the same torque (they will have a much lower rate of torque drop-off than the steppers, so can be geared down and run faster. Toothed-belt drive solves a lot of coupling/mounting headaches too)
They are not the most EMC-friendly device (you might end up adding your own incremental encoders) and I don't know of anyone using them though.

For tight EMC integration, look at servo drive systems from Mesanet.com and pico-systems.com and servos from eBay (or bought conventionally if you have the budget)

I

should say I don't know control theory that well, but having a loop within a loop sounds um, interesting - are there any redundancies or even possible failure states doing things that way ?

It is actually very common. For example a car engine might use a sensor to measure exhaust oxygen, use that and a PID to control the rate of EGR. The rate of EGR requested will feed into another PID that controls the EGR valve command position. The EGR valve command position will feed into a PID that controls the EGR valve motor velocity, and that in turn might feed into a PID that controls the motor PWM rate…

Does that prove fastech is just an expensive stepper? or is it quite good?

It is reasonable, the dmm-tech servos actually have a very similar spec (which seems odd)

I picked up three MDSKSRS036-23 motors eBay for UKP50
www.lenze.com/lenze.com_en_active/images...Catalog_Lenze_en.pdf

They have a torque curve flat to 3000rpm. The drawback is that they use resolvers for feedback. I have found a way round that, but it is not for the faint-hearted.

My problem with dmm-tech is that drive setup and tuning is windows-only and that there is no encoder feedback back to EMC2

EMC2 works best when it is in charge of the servo loop (it is a more advanced controller than most, has a lot more CPU power available, andHalscope lets you see exactly what is going on at all the points in the chain). EMC2 really wants totally dumb drives.

1:1 wrote:

-Does 1 parallel port have enough I/O for EMC2 to run 4 closed loop axes ?

4 axes is 4 x 3 channels for encoders, and 4 x PWM (and that is stretching it, using centre-biased PWM rather than PWM and direction, and no amp-enables).

I think it is _just_ doable, but wouldn't recommend it.

Mesa sell the 7i43. That converts your pararell port into 48 channels of IO, does the PWM, encoder counting and step-generation (if required) onboard. All for $79. The driver is well supported in EMC2 (in fact a new feature, support of the 8i20 2.2kW servo drive is happening right now).

For some other non-CNC projects I'd like my motors to run quite high resolution encoders, but for EMC2 I'd need a driver that can divide these before they hit the (likely too slow) PC ?

Do the mathematics, based on a base thread of 20nS, you might be surprised how fast you can count.

What are the best 'totally dumb' drives ? Like which Gecko models ?

I have not looked too hard at Gecko servo drives. I know a fair bit about the Mesa 7i39 (I wrote the EMC2 driver). It depends in if you are using DC (brushed) servos or BLDC servos. I don't think that Gecko make any BLDC drivers.

and is there a good 'these are the best servos to mate with this Gecko drive' website ? (why don't Gecko sell motors also?)

Brushed DC servos are easy to drive, and any motor will work with any drive, basically. If you get the power/current/voltages to far out you will end up with a weak system and/or melt things, but it will work.

BLDC is a bit more tricky as there are a lot of possible relationships between the Hall sensor signals and the correct stator excitation pattern (64, I think). Manufacturers tend to deliberately switch things about so that their motors only work with their drives. (The EMC2/Hostmot2 BLDC_Hall driver lets you try all the patterns until one works, the BLDC_Sine one uses the encoder instead)

1:1 wrote:

Maybe EMC2 isn't for me ?

It's possible, but let me show you two videos of very, very different things that EMC2 has done recently:



www.youtube.com/user/77dab77#p/a/u/0/4vd_S6mgyXk