Re: [CAD_CAM_EDM_DRO] How do you compare steppers to servos

On 26/02/11 04:04, Danny Miller wrote:

> Steppers don't have "springiness" like this. They lock on a step pretty
> tight.

Do the experiment I described. It is a matter of physics. Magnetic
fields are not solid. If you do the experiment, try it at one full step
and one half step. You will notice the stiffness varies quite a bit.

> A servo, if you put an arm on it, is probably less capable of holding an
> exact position, as the encoder would sense motion and try to drive in
> the opposite direction to correct it, but it would be confused as to how
> MUCH to drive it.

Not really. The drive system knows that if the motor is only one encoder
count out, it probably doesn't need to apply much torque to return to
center, whereas if you are lots of encoder counts out then you need to
apply more power. Integral helps overcome steady state loads.

> In fact servos are NOT 100% step-by-step accurate on the counts, as the
> control system must deal with issues of inertia and allows some leeway
> for correction.

This is true for both steppers and servos. As soon as a stepper has to
generate torque it will move away from the ideal position.

> IIRC, the servo system is inherently less accurate in
> this regard, but not on a level that matters for a router, really.

If you compare a 200 step stepper with a servo that has a 200 count (50
line) encoder then your statement is correct. However if you have a 2000
count (500 line) encoder then the servo will almost undoubtedly be
stiffer than the stepper.

> In any case, this is not how drives are characterized, by
> "springiness".

The correct term is stiffness and it directly affects accuracy so yes it
is important. Much of servo tuning is aimed at maximizing stiffness.

> There's primarily just torque and speed issues. Step
> accuracy typically does not come up. Even if you were to direct-drive a
> 2 TPI-lead ballscrew with a stepper, 200 steps/in is 0.002" accuracy
> which is pretty good

Actually for a 2TPI screw you will have 400 steps/in or 0.0025"/step.
for a lot of machines, 0.0025" is totally unacceptable. For a wood
router it is acceptable. It all depends on the end use.

> - and that doesn't even consider microsteps yet!

This is a common misconception. Microsteps improve smoothness but have
little effect on accuracy. The only way to change the stiffness of a
stepper is to increase the effective strength of the magnetic field.
This can be achieved either by increasing the size of the motor or
increasing the current.

> The closed-loop behavior of servos is great for reliability; stepper
> stalls are awful- but a properly tuned system will NOT stall (or lose
> steps).

While this is true in theory, getting the stepper system tuned correctly
is a lot harder than most people think. I do a lot of repair work on CNC
machines. The vast majority are servo but of the stepper machines, many
do sometimes lose steps. It may be only a few times a year or in the
case of one particularly horrible machine almost every day but it does
happen. This is on industrial kit that is presumably properly tuned. The
problem with tuning steppers is that the limit is a brick wall. Exceed
that limit by the tiniest amount and you lose a step. For reliability
you need to allow a huge amount of overhead to allow for unexpected
loads, cold lubricant etc. Servos are a lot more forgiving in this respect.

> I've heard the low-end torque (and thus acceleration) of steppers tends
> to be superior, but the servos spin faster. Of course, as long as it's
> not direct-drive, gear ratios can change things, but the fact remains
> that the servo's power will always be biased in favor of the high speed
> range, whereas steppers are biased in favor of the lower speed. Either
> one can cause design problems for the device, but that's just part of
> the design process.

Stepper torque falls of fairly sharply with speed while servos are
fairly linear. This means that a stepper rated to provide adequate
torque at high speed will have an excess of torque at low speed. For
pure acceleration at lower speeds, steppers are generally better as they
usually have a very low rotor inertia. As you say, matching torque to
the load is part of the design process.

Les