servos vs steppers

Hi, I noticed in a few places that the plan was to move from stepper motors to servos. I was curious what the advantages of doing so were and why? And by servos do you mean like are used in hobby RC craft or something else? is a repraped geartrain needed for such a switch?

The main advantage is cheapness. Servos are a lot cheaper than stepper motors. By servos, we mean continuous rotation servos, which maybe used in RC airplanes, cars, etc. A repraped geartrain may not be needed.

Servos also open up some interesting possibilities, a servo is nothing more than a motor hooked up to a measuring device. I believe it is feasible for reprap to make motors using threaded rod, magnets, copper wire, and ball bearings. If we can get reprap to work with servos, then we can get reprap to work with motors it made itself.

The main problem with servos is the inertia of the rotor. I used magnetic encoders plus cheap gearmotors from Solarbotics to run my Tommelise 1.0 repstrap. I found that they worked fairly well for printing straight lines of extruded plastic of 5 mm and greater length. If you needed to print very short line segments to emulate curved surfaces like, for example, gear teeth, however, you had to slow down the print speed enormously in order to avoid overruns caused by the internal inertia of the gearmotors. With the solarbotics motors I could max out at about 2.5 mm/sec on a single axis with straight line segments. When I tried curves like gear teeth, however, I was down to 0.5-0.75 mm/sec before the distortions got managable.

The problem with printing at very low transition speeds is that the extruder has a hard time extruding at those very, very low rates and the hot extruder tip tends to melt whatever it is that you are trying to print. It was very frustrating and caused me to set the idea aside after many months of trying to make that scheme work.

Actually, servo control is considered the more sophisticated method of control. Stepper motor control is open loop, in that you tell the motor what to do, and hope it does it.

Servo control is closed loop, in that you continually measure the motor response and feed back accordingly. In the CNC world, servo motors tend to be many times faster than stepper motors. They also require more sophisticated controls as well. Feedback circuits usually implement PID control (proportional, integral, derivative) to very precisely get the motor to go where you want it to go.

Usually when talking about improving from stepper control to servo control, RC servos are not what is being talked about. The motors are usually standard permanent magnet dc motors at the low end, but go to coreless 3 phase ac servos at the high end for speed of response and low inertia. You ususally find these in industrial robot drives, hooked up to harmonic gearboxes.

In the most straightforward servo setup for what we are dealing with here, I would imagine an encoder on the end of a dc motor to a control circuit to provide some trapizoid ramping up and down as you approach your target location. Usually you have to adjust the control circuit for your motor inertia, to adjust the ramping and prevent overshoot.

(looks at the price tag of all that)

I would think the inertia of the head assembly would trounce the inertia of the rotor, and would be equally an issue for steppers or servos. is this not the case?

Forrest: do you use magnetic encoders to encode an absolute position or just how far the motor moved? as in, do you still need to zero out your location occasionally? Hmm.. did you try to do any gradual speedup/slowdown to combat the issue of rotor inertia? If you have a full H-bridge controlling it, it seems you can also 'brake' the motor by bringing both motor inputs high (or low), though, I am not sure if that helps...

"do you use magnetic encoders to encode an absolute position or just how far the motor moved?"

The operant verb is "did". I'm not currently working with encoders. The encoders were rotary. I kept track of where I was via the firmware.

"as in, do you still need to zero out your location occasionally?"

Yes, though not all that often. The AS5035 is a pretty good chip.

"Hmm.. did you try to do any gradual speedup/slowdown to combat the issue of rotor inertia?"

Yes, amongst many other strategies.

"If you have a full H-bridge controlling it, it seems you can also 'brake' the motor by bringing both motor inputs high (or low), though, I am not sure if that helps."

Both half and full bridge (754410 and L298N). It didn't help. Mind, I am sure that somebody a lot better at this game could have made a better job of it than I did.

Given the precision of the current steppers is .9 deg and the precision of the Lego/Mindstorms - NI motors are 1.0 deg - has anyone considered an implementation with the Lego/Labview components. I know it not open source, but it may reach a larger community.

repost to get emails

John Meacham Wrote:
-------------------------------------------------------
> Forrest: do you use magnetic encoders to encode an
> absolute position or just how far the motor moved?

The word is "did", not "do" and the answer is yes.

> as in, do you still need to zero out your location
> occasionally?

Yes. Usually after each layer print. I expect that I could have easily improved on that.

> Hmm.. did you try to do any gradual
> speedup/slowdown to combat the issue of rotor
> inertia?

Yes, though when I got down to 0.8 mm/sec I decided that life was far too short.

The problem wasn't the printing of long roads of plastic (> 5 mm). It was pretty easy to accelerate and decelerate for that king of work. It was doing curves that killed me. When I had to start and stop an axis in a fraction of a mm, which is what happens when you do curves, life just became impossible.

I wanted to print things like involute profile gears, however, and the gearmotor/encoder approach just wasn't delivering me the curves that I wanted at anything like the resolution that I considered appropriate.

That's not to say that some clever boffin couldn't make it happen. It's just that I couldn't.

> If you have a full H-bridge controlling
> it, it seems you can also 'brake' the motor by
> bringing both motor inputs high (or low), though,

I was using the quadruple half-H controller chip SN754410. It doesn't do braking.

Nostalgia time...

In college I did a work term (1982, I think) at a company that had a bunch of DECwriter terminals. Basically a big dot-matrix printer with a keyboard. I was on a team responsible for keeping them running.

They used a shaft-encoded servo motor to position the head. It had a metal encoder wheel and photocells to do the quadrature encoding. The encoding circuit was all TTL. The photocell assembly was on a bracket with an adjustment screw. It was a mess. The encoder wheel would get bits of paper stuck in it. Vibrations would knock the photocells out of alignment. It was amazing these things worked for more than a few weeks at a time.

In order to adjust the encoder, you had to remove the head positioning belt, hook up a dual trace scope to the quadrature outputs and home the head. It was very tricky to get the phases just right. A couple techs liked to try to save time by adjusting the photocells without removing the belt. This would often cause the head to violently slam against the home position, causing even more alignment problems and more repair work.

By that time, stepper motors were already replacing servos in printers. Stepper motors were cheaper and easier to maintain. All of the new desktop printers seemed to use stepper motors.

Now it seems like servos are coming back into fashion. I guess the circuitry and optics are a lot smaller and cheaper now.

The NXT is great fun but it has a few handicaps for using the motors with the reprap.

The NXT motors achieve their 1 Deg. resolution using a course encoder driven from the motor before the large gear train that gears the output down. There is a nice picture here.
[www.philohome.com]

Playing with mine the backlash, between the output and the encoder is approx. 5 Deg which would be hard to design around.

The torque from the motors is about 0.2Nm, which is good for such small motors, but to achieve this the NXT motors are heavily geared and run at less than 100 rpm under load. Even worse if gearing were used to increase the torque to match the steppers, approx. 1Nm, it would be less than 20 rpm or 3 seconds per rev.

Neither of these mean it couldn't be done, just it wouldn't be simple.

One can always use an encoder on a stepper to gain the closed-loop advantage.
A few links I've run across,

A good intro to steppers;
[solarbotics.net]
A small, cheap, good, encoder;
[rotaryencoders.com]
That fits many of these steppers;
[anaheimautomation.com]
(they also sell some servos and accessories)
Easy Step/Dir microstep driver;
[allegromicro.com]

All of these sources sell singles ( [search.digikey.com] the IC) and publish their prices. $55 for a closed loop motor is not bad, tie it together with a PIC and laugh at the cost of a servo.