View Full Version : Mariss Freimanis Seppers vs Servos Advantages/Disadvantages of both

Lee Roberts
02-10-2007, 09:08 AM
Thanks to NC Cams for compiling the following info:

Everything you wanted to know and more about steppers and servos as put forth in a reply from Mariss Freimanis to Freak Brain on another post. I've found this to be a definitive reply that addresses the recurring "stepper versus servo" inquiry as posed by Freak Brain (with editing for spelling and apparent typo’s):

Step motors and servo motors service similar applications, ones where precise positioning and speed are required.

The biggest difference is that steppers are operated "open loop". This means there is no feedback required from the motor. You send a step pulse to the drive and take on faith it will be executed. Seems like a problem but it's not.

If you have a quartz watch with hour and minute hands, then you have a step motor on your wrist. The electronics generates 1 step pulse per second, driving a 60 step per revolution motor which turns at 1 RPM. It keeps nearly perfect time. Any errors are due entirely to the electronics timing accuracy (quartz crystal oscillator).

With apologies to David Letterman, here's some "Top 10" lists I came up with for choosing between steppers and brush DC servos. Others can add more, I'm sure.

Top Ten Stepper Advantages:

1) Stable. Can drive a wide range of frictional and inertial loads.
2) Needs no feedback. The motor is also the position transducer.
3) Inexpensive relative to other motion control systems.
4) Standardized frame size and performance.
5) Plug and play. Easy to setup and use.
6) Safe. If anything breaks, the motor stops.
7) Long life. Bearings are the only wear-out mechanism.
8) Excellent low speed torque. Can drive many loads without gearing.
9) Excellent repeatability. Returns to the same location accurately.
10) Overload safe. Motor cannot be damaged by mechanical overload.

Top Ten DC Servo Advantages:

1) High output power relative to motor size and weight.
2) Encoder determines accuracy and resolution.
3) High efficiency. Can approach 90% at light loads.
4) High torque to inertia ratio. Can rapidly accelerate loads.
5) Has "reserve" power. 2-3 times continuous power for short periods.
6) Has "reserve" torque. 5-10 times rated torque for short periods.
7) Motor stays cool. Current draw proportional to load.
8) Usable high speed torque. Maintains rated torque to 90% of NL RPM
9) Audibly quiet at high speeds.
10) Resonance and vibration free operation.

Top Ten Stepper Disadvantages:

1) Low efficiency. Motor draws substantial power regardless of load.
2) Torque drops rapidly with speed (torque is the inverse of speed).
3) Low accuracy. 1:200 at full load, 1:2000 at light loads.
4) Prone to resonance. Requires micro-stepping to move smoothly.
5) No feedback to indicate missed steps.
6) Low torque to inertia ratio. Cannot accelerate loads very rapidly.
7) Motor gets very hot in high performance configurations.
8) Motor will not "pick up" after momentary overload.
9) Motor is audibly very noisy at moderate to high speeds.
10) Low output power for size and weight.

Top Ten DC Servo (brush type) Disadvantages (besides higher relative cost):

1) Requires "tuning" to stabilize feedback loop.
2) Motor "runs away" when something breaks. Safety circuits required.
3) Complex. Requires encoder.
4) Brush wear limits life to 2,000 hrs. Service is then required.
5) Peak torque is limited to a 1% duty cycle.
6) Motor can be damaged by sustained overload.
7) Bewildering choice of motors, encoders, servo drives.
8) Power supply current 10 times average to use peak torque. See (5).
9) Motor develops peak power at higher speeds. Gearing often required.
10) Poor motor cooling. Ventilated motors are easily contaminated.


Subsequently another question was posed by Freak Brain:

One question....

On the Top Ten DC Servo Disadvantages you wrote, Requires "tuning" to stabilize feedback loop. Is this done in the drive or is this done in the encoder? and the safety circuits, are these in the drive or is this something I would have to buy separately. Like your drives, are these items in your drives?

Mariss’ reply to this inquiry:

Tuning refers to adjusting the PID coefficients to cause a critically damped response from the motor/load when adjusting to a disturbance. Sounds complicated but it's not.

PID stands for Proportional, Integral and Differential. The "difference" error (where you should be versus where you are) is separated into 3 channels (PID), then recombined. You perform this algorithm unconsciously when you drive a car.

Say you take a road trip from LA to San Francisco up I-5. Your task is to stay side-by-side next to the "command position" car on this trip. It can accelerate and stop instantly, you can't. Your gas pedal and brake (torque command) only adjusts acceleration and deceleration like in a real car.

At the start of the trip, both of you are stopped. The "command car" instantly accelerates to 85MPH (average to slow for what you see on I-5).

The first thing you notice is a lot of distance has opened up between you and the "command car". This is the Proportional component. You press on the gas and away we go. Your speed builds up and after a while the distance begins to close.

Your rate of closure is the derivative or Differential component. As long as the distance to the "command car' keeps opening, you press harder on the gas. As it closes, you ease up.

To close the distance, you have to go faster than the "command car". Otherwise you will never catch up.

You are now getting very near the "command car". Both the separating distance and the rate of closure decreases towards zero so they are no longer of use. You have come off of the gas enough to nearly match its speed.

This is where the Integral component comes in. You are now side by side. You now adjust your speed based on inches of separation. If you slightly edge into the lead, you ease off. If you slightly fall behind, you make up the difference. Rate of closure (differential) or separating distance (Proportional) are too small to use.

Using this method (PID), you will arrive at your destination simultaneously even though hundreds of miles and hours of travel have elapsed. You do it naturally and unconsciously.

A mis-tuned servo (not enough D or too much P) by this analogy would have you overshoot the "command car", hit the brake, fall behind, over-accelerate and overshoot the "command car" again, over and over (sometimes called "hunting" or "oscillating"). Tough on your passengers and car (400 miles, 4 hours) and equally hard on your servo motor.

All sorts of other stuff works with this analogy. Two things come to mind.

1) Feed-forward compensated PID servos. This is where you are told ahead of time what the "command car" will do. You don't have to sit and watch in surprise when it suddenly takes off or changes direction; you are fore-armed with its future intentions. This somewhat makes up for time otherwise lost in catching up.

2) S-shaped acceleration/deceleration profiles. This keeps the 2nd derivative of velocity (namely jerk) finite, minimizing the "jerk" factor. Again, you do this naturally driving a car.

(NC addition: The rate of change of the motion is velocity or first derivative of motion. The rate of change of velocity is acceleration or the first derivative of velocity and second of motion. The rate of change of acceleration is jerk or the first derivative of acceleration, or second derivative of velocity or third derivative of motion. Got that??? If you do, you understand elementary calculus.).

Imagine you are cruising down a boulevard when the light up ahead changes from green to red. In a simple CNC program, you stand on your brake until you come to a stop. This would be very uncomfortable in real life.

When you decelerate in a car, you tense your muscles to balance against the deceleration G-force until you just counteract that force. When you come to a stop, deceleration abruptly disappears along with its G-force. Your tensing against it does not though. The result is head and body bobbing back and forth until you find the new balance. Not comfortable.

What you actually do when coming to a stop is to tail-off on the brake pedal. Before coming to a stop you lessen the rate of deceleration (brake pedal pressure) to make it become zero as your speed approaches zero. This is an S-shaped deceleration curve.

Where it matters on your CNC machine is it eliminates ringing (head bobbing) at the beginning and end of acceleration and deceleration. This decreases wear and error.


Additions by NC Cams:

The safety circuit inquiry response is sort of a yes and no answer.

Limit switches are mounted externally on the machine. These cut power via drive disengagement should "runaway" occur.

Most servo amps can output S/D (shutdown) signals from built-in protection circuitry that tell the controller a fault occurred in the drive (IE: over voltage, over current, rpm runaway, loss of feedback, etc).

The machine tool builder is the one who is charged with installing limit switches. These are the last defense for stopping a runaway servo at some end limit point of the table/quill travel.

Further elaboration by Arvidb:

The way acceleration and deceleration is handled is up to the controller (Mach2 for instance). You won't get oscillations by setting these wrong (in the drive), but S shaped acceleration is easier on the mechanics than constant/linear acceleration.

PID parameters are set in the servo driver (e.g. the Gecko drive). Get these wrong, and you might get very nasty oscillations (vibrations or "buzzing") from the motor. These will wear out the mechanics quickly - and sound awful! This is the "not enough D or too much P" part described by Mariss.

(Mostly to Allen): PID adjustments are done to get the machine to respond as quickly and accurately as possible to a command, while still making sure the machine is stable (no oscillations).

First the difference between actual and commanded values are calculated (differences = "error"), then a control signal is calculated as the sum of proportional (= factor * error), integral (= factor * sum of all errors measured from machine start), and derivative (= factor * rate of change of error) terms.

The three factors are the things one adjusts (usually called Kp, Ti, Td). It can be quite difficult to get them right.

Response by TORSEN:

The difference between a stepper and a servo in simple terms is the fact that when not moving a stepper is held in position by current -- servo is not.

(NC Cams comment: this is technically true BUT if the cutting or other forces try to move the table against the servo, the encoder shows that motion is or has taken place and the servo will apply force via the application of current to return the table to the prescribed position. For a snapshot in time, there is no current flow, however this can rapidly change as the cut is made - read on)

The full current is flowing to hold a stepper motor in position this is why they are rated for a certain amount of holding torque. As the stepper starts to rotate, the torque diminishes as the speed increases.

(When) No current is flow(ing) on a stopped servo motor, its power is regulated by a error register in the driver. When a force is encountered (like somebody trying to turn the shaft by hand or cutter force tries to move the servo) the error of position on the shaft will cause a current to be send to the motor to correct the error.

The amount of current is in direct relation to the amount of error so a small error will produce only a small current. This is like a rubber band effect when trying to move the motor and should be taken into consideration when the resolution of a system is considered.

With the ever moving improvements of the technical community both technologies have been working its way towards each other. Servos can now be controlled just like stepper systems and Multi-stepping systems with current reduction at stand still make them behave much more alike.

Rogers Machine (???) makes a card that protects against lost steps win a Mach system which may be of interest. However, dealing with lost steps seesm to be an after the fact reactionary approach whereas preventing them via proper motor sizing is probably a more appropriate anticipatory approach.

Hope this helps!