[MAX711]

I've had a quick search but could find any postings on this subject, so here goes!

I am thinking of using DC Brushed Servo Motors on my project and had a question about holding position.

If I understand correctly, in order for a servo motor to hold its position against an external force, it needs to be repeatedly driven in opposite directions (from one pulse of the encoder to the opposite pulse in the other direction). So this would mean that even when stationary, the motor would be "vibrating" back and forth a tiny amount (equal to the encoder pulse spacing). Is my understanding correct? And if so, what does that do to the motor (heating, bearings, etc) and also to any downstream mechanical items like ballscrews and couplings?

Thanks

[Al_The_Man]

That is not necessarily true of sophisticated control anyway, where the loop is closed back to the intelligent control, most amplifiers today operate in the torque or current mode of operation where the control outputs a voltage to the amplifier which is directly proportional to the current out to the motor, the encoder feeds back the position to the control, if the control detects that the motor is not in position or is not moving to the degree programmed by the voltage, the motor current will be proportional to the error. If the motor is moved to a position and brought to rest the encoder tells the control it is in position and no further output is neccessary and will be zero, if the motor moves by virtue of external forces then the correction will be made, otherwise there is no need to output.
If you experience an amplifier that appears to be constantly active (sings) and exhibits fine oscillation even when it should be at rest, it is usually due to too high a gain adjustment of the amplifier itself.
Al.

[Mariss Freimanis]

"Dithering" is due to the simple fact there is no encoder feedback between encoder edges. The "I" (integral) in the PID alogorithm will take an error no matter how small and amplify it over time. This will cause the motor to move until there is feedback; an encoder edge. This feedback generates a miniscule error signal in the opposite direction and the process repeats, the motor ping-ponging between adjacent encoder edges.

Don't like that? Use sine-cosine output encoders. These provide continuous feedback between the zero-crossing (quadrature) locations. You also would need a servo drive that accepts and processes analog inputs.

Mariss

[HillBilly]

I have noticed what is considered balanced on a analog amp only applies to the time the adjustment is made (It never stays perfect.). This will also contribute to the ping pong effect.

The amp command signal is generated across a comparator in a velocity loop system, any difference in the commanded velocity and the actual velocity basically yields a full output. The I in PID basically makes the position loop behave in the same maner (Any difference in commanded position and the actual postion will yield a higher than linear output.).

Darek

[MAX711]

Thanks for all the great replys guys.

Some of it went straight over my head, but I think you've confirmed what I suspected. I guess what I was digging for was whether my machine would start "buzzing" the moment I switched it on. I saw a video on the web of someone who had converted their mill/drill with DC servos and the thing sounded like a mini chainsaw as soon as he put power on the controls (it wasn't the spindle). That got me thinking about how a servo works.

If I put the encoder on the ballscrew instead of the motor (with pulleys in between), would that make the ping-ponging effect worse? Can the ping-ponging be reduced by tuning the drives using an oscilloscope? Also, do the Gecko or Rutex drives accept a sine-cosine encoder input?

Thanks

[Al_The_Man]

you may want to watch an exellent in depth explanation of encoder fitted to the ballscrew and motor on the Galil site, the instructional classroom video's the one entitled 'Dual Loop Compensation'.
With Modern drives you should not get 'buzzing' at rest.
The encoders I use on the systems I am using are a minimum of 1kcounts/rev to 2k/rev and up, so if the 4 quad edges are used to increase the resolution, even if the motor moved between one edge to the next this represents a rotor movement of .09deg to .045 deg or 5.4 arc min to 2.7 arc min. Which is barely perceptable to the naked eye.
Al.

[Mariss Freimanis]

I don't know about Rutex drives but there is no "chain-saw" sound from the G320/G340 drives; you might hear a faint sound from our drives if you are in a totally quiet room. We eliminate that problem by using a modified dead-band technique. Gain and damping coefficients (PD) are knocked back to 10% of set value when the drive is within 2 encoder counts of the command position. This makes the motor silent when it is in position.

I have a working circuit that accepts sine-cosine encoder inputs and works perfectly. The sine function is nearly linear from -45 to +45 degrees; this circuit multiplexes the sine, cosine and -sine, -cosine components from the inputs into what looks like a sawtooth waveform. This is then summed to the position error node. The result is a motor that is very stiff and totally silent; if the encoder is accurate, an additional 2 to 3 bits of position resolution can be interpolated as well. The circuit overhead is a quad op-amp, quad comparator and a quad analog multiplexer (4052).

The rub is trying to find sine-cosine encoders; they simply are not commonly available. Sad thing is every incremental encoder has sine-cosine signals that are squared-up internally and presented as quadrature outputs. Kind of a shame really.

Mariss