Filter for Feedback Signal

First things first.

• Are you using a shielded cable to connect your potentiometer to the controller?
• Is your potentiometer in a shielded box?
• Are all these properly grounded?
• How long a cable are you using?

The fourth question might not be the problem, if I read your post correctly. You seem to have replaced your old servos while retaining the position feedback sensors. If you changed their location, it might be a problem if they're farther away than before.

There's another thing to check:

• Your potentiometer may be defective. The traces may have been abraded away, resulting in a noisy signal as the potentiometer moves.

Thanks for your interest. No, nothing is shielded. These are old electro mechanical systems and upgrading the wiring is a huge undertaking. These are linear pots and are also subject to a small vibration induced oscillation, or "bounce". The old servos ignored this as well as the line noise. The old servos were replaced primarily because we can't get them anymore.

I know that there are many ways to fix this. But the overall accuracy and response isn't critical. It just needs to be stable. There are a lot of these and the question becomes; should we just scrap everything and start over? So I was hoping that someone would just say, "oh, all you need is to put an xxx electrolytic capacitor across the input terminals". Wishful thinking I know but I thought I would ask.

Thanks again.

Putting a capacitor in there shouldn't present much of a problem. The size or value of capacitor will depend on how much noise there is and how big it is.

I can't give you any hard numbers but you can start with 1μF and go up or down from there. Capacitors of this size are cheap and if your "pack rat" box is well-stocked there'd be no expense.

I'm a little worried about the fact that your cabling is unshielded. Maybe the new servos are producing lots more of EM noise than your old ones so they may actually be the source of your problems.

I think it will be better to use a RC filter combination.

A resistor , let say 1000 ohm 1 watt in serie with this signal line and than a capacitor from 1, 5 or 100 micro farad connected to the ground.

You have to take away the signal at the connection between the resistor and the capacitor.

What you are doing, is charging and uncharging the capacitor, but the bigger the capacitor is, the longer it takes in time to charge and uncharge the capacitor, so your signal will be much smooter.

If you raise the value of the resistor the current to charge the capacitor will be smaller and (in the same way as with a bigger capacitor) it takes more time to charge the capacitor, so your signal will variate much slower.

good luck

Leurs Rudy Belgium

Remember that Zippy2 said that the voltage to the input is only 2.5 volts and a 0.01 volt change is enough to cause a reaction from the servos. A resistor in series will drop that voltage and possibly affect the performance or maybe even cause it not to react at all.

That's a pretty sensitive setting. You said that there are no parameter settings for hysterisis or dead band but didn't say if there was a parameter for gain (there should be). Is it feasible to lower the gain on the things? 'Shouldn't cost you anything.

You may have noticed that I'm reluctant to make any hardware changes unless they're really needed. I tend to avoid adding to anything to equipment or processes 'coz, in the end, I would be required to service these things. I'd rather keep things simple and have less to maintain and check.

Yes, that's also a problem from time to time and when they get too bad we change them. But again, the accuracy of the signal isn't that critical provided it's stable and does not fluctuate or bounce around a lot. We have over 40 of these, 5 with the new servos for a trial. We could buy servos with filters and dead-bands but for a long list of reasons would prefer to stay with these if we can get them to work. It boils down to this; if these can be made to work with a simple fix, the cost savings are enormous.

The noisy signal combined with the input sensitivity of the servo results in the servo firing almost continuously and overheating.

I appreciate your comments and suggestions. They help me review the details.

Hello,

In one hand you already have RC network by putting an capacitor. But because of the small resistance of potentiometer, the pole of RC network is very high. I suggest that you try to make LC filter which has attenuation of 40db instead of 20db as RC filter has. That should not be a hard, because I expect that the input current is low, so the small dimension of L (inductor) designed on the ferrite bead will do the job.

Regards

The best way is to use a low pass RC filter. See http://en.wikipedia.org/wiki/Low-pass_filter

If you know the frequency of the signal you want to keep, , you filter all the noise above that signal. If the good signal is 3600hz use a cut off frequency about say 4 khz. this will attenuate the signal starting around 4 khz and above

If you know the approximate length of the wire from the pot to the servo, you can calculate the inherent Resistance . see http://www.otherpower.com/cgi-bin/webbbs/webbbs_config.pl?noframes;read=6346 (Metric) or http://hyperphysics.phy-astr.gsu.edu/hbase/Tables/wirega.html (US)

Formula for calculating the size of the capacitor is:

Capacitance in farads = 1 / (2 X pi X Resistance X cut-off frequency)

No, it is not this case. The signal has DC level, so noise is out of DC signal - it is all AC component. If you use RC filter, you will attenuate the signal in the rate proportional to input impedance. Inductor has negliable resistance, that is better to avoid attenuation.

Regards.

Thanks all,

I'll be experimenting with the various recommended fixes today and I'll let you know what happens.