Servo Instability Due To High Frequency

One possibility is inductive pickup from the high frequency outgoing to the position sensing arrangements of the servos. A possible solution would be to de-energise the servos so that they cannot move while high frequency is present.

Thank you.

But in my case servo's are without brake, so if I de-energise the servos I would loose my welding position.

Is there any other method to prevent inductive pickup from HF?

• Apply brakes to the servos.
• Use fibre-optic positioning instead of electronic.
• Use low-impedance current as the position measurement instead of high impedance voltage.
• Use mechanical positioning instead of electronic.
• Dispense with the high frequency.
• Etc.

Please list the methods that have been tried that do not work.

You didn't mention what type of encoders or transmission cable you are using. Here are a few links that may help.

http://www.optoresolver.com/help/tutorials/encoder_electrical_interface.htm

https://en.wikipedia.org/wiki/Differential_Manchester_encoding

http://www.phoenixamerica.com/applications/electrically-noisy-environments/default.html

Proper grounding, and avoiding ground loops may help.

What are the position encoders attached to the servo shaft? e.g. TTL quadrature outputs?

That high frequency looks like a broad band RF transmitter signal. Cable shielding of the low level susceptible position encoder wiring to the electronics reading this is important, as well how this shield is terminated to ground, and the quality of the ground. As you not only couple noise to ground, but noise from the ground back to your signals. My guess is the zero crossing of the shaft encoders, has this RF noise on it, and the electronics reading this is seeing extra "clocks" that it counts. If you are powering your electronics from the same power source of the welding machine, this can cause a conducted EMI issue. How good is the common mode filtering of the servo control system to reject this, if this is the source of the noise? But again, the common ground of the welding machine and servo electronics can be the 'hard wired' connection from the high frequency arc ignition system. So common mode filters that use the ground as a place to decouple the noise is actually the source injection point to the servo system.

You need a clean ground, and proper shielding, and maybe some capacitors that do not damp the position feed back, but decouple any noise the wiring picks up as an antenna, and decouples this to a quiet ground.

Also how are the encoders powered? And are they grounded out at the servo, or is this kept isolated from the local servo chassis ground.

If you have an oscilloscope, you can look for where this noise is in the system, and with that, try 'bandaids' to fix the EMI susceptibility problem. The big thing to NOT TO DO, is kill the encoder system by installing REAL GOOD filters that not only takes out the noise, but the signal.

As PWSlack was indicating the impedance level of high vs. low of your position feedback can be where this problem is. RF noise loves to float on high impedance signals.

edit:

While typing this Mike Proctor provided several links of interest.

Excellent response. You 'bout covered everything.

The one thing I might add is that for shielded cables, some equipment manufacturers will tell you to ground the cable at only one end, but in my experience, when dealing with HF and up frequencies, grounding both ends of shielded cables is necessary. Lower frequency ground loops caused by grounding both ends of the cable is USUALLY not a problem.

But like you said, watching the signal of interest on an o'scope is a good way of evaluating the 'fixes'. A real time spectrum analyzer is even better but is an order of magnitude higher in cost. Use a differential probe if you are not sure about the signal path voltages or current paths in relationship to earth ground. This will help prevent the destruction of the front end of your o'scope.

You are getting some good answers - good in the sense of directly addressing your problem. The situation you're in is one of the classic problems encountered by Electromagnetic Compatibility (EMC) engineers.

First, you seem to be exposing sensitive control circuits to radio frequency energy at levels they are susceptible to. Second, unaware, perhaps, of that susceptibility, you seem not to have required or included a level of electromagnetic compatibility needed to prevent it.

A large number of EMC related incidents have been reported to the Cherry Clough EMC organization's "Banana Skins" page over the years. From that site:

"49. Failure to correctly specify EMC performance. A large manufacturer of industrial fasteners, negotiating with a major customer, agreed to install a packaging cell containing an automatic weighing machine, which filled plastic packets with fasteners and an RF welding machine to seal the packets. For cost reasons the two machines were purchased separately. No assessment of the electromagnetic environment took place, and the machine contract specifications included no EMC requirements other than "shall meet all legal requirements".

Both machines were supplied, installed, and tested successfully. Unfortunately when both were operated together the weighing machine suffered >25 % errors due to interference from the RF used by the welder (not an uncommon problem). In an 8 hour shift the cell should have packaged £20,000 of fasteners, but could have given away up to £4,000 of product in incorrect weights.

There was no comeback on the machine suppliers, whose products met specification. Both suppliers appeared willing to help, but when pressed blamed each other. Expert technical assistance was brought in and solved the problem. The fastening manufacturer lost 6 weeks production, suffered additional costs, and lost credibility with their major customer."

http://www.compliance-club.com/archive/old_archive/Bananaskins.htm