Encoder Problems need advise!

[judleroy]

Ok here is whats going on. I have quadrature encoders that output 4096 cts per rev. The longest cables are about 10 feet long. The cables are sheilded. They only have A, B, +5v, and Ground wired to the controller and read just fine. The counts want to float while the machine is on so the motors are constantly adjusting for errors that arent really there. That means nothing can be cut accuratly. At first they would float at idle when the machine was just sitting there. Then I connected the machine directly to earth ground and ran the sheilding from the motor power wires to the machine chassis near the motor end of the cables. This fixed the problem of them wanting to float while the machine was idle. This however did nothing to keep them from floating while the machine is running programs. The effects get worse the more motors running at the same time. The next step I have been told is to ground the sheilding from the encoder wires to the same place as the encoder ground wires go at the controller end. Then I am told to use a 1/4 watt 150-250ohm resister running from the encoder A to the same ground and also another one from the encoder B to ground. Each encoder has its own ground at the board. Does this sound like it will fix my noise problem. Are there any other suggetions to help with the noise problem? I will be trying all this tommorow so I guess we'll see how it works. Just wondering if anyone else has had problems with encoder noise causing them to float and how you fixed the problem in your setup. Thanks for any advise.
Judleroy

[Al_The_Man]

By 'float' I assume you mean there are spurious counts?
Or do you mean that the signal Level is varying?
What I normally do is to ground the encoder shield at the Earth Ground star point wherever that is set up.
Do you know the nature of the encoder output? Single ended encoders can be push/pull or open collector output, if open collector, you will definitely need a resistor from each output to 5v+.
Another thing that may resolve it is to ground the common of the 5v supply to the ground star point also.
Al.

[mike_Kilroy]

you dont mention a-not & b-not channels.... do u have them available to use? if using just a & b and not a'a-,b,b-, you are fighting an uphill battle. if you can use differential outputs with a,a-,b,b- your problems will likely be gone.

if you cannot get differential outputs, then you are at the mercy of the shielding, wire length, noise in your machine. assuming you do not have open collector outputs (that require external added pull up resistors to +5vdc) as Al said, then, yes, the suggestion to add those 200 ohm resistors to common is fair advice simply adds more low impedance to the lines to try to help reduce the noise.

so ur 10' encoder cable is shielded' good. latest grounding scheme is to tie the shield AT BOTH ENDS not just one end. and dont bother doing this by soldering a 6" #18 wire to the shield and taking it over to a screw - not worth the effort at the PWM turn on times used today. You get 400% better shield ground by carefully cutting back the plastic jacket over shield, exposing it, and wire tying it with a plastic wire tie directly against a ground point - at both ends.

since u know it is motor PWM noise (you improved it by changing motor shield ground & it gets worse with more motors running), do same for their shields - tye at both ends by wire tying to ground point.

last comment: is ur encoder wire run separate from the motor leads? it cannot run parallel to them in same wireway, or even parallel in wireway with any other 120 or 230v ac wires. if you cannot get it to stop counting wrong try pulling encoder cable out of machine away from everything across floor to motor and see if it fixes it - will give u ideas of how to rerun it in cabinet if helps/works.

really tho, no one uses or should use single ended encoders (a & b only) anymore - the days of being able to reliably get away with that ended when PWM drives replaced SCR drives.

[judleroy]

Thanks Al and Mike
I will try what you have suggested. If it will help you lead me in the right direction the motors are the ones asked about in this post. http://www.cnczone.com/forums/genera...se_servos.html
The encoder wiring is shown in post 2 and 3. The encoders have a ribbon cable with ten wires coming out of them. The wiring doesnt mention a- and b- though. Thanks again.
Judleroy

[Al_The_Man]

These look like older Renco encoders, they traditionally used either push/pull or open collector for single ended type, they show pull up resistor N/A, so they may be push/pull which does offer fairly low impedance drive, so you may get away with a run of 10'.
The other option is to convert to RS485 using a suitable driver I.C., if your system takes differential then you will not need a corresponding receiver IC.
Do you have the encoder 5v supply connected to Earth Ground point?
Al.

[judleroy]

Hi Al
The encoders are powered by the 5v from my k-motion board. It has 4 5v outputs and 4 grounds for the encoders built into the board.The actual power comes from a pc power supply connected to the board through the molex connector. The pc supply is mounted to the side of an industrial control box. The box is plastic and has a metal plate across the bottom that the boards and power supply are bolted to. The pc supply is not touching the plate at all. Would it help to connect the pc supply to the screw in ground points built into the plate by a wire. Also the control box is not connected to earth ground. Should I run a ground wire to earth ground from that same ground screw hub. Sorry but electrical is all new to me and proper grounding has me confused. If you would like or if it will help I will take some pictures of what I have going on. Thanks alot for the advice.
Judleroy

[Al_The_Man]

Your PC power supply should be grounded through the MB screws, so check with a meter, I bond all power supplies to the Earth Ground star point.
Ground Practices
I would ensure your encoder supply common is bonded to this point also.
It has been my experience that most noise related problems are due to lack of grounding and bonding all supplies to earth ground.
I prefer this over the 'isolate everything' method.
http://www.cnczone.com/forums/phase_...se_issues.html
Al.

[engineertony]

A few electronic basics. I usually define "floating" as not grounded, i.e. prone to pick up voltages induced by stray magnetic or electrostatic fields. These voltages can be read by a counter as pulses and therefore give a wrong position reading.
Shielding is meant to pick up the volts and short to ground. We normally only ground at one point to prevent the formation of a complete loop which would act as a turn on a transformer and cause current to circulate.
Somewhere at the counter there is a circuit which sees logic 1 or 0, and this has a differential circuit which needs to go past a certain voltage level to register as 1 or 0. For example encoder channel A is at 5v, (logic 0) then it must go below say 1 v to change to 0v, (logic 1). Then having established logic 1, it must go above say 3v to return to logic 0.
So to have counter interference you must have quite high voltage spikes, the system is immune to say 1 volt spikes.
I used to track these with a storage oscilloscope, you can see the spikes or wobbles superimposed on top of the encoder pulses, and experiment with various grounding configurations and resistor loadings to get rid of them.
Pretty standard grounding arrangement in my experience is a tree structure, all single ended and all brought to one point, with electronic ground well away from the power ground.

[Al_The_Man]

Originally Posted by engineertony Pretty standard grounding arrangement in my experience is a tree structure, all single ended and all brought to one point, with electronic ground well away from the power ground.

Agree with most, but I only use one common ground and that is one Earth Ground star point.
Your PC (electronic) MB is going to be at Earth ground anyway if it is secured the usuall way with MB screws.
And your Spindle motor is at Earth Ground via the Grounded neutral.
Al.

[mike_Kilroy]

also agree with most of what was said - well put!

One difference is the grounding shield at only 1 end; that is old school and has been updated to ground at both ends today. Old rule of thumb was ground at one end only to make an electrostatic shield (keeps noise OUT) of the cable, and ground at both ends for electromagnetic shield (keeps the noise IN from getting out). When we had SCR drives with slow rise times all this worked fine.... then came transistors and faster and faster switching times. A lot of today's IGBTs used in VFD and servo drives switch in a few nanoseconds; result is noise pulses with rise times making 100's of mhz up into ghz frequency noise now. I believe most servo drive manufactures have learned by now that tying the shield at both ends of both signal and power cables does a better job of shielding that the old fashion one side only. In addition, we found that the old common method of bunching the shield together at the end of the cable and soldering a 18 gauge wire on it 12" long to go to that ground screw was next to useless also: think of the 1/4 length antenna on your cell phone at 2.4ghz - a couple inches. A quarter wavelength means the noise voltage at the shield soldered point becomes 0 at the other end! Infinite impedance. So useless. And a lot of us still make that wire 6-12" long! My 2 meter ham radio uses 18" length for 1/4 wavelength at 146mhz! So a pigtail no longer makes a good ground. In addition, at these higher frequencies, there is something called "skin effect:" means the actual noise current is riding on the outside molecules of the shield wire itself and little is on the inside of the shield; so tying all that stuff together was a bit of a waste, in addition to only gathering up a small area of the OD of the shield cable at the bunched point of the soldered pigtail anyway.

So today's best practice is to cut back the plastic sheathing on the shielded cable and wiretie it to a good earth ground at both its ends. That makes the largest possible area touch ground to satisfy the skin effect problem, drains off more of the noise due to grounds at both ends, and prevents adding a nice pig impedance pigtail wire. At these high frequencies there is not so much concern over circulating currents as there is at low frequencies.

[Al_The_Man]

I try to keep up to date on what developments are taking place in the control world, but have not yet come across the reference for the change in shield grounding.
I have tried to find any definitive description with supported qualifying data related to switching to grounding shielding at both ends, the only thing I could find is a technical paper that mentions that the requirement is applied to shielded cables carrying signals in the GHz range??
Al.

[mike_Kilroy]

Al,

I have worked closely with the engineers at Kollmorgen over the years who have had to prepare our drives for CE certification. One of the first things they learned was the need for grounding ALL shields at both ends to meet CE. Here are some examples from their drive manuals:

page 81 shown of AKD drive manual:

http://www.kollmorgen.com/website/co..._en_07_10_.pdf