[balsaman]

I seem to have developed(?) a problem with my x axis. It faults randomly. I did some testing and it seems I am loosing some position as well. In 200 1" moves (back and forth) I lost 50 thousands of an inch.

I ran a new encoder cable directly from the encoder to the drive outside the cabletray/cat-tracks. Same problem. switched the x gecko drive with the A(rotary axis) drive. Same problem. I am starting to wonder if it's the encoder itself, but I want to eliminate some more stuff first.

Question: I use 8 conductor shielded cable for the encoder. 4 twisted pairs and a shield. I only use 4 of them. Should A and B from the encoder be a twisted pair? or should I use a wire from one pair for A and a wire from another pair for B with the other wires from the pairs grounded? Or just cut off? I use a twisted pair for the +5 and 0V. The shield is tied to ground at the drive end only.

Anyone else have an idea what's wrong?

Eric

[Mariss Freimanis]

Eric,

Never have 2 signals on the same twisted pair! The cross-talk between them will be terrible.

Send encoder channel A on one wire of the pair, send either Ground or +5VDC on the other wire. Use a seperate twisted-pair for channel B (ground or +5V on one wire, channel B on the other).

Mariss

[Al_The_Man]

Eric, I would not think it was the encoder at this stage, especially if the error is repeatable, the 50 thou error I suspect may be due to backlash?
If you use an encoder cable with independant pairs, it is customary to use a twisted pair for the A,/A & B,/B pair etc
The shield should only be grounded at one end, Usually the amp input end as this is usually closer to the main cabinet gnd.
Due to noise problems I had on one installation, I have been using belden 9891, I believe this is an ethernet cable, but it has one pair which is larger than the other conductors which I use for the power pair. I have had no problems since with any of the retrofits I have done.
Is your encoder on the motor or on the final motion actuator, i.e. ball screw or slide?
Al

[HuFlungDung]

Hi Eric,

I don't know right off if twisted pairs does much of anything for shielding low voltage wires. I know it cancels the field self induced by motor power wires, but that is the only place where I have heard that twisted pairs accomplish anything. But, you're the expert on that anyway.

You can maybe get ahold of cable with shielded pairs, too, all wrapped inside of a larger outer shield. But, I somehow doubt that encoder wires would crosstalk to each other anyway.

Can you swap encoders with another axis to find out if the problem follows?

Whatever is counting your encoder output is what I would suspect. Is that in your drive or in the computer?

[balsaman]

Hu, the drive counts the encoder pulses. I have eliminated that as the problem. I am no expert...just know enough to be dangerous.

Al, there is backlash, but the error accumulates the longer the machine is run. Backlash errors are not cumulative.

Mariss, I used one wire from a pair for A, another from a pair for B. The +5 and 0V are on a twisted pair. The rest of the wires (one pair, and the two left over from A and B) are cut off on the encoder end, folded over and taped on the drive end. Will tieing them to ground at the drive end help? If I make the +5 and 0V the wires twisted with A and B I have to open the encoders and resolder. I can do that if it's the best way.

Thanks,

Eric

[Mariss Freimanis]

Having the unused wire of a pair grounded helps shield the signal carrying wire. Ground them at the drive end only to "encoder minus".

You mentioned on the "geckodrive" group the LEDs are dimly "on" while the motors are running. I looked at your ERR/RES circuit but could not make out the resistor size. If they are 10K or less, the LEDs will be partially on. That should be no problem though.

A bad STEP/DIR cable can cause problems like that as well. The drive may be faulting because it is seeing a burst of noise pulses it stupidly thinks are steps. Try gently wiggling the cable. See what the effect is.

Mariss

[balsaman]

I will ground left over signal wires and try again.

The resistors are 4.7k.

Thanks for the help.

Eric

[Mariss Freimanis]

Let me know if that helps. By the way, are you using a new PC? Some of the newer PCs have 3.3V logic swings on the parallel port instead of the expected 5V. This can cause trouble.

Mariss

[balsaman]

Tied all the left over wires to enc- and same thing....I will have to try a new parallel cable I guess. Wiggling the cable has no effect. The computer is a Pll 233.

Funny thing is, this is the only encoder cable that doesn't run along side the DC motor cables (and noisy router motor cable) all around the machine and thru the cat tacks etc. It goes nicely all alone to the back where the encoder is. The motor is on the oposite side of the leadscrew. Yes the couplings are tight.

Eric

[Mariss Freimanis]

Eric,

What revision number is your drive? Later versions have the REV number screened on the board area covered by the main connector. Our current version is REV-7.

Another possibility is a badly misaligned encoder, an encoder wheel that has a bend in it (metal wheels only) or has some crud or foreign matter on it. This is where the CH-A and CH-B signals aren't in quadrature for a portion of a revolution, or a channel drops out (crud on wheel or bent disk). This drives the quadrature decoder nuts and will result in errors or a FAULT condition.

This is best tested for with a scope. Barring that, cause the motor to move at 5 to 10 RPM. See if the motor acts or sounds funny at the same "clock" location on every revolution. If it does, it's your encoder.

Mariss

[balsaman]

I have tried two drives. The one in there now is one that was replaced in September? by you when you recalled a few drives. I will check the encoder by your 10 rpm test today. I have access to a scope too so I will try that if need be.

Thanks,

Eric

[HomeCNC]

Eric,

How long is this cable. Is it the longest of your other ones?

[balsaman]

It's the same length as the y and shorter than the z.

I replaced the parallel cable, and took the cover off the encoder and blew air threw it. It's much better now. In 600 1" moves I was out 10 thousands. That is a lot better than 50 in 200 moves. These moves were all at 100 ipm. (I tried it at 10 ipm and was also losing position). The motors sound smooth. Get the odd hickup but I think that is turbocnc. I get the same error running Mach 1.

Question. My leadscrews are 6 tpi. That makes a crazy number for steps per inch with my 300 count encoders. Rounding errors should not acumulate should they? I would think not, since any error in one direction should cancel when the axis returns.

I am sure it's a problem with the encoder. I am going to take it apart, clean the glass disk, and put it back together. While I am at it I will see if I can find some cable with individually shielded pairs.

BTW, the drive isn't faulting anymore either. (at least not in the last 1500 1" moves)...

Eric

[balsaman]

Should the Step pin state be Active high or Active low in the Turbocnc setup for the geckodrives?

Eric

[sol]

Active high

[balsaman]

Thanks Sol.

Changing the setup from active low to active high seems to have eliminated the last of the error. I guess I was loosing one step every time it changed direction.

Now I need to put it all back together and see if I can maintain the accuracy. Still not sure which fix fixed it, but the biggest improvement happened when I blew air through the encoder at the glass disk while it was running. Funny, because it looked pristine inside the dust tight encoder enclosure. Never argue with success I guess....

Eric