Axis' don't return to zero - accumulating error

[Scott Kelley]

I've got an error with a servo system that I have not been able to figure out. Everything works properly, except that the axis' slowly accumulate errors. I can run a routine that includes multiple movements of the axis', and when I return to 0,0,0 the spindle is no longer over the 0,0 position, but has moved by tens of thousandths, always in the positive direction.

Here is the system info:

Bridgeport Boss mill.
OEM DC servo motors.
AMT-102 Encoders (same problem occurred with the OEM BEI encoders)
CNC4PC C11 Break-out board
CNCDRIVE Mammut Servo Drives
Mach 3


I recently tested the problem using the test described below, I find that the table consistently moves 3" in the positive direction when instructed, and it consistently moves about 4 thousandths shy of the 3" called for when traveling back to zero in the negative direction (backlash is taken out of measurement as described below).

I believe that smaller movements result in smaller errors.

Problem persisted after changing encoders from the original BEI encoders and cables to AMT encoders with different cables.

Movements directed to the servos via USB using the manufacturer's "servoconfigurator" application result in very accurate movement in both directions with no error accumulation.

So it would seem that the problem must be one of the following:
* Erroneous / Extraneous data sent from Mach3.
* Erroneous data received by servo drive from parallel input line (noise).
* Internal error in servo drive.
* Erroneous data sent by encoders (unlikely - same problem with different encoders).
* Erroneous data received from encoders (unlikely - shielded OEM cables with OEM encoders / twisted pair cables with new encoders - no ground loop).

Questions:

1. Am I overlooking any possible source of the problem?

2. Does this problem sound familiar to anyone?

3. What tools are readily available to test the outputs of the various components (to actually see where the erroneous count is occurring)?

4. Any thoughts as to additional tests that I may perform to narrow down the problem?


Test procedure used:
To measure error while table moves in X+ direction; Set up dial indicator to measure X position of spindle relative to table using precision gauge blocks. Move spindle to X.9, insert 1" gauge block, move spindle to x1, zero indicator, move to X4, insert 4" gauge block, read indicator.

To measure error while table moves in -X direction; Move spindle to X4.1, insert 4" gauge block, move spindle to X4.0, zero indicator, move to X1, insert 1" gauge block, read indicator.


Thanks,
Scott Kelley

Similar Threads:

• Need Help!- z axis return
• Need Help!- Refernce return error
• Position Accumulating
• Problem- VF-0E Z axis does Zero return towards table.
• Z return margin error

[mike_Kilroy]

good test running the 'mfgr servoconfigurator' getting commands via USB to help isolate the problem..... tell us who the mfgr is and what the device is that uses this servoconfigurator program. I assume it is the drive amplifier software but that is just a guess...... if you reply tonight I will use that to make a prediction of where you may want to look.....

2 additional questions: the encoders feedback is using differential sigs (a,a-,b,b-)?

and same question on parallel output from printer port; I assume it is step & dir commands to the drive - so u have step.step-, dir, & dir- hooked up?

3rd question: what is cable from par port to interface card - ribbon cable? shielded twisted pairs? how long? and is it near ANY other wires?

couple more hit me: u sure motor coupling is not loose? something else in drivetrain mechanically loose/slipping?

does the the servoconfig program uses a slower accel rated than the mach3 so it didnt mech slip?

is there a spindle drive on this machine? did u try test with it off (they often cause noise problems)?

[Scott Kelley]

My OP has a listing of the major components.

These encoders do NOT have differential outputs.

Cable from computer to BOB is about 2.5' long, shielded.

Definitely not anything slipping in the drive (as confirmed by the direct test of the servo drives, which test is done at a much higher acceleration rate than the Mach3 signals).

There is a VFD - it is ON, but not running during the test. I have not tested with it Off.

Subsequent to my OP, I have moved all signal wires away from high current lines of any sort, and away from the VFD. Re-ran the same test and got the same results. I am really having a hard time believing that it is a noise problem at this point - it is TOO CONSISTENT . . .

[mike_Kilroy]

yes, try with the vfd off - worth the test.

I understand your saying it is too consistent and so prob not noise - I am not there so cannot say but noise does this kind of thing on no differential encoder feedback systems- hence, why I asked about differential sigs or not. I've been sent all across the USA (they want me to go to Spain now and I am resisting!) to help solve this kind of servo noise problem; my experience says having home position drift is typical of non differential encoder feedbacks. Yes, it normally continues to drift with time and running so it gets worse and worse, but it can drift one way then the other also so just get off and stay off and not continue to get worse, just drift around the right spot. anyway, most anyone can usually get away with non-differential enc feedback is about 8 feet length.... and then it is a crap shoot still. If your enc has A- and B- outputs, can you hook them up to your breakout board? It may solve the issue. But if the driveconfig program thing u mentioned runs the drive itself and it does not drift and it uses the encoder also, this sorta says it aint the encoder huh?

That is why I asked if step/dir were differential: unless the encoder is NOT used during that other program test u did, it sounds like non diff step/dir lines may be picking up the noise.

so does the other test that did not drift use the same encoder or not?

[Al_The_Man]

If your tuning software commands the drives and there is no error, only when you are running Mach,.
Your encoders go back to the drive, so if they were reading incorrectly it would show up in the tuning also.
You also need to run a measuring test in both directions to confirm whether it is a unidirectional error.
Al.

[Scott Kelley]

so does the other test that did not drift use the same encoder or not?

Yes, it does use the same encoder

[Scott Kelley]

You also need to run a measuring test in both directions to confirm whether it is a unidirectional error.
Al.

My measuring test (described in the OP) indicated that it is wholly or substantially a unidirectional error.

I posted to the Mach forum to see if anyone has, or knows of a piece of communication code that would emulate the Mach3 output, in order to eliminate (or implicate) Mach as the problem source.

[mike_Kilroy]

so it is not your encoders causing the drift, no?

so, again, is your mach3 to drive connection step/dir DIFFERENTIAL? If not, I would bet a case of beer from your description that that is the problem. make it differential. hook up the missing 2 wires.

[Scott Kelley]

so, again, is your mach3 to drive connection step/dir DIFFERENTIAL? If not, I would bet a case of beer from your description that that is the problem. make it differential. hook up the missing 2 wires.

It is not differential. This is a parallel port output from a PC. Connections are made via the breakout board and a secondary board that connects to the drives via short ethernet cables.

PC to CNC4PC C-11 break-out board via parallel cable.
C-11 board to "Simple LPT port breakout board" (CNCdrive - Servodrives and CNC components).
"Simple LPT port breakout board" to drives via Ethernet cables.

[mike_Kilroy]

ok. looked at board and see simple step and dir. so. next questions:

1) did this USED to work? Usually we do something that causes it to stop working.... can u think of a change around the time it started to creep?
2) your shielded cable 2.5' is from computer par port to breakout brd.... away from noisey other wires no doubt. how about the rj45s to the drives? they are NOT shielded since they are office style ethernet cables right? how long? can you make shorter ones? shield them? (for a test wrap tin foil around them from board to drive and clip lead this to ground?) but how long are they?
3) only 1 axis creeping? can u change rj45 cable? drive location? anything to switch a good and bad axis one part at a time?
4) did u run with vfd unpowered? did it still creep? if not, it is time to consider changing motor leads from vfd to motor to 4 conductor shielded cable.
5) breakout says u can usually set step to be active hi or lo on mach3 & drive (to match) - urs match? worth changing from hi to lo or lo to hi on both and see if your creep reverses direction - would help confirm it is noise on the step line of this axis....

[arturod]

Scott,

I received your support ticket. Thanks for the links.
I know double optoisolation can be very tricky, and we have had problems with this in the past. I can assure you this has nothing to do with the problem you are experiencing now. Depending on the board, some can optoisolated at 100khz and others at 1Mhz.

I presume this is the board you bought back in 2008. Are you just setting this up now, or was it working ok and then it started having problems?

The problem you are describing makes me think that the board is underpowered. Possible causes for this are:
- Problems with the power supply.
- External short draining power from the system.
- Damaged chip on the board, which can create an internal short, leaving other chips underpowered.

I suggest you take a look at this guide for trouble shooting this with your board:

Please let me know how it goes.

Arturo Duncan
:: CNC4PC :: iNtRo

[Scott Kelley]

Hi Arturo, thanks for the reply.

Yes, this is that same board/project, and yes, I am just finally getting around to getting it working right (procrastination is one of my strengths). I have used it a little bit for things where high accuracy is not important, but those other items that do require accuracy are piling up and it's time to fix the problem .

I will plan to check out the items you mention this evening. When you refer to "this guide", did you mean the 3 items listed, or did you mean to attach a link to something?

Also, it would help me a lot if you could provide me a schematic of the board so that I can follow the signals in and out of the board without having to trace them out (normally no biggie, but in this case the board and all wires are tied down to a mounting plate).

Thanks for the help,
Scott Kelley

[Scott Kelley]

so. next questions:

Good troubleshooting info Mike - still working through the list.

[arturod]

Scott,

You can count that this will be fixed and your machine will be reliable.

Since this is an old board and I do not kept records of the serial numbers on the boards we shipped, I do not have the serial number so I do not know the revision. Check the guide for the revision you have. You can also send the serial # and I will pull the inspection record.

I do not have an schematic in a format that I can share. I do not think this will get to this, but if required will send the information needed for testing.

Thanks,

Arturo Duncan
:: CNC4PC :: iNtRo

[Scott Kelley]

Arturo,

What is the "guide" that you are referring to?

[arturod]

Scott,

Here: CNC4PC.

Arturo Duncan
:: CNC4PC :: iNtRo

[arturod]

Not sure why it did not show:
Buffers and optoisolators are the board components that take the hit when a power surge happens in the system. In this case they can blow protecting the PC and the rest of the system. These can be easily replaced because they are mounted on sockets. Now you can order Integrated Circuit Replacement Kit for most of the boards. Be careful to align the marks in the IC with the marks in the socket. Before ordering please inspect the board for damage on the PCB, or other components, as the damage could be extensive and affected other areas of the board. If you are not comfortable with this, or you see the damage is massive, you can email me for upgrading to a new board with 25% off the listed price of the new board. This is part of the policy of not making a profit out of user’s mishaps.

Here are some recommendations to trouble shoot a board to see if the problem is associated to a blown chip:

- Make sure the signal does come into the board, the signal at the cable that goes into the board.
- Make sure the board has the outputs enabled, check the signal at the EN terminal.
- Make sure the board is properly powered, make sure the board does provide at least the rated amperage.
- Read the manual for proper configuration and other jumper settings.

If all the conditions above mentioned have been checked, then check the following:
- When buffer chips blow, many times they create an internal short. This short could not just affect the specific signal, but could create other strange behavior, like deforming the waveform that goes through other pins.

- Look for:
* Hot spots on the board or chips.
* Voltage drop at the power terminals (it could also be a problem with the power source).
* Unusual power consumption, use a multimeter to check the amps consumed.
* Disconnect all the connections, and test pin by pin. The problem could be at a different pin.

- If you have doubts on the board, please email us with serial number and we can check the inspection record.

Arturo Duncan
:: CNC4PC :: iNtRo

[mike_Kilroy]

Good troubleshooting info Mike - still working through the list.

good lord scott; I missed the C11 board interface! I thought u just had serial port to that 'simple interface' board to drives. u got lotsa stuff on that c11 board - u gotta be talking to the c11 guy like u r in my opinion.....

but I will but in just in case, as an outside unbiased observer.....

that said, spec says c11 board is good for upto 150khz step commands: r u exceeding this??? do u know what DISTANCE each step equals? if u exceed this limit the c11 has u will miss pulses and thus come up short on moves. so it is worth double checking ur max speed in IPM and what a step equals to verify u do not have this basic problem.... tell us and we will calculate if u want to eliminate this as a possible reason for short actual moves.

u still shud go thru the other checks I mentioned earlier too..... remember, knowledge is POWER..... if u want me to back out and shut up just say the word.....

[Scott Kelley]

ok. looked at board and see simple step and dir. so. next questions:

1) did this USED to work? Usually we do something that causes it to stop working.... can u think of a change around the time it started to creep?
2) your shielded cable 2.5' is from computer par port to breakout brd.... away from noisey other wires no doubt. how about the rj45s to the drives? they are NOT shielded since they are office style ethernet cables right? how long? can you make shorter ones? shield them? (for a test wrap tin foil around them from board to drive and clip lead this to ground?) but how long are they?
3) only 1 axis creeping? can u change rj45 cable? drive location? anything to switch a good and bad axis one part at a time?
4) did u run with vfd unpowered? did it still creep? if not, it is time to consider changing motor leads from vfd to motor to 4 conductor shielded cable.
5) breakout says u can usually set step to be active hi or lo on mach3 & drive (to match) - urs match? worth changing from hi to lo or lo to hi on both and see if your creep reverses direction - would help confirm it is noise on the step line of this axis....

1. No, it has not yet worked correctly.

2. Cables are very short - ethernet cables are 18" - all cables have been re-routed with no improvement.

3. X and Y axis' are creeping - Z is not. As mentioned before, a few thousandths are lost over 3" test distance, in ONE direction only - other direction is always dead on. I connected the Z axis input (Z axis works without problem), to the X axis and performed the same test. Results are the same - creeping in one direction. Seems to implicate the servo drive...

4. Yes it still creeps with the VFD disconnected.

5. The servo has step hi/lo selection, but it does not seem to work - only works with one type of step input (don't remember which), no matter how the servo is set.

[Scott Kelley]

Arturo

Board seems fine - both 5 volt lines are at reasonable voltages (5.00 & 4.85 from computer).

Signal output looks great.