Fanuc 21i-ta gridshift issues

Fanuc 21i-a t controller
Absolute digital encoders
Cincom BL12 type V machine
2 dual axis controllers (z1/x1 and y1/x2)
2 single servo amplifiers for spindles

The software for the machine was reloaded a few years ago.
Supposedly, everything was working correctly except for the Z zero postion.
This issue has been programmed around for a few years.

The guy over this machine has stated they had the machine physically apart down to the z axis mechanics.
I wonder if they dropped the ballscrew nut??? down one side when they removed the top plate and pulled it up the other side during reassembly without realizing that it would put the encoder off one revolution. Then never did a 1815 APZ procedure.

The citizen guy claims that the 1850 gridshift value is always metric and is viewed as 0.000 mm. My investigations seem to support this when making small adjustments to the 1850 z1 value. When a large adjustment is made unexpected changes happen.

The factory 1850 value was set at -4076 in may of 2001. The 1850 value was set at 1000 on 12-13-2013. Currently the -4076 value nor the 1000 value are correct according to alignment procedures.

I investigated the possibility of the value being detect units. My thinking was that the multiplier may be close enough at small increments to make detect units and 0.000mm units yield close results. When doing the math on the detect units it still falls victim to the unexpected hard measurement values.

I have resorted to making small increment changes in the 1850 paramter to find a good place to for usability. I have pinned it down to a value of -650 or -649.

This is a fixed headstock cincom with a stopper for barstock feed control. In the Z1 axis - takes you closer to the main spindle while positive takes you further away. (backwards from most cincom machines).

At an 1850 value of -649 we are .085" from the procedural value of hard measurement.
At an 1850 value of -650 we are more than -.1181" from the procedural hard measurement.

Why is there such a large value shift when cincom claims that the value is only .001mm different?
Why would the abnormality happen at such a specific number and be repeatable?
I am afraid to do the 1815 APZ procedure because currently we have a usable if somewhat downgraded machine. At the same time though maybe that's all thats wrong with it?
My thought is that maybe and absolute reference mark is at the -649/-650 postion mark on the encoder and that throws it off some how.

I have tried homing the machine within one revolution of the manual zero return mark and it doesn't seem to make any difference. I have been through the entire parameter manual and can't link any bad values to the issue. Maybe it is something simple I have no knowledge of that someone with more expierience can answer easily. I believe I have done my due diligence with researching parameter manuals, looking at other 1850 threads on this site and any others google will pull up. I apologize if a similar issue has already been answered and I could figure it out from those answers.

I am not familiar with this machine but I am curious why you are changing the grid shift parameters on a machine with absolute encoders. I sounds like you are on the edge of the grid and jumping to the next one. I would physically position the slide at the home position, zero the grid shift in parameter 1850 and reset your 1815.4 APZ for that axis.

Maybe angelw/Bill, vancbiker, yardbird or some of the other smarter ones can help explain this. I have noticed this quite often on several machines lately where there are numbers in the grid shift and the axis also use absolute encoders or scales. I'd be interested in learning why they do this. I figured that it was an oversight as some builders use generic copies of parameters and logic when setting up similar machines.

Sounds like you are "splitting" the grid with the 649, 650 shift. Usually you need to change the relationship of the screw and the motor by approximately 1/2 turn then retry setting the shift.

Just for my own curiosity, is the ballscrew pitch 5mm?

Vancbiker said:

Sounds like you are "splitting" the grid with the 649, 650 shift. Usually you need to change the relationship of the screw and the motor by approximately 1/2 turn then retry setting the shift.

Click to expand...

Doesn't that only apply to dog and limit switch arrangements?

I appreciate what your saying for that style, get an axis were the switch point is horribly close to the encoder reference marker, and it will jump a turn, if the switch is just marginally slow to make / break. Its the one and only time I've seen the word "Hysteresis" used. (And it wasn't by me).

That shouldn't apply here were he has absolute encoders. Where you can make any encoder pulse the zero count.

JLaugh said:

The citizen guy claims that the 1850 gridshift value is always metric and is viewed as 0.000 mm. My investigations seem to support this when making small adjustments to the 1850 z1 value. When a large adjustment is made unexpected changes happen.

Click to expand...

I'd concur with that, the units are always in 1 micron, but I've never seen an axis jump like yours 0.2041", depending if your -648 or - 649. (.085" / -.1181"). That's really weird and wonderful. I spent a few minutes with a calculator, I can't see any rhyme or rhythm to the problem you are having.

It cant be a setting issue. I don't have a manual in front of me, but you can set that to either +/- 4095 or +/-999999

I wouldn't hesitate to do the Absolute Zero reset 1815.4 APZ. I've never understood why some guys leave the 1850 grid shift parameters as set. The axis is lost anyway. I zero them out, then do the APZ, fine tune it if need be with grid shift and drive on.

I read you are hesitant to do the Abs Zero Reset (1815.4). Its not that hard. I'd do it just because your grid shift seems to be so screwed up.

Other Brother said:

I have noticed this quite often on several machines lately where there are numbers in the grid shift and the axis also use absolute encoders or scales. I'd be interested in learning why they do this.

Click to expand...

I'd note this guy had a setting of -4076 as of May 2001. That surely must have been a Factory setting.

I have a theory, having visited several Japanese manufacturers. Systems / Fanuc are commissioned by the electrical / electronic guys. They get the system to function. And set a rough Zero, that ties in with the software over travel parameters.So the machine is safe.

Between the machine coming to life, and going out the door, the mechanical / Q.C guys have to make two spindle's look at each other, or a turret get to true centre line. Those guys use the Grid-shift, as its all ready in the ballpark. And less power downs at the Case level, as you would get with 1815. 1850 is virtually instant.

Down here, in the field, on a one-off machine, I cancel Grid-Shift and just get the zero position correct.

Regards Phil.

machtool said:

Doesn't that only apply to dog and limit switch arrangements?

Click to expand...

I'm pretty sure this can happen with absolute systems. This gets tied in with the reference counter parameter somehow. Here is a quote from a Fanuc manual I have about the reading of the absolute encoders at power up after setting a new grid shift.

"APC communication is performed to establish the current position. The distance from the origin is calculated at that time. The distance from the grid is also calculated. This value is assigned to a reference counter. This enables the grid to be established. The sequence of step 1a is therefore not required".

Step 1a describes a dog type reference return system.

machtool said:

I appreciate what your saying for that style, get an axis were the switch point is horribly close to the encoder reference marker, and it will jump a turn, if the switch is just marginally slow to make / break. Its the one and only time I've seen the word "Hysteresis" used. (And it wasn't by me).

Click to expand...

The reason that sometimes you "split" a grid with dog and switch/sensor reference can be hysteresis in the switch or sensor. The other reason is within the Fanuc system itself. Again another quote from my document.

"After a signal indicating indicating that the deceleration dog has been turned off is output to the hardware, it takes some time (up to 20ms) for the software in the digital servo to recognize the signal. In addition, the time required for the recognition varies. As a result , if the deceleration dog is too close to the origin, the software sometimes fails to capture the signal".

Vancbiker said:

I'm pretty sure this can happen with absolute systems. This gets tied in with the reference counter parameter somehow. Here is a quote from a Fanuc manual I have about the reading of the absolute encoders at power up after setting a new grid shift.

"APC communication is performed to establish the current position. The distance from the origin is calculated at that time. The distance from the grid is also calculated. This value is assigned to a reference counter. This enables the grid to be established. The sequence of step 1a is therefore not required".

Click to expand...

I dont agree with the part about having to move the motor / encoder half a turn.

"APC communication is performed to establish the current position

Click to expand...

I'm pretty sure that just means in my best Japanese / English conversion, The control just knows the current position, due to back up battery of the encoders. Despite the position where you hit the E.Stop. I.E it catches the position of encoder, even with the power off. Most tyipically with Z axis that drop. Absolute encoders still know where it's at.

Vancbiker said:

The reason that sometimes you "split" a grid with dog and switch/sensor reference can be hysteresis in the switch or sensor. The other reason is within the Fanuc system itself. Again another quote from my document.

"After a signal indicating indicating that the deceleration dog has been turned off is output to the hardware, it takes some time (up to 20ms) for the software in the digital servo to recognize the signal. In addition, the time required for the recognition varies. As a result , if the deceleration dog is too close to the origin, the software sometimes fails to capture the signal".

Click to expand...

It specifically mentions a "deceleration dog". Surely theres no dog's or switches, on an absolute axis?

Vancbiker said:

Step 1a describes a dog type reference return system.

Click to expand...

Bro's we need to get on the same page. Woof, Woof. No dogs on an absolute axis.

It is possible he is at the roll-over point of the encoder. Even on a 32,000 count absolute encoder there is a roll-over to zero. There may also be a rotation count pulse. My suggestion is to look at the encoder count in the parameters (if available on fanuc, I can't remember). However, I would also do the APZ reset. This will re-sync the encoder count to machine zero. I also never use grid shift on Fanuc 21i's as there is no need with the APZ procedure.

Tonytn36 said:

It is possible he is at the roll-over point of the encoder. Even on a 32,000 count absolute encoder there is a roll-over to zero.

Click to expand...

Wouldn't the case be, on a 32,000 count encoder, thats over the entire travel of the axis, I:e multiple turns? Fill me in, who's making a 32,000 pulse per rev encoder?

With due respect, he's having problems at - 649 or - 650, on a Fanuc 21i

I'll stroke your cock with this 32,000 pulse thing. That's really disingenous. This bloke would be lucky to have 4k to play with.

machtool said:

It specifically mentions a "deceleration dog". Surely theres no dog's or switches, on an absolute axis?

Bro's we need to get on the same page. Woof, Woof. No dogs on an absolute axis.

Click to expand...

I should have split up my reply better to not add confusion.

When I mentioned the dog type reference, it was an attempt to explain one cause of "splitting" a grid.

The Fanuc manual I am referencing this info from is outlining differences between behaviors of absolute and incremental encoders. That is why the end of the quote says "The sequence of step 1a is therefore not required". Earlier in the book, step 1a described dog type reference. It's just a roundabout way the manual tells us that abs encoders do not need a dog type reference but still establish a "logical" grid based on a calculation.

machtool said:

Wouldn't the case be, on a 32,000 count encoder, thats over the entire travel of the axis, I:e multiple turns? Fill me in, who's making a 32,000 pulse per rev encoder?

With due respect, he's having problems at - 649 or - 650, on a Fanuc 21i

I'll stroke your cock with this 32,000 pulse thing. That's really disingenous. This bloke would be lucky to have 4k to play with.

Click to expand...

We have lots of absolute encoders in our facility with 18-bit (262,144 steps) single rev with 32,768 revs available before roll-over. These are TR encoders. Obviously not Fanuc. Was just mentioning a possibility.

While I have not dissassembaled the machine myself, from what I know of the system involved and, from what the maintenance manual says I don't believe there are any dogs or limit switches.

The reason I am afraid to do an 1815.4 procedure is the massive jump at -649/-650 makes me think there could be an issue with the encoder itself. The machine is currently operating and has orders pending for atleast the next 3 weeks. If I do an 1815.4 and we find out that the encoder is faulty and must be replaced it could cause issues with getting our orders out.

Our staff including myself is very young compared to what I would think the industry average is. We are very methodical in our approach. I hope this is conveyed in my original post. Being prudent and cautious is how we avoid missteps and make sure we have the time to resolve issues before they effect our customers.

At the -649/-650 setting the "jump" is even more than .2041. The z1 axis setting procedure says to put the z1 axis at machine postion z1=0.0. The face of the main spindle (just to be clear I know this means the spindle bore's face and not the collet) to the top plate measurement should be 43.5mm (1.7126"). At -650 we have to back out to machine position .085" for the gauge blocks to fit. At -649 we move to -.1181 machine postion and are still around .100" from touching the gauge blocks.

Currently we are turning 304SS tubing into precision spacers at 13s per part so we want to finish that before we have any down time.

How would I figure out how many detect units or pulses are in a turn of the encoder?
Can an encoder just "go bad"?
This encoder is probably about 15 years old.
If the enocder is bad would the machine even work at all?
What kind of fault in the encoder could cause the jump?

If there was some kind of fault in the encoder that could show up during an 1815.4 procedure?

Could it be as simple as someone took the top plate off the ball screw "nut" letting the "nut" down the left side of the screw and then pulling it up the right side during reassembly, putting the encoder off one revolution? I'm leaning torwards doing the 1815.4 but, want to be sure before I do so.

Is this a belt driven axis?

If so - just jump the belt X amount of teeth.

If direct drive, it likely has no keyway, and you would want to scribe a line in the shaft and connector, and slip it to your expected location as closely as possible. Then you would be closer to ZERO and further from the possible "bad" spot in the encoder. (Kan't really be bad if it's making good parts w/o comping most other Z moves, so ....)


I wouldn't be scared of the ZRN procedure, but I have experience with it too, and I can appreciate your reasoning here.
The mechanical procedure of the belt of yoke trick may not cause as much hesitance. (in the older folks especially!)



--------------------

Think Snow Eh!
Ox

It is a direct drive axis.

Everytime we edit the 1850 we have to powerdown and then do what cincom calls a manual zero return (alarm page calls it a "Z axis ZRN")

It's on a special page in the maintenance section. It just dawned on me, surely this isn't a built in shortcut to doing a 1815.4? If so that rules out that fixing the problem.

I've never used BOTH 1850 and 1815 on the same machine before.


--------------------

Think Snow Eh!
Ox

JLaugh said:

It is a direct drive axis.

Everytime we edit the 1850 we have to powerdown and then do what cincom calls a manual zero return (alarm page calls it a "Z axis ZRN")

It's on a special page in the maintenance section. It just dawned on me, surely this isn't a built in shortcut to doing a 1815.4? If so that rules out that fixing the problem.

Click to expand...

So how do you reference the slide after editing 1850 and powering down? Does 1815.4 turn to a zero? Anytime you change a grid shift on Fanucs you have to power down.