[saabaero]

Has anyone had to adjust the calibration on any of the axes on their mill.

I checked my "X" axis today and it is off .0085" over 16" and .002" at 4".

I verified it with my $6000 B&S height gauge and digital caliper and they read within .0002" of each other. On the mill I used the electronic touch probe and a edge finder and also got within .0002" of each other.

Is there a parameter in the .XML file that needs to be tweaked?

[philbur]

I think you need to check number of steps per unit in the motor tuning configuration file. It should read 10000. I think if this is correct then your problem is most probably mechanical.

What exactly was your setup for measuring the error. Photo would be good.

Phil

Has anyone had to adjust the calibration on any of the axes on their mill.

I checked my "X" axis today and it is off .0085" over 16" and .002" at 4".

I verified it with my $6000 B&S height gauge and digital caliper and they read within .0002" of each other. On the mill I used the electronic touch probe and a edge finder and also got within .0002" of each other.

Is there a parameter in the .XML file that needs to be tweaked?

[Davo J]

This may help,it out seems easy enough, I am new to all of this (CNC)I have just bought my ballscrews/double nuts for a small knee mill (HM52 same modle as chich2 here on the forum)so I have a long way to go before I am calibrating
[ame="http://www.youtube.com/watch?v=XkO5tc-jSxw&feature=channel_page"]YouTube - Improved Mach 3 Axis Calibration[/ame]
Davo

[saabaero]

Thank you for the replies!
I wasn't aware of an axis calibration utility for the Tormach version of Mach so I went into the .xlm file and searched for the 'steps per unit' values for the axes. They were set at 10,000 for each linear axis. Since whe machine was moving a .0005" per inch less than it was supposed to I increased the value for the "X" axis to 10,005 (2000 steps per rev. * 5 pitch screw = 10,000 steps per inch * .0005 = 5 steps). The strange thing is that made the problem worse in that it was now .001" per inch short.

So I went in the other direction and changed the value to 9995. That brought the "X" axis travel to within .0002" at 16 inches!

In line with that I checked my "Y" axis also and it was short by the same amount my "X" axis was so I changed it to 9995 also and rechecked it over 8 inches of travel. It was then right on.

The "Z" axis checked out good!

[philbur]

I recommend you find the root cause.

Phil

Thank you for the replies!
I wasn't aware of an axis calibration utility for the Tormach version of Mach so I went into the .xlm file and searched for the 'steps per unit' values for the axes. They were set at 10,000 for each linear axis. Since whe machine was moving a .0005" per inch less than it was supposed to I increased the value for the "X" axis to 10,005 (2000 steps per rev. * 5 pitch screw = 10,000 steps per inch * .0005 = 5 steps). The strange thing is that made the problem worse in that it was now .001" per inch short.

So I went in the other direction and changed the value to 9995. That brought the "X" axis travel to within .0002" at 16 inches!

In line with that I checked my "Y" axis also and it was short by the same amount my "X" axis was so I changed it to 9995 also and rechecked it over 8 inches of travel. It was then right on.

The "Z" axis checked out good!

[benji2505]

I recommend you find the root cause.

Phil

I could not agree more with Phil.

The first question I have was, whether the machine was always off by that amount or whether it occured recently

[StephanWenger]

Hi,

could this be a temperature related thing? At room temperature, steel has a thermal expansion coefficient alpha of somewhere between 10 and 20 E-6 (typically, heavily alloyed steels are higher), based on the Celsius/Kelvin scale.

If the ball screws were machined to be precise at 40 degree C (around 100 F), and your shop is very cold (say, 0 C, which is freezing), you have a temperature difference of 40 Kelvin. Assuming an alpha of 15 E-6, your thermal contraction would be 40 * 15 E-6 == 0.0006. Over four inches, you would loose 4 in * 0.0006 == 0.0024 in.

Of course, one also has to factor in the measurement tools, which are also subject to thermal expansion/contraction. However, quality tools are made of materials with as low a thermal coefficient as possible.

Regards,
Stephan

[philbur]

The error might not be something you would spot unless you specifically looked for it. I think a description of how the error was measured would be a good starting point.

Phil

I could not agree more with Phil.

The first question I have was, whether the machine was always off by that amount or whether it occured recently

[philbur]

Getting a height gauge and a digital calliper to give the same reading within 0.0002" sounds either coincidental or spectacular to me. No offence meant.

Phil

[saabaero]

Getting a height gauge and a digital calliper to give the same reading within 0.0002" sounds either coincidental or spectacular to me. No offence meant.

Phil

What do you mean coincidence??? Wouldn't you expect my $6K Brown & Sharp height gauge to be as accurate as my import 24" digital caliper?!

Actually, I was surprised also. In addition, the Tormach touch probe and a mechanical edge finder were within 2 tenths on the CNC.

I would attribute the travel inaccuracies to either ball screw tolerance or ambient temperature variation. I expect that the machines are all sent out with the default 10,000 step per inch parameters for all linear axes. The interesting part is that the X & Y axes showed the same amount of error but the Z axis seemed good.

I noticed in the mach 3 setup that there is a calibration routine where you tell the machine how far to move and then measure and enter the actual travel. The software automatically calculated the correct step per inch value needed.

I have had the machine since last fall but have been learning the machine and have machined small parts where the errors would be difficult to see. This is the first oppurtunity I have had to check accuracies.

Over the weekend I did notice a random shift in my "Y" origin when returning to it after machining a part. Some times it was within tenths and other times it was as much as .004", .010" up to .040" off. I am currently worrking with Tormach tech support to get that issue resolved. (so far, I checked the motor coupling tightness and for loose electrical connections)

[philbur]

Yes maybe but not necessarily visa versa. I wouldn't expect you to measure accurately to 0.0002" with 24" digital calliper, hence the coincidence. Also with your spiffy height gauge on a surface plate 0.0002" is one thing, using it to measure x travel is quite another. Instrument resolution/precision of course has potentially little to do with the precision of a particular measurement.

I'm still interested in how you used your instruments to measure the x travel.

According to the Tormach documentation quote:

"P4 allows 16 um error within any 300 mm. In imperial
units, this is 0.0006" per foot. Cumulative error at 900
mm is limited to 22 um (0.0008" at 35")"

So if your measurements are truely ballscrew errors then you should request replacement ballscrews.

Phil

What do you mean coincidence??? Wouldn't you expect my $6K Brown & Sharp height gauge to be as accurate as my import 24" digital caliper?!

[philbur]

This is definitely not ballscrew tolerance.

Also messing around with steps per unit length and ballscrew mapping is a last resort for machines that have low quality ballscrews. This should not be the case with P4 and definitely not for fixing 0.04" errors.

You need to get your magnifying glass out and put your deerstalker on.

Phil

Over the weekend I did notice a random shift in my "Y" origin when returning to it after machining a part. Some times it was within tenths and other times it was as much as .004", .010" up to .040" off.

[saabaero]

This is definitely not ballscrew tolerance.

This should not be the case with P4 and definitely not for fixing 0.04" errors.

You need to get your magnifying glass out and put your deerstalker on.

Phil

Actually, the .040" error is random and I'm sure is another issue. The "X" axis which was moving short initially repeats fine.

As far as the accuracy of my measurements.. I also verified my results with gauge blocks and the numbers all jive so I'm sure the X & Y axes were moving short by .0005" per inch each.

I would agree that I would expect the accuracy of these ball screws to be better but I'm not sure what temperature they are rated at.

[BobWarfield]

Not surprised to find your machine may benefit from a slight adjustment in steps per inch, that's not uncommon.

Do be sure to measure your backlash. A lot of evil is possible when there is backlash present.

Cheers,

BW

[philbur]

As somebody pointed out you would need something like 40 degrees C above ambient to account for your error, so repeat the measurement with the screws cold.

Trying to trace your ballscrew error, while you have random errors of upto 40 thou, sounds like a futile task to me.

Phil

Actually, the .040" error is random and I'm sure is another issue. The "X" axis which was moving short initially repeats fine.

As far as the accuracy of my measurements.. I also verified my results with gauge blocks and the numbers all jive so I'm sure the X & Y axes were moving short by .0005" per inch each.

I would agree that I would expect the accuracy of these ball screws to be better but I'm not sure what temperature they are rated at.

[saabaero]

Not surprised to find your machine may benefit from a slight adjustment in steps per inch, that's not uncommon.

Do be sure to measure your backlash. A lot of evil is possible when there is backlash present.

Cheers,

BW

My original thought was that every machine built would be slightly different (due to manufacturing tolerances) and would require it's own individual tweaking to the parameters. With something that is supposed to be as accurate as a CNC you would think that would be done individually at the factory. However, I am not sure that is the case. Take for example someone that would purchase a machine without a PC and use their own. They would install the PCNC software which would contain a default ".xml" file. Likewise my guess is that even when the complete package is purchased from Tormach, the PC/software and machine never see each other until the package is shipped out the door so there is no individual calibration being done.

In the factory inspection sheet, I do not see where actual vs theorethical 'steps per inch' is even being checked.

[BobWarfield]

My original thought was that every machine built would be slightly different (due to manufacturing tolerances) and would require it's own individual tweaking to the parameters. With something that is supposed to be as accurate as a CNC you would think that would be done individually at the factory. However, I am not sure that is the case. Take for example someone that would purchase a machine without a PC and use their own. They would install the PCNC software which would contain a default ".xml" file. Likewise my guess is that even when the complete package is purchased from Tormach, the PC/software and machine never see each other until the package is shipped out the door so there is no individual calibration being done.

In the factory inspection sheet, I do not see where actual vs theorethical 'steps per inch' is even being checked.

They do say they run a laser interferometer check on 100% of the screws, so they should know if the screw was in spec. This one is not.

With all that said, if you establish there is no backlash or other lost motion (perhaps from a lose mount), and recalibrating steps/in works reliably, it probably isn't worth the pain of replacing them.

Checking for backlash would definitely be my next step. If the backlash is negligible, and you have corrected steps/inch reliably, the random error would seem to be more electrical perhaps? Another possibility would be a loose gib, which can lead to some unpredictable goings on.

It's good that you're measuring these things. I suspect very few hobbyists get down and dirty to carefully check the performance of their machine, particularly if they have a brand new turnkey like a Tormach. Yet, there is a lot to be learned there, and as you are seeing, some potential surprises as well.

Cheers,

BW