Odd measurement error - Cannot locate the source

The setup:
I recently rebuilt/modified a Tree Master Journeyman 425. I replaced the old servos with stepper motors (with encoder feedback to the drives). I had to setup the new Centroid ACORN controller I installed, among the settings I had to enter was the "steps per revolution" and "turns per inch". The steps per revolution I got off the stepper drives (in this case 3200). For the turns per inch I used a dial indicator with a 1 inch travel. I setup on the x axis and took a few measurements and tuned the turns per inch vaule before arriving at 10.113 revolutions per inch. I did the other two axis Y and Z the same way. Made sure it was repeatable and moved on.

The Problem
I was drilling some holes on a larger project 25 inches apart and noticed things were not lining up. I do not own a precision measuring device that will span 25 inches, but a simple tape measure was enough to show about 1/16th inch of error.

I made sure the vice was trued up correctly then I got out some gauge blocks, squared it in the soft jaws (yes, I know you shouldn't use them for this) and using the probe took a measurement on the 1 inch gauge block. It came out to 0.9997 inches and checking the turns per inch they were still 10.113 repeated this a few times and got similar results, sounds good, right?

Next I put the 4 inch gauge block in and preformed the same test to get 4.0128 inches. I repeated it several times with results varying only 0.0003 different.

Finely I recomputed based on this information for the 4 inch gauge block to read accurate, the turns per inch value would need to be set at 10.144 turns per inch. I entered the updated value and was getting measurements of 4.0001.

I have no idea what would be causing this type of error. I've never see something like this in half dozen or so machines I refitted, built, or worked on. How can for 1 inch the turns per inch need to be 10.113 and at 4 inches it need to be 10.144? The only thing I can think of is the turns per rev are incorrect, but I am unable to think a good way to check this?


My apologies this is a long post.
(I didn't post this in the Tree CNC Machine section of the forum as I do not think it is relevant to the machine)

Machine frame = Tree 425
Controller = Acorn by Centroid
Steppers = 34HS59-6004D-E1000
Drives = CL86Y (shielded control lines between controller and drives)
Probe = DrewTronics

Normally computing over a longer distance is better. I would set to 10.144, then spot 2 divots at the maximum distance your machine will allow and measure with a tape measure. It is possible that what you are seeing is ball screw wear. Or just a bit of error in the original calculation.

I assume that you have a 2:1 belt reduction on the drive, and a 5mm lead ball screw.

...is the Ball Screw 5MM(.1968) pitch or is it a .200 Inch pitch. An error of .0032" x 25" = .080"

DJ

Perhaps an assumption on my part, but the whole machine was loaded with metric fasteners (I believe Tree machine hardware was a product of Czech Republic and was outfitted with its electronics in the US) I guessed it was metric and a 5mm pitch.

Please forgive my ignorance, this isn't something I've encountered before. Why would it matter what the pitch progression of the screw is? As the pitch is constant, shouldn't that be covered in the "turns per inch" value?

the only part I put on the machine that was custom was the timed pulley for the stepper motor. I'm sure the diameter is not exactly what the original was, but again, if the diameter is constant would it matter?

Something is causing me to lose distance at what appears to be a linear progression.

1" gauge block = 0.9997 measured
2" gauge block = 2.0044 measured
3" gauge block = 3.0090 measured
4" gauge block = 4.0128 measured
The machine is traveling less distance with each increment, so so Jim's suggestion, I can take long cross bed measurement but that would only be a correct "turns per inch" value for that specific distance.

The only thing I can think to do would be to purposely alter the "turns per revolution" value to correct for error factor that appears to increasing in a linear way.

Since the measured distance is increasing, you can do a bit of math and get it really close

Let's take the longest measured distance 4.0128

4.000/4.0128 = .99681

So 0.99681*10.113 = 10.0807

And if that doesn't work, because I always forget which way to go :)

1/0.99681 = 1.0032

10.113*1.0032 = 10.1453