0.005 error randomly accumulating over 30.000 inch travel, solutions etc ??

short jogging results are consistent, plus or minus 0.0001 to 0.0003, within limit for a 50 yr old machine. The random accumulation of errors across long travel however have me concerned. At distances under 1-2 inches the error is non existent but over long runs it is accumulating. I have looked with a scope and the signals look really good, so I am unsure as to where I should look , it's a 3.00 tpi running at 300ipm with a 4096 ppr with index. 

Is is possible that the quadrature count is not fast enough to measure all pulses or is there  known math errors is with certain encoder type ? 

How did i measure the error, I put my Mititoyo Rapid tenths indicator on lath bed, zero'd out and then set g28, then I command out to z32 and call it back to g28, it stops at the dro indicated 10.5013 each time, but the indicator shows the error accumulating by 3-6 thousandths each time it comes back to g28 position. no noticeable Ferror showing in halscope. However moving the axis 1 2 or 4 inches and calling back to g28 does not produce such a large error I see no measureable error. I could understand variation in the length of the ballscrew, that's expected, this is strange as it is accumulating in a semi random way. 

It sounds like something is moving. Is it continuing to accumulate on multiple runs? If not, what about thrust bearings. The Cinturn is a box way machine, are their slippers under the saddle that could be binding?

it is accumulating every run, and the machine is tight and clean with excellent ways and screws. I am going to check the slipper tomorow, but the error is stacking up. short move testing indicates it is unlikely to be a lash or a way issue specific to the stopping area in the g28 command. however the exact number of encoder revolution may shed some light on this. I am looking over the motion control code, I suspect the floating point is being truncated, that might offer a plausible explanation to the error accumulation. if it was just randomly 5 7 3 8 3 I would accept that it is hardware, in fact some of that lash is to be expected, but when it's going up adding almost exactly 5.5 thou every time, that's not mechanical. was just outside remeasuring, it's 5.5 thou every time

this is the EXACT number of required encoder revolutions for this move

241.5686724609375 <

---

this is a very long number

I thought it was a 4 pitch screw in 1976.

5 mills out of 30 inches is a relatively huge error
Can you check the shaft position?

(a 5 mill error should be about a 5.4 degree shaft position difference)

Is it possible you have a loose encoder wheel/coupling?

I'll check again, I'm bringing my 4mhz scope home from work today as well. i the spindle encoder is dropping out around 2600rpm, encoder appears to be sending valid data.

4096ppr there as well , I'll call you later