V-grove & rollers, dimensional accuracy - in Z plane

Folks,
I was testing my BLtouch repeatability and discovered something interesting.

I think my results show z height changes due to the V-grove wheels being not perfectly circular.

For example, if I repeatedly probe a single point on the bed, I can measure very low standard deviation, in the region of 0.005 to 00001
But if I perform several G29 commands (grid leveling), then the calculated std dev for each point in the grid, the std dev falls to something like 0.07

In other words, in a static test where the wheels are NOT required to move, the deviation is very low.
But when the X or Y location change, the deviation increases dramatically.

Before anyone asks, yes I have factored in the bed profile being not flat.
The test measures multiple G29 output's (9 grid points) then measured deviation of each grid point set.
e.g. perform 5 runs of G29, then get deviation of grid point 1 using the 5 sets of data, then repeat for each grid point.

I just want to know if my findings are correct, do the v-grove wheels cause changes in Z height due to the imperfect shape?


Thanks

Ray

Try probing the points on the grid individually to see if there is any variation from what you get on the G29. Also, try repeating the same point but moving off and then back in each direction - say +20mm on X and back, -20mm on X and back and repeat for Y. I would expect different figures for each but somewhat less than 70µm - more like 10µm or less would be sensible. If you find a bad spot then see if you can feel slop between the wheels and rail or if the wheels can be turned easily by finger pressure at that point.
Although it is possible that errors of the order of 70µm could be caused by out of round errors in the V-groove wheels, other possible origins are cutting errors in the V-groove of the extrusion, cogging of the belt in a pulley - accentuated if the belt is incorrectly aligned.

The way to indicate if it is the wheels is to probe a number of points along one axis and see if the deviation maxima occur roughly a wheel circumference apart.

Mike

Excellent stuff Mike, I was worried that my post was not understandable!

I will indeed try your suggestions, and yes it could be the things you suggested.
Thanks for the advice, I particularly like the idea about test points on one axis to check for a wheel signature!

Come to think about it I recall the wheels had a divot or dent in them from sitting on the shelf for too long when I received the printer.
My solution was to swap them about to disguise the divot, but in hindsight perhaps that didn't fully fix the issue. There may still be deformity in the wheels as the factory had them way too tight!
I may need to order a new set.

Well Mike & all.. I finally solved the mystery!

I had followed the rabbit down the hold for too long, so I decided to do a full review of all my Marlin settings

I had eliminated V-Grove issues and such, nothing was wrong there.

Then I stumbled upon
#define RESTORE_LEVELING_AFTER_G28
and presto! like magic the results were perfectly repeatable.

I figure that G28 was homing just fine, but then I was executing G29 it re-homed and made a new set of records.
Then when I ran G29 again, it built upon the previous final homing point and started again, but was using the final homing point as it's 'base home'
I'm not explaining it too well, but at least it's resolved.