Ian Tolond wrote:
how so ?
If you measure all 6 distances between the anchors (including diagonals) then
something like
will let you enter the measurements and spit out the coordinates of the anchors
(and with the 2nd diagonal, give an estimate of how good your measurements are)
David Lang
what are the optimum frame anchor dimensions? Metric)
also the darker yellow and lighter yellow meaning?
the laser level was based off with the spoil sheet base (i.e we levelled the frame spoil sheet base with the horizontal line of the laser level -then proceeded to mark anchor centre points relative to this.)
update:
Set retraction force to 700
Changed Maslow_Acceptable_Calibration_Threshold. From 0.450 to 0.3
Ran find anchors - success
Set retraction force to 500
Ran find anchors - fail
Set retraction force to 600
Ran find anchors - fail
Set retraction force to 700
Ran find anchors - fail
Noticed when find anchors failed, the sled was often lower right and when find anchors is run it fails - this is not typical if the sled in pretty close to centred
Ian Tolond wrote:
what are the optimum frame anchor dimensions? Metric)
no strict numbers, but your frame is narrow for itās width
also the darker yellow and lighter yellow meaning?
just different causes for the limit.
the ones that bulge out from the edges are due to the limit of the angle of the
belt between two adjacent anchors (how close can it get to the side before the
arms hit the frame opening apart from each other)
the ones in the corner are from the opposite anchors (the area in the top left
is looking at the top right and bottom right belts, how close can they get to
each other
each band in these is 1 degree of belt error, and the table above the image has
a charge to show how much the belt length is affected by the inability of the
arm to swing into the right place
David Lang
700 is a great value, thatās nice and low.
What are the issues that you are seeing now? That sounds like a success to me.
Testing the parameters:
What is the nett effect of changing to 0.3 (from 0.45) in real terms?
I did observe a noticeable lift on the sled base on the lower left patternmostly during calibration
It means that the criteria for acceptance is lower, like a lower pass mark. So, the machine can go onto the next level of testing.
understood - but what is the manifestation of that change - are we talking a 2-3mm variation on cuts?
I donāt think i has been quantified. Did you try the test version of software for Find Anchors?
Ian Tolond wrote:
understood - but what is the manifestation of that change - are we talking a
2-3mm variation on cuts?
with the old code (i.e. anything thatās been released already) the fitness score
is 1/(average error in anchor positions)
so a fitness of 1 is saying that it thinks that the belt lengths it measured are
averagimg about 1mm of the anchor positions it calculated
a fitness of .5 is saying that it thinks they are avaraging around 2mm from the
reported anchor positions
what that variation results in as far as router accuracy, we donāt have
measurements.
the errors on a maslow-style machine positioning are not as neat and predictable
as they are on traditional gantry systems. they tend to be curves, and weāve
seen them be much larger than expected near the edges. finding someone who can
take the time do do a systematic test and report the results has just not
happened much.
David Lang
Uploaded the firmware and index. HTML (pre-release version)
Maslow-serial (4).log (24.6 KB)
Maslow-serial (5).log (57.6 KB)
Maslow-serial (6).log (38.6 KB)
Maslow-serial (7).log (54.7 KB)
Maslow-serial (8).log (83.6 KB)
try entering these values and then using the machine (skipping calibration)
TOP LEFT (TL)
X: 21.09 Ā± 1.52
Y: 2375.51 Ā± 4.08
TOP RIGHT (TR)
X: 2920.39 Ā± 3.91
Y: 2363.10 Ā± 4.41
BOTTOM RIGHT (BR)
X: 2914.97 Ā± 3.85
Y fixed at 0
this is the results from pasting the clbm line into
Maslow Levenberg-Marquardt Calibrator to match the
new logic. these are far from the least accurate results Iāve seen.
David Lang
