One more try then I am giving up

[Ian_Tolond]

**At the end of my patience with this device.
**
Calibration:

• Anchors on the vertical table are symmetrical and aligned by laser level centres
• The table has been fabricated within Table Spec, as shown on the Maslow website
• Anchors are printed as per the not-shop files in PLA
• Pins are as per not shop from Amazon
• Dust extractor is printed as per the not-shop files in PLA
• The router is the European version of the DeWalt router
• Assembled the sled as per the guide, checked and rechecked (appears no issues)
• Sled anchor points move freely with no excessive play
• Software/firmware updated to the latest version
• All tests on the Sled app appear fine and within tolerances.

However, I cannot seem to get this thing to calibrate, regardless of what I have tried, and it is excessively time-consuming compared to other equipment in my shop, which requires calibration.

Maslow-serial (5).log (42.2 KB)

Maslow-serial (6).log (81.2 KB)

[ian_ab]

The Find anchors is supposed to be able to work with any frame and shape, however I have found the closer the initial measurements are to the physical dimensions the more likely the find anchors will work.
So, measure the distances between anchor points and enter the results directly into the maslow.yaml file. I don’t recommend doing this in an editor, as the file format is critical, instead do this:
From FluidNC pull down, then Config Items, go down to /kinematics/MaslowKinematics/tlX and enter the height and width values, using blX,bly as 0,0 as the zero reference.
For a perfect rectangle, tlX is 0, tlY is the height, trX,trY is height,width and brX,BrY is width,0. Also make sure your Z values are consistent with their relative arm height. Set each value then save (at top). All measurements are in mm.
After entering these, you can try Apply Tension and see if it moves when you jog, or run the find anchors again.

[Ian_Tolond]

@ian_ab

Also make sure your Z values are consistent with their relative arm height

measured from the base of the sled to the (mlower/middle/top) of the arm?

[bar]

It’s looking really really close. From the log:

Calculated Fitness Too Low (0.3985303 < 0.45). Retry 3/10...
Find Anchor Values:
Fitness: 0.3985303

It seems like something might be slightly off still, but 0.39 is not too bad at all.

If you are frustrated and just want things to work you can lower that threshold in settings.

Try lowering that setting to something lower like 0.3 and it should pass.

The tradeoff is that you will probably not get as accurate of cuts, but at least it should get you unstuck

[dlang]

Ian Tolond wrote:

@ian_ab

Also make sure your Z values are consistent with their relative arm height

measured from the base of the sled to the (mlower/middle/top) of the arm?

from the bottom of the belt end when it’s fully retracted to the bottom of the
belt end when it’s attached to the anchor.

we have had a couple reports of people who said that the having the belt
extension lenth set too high had some effect on their calibrtion, you need the
belts to extend out enough to reach the anchors, but don’t want it to be a huge
amount beyond that.

David Lang

[dlang]

@bar what happens if you put his measurements into your calibration simulator

when I put it into one I have, it finds a bunch of outlier points, all showing
the bottom belts too long at those points

Index 35: {bl: 1366.27, br: 2092.94, tr: 2367.25, tl: 1754.18} (BR error: 10.91
mm, BL error: 2.59 mm)
Index 48: {bl: 2443.59, br: 1597.19, tr: 1312.93, tl: 2282.07} (BR error: 10.27
mm, BL error: 1.37 mm)
Index 49: {bl: 2501.27, br: 1553.46, tr: 1272.55, tl: 2344.22} (BR error: 16.85
mm, BL error: 4.60 mm)
Index 50: {bl: 2497.01, br: 1492.29, tr: 1312.34, tl: 2407.04} (BR error: 3.10
mm, BL error: 11.76 mm)
Index 52: {bl: 2419.92, br: 1353.85, tr: 1434.03, tl: 2507.01} (BR error: 16.62
mm, BL error: 8.64 mm)
Index 53: {bl: 2384.42, br: 1296.22, tr: 1495.51, tl: 2548.24} (BR error: 22.91
mm, BL error: 14.97 mm)
Index 54: {bl: 2350.58, br: 1236.74, tr: 1562.63, tl: 2591.60} (BR error: 28.60
mm, BL error: 18.13 mm)

with these points included, it estimates the errors at -+10mm or so, without it
estimates them at -+7mm or so

Ian T, I don’t remember what happened, but since you have positioned your
anchors so carefully, did you try to just enter the anchor distances and try
using the machine?

bl will be 0,0
br will be width,0
tl will be 0,height
tr will be width,height
wasteboard thickness will be -(thickness of the bottom of the plastic anchors)

with that set, try just apply tension and see if it works or throws an error.

David Lang

[bar]

Nothing looks out of wack to me:

I’m getting basically the same score that he got, something like 0.41 which is just below the 0.45 threshold

[dlang]

Bar wrote:

Nothing looks out of wack to me:

I’m getting basically the same score that he got, something like 0.41 which is just below the 0.45 threshold

is that the placement you expect for the points? are any of the point
measurements odd?

if you put the measured points into the new algorithm, what does it come up
with for coordinates and accuracy?

David Lang

[ian_ab]

Yes, but more they are in the right order.

[dlang]

I created a new version of my checking tool at

Maslow Levenberg-Marquardt Calibrator

paste in the point array and it will show you outliers, give you the calculation
that should match what the latest test would provide and show outliers (and what
happens if you leave off the outliers)

David Lang

[dlang]

hmm, according to this calculation, the frame is leaning left a fair bit more than we thought.

try putting these numbers in and see if it works
this tool is based on the new algorithm that Bar has tested and is in the post 1.21 code

[Bones]

@Ian_Tolond

I feel your frustration I’ve been trying to get mine dialed in for well over a month. Innumerable hours and i finally made some forward progress. Continue to post your results, and the folks on here will help you get over the hump. Ive put more miles on my belts than I care to count. I’m currently in the process of disassembly and inspection of components to cross T’s and dot I’s. Ive got a replacement board on it’s way.

@dlang
Would dropping back to 1.15 possibly work? I made progress once I dropped back to 1.15

[Bones]

Initial set up and calibration

I received my 4.1 about 3 weeks ago and immediately started the assembly and set-up. First try was anchored to the floor and tried to run calibration, failed. I built an 8x10 frame to run it at 30 degree horizontal and tried to run the calibration again, failed. I have uploaded all the recent firmware (I think). I have run “retract all", “extend all", “apply tension” tried lowering the Z and it doesn’t move. The Z readout changes value but it does not move. I have done all the “unplug it", “move…

There might be something in this thread that will help. I asked A LOT of questions, to understand the mechanics. Once you know the machine it may help you understand the problems.

[ian_ab]

Bar has been doing some development work on the Find anchors.
Try this version Interstitial Firmware Releases [quote=“bar, post:882, topic:20215”]
Go to bottom generated yesterday.
Here is another new version which is running the new math
[/quote]

[N8DGR8]

I never got calibration to pass, you don’t have to do calibration at all. You can manually measure anchor points and update the 4 corners in the yaml file.

[Ian_Tolond]

how so ?

[Ian_Tolond]

Hi David,

The floor is uneven where the frame is setup:

The whole frame as a result leans left (looking at the work surface) by about (a guess 15-25mm). We levelled the frame and then used a laser level to mark where the anchor points would be ( I expect due to the tropical environment there has been some movement, however not that much). In essence the anchor points align with the spoil back board and how materials will be mounted for cutting/routing.

as mentioned the frame leans left due to the floor

Previously I posted if this would cause an issue, but I was advised otherwise as its based upon the anchor locations. confirming this is still valid?

[Ian_Tolond]

I Measured the CL of the anchor points and the following is true:
Anchor points (centres)

Width 2920mm

Height 2373mm

[dlang]

Ian Tolond wrote:

The floor is uneven where the frame is setup:

The whole frame as a result leans left (looking at the work surface) by about (a guess 15-25mm). We levelled the frame and then used a laser level to mark where the anchor points would be ( I expect due to the tropical environment there has been some movement, however not that much). In essence the anchor points align with the spoil back board and how materials will be mounted for cutting/routing.

as mentioned the frame leans left due to the floor

Previously I posted if this would cause an issue, but I was advised otherwise as its based upon the anchor locations. confirming this is still valid?

you can probably get away with it, but in an ideal world, instead of using the
laser to level the ancors to the ground, you would treat the frame as ‘flat’ and
adjust (or at least measure) based on that.

so if you were to put a long straightedge out from the tabletop to the corner
(say a long level), the point that it hit the wall would be the 0 relative to
the table.

does this make sense?

David Lang

[dlang]

Ian Tolond wrote:

Width 2920mm

Height 2373mm

based on this, I think you will want to limit your cutting area to the middle
1880x1220 (6’x4’) see http://lang.hm/maslow/maslow4_frame.html

David Lang