Hardware issue: z axis slippage in horizontal configuration causes progressive z error

[Michael_Barrow]

I tried cutting out a job with my Maslow 4 that has edges close to the far edge of the work area (x is about 3cm).

My machine plunged into my floor and wrecked my mill.

After some debug (no mill in the machine and a marker at stock height taped above a z motor) I found the problem, although I don’t know how this one might be fixed.

The machine puts so much tension on the leading belts that the z motor stalls as the machine trys to reduce z depth whilst simultaniously increasing y (basically the y movement prys the sled onto its highest y edge so it would be doing a tiny endo if it were a BMX).

@bar, @dlang this is a tricky one. I think the top right arm having the steepest angle to the floor and the bottom left belt having the shalowest is contributing to this, so I will try to reduce angle of the top left with some stock height shims on the high anchor and re-calibrate. Otherwise, stumped. Video attached. You can hear the z axis groaning and failing. If my job depths were respected, the little stick would not bottom out. The warping in the tape is caused by an artificial offset being added progressively to z due to this problem. After this failure I saw about a 1/8" to 1/4" difference of exposed lead screw, but forgot to take a photo of that.

[dlang]

Michael Barr wrote:

I tried cutting out a job with my Maslow 4 that has edges close to the far edge of the work area (x is about 3cm).

My machine plunged into my floor and wrecked my mill.

unfortunantly the video you posted cannot be played on my browsers, so I’m not
sure exactly what the problem you had is.

But as you get close to the edge, you need to have additional material around
the workpiece for the sled to ride on to keep it from tipping off the edge

After some debug (no mill in the machine and a marker at stock height taped
above a z motor) I found the problem, although I don’t know how this one might
be fixed.

The machine puts so much tension on the leading belts that the z motor stalls
as the machine trys to reduce z depth whilst simultaniously increasing y
(basically the y movement prys the sled onto its highest y edge so it would be
doing a tiny endo if it were a BMX).

@bar, @dlang this is a tricky one. I think the top right arm having the
steepest angle to the floor and the bottom left belt having the shalowest is
contributing to this, so I will try to reduce angle of the top left with some
stock height shims on the high anchor and re-calibrate.

add skirting around the workpiece, it doesn’t need to be the full radius of the
sled, but a couple inches will help a lot.

Also, as you increase the Z height of the anchor, make sure there is no flex
being added (you are giving the belt more leverage)

And remember to change the Z offset values for the anchors you are changing.

Otherwise, stumped. Video attached. You can hear the z axis groaning and
failing. If my job depths were respected, the little stick would not bottom
out. The warping in the tape is caused by an artificial offset being added
progressively to z due to this problem. After this failure I saw about a 1/8"
to 1/4" difference of exposed lead screw, but forgot to take a photo of that.

see the thread “one more source of error” from friday, I identified an error in
the math that makes the effective length of the belts a bit shorter than the
should be as the belt angle gets steeper, I expect a fix for that will go out
either this week or next week (changing the math is a dangerous thing to do, so
there may be more testing than is possible before Wednesday). I would not be
surprised if you are running into this as well.

But raising the belts so they are closer to level will help this in any case.

David Lang

[Michael_Barrow]

@dlang, cheers for the fast response. Bummer the video does not play… Not sure what to do about that

Anyway its not possible to see on the video but I do have support skirts there in case of tipping.

I raised the top left anchor by 2.5" from the ground, approximately its height from the base of the arm assembly, re-calibrated and the issue seems to have gone.

I will mark the problem as resolved in the thread if I am able to complete the job that caused my issue

Another point, I noticed some slight bowing of my support skirt which I clamped down, and added some paste wax to the base of the sled for good measure

[dlang]

Michael Barrow wrote:

I raised the top left anchor by 2.5" from the ground, approximately its height from the base of the arm assembly, re-calibrated and the issue seems to have gone.

did you change the Z offset to match?

David Lang

[Michael_Barrow]

No, I only raised one anchor point.
I wanted to balance the downward force from the top left and bottom left cables since the uneven down force was causing a tip of the sled and stall of the z axis close to x=0.

I did not see individual offsets per anchor in the yaml, so left z at default. Fitness dropped after recalibration, but the results from the test job seem satisfactory.

[dlang]

Michael Barrow wrote:

No, I only raised one anchor point.
I wanted to balance the downward force from the top left and bottom left cables since the uneven down force was causing a tip of the sled and stall of the z axis close to x=0.

I did not see individual offsets per anchor in the yaml, so left z at default. Fitness dropped after recalibration, but the results from the test job seem satisfactory.

They are the Maslow_tlz Maslow_trz Maslow_blz and Maslow_brz variables

it it’s cutting well enough, you can leave it as-is, but if you have any other
issues, I would update the Z offset values and recalibrate (note that when the
math fix goes in, you will need to calibrate again for it)

David Lang

[Michael_Barrow]

Thanks, I should have looked more carefully.

Ill look forward to this maths update as I am hoping for as much accuracy over the full work area as possible

[Michael_Barrow]

This issue metastasized quite badly and I (think) I have only just been able to mitigate it.
The problem is that on the left edge of the work area, the bottom left belt goes quite slack when the sled travels up y. The sled cantilevers due to having little mass, and the z axis stalls.

@bar, does it make sense to adjust tension in the out-spooling belts (in this case bottom left)?

It seems to be a nasty case for me, maybe not others. The only solution (for me) was to

1. overdrive the z motors to 2 amps to overcome stalling. They get HOT now (maybe 65 Celsius)
2. add a 4.5kg counter weight platform to the bottom of the sled and flip dust collection to the top

On the bright side I can move on to trying the JST encoders now and tackle the 15mm accuracy issue.

[dlang]

Michael Barrow wrote:

This issue metastasized quite badly and I (think) I have only just been able to mitigate it.
The problem is that on the left edge of the work area, the bottom left belt goes quite slack when the sled travels up y. The sled cantilevers due to having little mass, and the z axis stalls.

If the belts are getting lose when moving, it means that the anchors aren’t
where it thinks they are. The machine doesn’t control belt tension, it controls
belt length (and sort-of figures out excessive tension based on the current
needed to try to move, which ends up trying to stretch the belt)

If the belt is hitting the dust collection, make sure you do the calibration
without the dust collection fitting so that it can do the detection without
interference (there are people experimenting with re-ordering the arms to not
have the lowest one on the bottom, to avoid interference with the dust
collection hose)

If tilting the sled is causing the Z axis to stall, that would have to men that
it’s flexing the linear rods, check how they are anchored (both into the sled
and the holders for their top brackets, including the towers) to see if there is
any flex there.

unless I am missing something…

David Lang

[Michael_Barrow]

Hi @dlang , thanks for a speedy response. Maybe some more detail will paint a clearer picture.

The sled is cantilevering, meaning the bottom (where dust collection exit port is typically installed) elevates off of the work piece by about 1cm.

It only does this at the left most edge of the work piece, whilst moving upward.

Here the primary contributor to travel is the top left belt. This belt has the highest installed height.

Due to the belt install height, the lack of tension on the opposing belt, high center of gravity of the machine, and friction of the work surface, the sled visibly cantilevers whilst traveling upwards in y.

The linear rods are not flexing when lead screws are not turning, but could if the z motors were to stall. The top z motor therefore obligingly stalls with the default drive current when the machine tries to cut 45 degree angled support tabs whilst moving up on the y axis.

This is the issue. I have taken steps to independently prevent cantilevering and motor stall. With these steps combined I have been able to work around the issue so far.

The belt is not hitting dust collection.

[Michael_Barrow]

@bar I think @dlang’s point raises a suggestion for hardware improvement. If the lead screws were to extend to the same height as the linear rails and have a ring bearing installed at the top, a second set of plastic supports could be added top y and bottom y of the sled as are present on left x and right x, supporting the linear bearing rails. The added supports would prevent flexing of the linear rails and probably prevent the z stall I am seeing. Will dump this in the hardware improvement thread also.

[dlang]

Michael Barrow wrote:

Hi @dlang , thanks for a speedy response. Maybe some more detail will paint a clearer picture.

The sled is cantilevering, meaning the bottom (where dust collection exit port is typically installed) elevates off of the work piece by about 1cm.

It only does this at the left most edge of the work piece, whilst moving upward.

Here the primary contributor to travel is the top left belt. This belt has the highest installed height.

Due to the belt install height, the lack of tension on the opposing belt, high center of gravity of the machine, and friction of the work surface, the sled visibly cantilevers whilst traveling upwards in y.

try sanding/waxing the bottom of the sled to reduce the friction. If that
doesn’t work, try tilting the frame back further (so that gravity is pulling
towards the workpiece more.

adding weight higher up to shift the CG up may help

The linear rods are not flexing when lead screws are not turning, but could if the z motors were to stall. The top z motor therefore obligingly stalls with the default drive current when the machine tries to cut 45 degree angled support tabs whilst moving up on the y axis.

I believe that the steppers are configured to run in stealthchop mode, which
makes them quieter, but reduces their power. Try switching the config to
coolstep (maslow.yaml)

David Lang

[dlang]

Michael Barrow wrote:

@bar I think @dlang’s point raises a suggestion for hardware improvement. If
the lead screws were to extend to the same height as the linear rails and have
a ring bearing installed at the top, a second set of plastic supports could be
added top y and bottom y of the sled as are present on left x and right x,
supporting the linear bearing rails. The added supports would prevent flexing
of the linear rails and probably prevent the z stall I am seeing. Will dump
this in the hardware improvement thread also.

I’m actually tinkering with the idea of replacing the linear rails with
leadscrews. It would solve the angle problems.

David Lang

[Michael_Barrow]

I tried pastewax but no dice there. I also tried various height offsets of the top left belt anchor without success.

Also tried switching to CoolStep but the z axis didn’t move at all. I don’t think FluidNC supports it properly. see here.

In fact, judging by how noisy the motors become when increasing the current threshold (in maslow.yaml) I would not be surprised if internally it is switching to spreadcycle anyway.

[bar]

@Michael_Barrow

flexing of the linear rails and probably prevent the z stall I am seeing.

Do you have a picture of this happening?

[Michael_Barrow]

I found one. Its frozen in relief since the lead screws basically lock the tipped over position.

This is the net effect of the stall. Router gets skewed and progressively sinks.

This is not the best picture maybe but one i had to hand.

Michael

[Brookeman]

I can’t play your video