A Tale of Two Days, Terminal Warfare, and Why You Should Never Touch the Wi-Fi Settings (Plus an 8x4 Setup Guide)

A Tale of Two Days, Terminal Warfare, and Why You Should Never Touch the Wi-Fi Settings (Plus an 8x4 Setup Guide)

​Hi everyone,

​I’ve just emerged from a grueling, two-day technical wrestling match with my Maslow 4.1 following the latest firmware update. Now that the sawdust has settled and I’ve successfully achieved a glorious 0.37 calibration fitness score, I wanted to share a few hard-learned lessons, a massive warning for new users, and a feature request for the developers.

​1. THE ULTIMATE WARNING: Step Away from the Wi-Fi Mode Settings!

​If you take away nothing else from this post, let it be this: Do not change your Wi-Fi Mode to “AP Only” in the settings. During a bit of standard troubleshooting, I swapped the mode. The firmware instantly trapped itself in a catastrophic boot loop, completely dropping the Maslow access point from the face of the earth. The UI was gone. My laptop couldn’t see it. The machine was booting up physically (fans spinning, standard startup sounds), but the radio was completely dead.

​To fix it, we had to go full-on “computer hacker” mode in Linux. It required installing native serial monitors (tio), tapping into the hardware pins via USB, and fighting a firmware interface that aggressively rejected shorthand commands with cryptic error:3 codes. We eventually had to blind-type the exact full path syntax: $WiFi/Mode=STA>AP followed by $System/Control=Restart just to force the access point to broadcast again. Save yourself the terminal warfare—leave the Wi-Fi settings alone!

​2. The Firmware Update: A Tale of Missing “Out-of-Square” Tolerance

​Before the most recent firmware update, my Maslow 4.1 calibrated quite happily on a massive, out-of-square 12\text{ ft} \times 7\text{ ft} frame bolted directly to my workshop wall. It was brilliant, versatile, and highly forgiving.

​However, the new firmware update appears to be wholly intolerant of out-of-square anchor points. It completely threw its toys out of the pram trying to run the math on my old rig. This strict intolerance ultimately forced me to abandon my large wall setup entirely and opt for a much smaller, dedicated 8\text{ ft} \times 4\text{ ft} benchtop frame. While it works beautifully now, losing that massive cutting area is a real shame and feels like a step backward for user-friendliness.

​3. A Suggestion for the Devs: Bring Back the Math Flexibility

​If the previous firmware could happily calculate and compensate for a slightly skewed or out-of-square anchor setup, it must be mathematically possible here.

My suggestion for a software revision: The dimensions calculating software could be updated to implement a Skew Compensation Factor or a secondary calculation pass. If the initial rigid geometric matrix fails due to out-of-square anchors, the software should fall back to a triangulation algorithm that treats the four anchor points as an irregular quadrilateral rather than a perfect rectangle. By allowing the software to calculate unique vector offsets for each corner based on the actual measured distances (rather than forcing them to conform to a strict orthogonal grid), the Maslow 4.1 would become infinitely more versatile for real-world workshops where walls and frames are rarely perfectly square.

​Quick Guide: Setting up a Fast 8x4 Benchtop Rig (UK Supplies)

​For those trapped by the new firmware constraints who need to set up a rigid, compliant 1800\text{ mm} \times 600\text{ mm} cutting grid on a standard 8\text{ ft} \times 4\text{ ft} sheet framework, here is the quick-and-dirty setup guide to avoid the pitfalls I hit:

  • The Frame: Build your base using standard C16/C24 structural timber (47\text{ mm} \times 100\text{ mm}) to ensure absolute rigidity. Ensure your sheet material sits completely flat.

  • The Software Safety Trap: When you first boot up after a reset, the UI might throw a terrifying “Low Z Position” warning, locking out your Z-axis because it thinks the bit is buried in the floor.

    1. ​Dismiss the warning.

    2. ​Click Define Z Home on the right panel to snap the readout back to a safe 0.00mm.

    3. ​Head to the Settings tab and temporarily toggle Soft Limits to OFF. This gives you full manual control to drop the axis to the bottom of the frame for anchor setup without the software blocking you.

  • Running the Grid: Use a standard 3x3 calibration grid. Be prepared for the console log to shout a few yellow warnings at you under the hood (like Retraction force exceeded limit). Don’t panic—during calibration, the sled is pulled into extreme corners where belt tension naturally spikes. Clear the logs and trust the final score.

  • Locking it In: Once you get a tight fitness score (anything under 1.0 is great, my 8x4 hit an incredible 0.37!), remember to immediately scroll down and hit Save Configuration so it writes permanently to your maslow.yaml file. Turn your Soft Limits back ON before cutting.

  • Mind Your G-Code Origin: If your CAM software (Easel, Carveco, etc.) is set to center origin, physically jog your sled to the dead center of your plywood sheet and hit Zero X and Zero Y in the Maslow UI. Your on-screen preview crosshairs will snap perfectly to the center of your design.

  • The “Clunk” Fix: If your router sounds rhythmic and clunky when changing directions, your clamp is slightly loose. If it bottoms out before fully gripping the router barrel, wrap a thin shim cut from an aluminum drinks can around the router body before tightening the clamp. It stops the rotation instantly and smooths out the cut.

​A massive thank you to the community members who helped keep me sane over the last 48 hours. Hopefully, this saves someone else a massive headache!

​Cheers,

David Lang

David Coster wrote:

​1. THE ULTIMATE WARNING: Step Away from the Wi-Fi Mode Settings!

​If you take away nothing else from this post, let it be this: Do not change
your Wi-Fi Mode to “AP Only” in the settings.
During a bit of standard
troubleshooting, I swapped the mode. The firmware instantly trapped itself in
a catastrophic boot loop, completely dropping the Maslow access point from the
face of the earth. The UI was gone. My laptop couldn’t see it. The machine was
booting up physically (fans spinning, standard startup sounds), but the radio
was completely dead.

using the FluidNC commands to change settings is not fun. I would have just
re-flashed via USB using the scripts that come with each release.

what settings did you set? I use AP or STA->AP on a regular basis

​To fix it, we had to go full-on “computer hacker” mode in Linux. It required
installing native serial monitors (tio), tapping into the hardware pins via
USB, and fighting a firmware interface that aggressively rejected shorthand
commands with cryptic error:3 codes. We eventually had to blind-type the exact
full path syntax: $WiFi/Mode=STA>AP followed by $System/Control=Restart just
to force the access point to broadcast again. Save yourself the terminal
warfare—leave the Wi-Fi settings alone!

just setting it to AP should not be a problem, the problem is what the other AP
settings were.

​2. The Firmware Update: A Tale of Missing “Out-of-Square” Tolerance

​Before the most recent firmware update, my Maslow 4.1 calibrated quite happily on a massive, out-of-square 12\text{ ft} \times 7\text{ ft} frame bolted directly to my workshop wall. It was brilliant, versatile, and highly forgiving.

what version of firmware are you working with?

are you talking 1.21 or the newer testing firmware? It should only be the newer
testing firmware that has the new math.

​However, the new firmware update appears to be wholly intolerant of
out-of-square anchor points. It completely threw its toys out of the pram
trying to run the math on my old rig.

​3. A Suggestion for the Devs: Bring Back the Math Flexibility

​If the previous firmware could happily calculate and compensate for a slightly
skewed or out-of-square anchor setup, it must be mathematically possible here.

My suggestion for a software revision: The dimensions calculating software
could be updated to implement a Skew Compensation Factor or a secondary
calculation pass. If the initial rigid geometric matrix fails due to
out-of-square anchors, the software should fall back to a triangulation
algorithm that treats the four anchor points as an irregular quadrilateral
rather than a perfect rectangle. By allowing the software to calculate unique
vector offsets for each corner based on the actual measured distances (rather
than forcing them to conform to a strict orthogonal grid), the Maslow 4.1
would become infinitely more versatile for real-world workshops where walls
and frames are rarely perfectly square.

the math does not assume a perfect rectangle.

Do you have your logs? can you provide the CBLM array so we can see what it’s
trying to do?

  • Running the Grid: Use a standard 3x3 calibration grid. Be prepared
    for the console log to shout a few yellow warnings at you under the hood (like
    Retraction force exceeded limit). Don’t panic—during calibration, the sled
    is pulled into extreme corners where belt tension naturally spikes. Clear the
    logs and trust the final score.

it is NOT normal for belt tension to spike to limits

a 3x3 grid may give you a high score, but it’s unlikely to give you good results

  • Locking it In: Once you get a tight fitness score (anything under 1.0 is
    great, my 8x4 hit an incredible 0.37!), remember to immediately scroll down
    and hit Save Configuration so it writes permanently to your maslow.yaml
    file. Turn your Soft Limits back ON before cutting.

ok, this implies that you re using one of the experimental snapshots. In the
releases, a higher fitness is better.

  • Mind Your G-Code Origin: If your CAM software (Easel, Carveco, etc.) is
    set to center origin, physically jog your sled to the dead center of your
    plywood sheet and hit Zero X and Zero Y in the Maslow UI. Your
    on-screen preview crosshairs will snap perfectly to the center of your design.

This is a common problem

  • The “Clunk” Fix: If your router sounds rhythmic and clunky when changing
    directions, your clamp is slightly loose. If it bottoms out before fully
    gripping the router barrel, wrap a thin shim cut from an aluminum drinks can
    around the router body before tightening the clamp. It stops the rotation
    instantly and smooths out the cut.

good callout

David Lang

Which version are you running?
Once you have good anchor points Back up your maslow.yaml file. There is no need to run Find Anchors so long as you preserve the Anchor settings.
On your points:

  1. Frame. Agreed, needs to be rigid. I am using anchors attached to the walls with 1 metre extensions, no problems.
  2. Low Z position will be fixed in next release, in the meantime just set with Define Set Home, Soft limits should be false by default.
  3. 3x3 grid will only provide minimal accuracy. 9x9 is optimal with an initial calibration grid X and Y of 0,0
  4. The higher the fitness score the better. 0.45 is the minimum by default. Anything over this will be saved to your maslow.yaml file automatically.
  5. Set Home to match the home you gave the design in your CAM software. Some like to have the origin in the middle of their cut, others prefer it bottom and left. It’s personal preference (and software determined).
  6. I like this shim fix but would also recommend the Turtle Clamp as an excellent alternative. one version of this lets you use 240V router with cord pointing up

Thanks for your replies… I’ve just installed firmware v1.21, but I rather stupidly also replaced the yaml file as well (Crazy man!), which is why a had to run the Find Anchors.

I knew my anchors (fitted to my workshop walls, with the Maslow running horizontal at workbench height) were out of square, as the walls are not parallel…but those same anchors had calibrated fine in the previous setup.

Once v1.21 was installed, the calibration timed out every time, as the fitness never reached an acceptable level…

How much ‘out of square’ tolerance can the calibration cope with… Why would this have changed since the previous firmware version, if it has, do you think? … or has something changed with my hardware maybe… but it’s now cutting fine anchored to the workbench, but has only a 1800 x 600mm cut area now…

I don’t have room for a floor layout, so the 2440 x 1220mm bench top has to be my workarea…

I didn’t save the serial log (again Crazy Man!) … but when I get a free day (…or two) I’ll save the current yaml file and try the wall anchors again…unless you have other suggestions.

Get TypeApp for Android

David Coster wrote:

Thanks for your replies… I’ve just installed firmware v1.21, but I rather stupidly also replaced the yaml file as well (Crazy man!), which is why a had to run the Find Anchors.

I knew my anchors (fitted to my workshop walls, with the Maslow running horizontal at workbench height) were out of square, as the walls are not parallel…but those same anchors had calibrated fine in the previous setup.

what are your z offsets? are the wall anchors at the height of the bottom of the
sled? or do you have them higher so the belts are closer to level. maslow seems
particuarly sensitive to this.

Once v1.21 was installed, the calibration timed out every time, as the fitness never reached an acceptable level…

How much ‘out of square’ tolerance can the calibration cope with… Why would
this have changed since the previous firmware version, if it has, do you
think? … or has something changed with my hardware maybe… but it’s
now cutting fine anchored to the workbench, but has only a 1800 x 600mm cut
area now…

nothing that is related to the anchors being out of square should have changed,
but it does seem to be more sensitive to Z offsets being wrong and what the
initial guess as to the frame dimensions (contents of the yaml file at the start
of the run). the default yaml file contains values that worked for a particular
frame that Bar had in place at one time, it’s not a good start for most people

I don’t have room for a floor layout, so the 2440 x 1220mm bench top has to be
my workarea…

that is a good plan, you aren’t the first person to do that.

I didn’t save the serial log (again Crazy Man!) … but when I get a free day
(…or two) I’ll save the current yaml file and try the wall anchors
again…unless you have other suggestions.

the topic “can find anchors find Z” or something like that has instructions for
how you can keep an extra yaml file on the device and switch between them (so
you can have one that you trust as your current best guess, and another to
experiment with

David Lang

Hi David, thanks again. With the original setup, I used a laser level placed the centre of the workbench, to project a line out across the bench top to the adjacent wall (off all 4 corners) and fixed an anchor, at bench top height, for each corner … so the z height of the anchors is the same height as the bench top, meaning the belts are pulling downward from the Maslow to the anchors, is that not correct??

Get TypeApp for Android

That is correct, what David is referring to is an alternative where the height is set so all the belts are parallel when the Maslow is fully down. You have to adjust the Z heights in the maslow.yaml file to reflect this (Set to 0).
Bar has made some improvements to the Find Anchors. An experimental version is on Interstitial Firmware Releases stream. Recommended.
You can have multiple versions of the yaml file setup for each configuration.
Set under FluidNC, Flash Settings, Config/Filename.

Thank you, Ian … sorry for being a numpty… but, does this mean that for optimal setup, each of the four belt anchors should be at a different height (on the wall), so that each belt is parallel to the bench surface when the Mazlow z-height is at its lowest (without a bit in the chuck, with the face of the chuck kissing the bench top)?.. So, fix each anchor at its own relative hight parallel to the bench top, so its corresponding belt is also parallel to the bench top??

Get TypeApp for Android

Yes, exactly
Then change the Z values in the maslow.yaml file to 0.

Thank you, I’ll give it a go and let you know how i get on!

Get TypeApp for Android

David Coster wrote:

Thank you, Ian … sorry for being a numpty… but, does this mean that for
optimal setup, each of the four belt anchors should be at a different height
(on the wall), so that each belt is parallel to the bench surface when the
Mazlow z-height is at its lowest (without a bit in the chuck, with the face of
the chuck kissing the bench top)?.. So, fix each anchor at its own relative
hight parallel to the bench top, so its corresponding belt is also parallel to
the bench top??

there is debate as to the ‘optimal’ setup, but I am in the camp that thinks that
having the belts level is better.

  1. less down force on the router means less friction, easier to move
  2. the closer the belts are to level, the less it matters if the heights are
    wrong

as a technical detail, in theory, the belts height is on the plane of the
workpiece, so if your desk isn’t level by your laser, the anchors should be
eqally un-level. in practice, it probably doesn’t matter much (the belts flex in
the vertical direction and the triangle that is the difference between the
perfect straight line and what is actually formed is probably so flat to not
matter, but this is one of those things that can make the heights just slightly
wrong for #2 above)

if you do set your heights to match the arms so the belts are level, then you
need to change the Z offsets in your maslow.yaml. the best way to do this is
through the fluidnc tab, right bubble and look for the tlz/trz/blz/brz entries
and set them to 0 if the belts are flat. the default values assume that if your
spoilbard was extended to where the anchors are, the anchors would be on the
surface of the spoilboard.

We hae had people shift things a bit and get good calibrations, and other people
who just cannot get things to word, fix the Z offset values, and suddently the
calculations work. we don’t know why the difference.

David Lang

I’ve just taken the Maslow assembly to bits and have discovered that the bottom clamp was over tightened and has deformed, pulling the clamp bolts through the plastic of the clamp ring, rendering the clamp wedge totally unsecurable… I am currently printing new turtle clamps!!!

For you maths geniuses… my 8’ x 4’ work-bench is skewed in relation to the walls of my workshop (the wall are also not parallel with each other). Putting a laser measure on the centre point of my work bench and measuring to each of the four projected anchor points (through each of the four corners of the work bench top) on the wall reveals quite a skewed indeed:

Measuring from the centre point of the 8’ x 4’ bench top out through the corner of the bench to the adjacent wall anchor:

To top right anchor = 2230mm

To bottom right anchor = 2226mm

To bottom left anchor = 2119mm

To top left anchor = 2077mm

(All anchors are currently level with the bench top, shortly to be moved individually up or down relavent to their respective belts)

My pre firmware upgrade setup happily calibrated on these anchor points as they are, whereas the current v 1.21 seems not to.

Should the current version of the firmware be able to handle the skew and achieve a satisfactory fitness level with that much of a non-square anchorage??

Regards

David

Get TypeApp for Android

David Coster wrote:

I’ve just taken the Maslow assembly to bits and have discovered that the bottom clamp was over tightened and has deformed, pulling the clamp bolts through the plastic of the clamp ring, rendering the clamp wedge totally unsecurable… I am currently printing new turtle clamps!!!

For you maths geniuses… my 8’ x 4’ work-bench is skewed in relation to the walls of my workshop (the wall are also not parallel with each other). Putting a laser measure on the centre point of my work bench and measuring to each of the four projected anchor points (through each of the four corners of the work bench top) on the wall reveals quite a skewed indeed:

Measuring from the centre point of the 8’ x 4’ bench top out through the corner of the bench to the adjacent wall anchor:

To top right anchor = 2230mm

To bottom right anchor = 2226mm

To bottom left anchor = 2119mm

To top left anchor = 2077mm

(All anchors are currently level with the bench top, shortly to be moved individually up or down relavent to their respective belts)

My pre firmware upgrade setup happily calibrated on these anchor points as they are, whereas the current v 1.21 seems not to.

Should the current version of the firmware be able to handle the skew and achieve a satisfactory fitness level with that much of a non-square anchorage??

it should, note that the X axis will be parallel to the line between the anchors
for the bottom

can you measure the 6 distances between the 4 anchors (including diagonals) and
post that?

David Lang

Brilliant, thank you both so much for your time and guidance :folded_hands:

Get TypeApp for Android

Thanks, again, David … I’ll do that tomorrow, as my wife is nagging me to find something useful to do! :laughing:

Get TypeApp for Android