Tension Jerks on manual movement

I do a lot of testing and some simple cuts by just using the manual controls.

You can see in this picture we have very nice corners created by my Maslow2.

Here you can see the divot and trail caused by the belts getting tensioned and released.

This happens every time you start a movement and will release if no command is receive in a second or two.

Maybe we could make that time delay a variable in the the maslow.yaml file?

I was thinking about making that a button, but then I will forget and I assume long term resting tension is not good for the belts.

FYI: This is a vertical mount.

1 Like

I know exactly what you are talking about. Generally I have the router bit out of the wood before starting to run gcode to avoid that, but if the router bit is in contact with the wood when the program starts it will jerk like that as it wakes up. Are you creating these squares using the jog command buttons?

Yes, in fact it might already be an option in FluidNC. It looks like this is what we want: Axes | Wiki.js

Unfortunately it seems like the maximum value that they support is 250ms or it can be disabled entirely. It shouldn’t be too hard for us to change that option to be longer if we would like.

My thinking exactly :grinning:

My thoughts generally on accuracy is that it was the next thing on my todo list a couple weeks ago and then I got sucked into fighting fires. It’s the top on my todo list again when I have a chance.

Here is my go-to testing file (three 600mm squares, just the corners):

SD_Test Squares.nc (6.9 KB)

There are a couple of known issues right now that are worth being aware of.

1st - The machine doesn’t have a good way to home the z-axis right now. When it powers up it believes that the spacing between the top of the z-axis motor and the bottom of the lower router clamp is 28mm regardless of if it actually is. The only way to get that right now is to lower it all the way down…raise it back up 28mm…power off…power back on which is a huge pain. I was working on making a button to do that automatically, but integrating it into FluidNC was turning out to be a little tricky. This is top of my ToDo list.

2nd - I had some code in there which was scaling everything by like 105%…basically my tests were coming out 5% smaller in both X and Y than expected so I scaled everything by 105% and it was good…and then I forgot about it. I came across that code later and it was such a hack that I took it out again so I wouldn’t forget. I think that what is actually going on is that the tooth spacing on the belts is not exactly 2.00000mm it’s like 1.9999mm or 2.0001 mm and we need to do some testing and put the right number in for that. I think that the only way to get that dialed in is to put in a new number…run calibration…cut test squares and measure them…tweak the number…repeat. At least that was my plan.

Yes, hense the title (sorry could not resist).

Thanks for the extended update, makes me feel bad about the dig above :slight_smile:

Oh, zeroing the Z. With the Kobalt I can’t run it all the way down because of the locking pin. See my note on zero setting… Maslow4 - how to do bit changes? - #23 by Ken32960

BTW I have been meaning to say what a great job you two have done to put this multiple discipline project together. It is an outstanding accomplishment!!! It even worked the first try. More than I can say about many of my projects.

1 Like

There are still a TON of bugs to get worked out, but we’re working them out a lot faster now that there is hardware out in the community.

It’s all a little chaotic at the moment but we’re getting there :grinning:

is this a lack of acceleration planning in the first movement?

David Lang

if you put a 5/8" thick spacer over the lead screw below the bottom bracket,
you can have it bottom out against that (I think you can get this down to 1/2"
or smaller if you cut the plexiglass and sled, but have not had time to try that

David Lang

No, when the sled is at rest the belts are let slack just a bit. When a command is received you can see the belts pull tight and the sled jumps up the project about 1/4". After a second it drops back down as the belts relax. If you can give the commands fast enough, there is some amount of queuing, all four moves will get completed with out the time out and relax. That is why the other corners look ok.
This is all in manual mode.

Yes, I was thinking about cutting the plexiglas but wasn’t sure I needed it with 60mm of movement. Might have to if I want to use 1/8" bits and get enough depth.

1 Like

Bar, I think you are correct. Might have to change an INT(8) to INT(16) giving a max of 32 or unsigned 65 seconds. That should be enough.

1 Like