There is a discussion going on this topic in another thread, but I want to start a dedicated thread for it.
One common issue that we’re seeing is that in horizontal orientation the first measurement isn’t very reliable. I think that this is happening because as we pull tight on the red belt we introduce some slack in the system like this:
This is happening because of the way that the angles work out. In order to extend enough belt to hook everything up we have to extend enough to reach the anchor points which means that we will be below the center in the Y axis.
Being below the center in the Y-axis means that when we pull tight on the red belt there will be a force acting to slack on the the other belts.
I’m playing around with pulling all four belts tight at the same time in horizontal orientation, but it will take a little bit of work (so I’m not going to include it in today’s firmware update).
Does anyone have any other thoughts or ideas on how we could better make the first measurement reliable?
Do you mean you think that one belt goes slightly slack but because that belt is already extended, we read a longer value for it?
I think @dlang suggested in the other thread tensioning twice to check, which should work.
That could be taken further by combining with a extension of the frame flex calc like I played with manually - so start with a low tension, and re-tension at +100/200 force a few times until you hit the calibration_force force. Dump all the result and their forces and offer the user some feedback based on it (or even in the future act on the result)?
We’re actually doing this already, but we’re only tensioning the lower two belts and because of the way the forces are that’s not fully pulling tight on the upper ones I think.
I like this idea. I think that there is some potential to explore there for sure!
Another open question could also be - what common setup issues could we try and identify at this point? I’m not sure which ones we actually could identify that early, but worth a bit of brainstorming.
Also think of cases where the frame is not exactly a rectangle, so even if the belt lengths do put you in the ‘middle’, the belts could still be uneven to generate the same result of pulling on one belt making another belt loose.
@bar you say that we already pull twice, but that’s pulling the same belt multiple times and taking the measurement on that one belt. using your diagram, if we were to go back and pull the top right belt tight after pulling the bottom right belt tight, it would not be a problem that the top right belt ended up with some slack after the bottom right belt is pulled tight
I think that if we’re pulling in on all of them at the same time then they should all end up tight no matter what because any belt that is lose will become tight when it gets pulled in, right?
The downside is that we’re not guaranteed to end up centered in the X-axis direction which the old approach gave us, but I think that is OK.
I think that if we’re pulling in on all of them at the same time then they
should all end up tight no matter what because any belt that is lose will
become tight when it gets pulled in, right?
I agree this should work. I was surprised to find that the belt commands were
only available one at a time rather than taking a set of what belts to operate
on.
The downside is that we’re not guaranteed to end up centered in the X-axis
direction which the old approach gave us, but I think that is OK.
we don’t have any assumptions of where we are in the x/y space when we are doing
this. we just may have to have people check that the arems aren’t hitting the
frame (or, as we start firming up the anchor locations, we can check the belt
lengths of a new point and compute if we are in a problem area and skip that
point???)
I think the ‘tighen them all’ approach should be done for all measurement
points, in both orientations, not just the first point in hoizontal orientation,
but let’s get it working for the first point first.
