TPU  

RE: TPU

Jokes aside, has anyone tried one of the following, on purpose or accidentally (sorry I have almost zero hobby time rn):

- what happens when the side and hotend filament sensors disagree (e.g. when filament breaks or is cut between them) ?
- what happens on clogs (both filament sensors triggered but filament not moving/moving too slow) ?
- what happens when the extruder sensor fails to detect (the flaky one seems to be the side sensor) ?

My point being, is the load cell actually involved in the logic yet? This was a major advertisment item the whole time but now we hear nothing about it anymore.

I've had a couple instances where some bad brittle filament snapped at the extruder gear. So both filament sensors still see filament but nothing getting extruded - print went on like nothing was wrong.

ATM, I don't think the loadcell does anything more than sense the pressure during exact instances: bed leveling for the model, bed leveling for the initial purge, tool offset calibration. I think all the 'cool loadcell things' are still in the works like so many other things.

RE: TPU

OK, so it's basically pointless at this stage except for levelling. If you think about it that makes sense, because it's actually the easy stuff. They are expecting a signal within a short time frame, they know its shape and what the onset looks like (because they know the speed and initital bed position). Also they can average a lot of points and repeat in case of strong outliers (although this is apparently not done or at least I haven't seen it yet). That's doable even with high precision and repeatability (which is where the nightmares start even with very straightforward sensors). I don't know much about loadcells, but probably at least environmental effects need to be corrected when you measure more longterm (such as heat changes in the hotend itself or ambient). Possibly also drift and response corrections and other stuff that tends to turn up when the more obvious corrections are done right. And then a ripped filament might be relatively easily detectable, but what about clogs or heatcreep that build up nonlinearly over an unknown stretch of time... I'm quite happy it's not my job to develop and test this kind of thing, so unironically hats off to them if they can manage.
I would have used a mouse wheel type sensor or some combination with a photo diode, period. When in doubt about the flow rate, I'd either stop the print and wait for user intervention or drive off the build area, try to purge and measure the response using all sensors. If possible without collision they could also move the build plate up and try to print a dummy object outside the print area, then check if the loadcell output looks nominal while printing (using stored first layer values to compare).