DeformedTree said:
What exactly do you think folks are building? 0.1mm over 2m (.004" over 7ft). If I called that out for a part I would have a lot of explaining to do, and that is making stuff out of metal.
Cabinets ?
It is not only about the square inaccuracy here. The rail interface surface not being exact parallel with the (slightly bent) rail full length adds about 0.1 mm per meter to that. Then the bending of the rail itself, another 0.1 or so for 1400. (yes I measured my rails when troubleshooting why the cuts were not square) Then there is the inherent inaccuracy from the reference edge not being absolutely precise/exact either which add another 0.1
At the end of the summation, you are easily 0.3 to 0.4 mm off over one meter length of cut before any inaccuracy of the square would come to play. At that point every 1/100 of a millimeter you can avoid being off is worth its weight in gold, as you have no practical way to reduce the other inaccuracies.
I my practice, 0.3mm off is already barely usable if you have multiple pieces that must meet and they need to be trimmed - which may not always be possible.
But that was not the point - the point was that the accuracy of the square needs to be about 2x or 3x better than the inaccuracy of the rail itself which is about 0.1 mm/meter (0.001" per feet) to make sure the cumulative error is still good-enough. I found that Festool has "calibrated" the max acceptable bend of their rails to be just about non-issue over typical lengths of cut.
When one uses a material reference surface, like a square has to, this about doubles this inaccuracy and moves into the "barely but still usable" territory. There is no space to triple it by the square inaccuracy and be "still fine".
I believe this is why GRS succeeded where others failed. It was able to produce spot-on square cuts by adding only insignificant own inaccuracy to the equation. In my kit case about 5x less than the rails (bends) add themselves. And it is no accident. Both my (Festool version) GRSes are the same precision.
EDIT: fixed math error
DeformedTree said:
...
Now, if they are just going thru a gauging fixture(s), that's much quicker/easier, but gauges are not cheap either.
If it was say Festool with volume production, gauges and so forth would be expected, but for a small shop, making a low volume thing, it would get crazy very fast.
I would expect a manual check against a known-good square against light. If well calibrated (I mean the check) it can give you sufficiently good results as long as you know your employees and can trust them.
That aspect is often overlooked, a small shop can produce very high quality stuff with minimal tooling by moving a lot to the "human factor". Makes it vulnerable for long term stability but works well.
A real/big company with multiple workers (above 2-3 mini-team) must invest in much more proper tooling to remove the human factor if it wants any kind of reliability.
This is actually how the Russians were able to produce crazy-precise stuff with almost no CNC tooling. The same way English could two centuries ago. All was based on good old craftsmanship. The problem is, one cannot scale that beyond low-volume production. but given squares are a relatively low-work-volume thingie and this is mostly a startup, I would not expect them to be too worried there.
