I was able to dedicate bulk of today for doing the autocad part of the new revision of the machine.

In parts it should be substantially more rigid than the first version in parts, though .. I'm not so sure. Managed to get it a bit tighter but it has cost me little right angle corner pieces I was using in the previous version for ensuring that all the angles are 90 degrees and for additional rigidness. Granted, all that is based on "gut feeling" as I have not bothered to do any calculations yet about .. well .. anything other than budget.

But this time I had an advantage of knowing that I want to build it first of acrylic. I tried to avoid the usual "hole and little tabs for bolted connection" I have seen used by other people for wooden machines and instead have decided to glue some of the stuff together. Obviously there are still little fingers going into square holes. In addition there is quite a lot threaded rod for pulling various pieces togehter ... I went with M5 for this version (previous one was using smaller amount of M6) - mainly because of weight considerations. M5 is 150g / meter while M6 was 250 g / m. There will be approx 15 meters of threaded rod in there in various places.

For X and Y it is possible to use max two steppers per axsis (one in both end) - initially plan is to use single Nema 17 per axis but I have added the mounting pattern also for Nema 23 should the Nema 17 be not sufficient (I expect it to be enough, considering the relatively modest dimensions of the machine). For Z-axis only a single Nema 17 is possible, which must be enough, considering the relatively short max travel distance and presumably rather modest loads on that axis.

There is an option to use four linear rods under the x-axis if using two is not rigid enough. By my napkin corner estimate the 16mm linear rod should have approx 0.2 mm bend through when the carriage is somewhere in the middle of it with two rods. I suspect in reality the first version will have more than that and I should not be worrying about the linear rods more about rigidity of the rest of the machine. I'll see about that, I guess, once the first version is functional. For a start I am going to order only enough parts for two linear rods under the X axis.

Should it be possible it might be easier / more economical to compensate for that kind of stuff in software instead of going for absolutely rigid structure. I have a rough idea in my mind about suitable coordinate transform for taking into account the bend through for z-axis as a function of location of x-axis (could be also easily extended to take into account bend through in both X and Y direction if needed). It is relatively simple (in my mind so far) so hopefully someone has already done something like that. Of-course in reality things that should be simple, in principle, can have some technical challenges associated with them ;) Presumably I will have plenty of time to play around with linuxcnc before my machine is properly functional so I'll think about it more when I reach that part.

The idea is just mapping the bend through with a dial either in 1D or as a 2D matrix, then fitting the results with a suitable function (at first probably just a second order polynomial either in 1D or 2D) and then just applying a correction term to the z-coordinate in G-code to take into account the bend through in that particular x or x/y value.

Almost none of the DIY CNC projects that are shown in the net talk about calibration of the machine. I assume that to be actually quite important part of the machine construction if aiming for any kind of reasonable accuracy.