Advise on new lathe project

I've been wanting to do a LinuxCNC project for a long time, and I finally have a lathe that I'd like to CNC.

I have some questions about MESA cards. Some people (like Rotary SMP who I've watched for years on yt) seem to go for the simplest interface board (7i92) and then add expansion boards. I'm not sure about previously, but it appears now that doing that costs almost the same as just getting one of the larger ethernet boards to begin with (7i95 or 7i96). Is there any reason to separate all the boards out?

I think I want to do my project in stages - scales first for a DRO, then motorize the z axis to add ELS functionality, and then motorize the x axis for full CNC. I could start with the Clough42 ELS, or Stefano's open source DRO that RotarySMP showed recently, or others (like this one that has some very nifty features that most other ELS don't seem to have - threading or surfacing to a known stop point, driving back to the start, and repeating, or doing "patterns" or multi-start threads: ), but it seems to make the most sense to just go right to LinuxCNC from the get go if that's the ultimate goal. (Let's not kid around though, I say "stages", but I've already bought an encoder for the spindle and a motor for the z-axis, but I haven't bought scales or the x-axis motor yet).

I don't see any issue using LinuxCNC as just a DRO to start with, but is there any ELS type functionality, driving one axis with a ratio from the spindle encoder (even if it requires some hacking some code together - I'm a software guy, so even if that requires writing and testing some cde from scratch, that's not really an issue)?

So because my goal is 3 encoders (spindle + 2 scales at least), and I want ethernet based MESA card, that kind of leaves me with 7i95 (basically can do it all in one board) or 7i96 with an expansion board for more encoder inputs because it only has one encoder input. Both options cost wise are almost the same. The only real difference I see is the 7i95 board doesn't have an analog out for the spindle VFD, but my VFD has RS-485 anyway, so I think I can control it directly, and don't need an analog output on the mesa card. Is there any other reason to go one way vs the other with the MESA boards?

Does anyone have any advice on using scales for feedback rather than changing the leadscrews to ballscrews? Is that an OK idea? LinuxCNC can compensate for the backlash in the screws with scale feedback, right? That's my plan for sort of both to save some money and effort, and to be able to still use the machine manually as well. As I understand it, ballscrews are not good for a manual machine because they can be easily back driven by the cutting forces. This machine is not exactly my "forever lathe", but it's the machine I have space for and was able to find at a reasonable price locally. I say now that I might want to use the machine manually, but I could just as easily see myself using lathe macros in LinuxCNC to do "manual" turning as well...

If you do intend for full cnc conversion I wouldn’t waste time and money on an ELS. Just go for it.
One thing I have found is that for some of the more simple tasks it can be a bit of a pain without a mpg for each axis.
Ball screws without question, for me since I didn’t have a decent Z leadscrew it was cheaper to do a ballscrew conversion, and somewhat easier. Make 2 plates to mount the bearing blocks and replace the apron with a section of square up angle iron.
My opinion is that it’s better to reduce backlash and any looseness mechanically first.
I have seen a couple of YouTube vids where the owner has said that if they were going to do it over they’d opt out from installing scales as they felt the cost didn’t justify any improvement.

There is one user that has a github page for an ELS type setup for Linuxcnc. If I remember correctly it does provide some manual support as well. I think if you search the forum for something ELS GUI you might get a hit.

github.com/tinic/ElectronicLeadscrew

Yeah, I was doing some googling, and found the manual ELS Gui for LinuxCNC posted here in another thread. It looks promising, but I think it's only in metric - I'd extend it to work in imperial and metric when I get to that stage.

I was also reading that people thought that scales were a waste of money too, if doing a proper conversion.

I think I also saw your posts about putting 2 MPGs on it to retain "manual" control - that's a good idea, and MPGs can be had pretty inexpensively these days! I should make sure whatever MESA card I choose has enough inputs for 2 MPGs - I know I saw mention of that in one of the manuals for one of the cards.

I will probably do the full ballscrew conversion eventually, but I think to do that effectively (or without paying $900 for a kit, which is almost as much as I paid for the lathe itself), I need my mill up and running, and not taken apart and covering my entire garage floor! I should really focus on that machine and get it running first before tearing into another (albeit way smaller / lighter / easier to manage) machine.

I invested too much time in repairing (and hacking) the Mitsubishi controller in the mill to convert it to LinuxCNC now, so the lathe will have to be my LinuxCNC based machine!

LinuxCNC can do up to 9 axis, so yes you can do a single axis CNC (which is what an ELS is). I did that on my Boley, for which I had no change gears. I just made up a bracket to mount a motor and pulleys to drive the lead screw, kludged up an encoder, and PSU, and used a PC to control it. Was never a permanent solution, I only used it to cut the nose spindle thread in a backing plate.

Backlash is less of an issue for a lathe than a mill, as you mostly don't change direction with the tool in the part (at least not on both axis), and you don't have the issue of climb milling. My mini lathe has pretty bad backlash on both axis, but can still turn somewhat precise parts open loop using LinuxCNC's backlash compensation.  I suspect that closed loop with linear encoders, with significant backlash in the screws, would be pretty tricky to tune the PID loops on.
However, as Cornholio said, it is best to simply engineer out the backlash. The Schaublin has no linear encoders, so I am DROing off the motor encoders. I would not buy a ballscrew kit with cheap rolled screws. Get some nice C3 or better ground ballscrews pulled from a fab in Korea, and make up you own mounts.

One of the big advantages of CNC is that you can put on enclosure, or at least better shielding around the chuck and constrain the mess.
You mentioned using an ELS to cut to a shoulder. Nicer is using two axis CNC, to cut to the shoulder, then retract out before it starts rubbing and chattering, and returning for multiple passes
However I also really like manual lathes for quick bodge jobs which need no precision, and like the "feel" of machining on a manual lathe.

As Cornholio has said, if you are going to end up with a full CNC, you are probably better off doing it all in one go. Doing a mixed CNC/manual interface is harder to do well, than one or the other.
With the Schaublin, I found my initial use of the machine, with a loose wireless keyboard, and touch monitor leaning aginst the tailstock, to be pretty sketchy and unsafe.  I felt a lot more comfortable with the safety once I built the control module with some physical buttons like feed hold, the encoder for feedoverride, and the jogging joystick etc.

The 7i73 is so cheap, having a single Cat 5 cable to you human interface can saves hour of discrete wiring.

I choose that set of diverse cards on the Schaublin for two reasons:
1/ when I went through and made a list of all the I/O, encoders, etc it would need, it was not going to work on one of the integrated cards. I currently have six encoders, (Spindle, X motor, Z motor, feed override, spindle override, jog override), with two more planned ( X Jog wheel, and Z jog wheel on a manual jog pendant). I also need a second 7i84, instead of the sketchy relay board to use available 7i84 5V outputs to switch 24V loads.
2/ I knew I would have a space challenge, and breaking up the cards reduced the need for a single big footprint in the control cabinet.

On the mini lathe, I just used a 7i96, and on the Maho a 5i25/7i77 (The 7i97 didn't exist yet), plus a 7i84 as it needed addition I/O.
Cheers,
Mark