LinuxCNC time again

Yes, call me a sadomasochist but the time has come for me to get involved in this again. The CNC Bantam sort of works but there are a few issues.

Why the Bantam? 

You might wonder why I'm bothered about the Bantam, given that I have a "proper" CNC lathe in the Tree UP-1000 but they are quite different machines. 

• The Tree can only sensibly handle work up to perhaps 6" diameter or so and it lacks a conventional tailstock. 
• The Tree's tailstock provided is only intended for supporting the work, not drills, taps etc.
• The Tree would struggle to do any heavy duty tapping or threading, as it has a fixed pulley ratio between the motor and spindle. While it's rated at 4kW, operation at low speed and high torque isn't really its strong point.
• The Bantam has a tailstock, a decent length (approaching 20") and a half decent swing (6.5"). It can also handle a faceplate and weird workpieces.

In short, there are some jobs that I need to Bantam for. However, last time I messed with the setup, there were some "unresolved issues":

• Andy Pugh's macros required Python 3, yet the current version of LinuxCNC (back in early 2023) only included Python 2. I got the macros installed in the GUI but they wouldn't actually do anything.
• I had grief with the RT default kernel and had to f*ck about a lot to get things running, eventually having to install a different kernel.
• I'd almost lost the will to live when getting the closed loop axes set up (X and Z axes have glass scale encoders).
• And the WMPG took some setting up, to say the least.

So, rather than leave this for another year, I hope to do a fresh installation of both Linux and LinuxCNC.

Fixing the intermittent errors:
Before getting caught up in updates, there's a problem with the cabinet. Sometimes the system works and sometimes it complains that the breakout boards aren't connected. That's classical "dodgy connector" stuff, requiring some investigation.

The lid will have to come off the PC. But that's got to happen anyway, as I need to swap the SSD so I can create a new installation while not losing the last known good (= working) installation.

Some photos for the record. 7i76 (top) and 7i85 (below). The 5i25 sits inside the PC on an ISA slot. It has an FPGA and talks to the external boards by means of ribbon cables via a couple of DB25 connectors:

7i76 closer up:

7i85 closer up:

The encoders for the X and Y axes and the spindle come in through the DB9 connectors at the bottom:

Here's the machine itself. Ground ballscrews, glass encoder scales, AC servos etc:

Installing LinuxCNC:

Currently, LinuxCNC is at v2.9.4 and it comes as an ISO image along with Debian 11 "Bookworm". Note that although Mint is built on Debian, this build uses XCFE, which is more "functional". Mint looks almost pleasant and almost Windows-like in appearance. In contrast, XCFE look like a clunky GUI from over a decade ago. Ho hum - I'm not going to make any further work for myself, so will have to swallow that.

I struggled with the install. I think this is because I used a 256GB UFD for the ISO file. Finally, having tried both Ventoy, Rufus and Etcher, I changed to a 16GB UFD and all was well. That was a couple more hours of my life I won't get back.

"New" (ie previously used) SSD:

Linux XCFE up and running finally:

Configuration challenges:

Having installed Linux (and LinuxCNC), the next challenge was to copy the various original config files over from the NAS (network drive), where I'd copied them. Naturally, the file manager that comes with XCFE is incomprehensively shite. Mapping a network drive in Mint was a piece of piss but XCFE's "Thunar" file manager boasts that it isn't really intended to be a network file manager. WTF? So you have to install the Gigolo app ("it just mounts what it is told to") and connect to the Samba (SMB) server in the NAS.

One suggestion on the LinuxCNC forum on how to do the update was to copy the Gmoccapy folder across into the new installation. Well that didn't work.

My approach is to try to create a functioning installation of Gmoccapy using the Pncconfig utility, then edit the resulting files by swapping out the relevant contents of the originals.

Naturally, the Pncconfig utility doesn't list my combination of Mesa boards (5125, 7i76 and 7i85). So although the 5i25 FPGA is flashed with the right bitfile, I can't select it from the Pncconfig utility. Consequently, I can't set up the HAL entries - or more precisely, I can't use the utility to configure them and generate the appropriate HAL file. Hopefully somebody on the forum will point me in the right direction.

Doubtless more to come....

Skates on, boys. Steerman SCS-20 caterpillar skates.

Yes, having battled with the Chinesium machine skates while moving the CNC lathe in the workshop, I've been on the lookout for something a bit more capable. For one thing, I will need to move The Shiz at some point, assuming we finally move into a more spacious and better located home at some point. It was moved here using two 2 tonne pallet trucks, something I'm not going to be able to repeat easily without buying, well, two 2 tonne trolley jacks. Better to try to obtain some professional machine moving gear.

So, another day, another accident with a mouse while surfing the internet - and ebay in particular. This time, I found a set of Steerman SCS Caterpillar Skates of the 20 tonne variety for £208. These were obviously ex-hire equipment that hadn't simply sat on the shelf, unwanted and unused. A bit battered but still intact and possibly even usable.

There's not a lot of information on how to use them but with a bit of digging I found a sort of user instructions called "Operating Instructions and Parts List" (which actually contains little more than safety warnings and zero parts lists) and a "Technical Specification" which simply comprises a few outline drawings and dimensions.

There's a large blue, wheeled metal trunk for holding and moving the gear. These are the main components - 4 skates, with 2 fixed, mountable pads and 2 swivel pads:

There's also a couple of angle iron braces to force the fixed pads to remain parallel and a couple of long handles for steering and pulling the front / steerable skates. And a hook with a log handle for general pulling duties.

They were a bit scrappy and dirty. oddly enough. So it would be wrong not to dismantle them, clean them up, possibly paint them and reassemble them.

The "caterpillar" tracks are essentially just chains with very wide rollers. You can see that some of the link pins will be easier to remove than others, possibly including the original joining pins.

Each part has a hire company sticker and a stock number. It's Brandon Hire, which is a nation wide company. They don't seem to keep them any more but Speedy Hire hires them out at £78/week. 

After some degreasant and a bit of masking tape, I slapped some yellow Nitromors spray paint on them.

Don't look too bad. Now I just need to reassemble the tracks and possibly lubricate the rollers.

Laser tool setter for The Shiz!

Wow, lasers in da shop?

For a few years now, I've been using a Renishaw TS27R tool setter for touching off tools in The Shiz. 

This is a massive timesaver and allows pretty accurate tool length measurement, which is rather important if you want to be able to change tools in the middle of a job without incurring errors due to a change in tool length - touching off against a fag paper or an accurate gauge (eg a 10mm broken end mill) is OK but takes a bit more messing about.

In Centroid CNC12, there's a pretty nifty extension called CHIPS that was developed by "Swissi" (Roland Kunz) that automates the process of measuring tool offsets and saving them to the tool library. In conjunction with a tool setting probe such as the TS27R, the process is wonderfully simple and quick.

One drawback of CHIPS to date has been that it only measured tool length offsets, not tool radius offsets (due to wear, tolerance and runout). However, a recent update has extended the CHIPS capability to include automatic tool radius offset measurement. This is generally done with a rotating tool, as many (most) tools have multiple flutes or inserts. Although the TS27R is intended to be capable of this kind of operation, it requires the CNC controller to have that functionality. Now it seems we have that available in the latest version of CHIPS when combined with the massive update of CNC12 to version 5.4.

Enough. Show me the lasers.

Rather than require an exotic (= expensive) toolsetter such as the TS27R, the recommended solution developed by Uwe (another CNC12 user / developer) is a "laser fork" sensor. This uses a laser beam to detect the tool, rather than contact with a setter. 

I have to say, approaching the TS27R with a spinning face mill would be a bit of an act of faith. Spinning it within a reasonably large (80mm) gap feels a little less of a buttck clencher, so there may be a lot to be said for this approach.

Uwe's solution was one of these Wenglor "laser fork" sensors.

It's pretty straightforward - once the beam is interrupted, the output goes high. The power supply voltage is 10 - 30V, ie compatible with the 24V used by Acorn's inputs. There is a slight compatibility issue, namely that the Acorn expects a tool setter to be normally low (normally closed to ground), going open circuit (pulled high by Acorn's input) when an interruption is seen.

Got a couple of these from German ebay, with 30mm and 80mm widths.

My machine is already set up for the TS27R and I'm loath to change the setup radically at this stage. I'd like the additional radius offset measurement features of the new CHIPS release, so perhaps I can configure the fork to work in the same way, so that I can simply swap them over as required.

Uwe's solution is to fit a standalone Chinesium "PNP to NPN converter". This is an opto isolator with a normally closed / low output that goes open circuit when presented with a high signal at its input. That's a bit of a PITA, so perhaps I can do something different and possibly even better.

NP to NPN converter:

What's needed is a normally on NPN transistor that is turned off when the laser fork output goes high. That requires 2 switches that could be either NPN or N channel. I have a bag of ZTX451 transistors that should do the trick. These are from Diodes Inc, which used to be Zetex (and before that, Ferranti), with a 60V rating and a decent current gain.

I'm going to make a "rat's nest" assembly, rather than fuck about with strip boards or even PCBs. Here's the guts of it:

This will sit inside the control cabinet at the end of a shielded flex cable.

At the other end, I've grafted a Chinesium connector that mates with the laser fork. This seems to be a common connector that I've not encountered before, called an "M8" connector, due to having an M8 x 1mm thread. This one's a 3 way version. Of course, the colours are all to cock. I figured out that the ground connection is yellow, the +10-30V is red and the output is black. I will correct that by the time the wiring emerges in the control cabinet.

With some heatshrink to protect each wire and a piece of adhesive lined heatshrink over the whole thing, it should hopefully be fluid proof like the IP67 sensor itself.

As the rat's nest isn't going to be very robust against shorting itself out internally, some rapid setting epoxy adhesive should do the trick. Obvs I tested that it worked before slathering this stuff on, as it's a one time event. If it doesn't work under the epoxy, I will have to start all over again:

And here it is finally, with the smaller (30mm) fork for testing. The final installation will use the 80mm fork.

Does it work, Fatty?

Yes it does - TFFT. Next steps will be to configure the tool setting function in CHIPs. I've got CNC12 5.4 working after the big update but CHIPS wanted a configuration setup focused on its own features, which I chose to leave until later.

First, I'll check that the TS27R works as expected, then fit an additional cable gland on the control cabinet for the laser fork. That will be the next workshop session...