Tuesday, November 15, 2016

A Little Shop Update

It's time to do my more-or-less weekly update to the shop mods I've been doing.  I'm surprised to see it's two weeks since my last real update, I mean with pictures of what I've been working on and all.

Most of this week has been taken up with shuffling computers around.  That always seems to take much more time than I think it would.  I spent a lot of Friday and Saturday trying to convince the Linux installer software to not commandeer the entire drive on the shop computer, but to leave me a partition for Windows 7, just in case.  The Linux distribution that comes with LinuxCNC is from Debian and it's called Wheezy.  This isn't exactly what Debian distributed as Wheezy, which was replaced about 18 months ago (in the same sense Windows 8 replaced 7 and 10 replaced 8).  Driving the parallel port properly is no simple trick and requires a separate low level hardware control called the RTAI kernel, so this Wheezy contains the RTAI kernel.  Confused yet? 

After several unsuccessful attempts to be sure Wheezy was going to allow me to keep the Windows partition, I declared defeat.  While Wheezy wasn't cooperative, I know that I've done this with Ubuntu several times, and I found a possible way to do this with an old version of Ubuntu called Precise.  Getting Precise to install dual boot was easy, and within a half hour I was back trying to get LinuxCNC installed.  I thought I could get it to support the RTAI kernel so it could run LinuxCNC but it kept telling me I couldn't do what I wanted.  Trips to the LinuxCNC Wiki and forums were fruitless.  After a few hours of that, I declared defeat and thought I'd try to get Wheezy to install itself over Precise.  When I did that, it promptly commandeered the entire hard drive for itself, destroying not only my bare installation of Ubuntu, but the Windows 7 OS I wanted to keep "just in case". 

I know of no way to recover that, so I said "Uncle", and committed to getting LinuxCNC running.  That was Saturday night.  Since then, I got the system to run my Sherline/A2ZCNC mill properly, and today got backlash compensation set properly on all four axes. Four axes may surprise some of you, so I grabbed a photo of the rotary or A axis. 
I came up with this way of checking backlash and it seemed to work, although I'd like a better way.  The square lathe cutting bit is held in a four jaw chuck and the rotary table initially set to zero degrees.  The magnetic holder for the indicator isn't doing anything (the table is aluminum), so I can slide the indicator across the half inch wide bit and ensure it doesn't move (it did at first, and I tweaked the angle until the bit was level and the indicator not moving - I made that my new zero).  Then I rotate to 90 degrees, which brings up another square face which should be level, and it is.  Finally, I go back to zero and make sure it's level, which it is.  Backlash would show up as 0 degrees not being level when I go from 90 back to 0.  A wider bit would be more sensitive, but I didn't have one lying around. 

None of this is a big deal, but my mill is usable again and I expect to be using it as I go forward with the work on the Grizzly conversion.  I spent entirely too much time with trying to install the system the way I wanted, though.  It would have been far easier to buy a second drive and keep Windows on that.  Instead, I just have my installation CD. 


  1. That is some good out of the box thinking.

    Digital angle gauge? If the gauge is accurate enough, and you stick the digital angle gauge to one of the jaws with the back of the gauge against the chuck, and then zeroed the gauge, would the gauge read 0.0, 90.0, and so on as the chuck turned. And if you turn the chuck in both directions during the test do you get the same results.

    Simple laser pointer? Laser weapon sight? My first thought was to bounce the laser beam off of the flat of the square bit and see if the spot returned to the same point. But then I thought that a longer action arm would magnify any backlash errors. If you stick the laser emitter to the chuck, so that the beam radiates outward, parallel to the shank of the chuck key, you could mark the laser spot on the ceiling, and two walls. If the laser spot returns to the marked spots then the backlash is good. (And you have invented the first CNC Cat Amusing Device) I am sure that by calculating the distance from the emitter to the walls/ceiling and applying the math you could figure out the accuracy of the backlash.

    1. I'm kind of surprised I didn't think of using a laser boresight finder, considering (1) we did that at my last job to verify that our radars were sweeping properly and backlash was cranked out, and (2) I have a boresighting laser about 4" from my left hand, and several other lasers around the house. Likewise, I have a digital angle indicator that has magnets on it and would stick to the chuck, or a piece of sheet steel.

      On the other hand, what I was doing would allow me to see a backlash of under a quarter of a degree fairly easily. (1/4" * sin(1/4 degree) = .0011" - pretty easy to see on a dial indicator) 15 minutes of arc? Is that good? How good does it have to be? It's probably acceptable for the size things I work on the Sherline. The table is 4" in diameter; if I worked on something that size, and really had 15 minutes of backlash, that's about .009". I don't like that amount of backlash, but it depends on exactly what the work is.

      We all like the idea of zero backlash, but I think there's almost always some wasted motion.

  2. Are you sure you want to go to Linux CNC? Those little all in one Mach3 boards are working great for me.

    On my third machine with them here.

    Delivered from China I got 4 axis board, 4 steppers, power supply and cable for $188 all in.

    Btw those Nema23 stepper spacers you was making.. I see that there are extruded models for about $6. I am thinking thats a better deal than 3d printing mine.


    1. Well, my fault for not investing in a 3D printer first. It's easier for me to make mine out of aluminum, all things considered.

      Could you drop a link for one of those Mach3 boards? I think you have information that I don't. I did some searching for "known good hardware" on the Linux forums and the only thing I could find were PC motherboards that didn't even have a processor.

      The steppers I'm using are Automation Technologies, 570 in-oz. I think the next higher torque are NEMA 34.