Advanced Search

Hi Mark,
I watched one of your Youtube videos again on the wonderful tool made for you by the guy in Malta. Nice to have friends like that.
A lot of good comments there. But I don't think you can add photos to the comments. So I'll stick my nose in here instead.
It's and area that makes my head swim. But it seems to me that the method is really just about making the quill square to the table. The vertical head on our machines only rotates around the Y axis, but on many Bridgeports it can also  rotate around the X axis. In that case you could use the device to make the quill perfectly square to the table. But you could also tilt their table to any angle (some can do that) and tilt the head to be perfectly square to that using  your device. Then you could use the quill to drill nice holes. You might do that to fit an oversized object perhaps? But it would do nasty things to the drill if you moved the knee while drilling. An extreme case example. Feel free to call me and idiot if I'm seeing it wrong.
My machine has that nice taper roller grinder style quill, which has a slight preload. So no looseness, but it will still have non repetitive run-out errors due to the various manufacturing errors in rollers etc.
On my machine you can see there is a detent knob on the head tilting. It clicks home at zero degrees on the tilt dial. When I put my dial clock on it then and swing it I get almost perfect alignment in the X direction. Big surprise! The head rotation clamps just like yours but with 4 bolts. You can see they scribed the zero mark on the head after they did the alignment by the missing paint.
I've added a photo of when my rotary table is removed. You can see the 8 bolts that clamp it on (so I don't misplace them). There are keys on the rotary table that fix the tilt when it's assembled. You can see it's not a dovetail system. I suspect I could align the table sag by removing that array of bolts along the bottom and putting shims between that X slide and the bearing rails.

Cheers,
David
 

OK, thanks. I'll give it a try. I'd like to add a button instead of macro buttons in qtdragon. thanks in advance.

It's not necessary, but if you have an encoder it's useful for verifying spindle rpm. The encoder signal does need to be connected in Hal for threading.

I have also your ic-nv circuit and had noise problems and voltage drop to 5v line that resolved by lowering the triac gate resistor.
For fuse can use a fuse with soldered legs because fuse pcb base can be problematic.
Noise problems resolved 90% by using other 5v power supply.
I think i will use ferrite beads to ic-nv boards 5v power cable or common mode choke and put
these boards inside grounded metalic box for the remaining noise problems.
Noise seems to appear when servo drives enabled that emit noise.

I'm browsing through your build of the lathe and I think somewhere there you got into a discussion of ChatGPT.
I've been feeding AI's G code for a while now to see how it goes. It's still hopeless. And I noted you mention it didn't understand subprograms. That also applies to program flow in general.
It occurred to me that in most cases where it's scraping the entire knowledge on the planet, it usually gets to see a coding  problem and attached it sees a printout of the answer.
In our case the "printout" is actually the movements of the machine itself. So I suspect what it needs to train on is a video of the machine movements with a picture in picture of a block of execution concurrent with what that block does on the machines move.
Not likely to scrape that currently off the internet I suspect.
But there are AI programs that now do very realistic art, movies and presentations of people who are dead saying modern speeches etc.
So it should be possible to combine both types of AI to make really fantastic CNC programs?.

You've seen how I usually put a lot of comments that explain every line of my G code I think?
If I remove the comments the AI is hopeless. It will give a vaguely generic statement of what some of the lines mean. If I leave the comments in it does better, but if you look you see that has just cleverly reworded the comments. It does no better with the code at all.

Maybe someone in the LinuxCNC community knows someone in the AI community who could run with the concurrent video/ block by block idea. Nearly every CNC machine on earth uses G code and it would be great if we could just get an AI to program them all.

Cheers.
David

Here is the big one. I split it using the print to file and custom page numbers. Hope it works

 Sounds like you spot on with you home made lubricator.Since the mechanical logic in the Willy Vogel system is dumb there is the possibility that the relief pressure to the tank may end up at a lower pressure than the controller pressure switch. It hasn't happened on mine though.The Vogel controller is very simple on mine. It only allows you to set the time between lubrication events. And it always does a cycle on power up. During commissioning my machine it dumped and the entire tank because of that. So I put a delay on timer on it to delay the startup cycle.It also has a mode where it will lube after a certain number of pulses are sent to it. So you could put micro-switches on the drive motors to count the slide movements and oil accordingly. My Heidenhain has the ability to count slide distance traveled, but I haven't figured out how to invoke that.Plus I would need to have 4 individual solenoids to deliver to my 4 axes individually as they reach the distance.BTW: the pump on the lubricator is a a cycloidal one typical of many cars. That's so the pressure pulsations are smaller and don't confuse the pressure switches.

Got the power to the screen and pc wired:)