Setting Z Offsets

Hey guys-
I’m an experienced manual machinist but 100% CNC neophyte. There’s probably an obvious answer to this, but I can’t for the life of me sort it out. I'm trying to program my Z tool offsets (3 axis mill) and am pulling my hair out. I've got 15 or so tools in holders that should handle 99% of the work I'll do. I set all the Z heights in the tool table by doing the following:
-Load the tool & MDI a T(n) M6 G43
-Selected "touch off to fixture"
-Very carefully jog down to a machined flat piece of stock and take a whisper cut
-Hit tool touch off button and leave at 0
I repeated that process for all the tools on the same flat surface. As I understand it, the Z offsets for each tool are now set relative to one another, but not in space.

Then I made a fairly simple program that uses all the tools, with all of them taking a facing cut. My process was to edgefind my origin and hit the "touch off" (not "tool touch off") button in X and Y, load the tool, jog down & take a whisper cut in Z, and "touch off" in Z. I was expecting this process to work (seems pretty straightforward) but it definitely didn't. Using the above it'll cut air, somewhere around 2" above the stock top.

I reloaded all my tools 3 separate times just to make sure I didn't make a math error, or change something inadvertently, but no luck. feel like I'm missing something simple, but can't get it. Anyone see something obvious I left out?

Edit to this post because I didn't read the question properly. The manual for Axis (if that's being used) reads:
"Tool touch-off to fixture is relative to the ninth (G59.3) coordinate system"
linuxcnc.org/docs/2.8/html/gui/axis.html#_a_typical_session

I leave the original post anyway because it's a well done video.


I use method #3 mostly because it's the way I was trained.
Method #2 looks like a good option.
From what you posted it sounds like method #! is being used.

I played a bit with this a little bit while in my shop this evening.
fwiw what you describe above appears to work, or at least it did with the single tool tested.
It's like method #2 above and works based on G59.3 Z offset being zero.
I didn't figure out how to make it fail but then again negative tool length offsets aren't my comfort zone.

That's a great video. If I'm understanding correctly though, I don't have the spindle to table number with any level of accuracy (16" in the video) on my machine. It's a knee mill (manual knee) with no limit or homing switches.

I couldn't tell you how far it is from Z Home (top of travel) to the spindle face touching table on any of my machines. A move that drops Z that far would be a bad thing since there's almost always a tool sticking out That video was meant more for the concepts rather than a how-to.
In the shop today I repeated the procedure from post #1 using 4 tools. No problem. After doing the second touch off (the G54) what is shown in the Var table in the home/linuxcnc/machine directory ? Item 5223 in that file is entry for G54 Z. That number should be the Z distance from the first (tool setting) surface to the second (work piece) surface. Or at least that's what I observed. I'll assume the cutting moves were in G54 since that's default

Being a knee mill shouldn't matter until the knee is moved, then Z is lost. Hopefully someone familiar with cnc on a knee mill will speak up. How the heck is does this work when you run out of quill stroke ?

Your reference point doesn't have to be the spindle nose. Often it's a reference tool that is used, I use my 3D edge finder as my reference tool and use a touch plate to measure the other tools lengths in relation to that . All the other tools z-offsets are then a positive or negative number depending if they are longer or shorter than the reference tool. The down side of that is that if the 3D-edge finder ever needs to be replaced I need to change all my tool offsets.
Many shops use an external tool setting setup to measure Z-offset of a new tool in relation to some special reference tool (say the ground front of a particular tool holder). That way the 3D-edgefinder also gets a z-offset and is handled just like any other tool and the tool offsets can be used on different machines.

I started out measuring tools on a bench fixture with height gauge (Vernier ) before moving on to more automated machines. On the home brews I use now the machines do double duty as a height gauge. Old dogs hate new tricks or something.

Why does the procedure outlined in post #1 work for me and not Jmcghee ?
One thought is a G92 Z is active.
Or the cutting program is not in G54 or using G43 improperly.
Another thought is what happens if a machine is defined 'upside down' ? Meaning Z axis 0 to some positive distance instead of 0 to some negative distance. It's the kind of thing that can make a person dizzy, which is why I set up my machines and offsets like a proper Luddite

And does this work the exactly the same across versions of linuxcnc ?

I don't use my mill all that often so I'm probably not the best to give advice on how to handle tool lists and setup. Here is a video for how it can be done in LCNC:

Another thought is what happens if a machine is defined 'upside down' ? Meaning Z axis 0 to some positive distance instead of 0 to some negative distance.

Not quite sure what you mean there. The convention is that the tool points in the negative Z direction.

[edit]

Also sometimes one forgets to activate G43 after a toolchange. e. g."T1 M6 G43"

The guy in the video is a developer from HeeksCam/FreeCAD , you probably already knew that but he deserves a shout-out. I was happy when my workplace purchased machines with tool setters. I was equally happy there was a button to call the maintenance department if the thing ever malfunctioned. As retirement hobby I'm into the KISS principle.

re the "upside down" comment, a bit of google shows I'm treading into a religious argument. In my mind Z exists in negative space or in linuxcnc.ini terms Max is zero (top) and Min is expressed as a negative number. Without bothering with the math I suspect this is a red herring for OP's problem since offsets are relative. Thinking through this I now realize the Z min should be a precise measurement to the table instead of tape measure close. In turns out I got something useful from this conversation while possibly not helping OP at all.

btw numbers can be negative as a vector but numerically less positive. That doesn't confuse others but ties me into knots sometimes.

The guy in the video is a developer from HeeksCam/FreeCAD , you probably already knew that but he deserves a shout-out.

No idea who the guy is but the video seemed informative. But then I'm a sloppy reader and I may very well have missed the point of the original question entierly.