Re: Re: Auto tool changers

Ballendo

I snipped a bunch from your interesting reply. I hope this post gets to
the point that I was trying to make with respect to your suggestion that
tool change Txx words should be coded several blocks before the
corresponding tool change M6 is programmed.

Disclaimer -- This post applies primarily to machines that use a "next
tool ready" type of tool changer.

>    From: "ballendo" <ballendo@...>

<s>

> Most good programmers will list the "T" word at least A FEW BLOCKS
> AHEAD OF THE M06, so the toolchange turret can "get ready"(rotate;
> talking 'bout mills here) so as to have the next tool in position for
> a quicker tool change WHEN THE M06 ACTUALLY INITIATES THE PHYSICAL
> CHANGE to the new tool.

<s>

> Lookahead exists to allow cutter radius comp, and to prevent crashes
> and gouging from accel/decel issues.

<s>

I did a little testing on the EMC. Four tools is probably overkill but
here is the program I ran.

-----tooltest.ngc-----
n1 g0 x0 y0 z0 (Tool change test program tooltest.ngc)
n2 t1 m6 (Preset and load tool 1)
n3 g1 f5 z-1
n4 x1
n5 y1
n6 x0
n7 y0
n8 g0 z0
n9 t2 m6 (Preset and load tool 2)
n10 g1 f5 z-1
n11 x1
n12 y1
n13 x0
n14 y0
n15 g0 z0
n16 t3 m6 ((Preset and load tool 3)
n17 g1 z-1
n18 x1
n19 y1
n20 x0
n21 y0
n22 g0 z0
n23 t4 m6 ((Preset and load tool 4)
n24 g1 z-1
n25 x1
n26 y1
n27 x0
n28 y0
n29 g0 z0
n30 m2
-----end of tooltest.ngc-----

I asked for full debugging so that all of the EMC commands would echo to
the terminal. Debug shows read ahead, look ahead, commands, and task
completion in the order that they occur.

The EMC uses an internal command EMC_TOOL_PREPARE to indicate that the
tool in pocket #xx needs to be moved into place. It uses a command
EMC_TOOL_LOAD to indicate that it is time to change the tool in the
spindle.

What we are looking for is the interval between the issuing of a
EMC_TOOL_PREPARE command and the EMC_TOOL_LOAD command. I'll ignore the
first tool change command since it can't be programmed to make any
difference for our discussion. The terminal output shown below is for
the second tool. Immediately before this quoted block there is a series
of motion commands as the interpreter looks ahead and begins to execute
lines numbered n3-n8.

-----debug output-----
...
Issuing EMC_TOOL_PREPARE -- (+1104,+16, +0, +2,)
Motion id 3 took 11.980118 seconds.
Motion id 4 took 11.979220 seconds.
Motion id 5 took 11.969769 seconds.
Motion id 6 took 11.959347 seconds.
Motion id 7 took 11.969727 seconds.
Motion id 8 took 0.949585 seconds.
Motion id 0 took 0.000005 seconds.
Issuing EMC_TOOL_LOAD -- (+1105,+12, +0,)
...
-----end-----

There it is as clear as you could want. While running this program the
command to prepare a tool is issued nearly a minute before the tool
change command even though they exist in the same program block. This
timing depends upon feedrate but it illustrates is that as soon as the
interpreter loads the tool change line into it's look ahead buffers, it
issues the tool prepare command! This timing is typical for subsequent
tool changes as well.

Now this brings up the question, "Okay so the command is issued early but
what does the machine do with the command." The answer is all PLC and
only PLC.

> PLC's exist as a semi-
> intelligent "servant" OF the control, which distributes the "brain"
> power needed to run the machine. But I've not seen them work together
> to "predict/discern" the next tool and get it ready, tho' they
> could...

I looked in a Meldas 50 control book that I have on hand. It says,
"Processing and completion sequences must be incorporated on the PLC side
for all T commands."

Your servant notion for the plc really denegrates a complex, symbiotic
relationship between these two systems. While there are many "hobby"
level machines that use little or no PLC, most CNC machines have
extensive PLC logic that intertwine Machine and motion. For example, in
this tool context, once the change tool (m6) command is given by the CNC,
it doesn't do jack until the PLC says, "I've got this tool in place and
it's ready for you to use." In that case who is the servant of whom?

Unless a CNC control has no look ahead or uses look ahead in the very
narrow, motion only way that you describe above, I can't see any value to
programming the way these experts suggest. In fact my initial thought
when I read your suggestion was that this was an urban legend that grew
from programming the first ever tool hive fitted to a 60's generation NC
controller running from punched tape. I really hope that even the hobby
level CNC units have moved past that.

I see your disclaimer "most good programers" but I guess that I program
using the exception 'cause a single line of tool change code is a lot
easier for a human to understand that a Txx block followed by a bunch of
blocks that RUN USING A TOOL OTHER THAN THE ONE indicated.

At the very least we can use the result of the test described above to
show that this programming technique has little value for the EMC.

HTH

RayH -- U.P. Michigan