Tool Change: Moving in axis without calling them

smoki3 · 15 Nov 2019, 13:41

It seems that 2.xx (2.02 is fine) and 3.xx have an issue with the tool offsets if you use a tool changer.

If I do a tool change directly from T1 to T2 for example then the printer crashes because it want to recover his old position.
If I do a tool change from T1 to T-1 and then to T2 everything is fine.

It looks like if I do a direct change there is a issue when the tool offsets is deleted or applied. It seems that between the last move of tfree0.g and the first move of tpre1.g it is doing a recovery of the old position ( G1 R2).

@dc42
It was definitely not like this on 2.02 and the older 3.0.

And its a really wired behaviour. Because I use:

; tfree0.g
; called when tool 0 is freed
;
; generated by RepRapFirmware Configuration Tool on Wed Sep 19 2018 21:12:53 GMT+0200 (Mitteleuropäische Sommerzeit)
G91
G1 Z10 F7200
G90
G53 G1 X0 F25000 ;XPOS
G53 G1 Y190
G53 G1 Y195 F5000
M400
M98 P"/macros/Toolhead/1. Toolhead unlock"
G4 P320
G53 G1 Y207 F5000
M400
G53 G1 Y165 F25000
M400

; tpre1.g
; called before tool 0 is selected
;
; generated by RepRapFirmware Configuration Tool on Wed Sep 19 2018 21:12:53 GMT+0200 (Mitteleuropäische Sommerzeit)
G1 X76 F25000
M400
G1 Y192
M400
M98 P"/macros/Toolhead/1. Toolhead unlock"
G4 P400
G1 Y207 F5000
M116 P3
M400
M98 P"/macros/Toolhead/2. Toolhead lock"
G4 P260
G1 Y150 F15000

The I only expect that only a X movement is done while the start of the tpre1.g but it first moves in X and Y.

I copied something together to track this issue here:

@dc42 said in New RepRapFirmware 3.0 early beta:

If you read the firmware update motes for 2.03 you will see that tool offsets are now applied during tfree and tpost (but still not in tpre). Does that explain the issue? You can use G53 to prevent tool offsets being applied.

That change I now, as you can see in my script they already include G53 commands in the tfree.g

But the problem is:

Look at the last movement of my tfree.g: It move to G53 Y165 (no tool offset)
Then the script is finished. Then it starts with the G1 X76 in the tpre (still no tool offset)

What I then expect is that the printer is just moving form G53 X0 (from the tfree.g) to the G1 X76.
But it is also moving in the Y axis.

For me it sound like more that it is a bug of this change:

What new:

Restore points (created by G60 and created automatically at the start of a pause or a tool change) now have their coordinates stored independently of any workplace offsets. So if you create a restore point and then change the workplace offsets, when you go back to the restore point it will go back to the same machine position regardless of the change in workplace offsets.

It seem like it is restore the old positions from the tool coordinate system during a tool change. Because also tries to restore the Z position during the tool change.

Means if you positioned the tool with G53 Y165 in the machine coordinate system then the tool stays at Y210 (depending on the tool offset) in the workplace coordinate system. Then you drop the tool! With the next movement of the tpre.g it tries to move with the machine cordnatesystem to Y210. Which is wrong at this point because for the tpre.g no tool offset is applied and it should just move in the axis you are calling.

There is a workaround for this issue:
If you put in the tpre.g the same command (G53 G1 Y165 F25000) on the first place then the printer is just restoring the Z position and the you can start with the tool change

We already discussed in detail about this issue long time ago: https://forum.duet3d.com/topic/11343/new-reprapfirmware-3-0-early-beta/30

smoki3 · 15 Nov 2019, 13:49

Your post from 15.08.2019:

@dc42 said in New RepRapFirmware 3.0 early beta:

@smoki3 said in New RepRapFirmware 3.0 early beta:

That change I now, as you can see in my script they already include G53 commands in the tfree.g
But the problem is:
Look at the last movement of my tfree.g: It move to G53 Y165 (no tool offset)
Then the script is finished. Then it starts with the G1 X76 in the tpre (still no tool offset)
What I then expect is that the printer is just moving form G53 X0 (from the tfree.g) to the G1 X76.
But it is also moving in the Y axis.
For me it sound like more that it is a bug of this change:

Please clarify. Are you saying that the G1 X76 command in tpre causes a move in the Y direction too? If so then this is certainly a bug. The coordinates stored in the tool change restore point should only be relevant at the end of the entire tool change sequence.

My Post from 15.08.2019

@dc42 Yes the G1 X76 also do a move in the Y direction. Also the Z axis is moving.

It restores the coordinates where the nozzle was before. The reference point of the printer tries to reach the point where the nozzle was before.

The only way actually can avoid this behavior is to put also a G53 G1 Y165 into the tpre.g script. Then only a Z move is done (but also this movement shouldn't be done).

@dc42 I switched now back to 2.02.

Here is everything fine. Also 2.04 and 2.03 has this issue. I tried also the old v3 build from the dropbox. Here also everthing is fine.

I assume that it has something to do with the commits from May 25th.

My post from 20.08.2019

@dc42
I have done some further testing:

If I put not only "G1 X76" at the beginning of the tpre rather a "G1 X76 Y165" (Y165 same Y position as tfree) then just the X axis and Z axis is moving and Y axis stays as it should. Anyway the Z axis shouldnt move at this point.

So if you don't but every single axis into the first command of tfree it will restore the old position of the tool head

Your post from 23.08.2019

https://forum.duet3d.com/topic/11343/new-reprapfirmware-3-0-early-beta/51

My config from 29.08.2019

@dc42 said in New RepRapFirmware 3.0 early beta:

@smoki3 said in New RepRapFirmware 3.0 early beta:
@dc42
It was definitely not like this on 2.02 and the older 3.0.

And its a really wired behaviour. Because I use:
; tfree0.g
; called when tool 0 is freed
;
; generated by RepRapFirmware Configuration Tool on Wed Sep 19 2018 21:12:53 GMT+0200 (Mitteleuropäische Sommerzeit)
G91
G1 Z10 F7200
G90
G53 G1 X0 F25000 ;XPOS
G53 G1 Y190
G53 G1 Y195 F5000
M400
M98 P"/macros/Toolhead/1. Toolhead unlock"
G4 P320
G53 G1 Y207 F5000
M400
G53 G1 Y165 F25000
M400
; tpre1.g
; called before tool 0 is selected
;
; generated by RepRapFirmware Configuration Tool on Wed Sep 19 2018 21:12:53 GMT+0200 (Mitteleuropäische Sommerzeit)
G1 X76 F25000
M400
G1 Y192
M400
M98 P"/macros/Toolhead/1. Toolhead unlock"
G4 P400
G1 Y207 F5000
M116 P3
M400
M98 P"/macros/Toolhead/2. Toolhead lock"
G4 P260
G1 Y150 F15000
The I only expect that only a X movement is done while the start of the tpre1.g but it first moves in X and Y.
@smoki3, I am looking into this but I need your config.g file, also the toolhead unlock files that you call from the tool change files.

I uploaded my config files and my tool changer macros:

Config:
https://drive.google.com/file/d/1xQ5j9E2WvJNFzIum62W02iligk8rOv4T/view?usp=sharing

Macros:
https://drive.google.com/file/d/1QcHqPQBpafr3AlDOjGM3XtQ7nB4VcalO/view?usp=sharing

Danal · 15 Nov 2019, 19:39

May we also see the tfre/pre/post macros?

smoki3 · 15 Nov 2019, 19:41

@Danal said in Tool Change: Moving in axis without calling them:

May we also see the tfre/pre/post macros?

You can see them in the first post and also can download it from the drive links from the second post.

Danal · 15 Nov 2019, 19:42

Ah, they are in the config zip. Got it.

Danal · 16 Nov 2019, 16:45

Running:

Board: Duet 3 MB6HC (MB6HC)
DSF Version: 1.1.0.5
Firmware: RepRapFirmware for Duet 3 MB6HC v0.6 or 1.0 3.0beta12 (2019-11-02b1)

I see no evidence that the system attempts to restore the position of the carriage at the end of the toolchange, that is at the end of the 'post' macro.

If I have a G0 or G1 with R2, near the end of 'post', I can get it to restore. Without that, it stays wherever the last move in Post puts it, no matter where it was prior to the Tn command.

smoki3 · 16 Nov 2019, 16:52

@Danal

I don't know if there is a difference but I use a Duet 2 WiFi with duex5 on a corexy.

It has nothing to do with the tpost script. The problem is between the tfree and tpre and if you have defined tool offsets for both tools in all 3 axis.

And it also only happens if you do a direct tool change from Tx to Ty without a T-1.

With the first G1 X76 in the tpre script I expect that the printer is just moving in the X Axis but also the Y axis and the Z axis (like it will restore the old position of these axis just without the tool offset)

My assumption is that the tool offsets is not correctly deleted and the printer have the wrong coordinates. Because after the tfree tool off set is inactive.

Danal · 16 Nov 2019, 17:11

I believe I see what is causing the move. I do not believe it is related to "G60 S2" or "G1 R2..." in any way.

The way the macros are written, and given these G10s in config.g, I'd expect a G1 move during a T1 T2 swap to move Y by about -73.8, even if Y were not specified in the move Gcode itself.

G10 P0 X0.0 Y74.5 Z-16.22 ; Set tool 0 axis offsets
G10 P1 X0.55 Y74.20 Z-16.28 ; Set tool 1 axis offsets
G10 P2 X0.06 Y74.00 Z-16.35 ; Set tool 2 axis offsets
G10 P3 X0.25 Y75.00 Z-16.10 ; Set tool 3 axis offsets

During 'tfree1' the offsets for tool 1 are active. At the end of 'tfree1' the machine does a state change to "T-1" (i.e. no tool mounted) and all tool offsets go away at that state change. This continues to be true during the execution of "tpre2", there is no tool mounted and therefore no offsets in force. That tool is only considered mounted at the end of "tpre2", and only then would the offsets (for tool 2) be in force.

Therefore, any G1 or G0 move will cause Y to move to where it would be with no (tool) offset applied. Whether Y is specified in the line of gcode or not.

Danal · 16 Nov 2019, 17:13

This can be fixed two ways:

Make every move in 'tpre' macros have a g53.
Do very little in 'tpre' and do the actual mount and position of the tool in 'tpost'. During 'tpost', tool offsets will be active.

There is no right or wrong between these two... still I personally prefer (2).

smoki3 · 16 Nov 2019, 17:17

@Danal said in Tool Change: Moving in axis without calling them:

This can be fixed two ways:

Make every move in 'tpre' macros have a g53.

Do very little in 'tpre' and do the actual mount and position of the tool in 'tpost'. During 'tpost', tool offsets will be active.

There is no right or wrong between these two... still I personally prefer (2).

Yes this is the workaround that I am actually using. But it is not possible for the Z axis because in a print the Z height is variable.

But this a definitely a bug because in tpre.g tool offset is inactive so the should be no difference if I use a G53 or not. If you do a T-1 between the tool change you get a completely different behavior.

Danal · 16 Nov 2019, 17:18

Just for reference, here is a set of files that does minimal work in 'pre' and does most of the work in 'post'.

http://danalspub.com/wp-content/uploads/2019/11/Backup-Toolchange-Jubilee-Nov-16-19-0945.zip

Keep in mind these files are NOT "golden" in any way... they just work on my machine.

Oh, and they are designed to move the new tool to the same position the old tool had before the tool change. If you don't want this, just take out the three G1 (or G0) commands at the end of 'post' that have the "R2" in them.

smoki3 · 16 Nov 2019, 17:25

@Danal
Why do you expect a Y movement?

With the end of the tfree.g the tool offset is gone. So every move you do after this is in the global coordinate system, no matter it was before with the activated tool offset. And the tool offset of the new tool get activated after the tfree.g (before the tpost.g)
And if I place my toolhead G53 G1 Y165 then it have to stay there until it gets another Y axis command. (same for the Z and X axis)

smoki3 · 16 Nov 2019, 17:26

@Danal said in Tool Change: Moving in axis without calling them:

Just for reference, here is a set of files that does minimal work in 'pre' and does most of the work in 'post'.

http://danalspub.com/wp-content/uploads/2019/11/Backup-Toolchange-Jubilee-Nov-16-19-0945.zip

Keep in mind these files are NOT "golden" in any way... they just work on my machine.

Oh, and they are designed to move the new tool to the same position the old tool had before the tool change. If you don't want this, just take out the three G1 (or G0) commands at the end of 'post' that have the "R2" in them.

I know there are many workarounds but it does not change the fact that this behavior is normal

Danal · 16 Nov 2019, 17:35

I'd expect that move because that's my understanding of how g-code works, per the NIST standard.

The NIST spec defines the exact sequence of how coordinates are calculated for every move. "Running" that sequence of events for a G1 after a tool offset is removed WILL result in a move of an axis that is "no longer where it should be".

I could be wrong... but, given that the machine behaves the way the spec reads...

Danal · 16 Nov 2019, 17:45

Of course, now I'm sitting here issuing sequences of M114, G10, G1, etc, etc, and I can't reproduce what I'm theorizing... so I probably am wrong...

I'll keep looking.

smoki3 · 10 Dec 2019, 21:43

@dc42
I found a video on the e3d forum where you can see the issue:

https://youtu.be/5RXE5qRVJgI

At 0:33 you can see the tool change and the moving Z-axis.

Danal · 11 Dec 2019, 15:02

It seems likely to me that all the Z moves from between about :27 and :37 are coded into their pre/post/free macros. I have Z moves in mine, and they have to produce a good outcome no matter what the sequence of events. Therefore, moves occur during a regular single change that look a little weird. They are important when starting with no tool, or ending that way, or etc.

Danal · 11 Dec 2019, 15:03

And, I have not had time to research this any further. I will be getting back to my toolchanging printer soon, to work out the final kinks. I will report back when I know something more about this issue.

smoki3 · 13 Dec 2019, 11:22

@Danal
no they are not coded. Just the lift of the nozzle before the tool change starts. Anything else is due to the fact that there is an issue after the tfree.g script with calculating the position.

If you do a T-1 and then a T1 you get a different behaviour. Then you dont have the Z-movement even though are running exactly the same scripts.

When you go back to 2.02 there you see you dont have this movements.

smoki3 · 13 Dec 2019, 12:31

@smoki3 said in Tool Change: Moving in axis without calling them:

@Danal
no they are not coded. Just the lift of the nozzle before the tool change starts. Anything else is due to the fact that there is an issue after the tfree.g script with calculating the position.

If you do a T-1 and then a T1 you get a different behaviour. Then you dont have the Z-movement even though are running exactly the same scripts.

When you go back to 2.02 there you see you dont have this movements.

I found this comment in the GCodes.cpp, which sound a bit like this bug @dc42

else if (rp != nullptr)
			{
				currentUserPosition[axis] = moveArg + rp->moveCoords[axis];
				// When a restore point is being used (G1 R parameter) then we used to set any coordinates that were not mentioned to the restore point values.
				// But that causes issues for tool change on IDEX machines because we end up restoring the U axis when we shouldn't.
				// So we no longer do that, and the user must mention any axes that he wants restored e.g. G1 R2 X0 Y0.
			}