Y Adapter filament switching

[Yonkiman]

I totally get the basic principles. It's just that your T0.g script seems incomplete:

You have to pull a bit back, cool it, extrude it, then yank it back again.

Where is the "yank it back again" part in the T0.g procedure? Does another macro start that yanks it back? Does control go back to the object's gcode file where the Tool Change script yanks it back?

Maybe I just don't understand when/how T0.g and T1.g are supposed to be used. To me they look like scripts that retract the filament, then shove it back in (minus 1/10 of a mm), and then things go idle, almost as if Tx.g had never been run.

Thanks again for bearing with me!

[kraegar]

Ok, now we're on the same page. T0.g does the "prep" work of priming the filament, so to speak, then at the end of it is the command "T0"

So, if your current extruder is T1, and you run my "T0.g" it will do the following:

Run T0.g
Run tpre1.g
Run tfree1.g
Run tpost0.g

Those last 3 are called when the T0 command is interpreted by RRF.

[Yonkiman]

That's it! Should have realized that a tool change will call those three .g files regardless of where it comes from. Thanks!

[Yonkiman]

I got it (mostly) working! Can swap filaments with T0.g and T1.g doing your packing trick. Dealing with some bed-leveling issues (the two Bowden tubes are causing some flexing of my printhead) and working on the same trade-off between putting the code in S3D vs in the Duet's .g files. I also think the swapping might be easier with my Lite6 (bowden tube all the way down to the nozzle) than with an original V6. Before I got your packing working reliably, I saw some pretty ugly, long filament trails, but it never jammed. Also, I didn't appreciate how useful it is to have a transparent Y adapter! Was originally going to try this with a printed Y, but the ability to see what's happening is alone worth the $20.

I also had an entertaining accident: S3D writes the extruder commands in absolute mode, but somewhere an M83 was accidentally inserted (or an M82 was missing), so my extruder was interpreting absolute commands as relative, and my 4:1 geared extruder sounded like a jet's turbine engine as it started spinning up with no end in sight. Apparently the right commands in the right order is sorta important.

Thanks again for your help, kraegar. Will post my best config once I dial things in a little more. Please let me know where you settle on the S3D/macro split.

[kraegar]

Awesome, glad it's working. I suspect the lite6 will indeed do better - it should suffer from heat creep less. Depending on what extruder you're running, if you can retract at 100mm/s you might want to try that method, as it was by far the cleanest.

Playing with the pause and such can make a big difference, too.

[DjDemonD]

This seems very relevant to the latest edge firmware:

Known issues

If you enable tool mixing, you should use relative extrusion only. If you use absolute extrusion, then if you pause and resume the print, the extruder is likely to extrude the wrong amount of filament in the first move after resuming.

[kraegar]

I only use relative extrusion during my prints, anyway, but I've wondered what would happen if you were switching from absolute to relative & back within your free/post scripts. I was guessing it probably wouldn't go so well.

[Yonkiman]

Ah…didn't know there was an S3D option for relative extrusion. That should make things a lot safer!

Here's my current setup/scripts:

S3D Tool Change Script (at this point I'm doing it all in tpost and tfree):

T[new_tool]		; Change tool

tfree:

; tfree0.g
; called when tool 0 is freed

M83			; Make sure extruder is in relative mode (prob not needed anymore)
G1 E-5.0 F3600		; Retract filament as soon as possible...
G91			; Relative
G1 Z2 F6000		; ...Then lower the bed...
G90			; Absolute
G1 X115 Y232 F18000	; ...Then get the hell away from the print
G4 S2			; Pause two seconds to cool the retracted filament
G1 E4.9    F3600	; Pack the filament to eliminate the tail
G1 E-150.1 F6000	; Yank the filament past the Y adapter and into its Bowden tube

tpost:

; tpost0.g
; called after tool 0 has been selected

G1 E145 F6000		; Extrude 145mm @ 100mm/s
G1 E5   F1500		; Extrude   5mm @  25mm/s
G91			; Relative
G1 Z-2  F6000		; Raise bed
G90			; Absolute

The retracted filament end doesn't look perfect (needs to be optimized for PLA with an E3D Lite6):

But it was working well enough that I could print this:

It's pretty ugly because I just built this printer and have a lot of single-filament fine-tuning to do, but it's a solid piece of plastic in two colors! Thanks for all your help, kraegar!

Now back to the fine-tuning…

[kraegar]

Off to a good start!

To test my switching, I just made a macro to retract a single extruder, and ran it over and over, slightly changing things until I got the cleanest break I could. I literally ran hundreds of tests. That break isn't bad, but for the lite6, I'd guess you can improve it

[kraegar]

Ok, so I couldn't figure out why my single color s3d prints with firmware retracts looked fine, but my dual color prints had stringing and super blobby retraction points… but only on the retraction points before filament change. Going to software based retraction didn't fix that - it got rid of the blobs, but made the stringing worse.

Turns out, s3d does some odd things on tool change, particularly if you use z-hop with retraction. Before a tool change, s3d will trigger a Z-Hop without retraction.

See here: http://imgur.com/p4WAb1k

The line after the highlighted one is the hop, but there's no retract before it. That'll cause a string.

You can stop that by going to Firmware retraction, but I was getting strings with it with the titans - dc42's response to me about how firmware retraction works - it spreads out the z-hop speed so it takes the same time as the retraction - was the key. I have to do very short z-hops (0.1mm) or I get very blobby retraction points.

I think this is due to the extra hysteresis in the bowden from the Y adapter - it's hard to clear the pressure on it fast enough, and during a z-hop on retraction the filament oozes, causing a blob. Going to a very short z-hop takes care of that.

Of course, short z-hops cause other problems, like the nozzle dragging over curled parts. Solve one thing, cause another issue.

So, for now, I'm back to firmware retracts along with some pressure advance like this:
M207 S3 R-0.2 F1500 T1200 Z0.1 ; y adapter
M572 D0 S0.1
M572 D1 S0.1

And it's yielding decent results.

[Yonkiman]

So the dual filament switching is working pretty reliably (though if kraegar has any new tweaks I'd love to see them, particularly since I upgraded my Lite6 to a full E3D V6 so our setups are that much more identical). But I'm running into a problem when I try to run SINGLE filament that I'd like to see if anyone else has run into.

I use T0 exclusively for single-filament printing. When I cancel a print from Duet Web Control, something (not sure if it's the firmware or the web interface) releases T0. So to start a new print, I have to Select Tool T0 from the Heater 1 pulldown. When that happens, the tpost0.g file is run, which is a big problem since that script tries to feed 150mm of filament at high speed, when the filament is already loaded into the hotend.

So is there a way to make sure either:

1. T0 is not released when a print is canceled, or
2. tpostX.g is not run if the previous tool was TX

I think 1 is probably the cleanest answer since 2 might be the right thing to do in other use cases.

[dc42]

If you create a cancel.g file then that will be run when you cancel a print, and it won't release the current tool unless you put T-1 in it.

[Yonkiman]

That was easy - thanks, David!

[DjDemonD]

Hi Kraegar thanks for this info finally got around to trying swapping filaments, I am using ABS but your tfreeX and tpostX macros for the titan work, and I can swap filaments. Now its a case of trying to work out how using slic3r to generate a wipe and prime tower in order to print some two coloured objects or just some rough and ready method to incorporate a prime of some sort into the tpostx file. Otherwise I am going to have to grapple with a new slicer which I'm not sure I have the time to do.

[kraegar]

I never found a good method for slic3r, and the last I knew prusa hadn't released his post processing script for adding a prime pillar yet. Cura does well for a lot of people, but I hit some weird oddities with it printing layers in the wrong order / swapping colors. Never figured it out, but it's probably the best free solution… though then you may hit bugs with absolute extrusion values, and switching to relative extrusion in the free/post scripts.

Currently I'm still using simplify 3d, as it's given me the best results. I haven't been dual color printing in a while, I was working on some printer changes, and now have a small backlog of prints to do. When I do get back to it, I have some things to try that might make switching even cleaner yet.

[deckingman]

It's pretty easy to use one of the Duet tool change macros to move to a corner of the bed, dump some filament, then move back again. If you have something like an old toothbrush strategically placed at a fixed position relative to the nozzle in Z, it'll wipe off any ooze as well. Works best with fast non-print moves.

[kraegar]

I played with that some with my delta, and it's one of the options I wanted to try - putting a "dump bucket" up high between two of the towers, so I could move up to it, switch filaments, purge a bit, and then move back. Has to be up high on a delta so the arms don't hit your print… or needs to follow along in Z as your print gets higher.

[deckingman]

Ah yes - hadn't considered what you would do with a Delta.

[DjDemonD]

I was thinking about just printing an ooze wall, purging some at the edge of the bed after a tool change, then coming back across the wall to wipe the nozzle. I tried bodging some gcode to print a tower but this won't work under all conditions such as layers which have only one colour etc..

Deltas do make it harder, I could devise a dump area but it would have to do some sort of relative move to lift away from the print, move towards the edge, descend to the dump area, then lift back up etc… to avoid collision with the print. It would be satisfying if I can work out the gcode to do it.

[kraegar]

I think for a delta an absolute position move to a dump bucket near the max Z for a perimeter would be the ideal. it's a LOT of travel, though, and I'm sure you'd want an ooze shield.

Originally I was just purging and using an ooze shield. You get tons of buildup in a hurry, though, and eventually I'd always end up with filament on my part (usually stuck to the nozzle and then it sticks to the part)