Temp Tower & Retraction Tower G-code Generator

What the generator prints

Two square pillars, two walls thick, with a brim on the first layer for adhesion. The nozzle prints one pillar's walls, travels across the gap, prints the other, and comes back — once per layer, all the way up. The tower is divided into equal-height sections and the test value changes at the start of each one:

• Temperature tower: an M104 at each section boundary sets the new nozzle temperature. The print continues while it changes, which takes a few layers for a 5 °C step — read each section from its top half.
• Retraction distance / speed: every travel across the gap retracts with that section's value. Everything else is held constant.

The crossing is the whole point. Stringing, oozing and blobbing all happen on travels, and a 40 mm unsupported gap between two small pillars is the harshest travel a print ever makes. Whichever section has clean pillars and a clean gap is your answer.

Reading the result

1. Note the section values from the table above (they're also in the G-code header). Section 1 is at the bottom.
2. For a temperature tower, look at four things per section: stringing across the gap (worse when hot), layer adhesion (try to snap a pillar — worse when cold), the gloss and smoothness of the walls, and any tiny bulges at the corners (over-extrusion when hot). Pick the coolest section with good layer bonding; for strength, go 5 °C hotter than pure looks would suggest.
3. For retraction distance, find the first section where the strings disappear; that's your distance. Going further doesn't help and risks pulling molten plastic up into the cold end, which causes jams on Bowden printers.
4. For retraction speed, faster usually helps up to a point where the extruder skips or the filament grinds; use the fastest section that still looks clean.

Suggested sequences

Do temperature first, then retraction distance at the chosen temperature, then retraction speed at the chosen distance. Flow should already be calibrated with the flow rate calculator; if it isn't, over-extrusion will masquerade as "too hot".

Before you print

• Preview the file. Drag it into PrusaSlicer, OrcaSlicer or Bambu Studio (File → Import → G-code), or any online G-code viewer, and check the pillars sit where you expect on the bed.
• Klipper users: if you rely on a PRINT_START macro for homing, mesh loading and heating, paste it in the custom start box with {hotend} and {bed} placeholders. The built-in start does a plain G28 and heats, which is enough for most setups.
• Bambu users: printing raw G-code works from the SD card or by importing the file into Bambu Studio and sending it as-is. The built-in start code skips Bambu's purge and calibration routine, which is fine for a small test print.
• Deltas: tick "origin at bed center" so the pillars land around 0,0.
• Watch the first layer. The generator can't know your Z offset. If the brim isn't sticking, stop and adjust it as you would for any print.

What it deliberately doesn't do

No infill, no overhang features, no bridges — those belong in their own tests, and combining them makes each harder to read. There is no pressure advance or linear advance tuning; use the routines built into Klipper and Marlin for that. The print settings are conservative on purpose: a calibration print should isolate one variable, and a 40 mm/s wall speed keeps the hotend well inside its volumetric limit so what you see is the temperature or retraction, not a flow problem.