Lesson 5 of 5 · 22 min
Calibration and troubleshooting
A freshly built or unboxed printer is not yet a measuring instrument. Its extruder may push 5 percent too little plastic, its bed may be 0.3 mm off level, and the spool may be damp. Each of these shows up as a specific kind of bad print, and each has a known fix. This lesson gives you the calibration order, the maths behind the two numbers you can compute, a symptom table for the common faults, and a short test plan that finishes with two real robot parts.
Calibrate in the right order
Problems stack, so fix them in the order that each depends on the last:
- Mechanical check. Tight frame screws, belts firm but not guitar-string tight, wheels turning smoothly, nothing wobbling.
- Bed levelling and first layer.
- Temperature (a temperature tower per filament).
- E-steps (the extruder's own accuracy).
- Flow (the filament's real behaviour).
- Retraction and speed.
- Dimensional accuracy with a calibration cube.
If you tune flow before the first layer is right, you will chase a problem that is really the bed.
First layer and bed levelling
The first layer decides whether anything else works. The nozzle should sit close enough to the bed that the plastic is gently squashed into a flat ribbon that bonds with its neighbours, and not so close that it scrapes. A gap of about 0.1 mm for a 0.2 mm first layer is typical.
For a manual bed, heat both bed and nozzle first, because the metal expands. Slide a sheet of ordinary paper (about 0.1 mm thick) under the nozzle at each corner and adjust until it drags slightly, then check the centre. A bed with automatic mesh levelling measures this for you, but you still need to set the Z offset, the final distance, by watching a test print.
Read the first layer as a diagnostic. Lines that are round and separated mean the nozzle is too high. Lines that are translucent, scraped or ridged at the edges mean it is too low. A good first layer is flat, smooth and slightly shiny with no gaps between lines.
E-steps: is the extruder honest?
The firmware assumes a certain number of motor steps pushes 1 mm of filament. If the real value differs because of the gear diameter or tolerances, the extruder delivers more or less plastic than commanded. The method:
- Heat the nozzle to printing temperature.
- Mark the filament 120 mm above the extruder entrance.
- Command a slow 100 mm extrusion, for example
G1 E100 F100, with the hotend free so it does not clog. - Measure from the entrance to the mark. It should be 20 mm. The difference is the error.
new e-steps = old e-steps x 100 / actual extruded length
Worked example. The old value is 93 steps/mm. After the test the mark sits 23 mm from the entrance, so the extruder pushed 120 - 23 = 97 mm instead of 100.
new = 93 x 100 / 97 = 95.9
Enter it with M92 E95.9, save with M500, and repeat the test until you get 100 mm to within 1 mm. Note the direction: the extruder under-delivered, so the steps per mm must go up, which makes it push farther per command.
Flow: the filament's behaviour
E-steps fixes the extruder. Flow (extrusion multiplier) compensates for the filament itself: its real diameter, which can vary between 1.70 and 1.80 mm, and its tendency to swell. Print a single-wall cube, 20 mm wide with a 0.4 mm line, and measure the wall with calipers at several points. If the wall measures 0.43 mm and the target is 0.40 mm:
new flow = old flow x target / measured = 100 percent x 0.40 / 0.43 = 93 percent
Recalibrate flow per spool and per colour, because pigments change melt behaviour.
Symptom, cause, fix
| Symptom | Likely cause | Fix |
|---|---|---|
| Fine strings between parts (stringing) | Wet filament, nozzle too hot, retraction too short | Dry filament at 45 to 55 C for 4 to 6 hours, lower temperature by 5 to 10 C, retraction up in 0.2 mm steps |
| Corners lift off the bed (warping) | Part cools unevenly, bed too cool, draft | Raise bed temperature, add a brim, enclose the printer, apply glue stick |
| Layers shifted sideways partway up | Loose belt or pulley grub screw, speed or acceleration too high, motor current low, nozzle hit a curled edge | Tighten belt and grub screws, reduce acceleration, check driver current |
| First layer will not stick | Bed dirty, nozzle too far away, bed too cold | Wash the bed with soap or isopropanol, lower Z offset, raise bed temperature |
| Gaps, thin walls, rough top (under-extrusion) | Partial clog, flow too low, nozzle too cold, worn extruder gear | Cold pull to clear the nozzle, raise temperature, recalibrate e-steps and flow |
| Blobs and zits on the surface | Retraction restarts, seam on show, too much flow | Lower flow, move the seam to a corner, calibrate retraction |
| Bottom edge flares out (elephant foot) | Bed too hot or nozzle too close | Lower bed temperature a little, raise Z offset slightly |
| Layers split apart | Printing too cold or cooling too aggressive | Raise nozzle temperature by 5 to 10 C, reduce fan |
| Sizes consistently off by the same percentage | Wrong steps per mm | Recompute steps per mm as in Lesson 4 |
| Extruder clicks | Flow demand too high, clogged nozzle, temperature too low | Slow down, raise temperature, clear the nozzle |
Change one thing at a time, and write down what you changed. If you alter three settings and the print improves, you will not know which one mattered.
A test print plan for robot parts
Do these in order, and stop at the first one that fails, fix it, and only then move on.
- First layer sheet. One flat 0.2 mm square, about 100 x 100 mm. It should be uniform with no gaps. This takes about 15 minutes.
- 20 mm calibration cube. Measure X, Y and Z with calipers. Anything within 0.1 to 0.2 mm is fine. A consistent error on one axis points to its steps per mm, and an error in Z alone points to the lead screw or its coupler.
- Tolerance coupon. From Lesson 3, with holes from 3.0 to 3.6 mm. Note which size an M3 screw slides into.
- Servo horn adaptor. A disc 25 mm in diameter and 5 mm thick, with a recess for the servo horn and two M2 screw holes. Print three copies with the horn recess at 0.2, 0.3 and 0.4 mm of clearance per side, in PLA at 0.2 mm layers. The right one pushes on by hand, with no wobble, and does not fall off when you turn it. Note the winning value, because it is your personal offset for press fits.
- Servo bracket. An L bracket 40 x 40 mm with 3 mm walls, a servo pocket sized from your measurement, four M3 holes and one boss for a heat-set insert. Print one lying flat and a second standing up. Clamp the base and hang a bag of weights from the arm, adding load slowly. The standing one will split along the layers at a much lower load. That result is the strongest evidence in the course for why orientation matters. When it passes, reprint the winner in PETG for the robot.
Wear eye protection for the load test, since a snapped part can fly, and keep a hand under the weights.
Check yourself
Your extruder was told to push 100 mm of filament but only 96 mm moved. The old setting is 93 steps/mm. What is the new e-steps value?
Check yourself
A print shows thin strings between two towers, and the spool has been open on a shelf for months. What should you try first?