YouTuber calibrates 3D printer with computer vision and precise time measurement

A maker project involving a Voron 0 3D printer shows how far the calibration of an FFF printer can be pushed technically. Dennis from “Made by Dennis” systematically measured and readjusted the movement mechanics of his printer in a video. The focus was on the travel distances, the perpendicularity of the axes, and the exact position of the print bed. His goal is to achieve an accuracy of better than 100 micrometers over a distance of 100 millimeters. To get there, he combines computer vision, precise time measurement, and linear algebra.

Dennis calculates a position correction matrix for corrections in the case of axes that are not exactly vertical or distorted. This is based on a camera on the toolhead and a ChArUco board on the print bed. Image recognition can detect the corners of the board’s tiles and determine their positions. If the focal length of the camera is known, the camera position can be derived using simple trigonometric calculations. By taking images at many points across the working area, a matrix is created that maps the position of the print head reported by the printer to its actual position.

Leveling the print bed is also a precision task in the setup. Dennis uses a PZ probe to detect contact points between the hot end and the bed at several locations. He also uses a wiper to remove plastic residue from the nozzle so that it does not distort the contact. Nevertheless, he was bothered by the fact that the bed still gave way slightly after contact. According to his estimate, this could result in errors of up to five micrometers.

As a countermeasure, he has the microcontroller in the hot end record the contact time and send it together with the contact signal to a Raspberry Pi controller. The Raspberry Pi logs times and positions so that the actual contact position can be determined retrospectively via the time stamp. This requires clock synchronization between the microcontrollers in the hot end and in the printer to within about one microsecond. The Raspberry Pi achieves this synchronization via USB start-of-frame packets.