Closed-Loop Servo Control for 3D Printers Using ESP32 and Klipper

This tutorial demonstrates how to design a custom ESP32 S3 PCB using KiCad, covering schematic creation with power regulation (5V to 3.3V linear regulator), USB differential pair routing with ESD protection, decoupling capacitors, and power-on reset circuitry; the process involves converting the schematic to PCB layout, strategically placing components, routing traces with proper impedance matching for USB lines, creating ground planes, and performing design rule checks to ensure manufacturability.

For controlling mechanum wheels with an ESP32, you need dual H-bridge motor drivers like TB6612FNG for four motors. Power connections go to VM pins on drivers, with negative connected to ground pins. The 3.3V output powers the H-bridge VCC and standby pins (must be high to activate). Input pins connect to ESP32 GPIOs: for front module, PWMA to GP19, AI2 to GP23, AI1 to GP32, BI1 to GP33, BI2 to GP25, PWMB to GP26. Rear module uses different GPIOs. Motors connect to A01/A02 and B01/B02 outputs, noting the mirrored output arrangement.

This section covers installing the motor driver board and setting up the ESP32 microcontroller. Secure the motor driver board to the train base using M3x8 bolts, connect motor wires to the 'motor A' terminals, and swap wires if the train moves in the wrong direction. For the ESP32: upload the provided code, print the board retainer, remove protective film from the camera lens, create a slight fold in the camera's ribbon cable, secure the ESP32 so it nests against the print, hold in place with the retainer, and attach to the model using additional M3x8 bolts from underneath. This integration combines motor control electronics with wireless communication capabilities.

This video demonstrates how to connect a VNH5019 dual motor driver (12A per channel) to a Keyes ESP32 board using Arduino IDE, requiring connections for common ground, VDD (5V/3.3V) for logic power, and three control pins (Motor 1A, Motor 1B for direction, and PWM for speed), with the ESP32 running at 240MHz requiring LEDCAttachPin and LEDCSetup functions instead of Arduino's analogWrite for PWM control.

Effective PCB layout requires systematic component segregation based on functionality—grouping analog, digital, power supply, low-speed, and high-speed circuits separately with designated track areas. Filters must be used at subsystem boundaries for signals flowing between groups. Layer stack-up arrangement is critical: multi-layer boards should dedicate one complete layer as a ground plane with the adjacent layer as a power plane, ensuring the ground layer sits between high-frequency signal traces and the power plane. For two-layer boards, ground grids are recommended. Digital circuit routing demands special attention: high-speed signals and clocks require short traces adjacent to ground planes, avoiding edges and connectors. Return current follows the least reactance path, so ground traces should stay close to signal traces to minimize loop areas. Differential signals benefit from running close together for magnetic field cancellation. Clock signals require proper termination (source, parallel, or AC) to prevent reflections that cause radiation. Analog circuits need guarding with ground traces, isolation from digital sections, and low-pass filtering to reject high-frequency noise.

This comprehensive process covers understanding 3D printer vibration sources (movement systems, loose belts, spool holders, structural issues), identifying ghosting and ringing artifacts from high-acceleration printing, and implementing input shaping as an open-loop algorithmic control system. Klipper firmware offers five input shaper algorithms: zero-vibration (ZV/MZV) for single-frequency spikes and extra-insensitive (EI variants) for broader frequency ranges. The workflow includes configuring Klipper host devices with Python dependencies, flashing accelerometer firmware with proper wiring (RX-TX crossover, 3.3V power), creating configuration files with correct serial addresses, and mounting accelerometers using rigid or passive methods. Pre-test preparation involves measuring ambient noise and ensuring stable mounting. Resonance tests sweep frequencies from 5-133 Hz, generating vibration response graphs that determine optimal input shaper selection based on single or multiple peak patterns.

Input shaper is a Klipper firmware feature that compensates for resonance/ringing in 3D printers by modifying velocity input commands during direction changes, effectively reducing oscillations that occur when printers accelerate quickly; the technique involves printing a test pattern to measure the natural resonance frequency of the printer's axes and then configuring the input shaper parameters with those frequencies to eliminate the ripple artifacts while maintaining high acceleration settings.

3D printer vibrations create ringing artifacts at regular intervals corresponding to the machine's resonant frequency—the natural frequency where vibrations become strongest. Input shaping addresses this by modifying stepper motor inputs to cancel vibrations, transforming the traditional speed-quality trade-off into a solvable engineering problem. The manual calibration process involves printing a ringing tower while gradually increasing acceleration, measuring oscillation distances between ghosted patterns, calculating resonant frequency using provided formulas, and selecting from six available input shaper algorithms. Accelerometer integration enables fully automated calibration, where the firmware shakes the carriage at increasing frequencies while monitoring real-time vibration feedback to identify optimal settings. Practical demonstration shows that input shaping enables dramatic speed increases without quality compromise: benchy tests showed pressure advance significantly reduced layer transition imperfections, complex model assembly at 140mm/s produced perfectly fitting parts, and decorative prints maintained exceptional surface quality. This represents a fundamental advancement in 3D printing technology.

Input shaping is a Klipper algorithm that reduces echo (ringing) by counteracting print head vibrations. Automatic calibration using accelerometers fails because it only considers relative vibration amplitudes, not absolute values, leading to incorrect shaper recommendations. The Klipper documentation's manual calibration method is fundamentally flawed because it measures combined vibrations from both axes during cornering, not pure single-axis frequencies. A reliable manual method involves printing a test model with maximum settings, using Tuning Tower commands to sweep frequencies, and finding the height with minimum echo on all sides. Iterative refinement with smaller steps achieves 1-2 Hz accuracy. If MZV fails, try ZV or HAM shapers. Mechanical issues like loose components or insufficient frame stiffness may require reinforcement. PR Advance calibration is unreliable because gap size depends on material and hardware factors. Optimal print speeds should be determined empirically through testing on corner-heavy models.

Input shaper is an open-loop control technique in Klipper firmware that reduces ringing (ghosting) in 3D prints by analyzing and canceling resonant vibrations through accelerometer measurements; the tuning process involves attaching an ADXL345 accelerometer to the printer's moving parts, running resonance tests to identify peak frequencies, and configuring input shaper parameters (shaper frequency and type) in the Klipper configuration file to enable faster printing with improved print quality.

Closed-loop control systems enable automatic process regulation through continuous feedback comparison between reference inputs and actual outputs. Modern mechatronic systems utilize PLCs as central controllers, power converters scaling signals to actuators (electric, pneumatic, hydraulic), and advanced sensors including vision systems. The industrial automation pyramid stratifies functions: field level (hardware devices), control level (PLCs and field devices), supervision level (HMIs and monitoring), and enterprise level (ERP systems). Human-Machine Interfaces enable operator interaction while predictive maintenance applies data analytics for equipment health forecasting.

Closed Loop Control System (폐쇄회로 방식) uses a scale (scales) attached to the machine table to detect position and provide feedback. This is the key characteristic that distinguishes it from other control systems. The scale is crucial for high-precision machining operations.

Industrial closed loop systems include hopper level control using level probes and Variable Frequency Drives to maintain consistent material flow rates. Robotic arms require dual closed loop control for position (joint angle) and velocity, with the controller generating errors for both and processing them to drive servo motors. These applications demonstrate how closed loop principles apply across different industries and equipment types.

Additive manufacturing systems face unique motion control challenges: standard CNC controllers cause excess material deposition at edges, may go out of CAD areas, and experience sudden jumps. Solutions include hardware modifications (safety measures, electric brakes, dynamic braking registers) and software limit configurations. Achieving dimensional accuracy requires closed-loop control systems with height sensors for real-time monitoring and feedback control loops. Industrial applications include repairing expensive turbine blades (70-75 lakhs) where small cracks would normally require complete part rejection. Powder characteristics significantly impact success: particle size (30-40 micron) and sphericity affect capture efficiency (80-90%). Smaller particles reduce energy requirements but may decrease capture efficiency as more powder is carried away from the substrate.

Closed loop control systems are used in more sophisticated CNC machines. In this system, sensors continuously measure the actual position of the machine axes and feed this information back to the control system. The control system compares the actual position with the commanded position and makes adjustments if there is any error. This provides much higher precision, typically achieving tolerances in thousandths of a millimeter. The system can compensate for any disturbances or errors that might occur during operation.