Underbed load cells, connected in parallel to a single ADC, will have areas where probing may fail. The positive magnitude response from one set of load cells will be perfectly canceled by negative magnitude response from the remaining load cells, resulting in no net load change detected by the probe force. This behavior was discussed in main load cell thread. Problem areas will be unpredictable with environmental dependencies on the electrical and mechanical response of the system.
Goal of this thread was:
- to capture real world performance of the load cell probe for underbed sensors
- guidelines, limitations, & best-practices for implementing underbed sensors
- document commercial load cells suitable for underbed mounting.
My experience with the current @garethky community testing branch used with 3 underbed sensors on a delta printer have been superb. I have a FLSUN T1 Pro printer, this is a newish delta printer and comes from factory with 3 underbed load cells for probing. The stock system has a separate board using an HX717 and STM32F103 to toggle a conventional probe signal to the main mcu. 3 full bridge load cells are connected in parallel to the HX717 and located on opposite sides of the bed from the towers. The STM32F103 did nothing more than read/process the HX717 data and toggle the probe pin wired to the main motherboard. I’ve modified the system to run from a raspberry pi 4 with the STM32F103 as a proper klipper mcu.
Probe_accuracy with 5 samples at each point was performed over a densely spaced grid covering 95% of build radius. Results of the probe_accuracy range were plotted to identify any problem areas. All of the sets of sample ranges were better than 15 microns over the entire bed, 97% had range of 5 microns or less, over half the bed was 1 micron or better.
The largest ranges are grouped across from the B tower where one of the load cells is mounted. The AC also comes in on this side and the power lines run under this sensor. The power lines and the sensor leads are all wrapped with shielding, but performance still seems to be impacted.
The probed heights are were also plotted. No abnormal values found, with pretty smooth profile aside from the region above the A tower. The algorithm for detecting the nozzle contact with bed seems to be robust, even for the areas outside of the ABC triangle.
