Loading files…
A four-point weighing platform built from scratch: mechanical frame, load cell integration, a custom-designed summing junction board (KiCad), and transmitter calibration.
The system measures weight using four load cells mounted at the corners of an aluminium plate. The cells are wired in parallel through a custom junction board, which sums their outputs into a single differential signal. This signal is read by a Baykon TX11 analog weight transmitter, which provides the digital readout and process output.
Platform capacity: 4 kg (4 × 1 kg)
Calibration references: 500 g and 1 kg
[Load Cell 1] ──┐
[Load Cell 2] ──┤
[Load Cell 3] ──┼──► [Junction Board] ──► [Baykon TX11] ──► Display / Process Output
[Load Cell 4] ──┘ (parallel sum, (excitation +
corner trim) signal cond.)
| Component | Detail |
|---|---|
| Load cells | 4 × single point load cells, 1 kg each, corner-mounted |
| Transmitter | Baykon TX11, analog input |
| Junction board | Custom 2-layer PCB, designed in KiCad |
| Platform | Aluminium plate, 200 × 100 mm |
| Supply | 24 VDC |
The junction board handles three jobs:
Schematic and PCB source files are in /kicad.
Calibration was performed on the TX11 using 500 g and 1 kg reference weights:
Corner balancing on a four-cell platform requires trimming resistances in the 10–20 Ω range — the correction needed is a small fraction of the load cell's bridge resistance.
Every trimmer potentiometer available on the local market started at 470 Ω minimum. Fitting one would have made the correction range roughly 25× coarser than required, making fine balancing impossible and risking a large unintended signal offset.
Resolution: the trimmer footprints were kept on the PCB but bridged with 0 Ω links in the first revision, so the board could be tested immediately without waiting on parts. The footprints remain available for a future revision using either low-value multiturn trimmers or fixed precision resistors selected per corner after measurement.
Takeaway: component availability is a design constraint, not an afterthought. Reserving the footprint kept the option open at zero cost instead of forcing a board respin.
Bringing the platform from assembled hardware to a trustworthy reading surfaced several issues that had to be resolved in sequence:
Corner deviation. The same reference weight produced different readings depending on which corner it was placed on. This is what drove the corner trimming requirement described above.
Platform rigidity. The plate flexed under load because the mounting surface was not properly fixed. The assembly was taken apart and rebuilt with the mounting positions measured and aligned, which removed the flex and made corner readings repeatable.
Zero drift. The empty-platform reading was not stable. The cause was mechanical stress introduced during assembly — the plate was being loaded by the mounting itself rather than sitting free. Releasing and re-seating the corners removed the residual stress.
Unstable readings. The least significant digit fluctuated continuously under no load. Resolved by adjusting the TX11 filtering and averaging settings.
Transmitter configuration — fault isolation. At one point the excitation voltage parameter would not respond to changes on the TX11, while the mechanical assembly had already been verified. Rather than continuing to troubleshoot the full four-cell system, the platform was stripped down and the transmitter was tested against a single load cell. Isolating the system to its simplest working configuration made the fault visible and allowed the setup to be rebuilt from a known-good baseline.
Working through these in order made one thing clear: on a load cell system the mechanical installation determines the achievable accuracy long before the transmitter settings do.
├── kicad/ KiCad schematic and PCB source files
├── platform.jpeg Assembled platform
├── schematic.png Junction board schematic
└── README.md
Habip Can Durmuş — Mechatronics Engineer LinkedIn · habipcandurmus07@gmail.com
No comments yet. Be the first to ask about this board.