Project advice needed: long wires, multiplexing, load cells, HX-711 ADCs

Hi everyone,

I’m planning a project in which 36 load cells are laid out in a 6x6 grid. Every four cells are carrying a 60x60cm wooden boards, the whole setup spanning 3x3m of pressure-sensitive floorboards.

The load cells are connected to a HX-711 ADC each, then will run to some multiplexers and from there to an MCU that’s positioned at one of the outer edges of the whole grid for better accessibility. I will be calculating things like the center of mass from each of the pressure-sensitive boards.

I’m attaching a sketch in order to help visualize it.

Now, I have a few questions about feasibility and would really appreciate some input from people with more experience doing these kinds of projects. I’ve worked with sensor grids and multiplexers before, but the distances between the sensors and from sensor to MCU were much shorter in those.

My questions are:

  1. Power: According to my calculations, the longest cables running from the HX-711 to the multiplexers/MCU should be around 4.5m … what kind of AWG would I need to use at this distance and would I need to be concerned about voltage drops at that length? Do I generally be concerned about voltage drops with this many HX-711s?
  2. The project is supposed to be displayed outside (albeit under a roof) at January temperatures (Europe), so somewhere around the 0 degree mark, possibly down to -10 (if it’s a particularly cold winter). How will that affect my whole setup? What should I be wary about?
  3. Multiplexing: I’m not sure how to think about this problem and would appreciate some input; can I just multiply the clock for all HX-711 and use an MCU with lots of pins (like an Arduino Mega) or do I need a multiplexer? If I can multiply the clock, do I need to boost that signal somehow? If I use a muxer, which one is best?
  4. Is there anything else I should be thinking about that I am missing here?

Any help and pointers in the right directions are much appreciated. Thanks!

Sketch:

For various reasons including external interference, communication over long wires presents a very serious problem, especially with the I2C bus.

Experiment with a simplified but full scale setup mimicking the larger installation before settling on a workable design.

Best if the test setup adequately represents worst-case environmental conditions of the final installation.

4 wire load cells?

yes!

thanks for getting back. could you elaborate on the other reasons apart from external interference?

how would I simplify this setup? I was thinking of adding sensor after sensor, starting with one at 5m cable length, then adding another etc. …

thanks again!

Hx is not bus device, you need separate data wire for every sensor.
Use some twisted/shielded wiring, ethernet or alarm cable for example. Voltage drop is likely not issue, but interferences might be.

1 Like

You could connect 2 loadcells to one HX711 so you would only need 18 HX711s. You can also clock them at a slower data rate like 10kHz so working with long wires may not be a problem.

The amount of current required by each HX711 with one laod cell is roughly 5V/Loadcell resistance. So for a 350 ohm cell the current is 5/350 ~ 14.3mA. Depending on how you wire the grid, AWG 20 for power and ground could be OK

1 Like

Signal shape degradation by capacitance between adjacent wires and conductive structures, combined with improper line termination, compounds the interference problem.

An oscilloscope will reveal the issues.

Interference would be my concern, not volt drop.
I would try twin shielded audio cable.
The type you see on audio RCA leads.
The two grounds can be used for the excitaion voltage E+, E-
And the two cores can be used for A+, A-

Are you sure about that multiplexer. I would think a Mega would be an easier solution.
Leo..

1 Like

you might want to look at this library

1 Like

thanks for all the answers everyone!

so my plan right now is:

  • 4-core shielded cable per sensor, AWG20, shield grounded at MCU/hub.

  • Bulk cap at hub (47–100 µF, 25–50 V electrolytic), across +5V–GND.

  • 0.1 µF ceramic + 10 µF electrolytic in parallel, close to the module pins, for decoupling at each HX711

  • One shared CLK, buffered. SN74HCT244 placed at the hub, 33 Ω (or higher) series resistor at each buffer output (driver_pin → 33Ω → cable).

  • Optionally lower clock rate (10 kHz)

and I will probably use a Teensy 4.1 as my MCU

That's a 3.3volt-logic board.
Cheap HX711 boards are 5volt-logic only, so you need to level-shift the data pins.
Or use the more expensive HX711 from Sparkfun, which can be cofigurated for 3.3V logic.
No idea why you chose a high-end processor.
Things with a HX711 happen slowly. The HX711 defaults to 10Hz.
They can be modded to 80Hz at a performance cost, but that is still slow.
Leo..

This is not the right choice. In terms of voltage, the capacitor must be rated for the operating voltage. In your case, it should be rated for 6.3 volts, or at most 10 volts. At such voltages, the restoration of the oxide layer destroyed by the electrolyte is guaranteed if the capacitor is electrolytic.