I need to read 4 load cells with relatively low resolution (12 bit is OK) and frequency (10 SPS is sufficient). The prototype used 2 ubiquitous HX711 breakouts with each powering two load cells in parallel (I've asked about this arrangement here and can confirm that it works).

In the final product I would like to use one 4-channel ADC and I've narrowed the choices down to two chips, AD7794 and MCP3428. I am also thinking about using current excitation sources to reduce the effect of 1.5m wires.

Now, AD7794 has two differential reference inputs that can be connected to excitation voltage to negate the noise in supply. But this means at least two load cells will be on a same current source. They do work just fine with constant voltage excitation. What I am worried about is rather big variance in load cell input impedance (8% per datasheet) and how they may affect each other on constant current excitation.

I am looking at the following choices:

  1. Use 2 current sources, each supplying a pair of load cells and providing voltage for one reference input;
  2. Use one current source for all 4 cells and a single reference input;
  3. Use relatively stable board supply voltage for excitation and a single reference input;
  4. Use much simpler MCP3428 chip with its internal reference.

Any advice?


1 Answer 1


Ratiometric conversion is the 'usual' technique for load cells. The reference and the excitation to the cells is the supply. The source of error would be voltage drop on the excitation wiring, but since load cells are generally low, that should not be a problem unless you have long wiring.

  • \$\begingroup\$ And your advice would be? \$\endgroup\$
    – Maple
    Commented Aug 17, 2021 at 14:06
  • 1
    \$\begingroup\$ Use ratiometric conversion. Do you have a specific reason not to? \$\endgroup\$
    – Kartman
    Commented Aug 17, 2021 at 21:49
  • \$\begingroup\$ The question was not as much about conversion as about current excitation of two (or 4) load cells in parallel. Due to manufacturing discrepancies one cell can be e.g. 320 Ohm, another 380 Ohm. If I can't run them in parallel I cannot use external reference inputs, so might as well switch to cheaper and smaller ADC \$\endgroup\$
    – Maple
    Commented Aug 18, 2021 at 0:31
  • 2
    \$\begingroup\$ Don’t use current excitation. If the excitation voltage is the same as the reference voltage then the absolute resistance on the load cell is immaterial. The assumption is the excitation supply has enough current for all the load cells and there is not significant voltage drop across the wiring. The other assumption is that your load cells are wheatstone bridges. \$\endgroup\$
    – Kartman
    Commented Aug 18, 2021 at 2:12

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.