I am currently having a lot of trouble with noise being picked up by load cells.

The strain gauges on the load cell are configured in a full-bridge pattern. In a noiseless environment (ie far away from any micro/power supply lines) the system performs admirably.

However, the four wires that come from the load cell I believe are picking up noise. If these wires lie close (within 75mm) of a USB cable they pick up noise, or there is a WIFI/3G module nearby the wires also pick-up the EMI noise.

In an attempt to mitigate these problems we have twisted the wires coming from the load cell to reduce the inductance loop and therefore reduce the noise problem. This has no helped.

Therefore we are now looking into types of shielding, and thus brings me to the question.

How effective would copper (or aluminium) tape be if I were to wrap it wrap the end of the load cell, all the way along the wires and over the PCB that has the ADC chip mounted on it?

Before you ask about is it coming from other noise sources, I have eliminated the power supply as the noise as I am using a heavily filtered vref design specifically for the low noise applications.

The signal level that exists between the S+ and S- lines is at a maximum 3mV therefore any noise very quickly ruins the signal.

When I twisted the wires coming from the load cell I twisted all four at once and not as twisted pairs, would this have any impact on the result?

  • \$\begingroup\$ How are you amplifying? \$\endgroup\$ Aug 12 '14 at 1:43
  • \$\begingroup\$ Are you filtering the signal at all? \$\endgroup\$
    – Phil Frost
    Aug 12 '14 at 2:05
  • \$\begingroup\$ See if you can get a sense of the noise in the frequency domain - that may help identify the origin. Also don't rule out the ADC. \$\endgroup\$ Aug 12 '14 at 3:59
  • \$\begingroup\$ Scott I am not using an amplifier at all as we have a 24bit differential ADC. Phil Yes we are. Chris Our update rate is limited to 80Hz therefore aliasing I believe is causing the signal to get skewed. \$\endgroup\$
    – Lhh92
    Aug 13 '14 at 6:27

The typical signal from a typical full-bridge load cell can be in the microvolt range and sometimes quite possibly as high as several hundred millivolts. Noise can affect all levels but is obviously going to be more problematic when the signal is small.

Screened cable with quad twisted cores is part of the solution to noise pick-up but, it sounds like the interference you are picking up is RF and might be also dealt with by appropriate low pass filters before the amplifier used to interface to the load cell.

If you are using an instrumentation amp, these are commonly affected by RF interference and so there are several well-documented options open to you.

  • \$\begingroup\$ Would a low pass filter not eliminate the entire signal and make it not measureable? \$\endgroup\$
    – Lhh92
    Aug 13 '14 at 10:09
  • \$\begingroup\$ You design the low pass filter to have a cut-off frequency above the maximum frequency of signal you wish to measure. \$\endgroup\$
    – Andy aka
    Aug 13 '14 at 13:34
  • \$\begingroup\$ A true analog filter will still have decimation due to the resistance loss. Additionally the temperature dependency of that resistor used in the filter would change the output response of the load cells? \$\endgroup\$
    – Lhh92
    Aug 14 '14 at 0:07
  • 2
    \$\begingroup\$ @LanceHenderson Don't be silly now. A resistor cannot cause decimation and as for the filter response changing with temperature, this will be small and irrelevant if you pick a high enough cut-off. \$\endgroup\$
    – Andy aka
    Aug 14 '14 at 8:34

You should be using a differential input ADC and a cable with an overall shield wound around individually shielded twisted pairs.

One pair should be used for bridge excitation and the other for the bridge signal output with the overall shield and the individual drain wires grounded to a single point, like this:

enter image description here

Something like a Belden 3084 should work, if the resistance of the 22 AWG pair didn't interfere too much with the bridge excitation.

On the other hand, since your "data" is just slowly varying DC, plain old PVC insulated, individually foil shielded twisted pairs would probably work fine. Try it. :-)

  • \$\begingroup\$ I´m currently working with load-cells in combination with raspberry pi and just read your wire diagram and overthinking if this is also required for a setup like in this tutorial (circuitdigest.com/microcontroller-projects/…)? Additionaly, i don´t understand why do you connect all ground cables together? \$\endgroup\$
    – goblingift
    Jan 8 '20 at 10:44
  • \$\begingroup\$ @goblingift: In that tutorial the weighing system is used to detect whether somebody stands in front a a gate. The weighing accuracy required to be able to detect that is minimal. Therefore noise can be neglected. - It is common practice to connect all earth cables to one common earthing point to prevent ground loops (currents flowing through earthing wires because of minute differences in voltage levels - because of the low wire resistances these currents can be quite high and can disturb the proper working of the circuit involved). \$\endgroup\$
    – PimV
    Apr 7 at 12:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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