1
\$\begingroup\$

I'm interfacing a load cell using a HX711 ADC to read its results into a microcontroller.

Many of the code samples take numerous readings (about 10) and average the result, but I was wondering if the readings could be improved by simply adding a capacitor across the load cell's sense leads to form a low-pass filter? Then I could take a single sample and use less energy.

Is this a worthwhile endeavour? Would a ceramic capacitor be too small, or would a bipolar electrolytic have too much leakage and affect the results?

\$\endgroup\$
  • \$\begingroup\$ How far from the load cell is the ADC? \$\endgroup\$ – EM Fields Feb 19 '16 at 21:14
  • \$\begingroup\$ With instrumentation it is better to figure out where your noise source is in the system. Is it coming from the ADC? Is it from the sensor? What is the impedance of your sensor an what resistance range does it have? I'd imagine that when you say load cell you mean Wheatstone bridge, how is the power connected? You could just "try it" but its a bad design practice if you don't know why it works. \$\endgroup\$ – laptop2d Feb 19 '16 at 23:13
  • \$\begingroup\$ @EMFields the load cells are around 20cm away from the sensor with unshielded wires. It's basically a digital bathroom scale. \$\endgroup\$ – Michael Feb 22 '16 at 1:07
  • \$\begingroup\$ Michael, The first thing I'd do would be to replace the unshielded wires between the load cell and the ADC with a twisted shielded pair with the shield grounded at either the sensor, the load cell, or both, determined empirically for minimum noise into the ADC. Can you post a picture of the load cell wiring? \$\endgroup\$ – EM Fields Feb 22 '16 at 12:55
1
\$\begingroup\$

Adding a capacitor across the sense lines can help reduce noise and smooth the signal. Note that in the HX711 spec sheet (page 6) there is even a reference schematic/PCB that shows an RC network on the sense lines, it includes a 0.1uf directly across the sense pin inputs.

If used the capacitor should be a non-polarized low leakage type (for e.g. Poly type), as any leakage across this capacitor could cause some amount of error. A high quality ceramic type may work well if you don't really need the highest performance. A higher value capacitor could help but at a certain point your response time to load changes will suffer. https://cdn.sparkfun.com/datasheets/Sensors/ForceFlex/hx711_english.pdf

It would still be a good idea to do some averaging of multiple conversions as there can also be noise coming from the rest of the system.

\$\endgroup\$
  • \$\begingroup\$ Thanks @Nedd, I didn't notice their reference circuit and I will give this a try. I just want to avoid oversampling if possible because this project runs off a coin cell so the more time in sleep the better. \$\endgroup\$ – Michael Feb 22 '16 at 1:14
2
\$\begingroup\$

Is this a worthwhile endeavour?

You may be missing the whole point here....

There is a subject matter called dithering and it's useful when sampling a noisy signal. It gives you more ADC resolution. Each 4 samples taken and added gives you a resolution increase of 1 bit.

Decibel increase in resolution = 10log(oversampling rate) and if the oversampling rate is 4 then the resolution increase is 6 dB.

On the other hand smoothing a really noisy signal is a good idea but, if you are trying to get resolution increases by over sampling then don't over-smooth the signal or you won't be able to make dithering work.

Consider a 1 bit ADC (a comparator) then consider the comparators output when the signal is noisy: -

enter image description here

Here's another example: -

enter image description here

\$\endgroup\$
  • \$\begingroup\$ Thanks Andy, currently I'd be very happy if my readings were anywhere near the HX711's 24bit accuracy as they jump a few hundred LSB between readings. \$\endgroup\$ – Michael Feb 22 '16 at 1:12

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.