Recently, my team and I made the decision to swap microcontrollers in one of our designs that interfaces with an ADS1234IPW to read the signal from a 1KG loadcell. After designing around the new microcontroller, we've noticed there has been an increase in noise on the analog signal that's about three times greater than the original design.
There isn't really a great alternative to the microcontroller used in the new design because of part availability so I'm looking for ways I can mitigate noise on this design through either better filtering on the analog input to the ADC1234IPW or better layout. It's really important that the analog signal is as noise-free as possible.
I can send the board and schematic files in a private email but here's a basic explanation of the difference:
Before: Used a PIC18F2580 microcontroller. The loadcell wires were soldered directly to the board and went through a low pass filter before entering the analog input of the ADS1234.
After: Used an ESP32 microcontroller. The loadcell is attached to the board through a 4 pin connector. An ESD protection diode is placed between the low pass filter and the loadcell connector. Because of the connector for the loadcell, traces into the ADS1234 are just slightly longer. The rest of the board is essentially the same except for the addition of a 3.3V regulator for the ESP32. The low pass filter is the same.
I could see a couple reasons why there might be an increase in noise. The ESP32 operates at a higher clock speed than the PIC18F2580. From that, I would expect more noise on the digital side of the circuit. However, the design uses a split ground plane between the analog side and digital side of the board.
Another reason could be the connector for the loadcell. However, data has suggested that the connector does not contribute to noise. I am not wholly convinced one way or another.
I wanted to see what your recommendations are for mitigating noise. My approach would be to tweak the RC constant of the low pass filter to have a more aggressive low pass filter. I'd also like to move to a 4 layer board to separate the analog and digital signals even further. I'm also debating removing the connector and going back to soldering the loadcell wires directly to the board which would allow me to shorten the traces, however, I can't imagine that the connector or the slight increase in trace length is causing the noise to triple (I could be wrong though!). The other thought is moving to a more complex filtering network, like a second order low pass filter but ideally I'd like to maintain unity gain and I'm not sure that is possible with a second order filter with the presence of the op-amp in that configuration.