We're using the FDC1004 to do some sneaky liquid presence detection.
I very closely followed all of the available app notes and datasheet info when I did the layout, and the solution functionally works.
We are streaming out raw counts and using our own algorithm in an MCU to detect liquid droplets.
That said, we are seeing small error differences from one electrode to the other: one has an error rate of nearly zero, the other around 3% in terms of successful droplet detection.
To be clear, I'm not talking about error in what we expect from the direct raw counts, but error in our identical drop detection algorithm.
The electodes are identical in design, but the traces from the IC to the electrodes are VERY slightly different - we are talking roughly 4mm versus 8mm.
Our schematic of just the IC portion:
Below, you can see some important layout features:
Red is component side and traces, yellow is layer 2 signal ground, pink is layer 3 shield.
You can see we are actively using channels 1 and 2 which drop through the board via some headers (this was done for prototyping purposes). We have additional shield traces between each of the electrode traces as well.
On the bottom side in green you can see our 2 electrodes:
My question: is there any reason we would expect the trace length to impact our noise or error rate? Should we attempt to length match these in a future design?
In a future design we'll very likely be eliminating the through hole headers entirely so the layout will become much simpler and with better shield coupling.
