Strategy with many buttons

[rlogiacco]

On behalf of a user who contacted me in private:

I've been using the library for a few years in a project utilizing a Teensy 2.0.
I had 10 buttons on one analog pin, and 6 other buttons on 6 digital pins using the BetterButtons lib. I'm moving to a Mega2650, and I need 24 buttons in total (3 rows of 8).
I think you had recommended not exceeding 8 buttons for memory reasons?
My project was working fine with 10, but at some point of increasing buttons, I would think the (smaller) voltage differences between the buttons might become an issue.
Is 24 buttons on one analog pin on a Mega2560 realistic? If not, if there were a way to instantiate your library three times(3 pins), would that be less CPU intensive than something like the BetterButtons library using 24 digital pins?

[rlogiacco]

The library is very well capable of handling 24 buttons on a single analog pin but, as you correctly stated, the voltage differences are going to be very small and that will be the tricky part. Memory consumption is not going to increase much, I went for a max 8 buttons default value for good measure (24 buttons, you must admit, is not the most common case): just add #define ANALOGBUTTONS_MAX_SIZE 24 before including the library.

You might want to test something like the circuit simulated at this link which, using a network of stock resistors, will provide more than 50mV difference between each button: that's around 10 points on the 12bits 1024 scale of ADC provided by AVR microchips. In other words, the resistors selection is going to be crucial to ensure you have enough gap to have accurate readings, not to mention the potential collision when multiple buttons get pressed simultaneously (don't know if that has to be taken into consideration in your use case).

You can also use 3 separate instances, which will solve your issue of having a small interval, but you'll end up having to check three separate pins, which can introduce a totally different problem: when checking one line you will not be checking the other twos...

A third approach might exist: multiplexing. You can still have 1 analog line plus 3 digital row activation pins with 8 buttons each, with plenty of gap between each button value, activating each one of the three digital pins in turn to scan each button group. Not sure how reliable this solution might be, but it's worth a try.

My advice is to play with the simulations linked above to maximize the minimum interval and go for a single instance (either multiplexed or not) as it will be easier software wise.