SX1509 and Long Press issues

[mamoru-miyagawa]

I have connected my Waveshare 3.5B to a SX1509 breakout board with 4 switches.
They all work as intended and can be easily setup from the web configurator.

Since then I have seen this issue where the touch screen does not accepts long presses and reacts to any sort of holding as multiple taps.
That is seen when I drag sliders around.

Not sure if something is weird with my board but the issue is solved if I disconnect the SX1509 and doesn't happen if I wire the switches directly into the Screen

[Builty]

Hi. Sorry for slow response. I'll test this combo here and see if it happens to me also.

[Builty]

I ran some tests. For me long press seems OK, but yeah dragging sliders is terrible with the SX1509 connected, and OK without it.
I will debug.

[Builty]

Hi. Quick follow up, I can see that when dragging sliders, if the SX1509 is connected, the touch controller chip is reporting that the finger has been released when it hasn't, which causes the value to be sent multiple times and slow things down.

I haven't found out why this controller is doing this or how to fix it yet.

[Builty]

Hi. Just wanted to mention that my progress on this has been slow. An urgent work project has been leaving me exhausted at the end of the days. I will try to get back onto it this week.
I can see that its something strange with the touch chip, and it only happens if the SX1509 is read frequently. Slowing down the reads makes the touch better, but of course makes the footswitches worse.

[mamoru-miyagawa]

Hey, no worries about that! I have worked around that by connecting the switches directly to the screen, I only wanted 4 and it is working completely fine like that.

[Builty]

I spent a few solid hours on this issue. As far as I can tell, its a quirk of the AXS15321B touch chip, in that it doesn't like there being other devices on the same communications bus that are being read rapidly (footswitches need this.)

All the other touch chips are fine with this but this AXS reports back as no touch intermittently, which makes the UI think a dragged slider was released and can confuse the long press handling.
I've tried a heap of work-arounds in code but best I could do was a small improvement, which also made the footswitches worse.

I think the only practical fix is to move the footswitch expander to a different set of pins and create a second communications bus so they are independant.
The only way to do that appears to be also moving the wired footswitch 1 to a new pin, which I think would catch out a lot of users when they update and suddenly find there footswitch is broken.

I think for now I will note in the platform file that these two don't play nice together and suggest that if external footswitches are required, maybe choose a different platform.
Maybe some other solution will arise in the future.

[mamoru-miyagawa]

Interesting, maybe a different platform would behave differently. I don't plan to change anything right now but if I ever find anything or test another platform I will let you know.

Thanks for the hard work!

[Builty]

Yes each platform is different here. It depends on which pins are available on the outside, and also what type of touch screen controller chip.
The JC3248W535 uses the same touch chip, but the pins on the outside are already a separate bus to the touch pins, so it doesn't suffer from the same issue.
Other touch chips generally don't care about other devices on the same bus, its something with AXS15321B chip in particular,