Touch not working

[BeanieBots]

I have 3 displays all with XPT2046 touch controller.
The 2.8" ILI9341 all works fine.
The 3.5" ILI9481 display works, no touch.
The 4.0" IL9488 display works, no touch.
They all have the same pinout, so swapping is easy.

I have tried with ESP32 firmware 1.47 and 1.51.6 with the same results.
Any advice?

[lyizb]

I don't know if this is related, but most (if not all) ILI9488s have a corrupted MISO which confounds the use of other SPI devices on the same pin. One solution is to wire MPU MISO to touch DO, then connect touch DO to LCD MISO with a 680R resistor.

Another option is a Schottky diode from LCD_DO to MISO (anode to LCD_SDO) after cutting the track from LCD_DO to MISO. Then add a 10k resistor from LCD_CS to MISO.

[BeanieBots]

Thanks lyizb, I'll hold fire making any circuit changes for now.
I have a vague recollection of reading something along those lines a while back so I think you could be on to something.
I'll try and get a 'scope on it to look for dodgey logic levels. Then maybe have a hack.

Would there be any point in trying to use the ILI9341 driver with the ILI9488 display just to see if the touch then works?

[cicciocb]

Thanks lyizb, I'll hold fire making any circuit changes for now.
I have a vague recollection of reading something along those lines a while back so I think you could be on to something.
I'll try and get a 'scope on it to look for dodgey logic levels. Then maybe have a hack.

Would there be any point in trying to use the ILI9341 driver with the ILI9488 display just to see if the touch then works?

There is an undocumented (and "temporary") possibility to modify the sensitivity of the touch controller.

int TOUCH_THRESHOLD = 350;
int TOUCH_AVERAGE = 50;
int FIRST_TOUCH_DELAY = 30;

These are the default values that can be changed with

debug 10 threshold
debug 11 average
debug 12 delay

(attention! no commas between the arguments!)

try to reduce the value of threshold and see if this has an effect for you

example :
debug 10 200

[BeanieBots]

Thanks both. I've got family staying at the moment so it would be rude to spend any time on this right now.
I'll be able to devote time at the weekend.

[BeanieBots]

Managed to spend some time on this.
Tried the simple first. Unfortunately, using debug to change values did not help.
Connected just the touch cables and it responded OK.
Put a 'scope on LCD MISO and it was clear that there is a logic level fight going on!
Tried the 680R suggested by lyizb et voila. All is good
So hopefully anyone else with these screens will find this thread.
Will probably end up with the diode/pullup version as I think that is better than having a scrap between them via 680R.

cicciocb, have you considered making those values configurable via a command? They might be useful even just as a preference for users.

[lyizb]

Will probably end up with the diode/pullup version as I think that is better than having a scrap between them via 680R.

Glad it worked for you.

I haven't done it yet, but that seems the superior solution to me--fix the LCD rather than having to fix each and every PCB--though I have a number of PCBs with that 680R hack.

[BeanieBots]

By accident rather design I found that things appear to work OK by simply not connecting the LCD MISO to anything.
Not sure why/if reading the LCD would be required by Annex but reading up about Bodmer's librarys it seems he uses a read in order to produce anti-aliased graphics and maybe something to do with sprites, so we'll wait and see. More research required
For now, thanks for the solution. At least I'm up and running.

[cicciocb]

By accident rather design I found that things appear to work OK by simply not connecting the LCD MISO to anything.
Not sure why/if reading the LCD would be required by Annex but reading up about Bodmer's librarys it seems he uses a read in order to produce anti-aliased graphics and maybe something to do with sprites, so we'll wait and see. More research required
For now, thanks for the solution. At least I'm up and running.

The MISO from the TFT is not required, except if you want take a dump of the image from the display.
SO you can simply remove it

[cicciocb]

cicciocb, have you considered making those values configurable via a command? They might be useful even just as a preference for users.

Probably not required, I just used them during the debug of the touchscreen.