PTC anomaly with channels above 8

[ZimM-LostPolygon]

I have been using an Attiny1616 (VQFN 20-pin variant) with the PTC library from mTC. I only needed a single selfcap touch button, and it worked great immediately! That is, until I had to change the pin. I was initially using the PB0 pin, which corresponds to PTC channel 5. However, after switching to pin PB4 (PTC channel 13, according to the datasheet), ptc_add_selfcap_node suddenly started PTC_LIB_BAD_ARGUMENT. After looking into the code, I think I've found the problem:

megaTinyCore/megaavr/libraries/PTC/src/ptc.c

Line 272 in 2caece5

ptc_ch_bm_t xCh = (ptc_ch_bm_t)pCh[0];

  ptc_ch_bm_t xCh = (ptc_ch_bm_t)pCh[0];
  ptc_ch_bm_t yCh = (ptc_ch_bm_t)pCh[typesize];

ptc_ch_bm_t is defined as uint16_t on parts with more than 8 PTC channels, however pCh is uint8_t*, so what happens, is that only the lower byte of the channel mask gets casted to ptc_ch_bm_t, with the result being 0 for all channels above 8. Changing those two lines to

  ptc_ch_bm_t xCh = *(ptc_ch_bm_t*)&pCh[0];
  ptc_ch_bm_t yCh = *(ptc_ch_bm_t*)&pCh[typesize];

... makes the library return PTC_LIB_SUCCESS, and seemingly that should be it, but... the node always immediately fails to calibrate with PTC_CB_EVENT_ERR_CALIB_LOW no matter what I try. There is nothing connected to that pin. So I suspect there might be a deeper issue which I don't understand.

[SpenceKonde]

@MX682X ? this is his work,

[zbn10]

@ZimM-LostPolygon
Great inspection.
I've been having a problem that X lines except X0 are malfunctioning on mutual capacitance sensor.
Your fix made a big step forward in problem analysis.

[MX682X]

I've taken a note, but I'm currently busy, Not sure when I will be able to tackle it. Maybe somewhen in December

[MX682X]

@ZimM-LostPolygon
PTC_CB_EVENT_ERR_CALIB_LOW usually means that the capacitance of the pin is too low. Try connecting a (long) wire or electrode (copper pour)

[ZimM-LostPolygon]

@MX682X Yes, that was my thought as well. However

1. I get this error for every channel above 8, including pins where I do have some metal connected (including a long wire I specifically had for PTC on channel 13, which worked on channel 5)
2. On channels <= 8, PTC works even by touching bare unconnected MCU pins

[zbn10]

@MX682X
I also use your PTC library and would like to thank you.
I have also encountered the situation that ZimM-LostPolygon describes.

In Attiny1616, ptc_ch_bm_t is 'typedef uint16_t'.
the following code loses pCh[1].
uint8_t ptc_add_node(cap_sensor_t *node, uint8_t *pCh, const uint8_t type) {
...
ptc_ch_bm_t xCh = (ptc_ch_bm_t)pCh[0];

I found other two code errors that I think are incorrect. I'd be happy if you could check it out when you have time.

1. function name typo. uint16_t ptc_get_node_cc_fempto(cap_sensor_t *node) in the ptc.h. 'femto' is correct right?
2. ptc_ch_to_pin[] in the ptc_io.h has incorrect conversion. The 14th parameter PORTB_ISC(6) should be PORTB_ISC(4)? I refer to the Microship's official data sheet, Table 5-1. https://ww1.microchip.com/downloads/en/DeviceDoc/ATtiny1614-16-17-DataSheet-DS40002204A.pdf

[MX682X]

Ok, so I've pushed some commits to my Fork. I already had some other bug fixes there that weren't comitted yet. A result of having three copies of the library...
@zbn10 thanks for pointing out the two code errors aswell, fixed

@ZimM-LostPolygon Do you have access to a Curiosity nano board or another debugger compatible with Atmel/Microchip Studio?
As I don't have a tiny with more then 8 PTC Pins I can't really debug the PTC_CB_EVENT_ERR_CALIB_LOW myself. I can give you instructions on how to debug the Arduino code to figure out where the reason for that error come from.

(I prefer Studio when it comes to reading out unofficial registers)