PDA

View Full Version : touch sensor pins not all created equal



JBeale
11-07-2012, 08:22 PM
Just FYI, if you are using touch sensing on the T3, there are 12 different possible pins to use, but not all pins are equivalent. If you want the lowest noise, your best bet is pin 15 (A1). The noisiest touch values are on pin 22 and 23, I presume from adjacent-net noise coupling from chip internal layout and/or board layout. Below values were obtained by running the attached program to read and average together a set of touch sensor readings, for each of the 12 possible touch pins. This is on a T3 board with edge pins soldered, but held in midair by the USB cable: not plugged into a breadboard and with nothing else connected. The Average, Peak-to-Peak (max - min) and Standard Deviation are sent out from the program, and in this case each number shown below is the average of 181 sets of readings, over about 20 minutes.

Note, pins 25-33 are on the bottom side of the board (access by SMT pads only). Pin 25 has the lowest absolute reading, indicating the shortest net with least coupling to the ground plane.

Incidentally, the calibration of the touch sensor is as advertised, 1 count = 0.02 pF at least within 10%. A 10 pF test capacitor I have showed a delta of 496 counts when connected.


Pin # Avg Pk-Pk St.Dev.
----------------------------------
0 415.92 1.52 0.23
1 424.79 2.96 0.40
15 401.00 0.01 0.00
16 372.27 1.76 0.42
17 449.68 14.42 2.01
18 604.84 6.53 0.85
19 618.47 3.38 0.56
22 516.41 48.76 7.08
23 626.27 52.35 8.54
25 282.78 1.00 0.37
32 305.06 6.65 1.06
33 315.89 20.36 2.48

From this chart it actually looks like pin 15 isn't even working, since the reading almost never budges even a single count. However I verified by connecting a short wire to the pin, that it does respond as expected. Not sure why it has so much lower noise than the rest, maybe specific to operation of my code (?)

Frollard
11-07-2012, 09:22 PM
Very helpful thread. Seems it is wise to have a rolling average or power on self test to calibrate each run - Do you see similar results between teensies?

JBeale
11-07-2012, 09:40 PM
So far I've only done this test on one unit- but feel free to run the test program on yours, and report your results!

Frollard
11-07-2012, 11:28 PM
Pin Average PkPk Std
0 410.7 1.039 0.426960784
1 416.1 2.000 0.314607843
15 396.5 2.980 0.512745098
16 373.0 2.000 0.314588235
17 451.9 12.235 1.768098039
18 616.6 7.431 1.082039216
19 625.9 3.431 0.474
22 512.6 45.196 6.760215686
23 618.1 54.412 7.305901961
25 297.6 1.020 0.460960784
32 319.0 10.392 1.632
33 329.0 23.804 3.129764706



Counted about 50 readings from each pin...playing with excel I can't for the life of me figure out how to have more than one vertical axis scale (so the numbers jump out a bit more, otherwise the stdev and pkpk are less than a pixel high)

JBeale
11-07-2012, 11:51 PM
Thanks Frollard, very interesting to compare results! The worst performing pins on your device match up with mine (22, 23, 33) although your best pins are different.

Frollard
11-08-2012, 04:58 AM
It was sat on my glass desk, but was near my mouse cable, and I was using the computer at the time, so that might have affected it a bit.

My numbers are without pin headers attached.

JBeale
11-08-2012, 06:24 AM
Just for curiosity, here is a different Teensy 3.0, again held up in the air by the USB cable. Board has pins soldered, but nothing connected. This time it is pin 0 which has no apparent noise (although the signal does go up when touched with a finger). Again it is pins 22,23,33 which have the most noise.


Pin # Avg Pk-Pk St.Dev.
0 440.00 0.00 0.00
1 443.12 2.35 0.45
15 432.68 1.15 0.45
16 408.54 1.54 0.49
17 491.33 9.63 1.78
18 649.49 5.35 0.81
19 660.22 4.06 0.59
22 550.56 45.70 6.96
23 668.00 47.78 8.32
25 310.43 2.00 0.55
32 333.64 8.13 1.33
33 342.02 15.44 2.26


...actually, any pin which has "Peak-to-Peak" noise near 1.0 might go to apparent "zero noise" if the DC level happens to be right in the middle of the ADC value. Or, if the average value happens to be at or near the threshold, it will toggle by one bit. Looks like all three samples agree than pins 0,1,15,16, and 25 are the best, and the others are significantly more noisy.