Bogus illuminance and moisture readings

[drspangle]

Hey all, been a while. I recently came back to working on this project and decided to bring up one of my v1.2.0 b-parasites. Everything looks good, I've flashed it with the ble firmware sample using bparasite_nrf52840@1.2.0 and it was detected immediately by Home Assistant. Great! I am getting good humidity and temperature readings, and the voltage readings seem fine as well. All three values look nominal when powered by coin cell or the debugger using pogo pins.

However, the illuminance values are stuck at 0 lx and the moisture values are stuck at 100%. It looks like all the components on the board are in good shape and I'm testing on my bench, so there's no grime or corrosion to speak of which would affect this. I expected the moisture readings to be 0% and the illuminance values do not vary regardless of lighting conditions.

I'm at a bit of a loss here. Maybe there's a calibration routine I've missed on, and maybe some simple probing on the board will give me some hints as to what's wrong? Any suggestions?

Edit: Never mind the illuminance readings, they're bogus because the component appears to have been installed backwards. I'm really just trying to diagnose the wonky moisture readings now.

[drspangle]

I've managed to get my debugger working and am reading output from the soil_read_loop firmware sample, here's what it looks like for a few seconds' worth of output from having the sensor suspended in open air, powered by the debugger's 3v3:

[00:03:54.424,407] <inf> main: 3.29;0
[00:03:54.924,560] <inf> main: 3.30;0
[00:03:55.424,713] <inf> main: 3.30;4294967295
[00:03:55.924,835] <inf> main: 3.30;4294967295
[00:03:56.424,987] <inf> main: 3.30;4294967295
[00:03:56.925,140] <inf> main: 3.30;0
[00:03:57.425,262] <inf> main: 3.29;4294967295
[00:03:57.925,415] <inf> main: 3.30;4294967295
[00:03:58.425,567] <inf> main: 3.30;0
[00:03:58.925,689] <inf> main: 3.29;4294967295
[00:03:59.425,842] <inf> main: 3.30;4294967295
[00:03:59.925,994] <inf> main: 3.30;4294967294
[00:04:00.426,116] <inf> main: 3.30;0
[00:04:00.926,269] <inf> main: 3.31;4294967294
[00:04:01.426,422] <inf> main: 3.30;0
[00:04:01.926,544] <inf> main: 3.30;4294967295
[00:04:02.426,696] <inf> main: 3.30;0
[00:04:02.926,849] <inf> main: 3.30;0
[00:04:03.426,971] <inf> main: 3.30;0
[00:04:03.927,124] <inf> main: 3.31;4294967295
[00:04:04.427,276] <inf> main: 3.30;0
[00:04:04.927,398] <inf> main: 3.23;4294967295
[00:04:05.427,551] <inf> main: 3.25;4294967294
[00:04:05.927,703] <inf> main: 3.26;1
[00:04:06.427,825] <inf> main: 3.22;0
[00:04:06.927,978] <inf> main: 3.23;1
[00:04:07.428,131] <inf> main: 3.25;4294967294
[00:04:07.928,253] <inf> main: 3.23;4294967295

This seems pretty bogus to me. The "battery" readings from the 3v3 seem nominal, but the soil probe values seem to be totally garbage. I'm not sure what to do with this. Does this mean I've got a bad board (since the sensor is ostensibly the board itself)?

Edit: I also updated the soil_read_loop firmware example: https://github.com/rbaron/b-parasite/blob/37ea3a10622728b989bb62faf0f8d2a456f2aebd/code/nrf-connect/samples/soil_read_loop/src/main.c#L67 to:

LOG_INF("%.3f;%.3f", batt_val_mv / 1000.0f, soil_val_mv / 1000.0f);

But when I did this all I get is zeroes output for soil_val_mv, no matter whether is wet or dry. It's always zero:

[00:02:27.795,227] <inf> main: 3.311;0.000
[00:02:28.295,379] <inf> main: 3.311;0.000
[00:02:28.795,532] <inf> main: 3.304;0.000

Hopefully this information will help with debugging this...

[rbaron]

Hey @drspangle,

I would try, sorted by likelihood and difficulty:

1. Double check the module you have and the module you compiled for. The 1.2.0 hardware has a fast discharge circuit whose pins are different between modules. That could explain the error, since the capacitive pads never discharge
2. Look for shorts/cold solder joints, specially between adjacent module pads. I would pay special attention to pins 2 & 3, as they are related to the moisture sensor in the 1.2.0 version (FAST_DISCH_EN and SENS_OUT -- these are the second and third physical pads on the nRF52840 module):

A few follow up questions:

1. Did you assembly this board yourself?
2. Do other boards from the same batch work with the same code?
3. What module are you using?
4. Any coating?

Please feel free to upload a photo of the board. Maybe we can spot something wrong with the components.

[drspangle]

Thanks for the reply @rbaron.

Did you assembly this board yourself?

No, I used JLCPCB and installed only the processor module. I resoldered the pins along the left side of the chip paying close attention to pins 1 and 2, and the joints looks good but still no luck.

What module are you using?

Here is a picture of the module (prior to cleaning and resoldering). I am pretty sure this is an nRF52840.

Do other boards from the same batch work with the same code?

I have not attempted to bring up any other boards from this particular batch, so I can't use this to rule out anything sadly.

Any coating?

Not yet. I wanted to test a couple from this batch before I began coating them using your tried and tested method.

[drspangle]

@rbaron I think I may have fixed a cold solder joint on pin 3 as it now has good continuity with R6. I also seem to have good continuity between pin 2 and R3. I also checked pin 15 and it has good connectivity with R4. That said, I'm still getting confusing readings from soil_read_loop (though they aren't zero constantly):

*** Booting Zephyr OS build zephyr-v3.5.0-3496-g87d056bd31b1 ***
[00:00:00.347,015] <dbg> pwm_nrfx: pwm_nrfx_set_cycles: channel 0, pulse 16, period 32, prescaler: 0.
[00:00:00.347,015] <inf> main: input_voltage;soil_adc_output
[00:00:00.847,167] <inf> main: 3.30;2
[00:00:01.347,320] <inf> main: 3.30;2
[00:00:01.847,442] <inf> main: 3.31;2
[00:00:02.347,595] <inf> main: 3.31;1
[00:00:02.847,747] <inf> main: 3.30;2
[00:00:03.347,869] <inf> main: 3.30;3
[00:00:03.848,022] <inf> main: 3.30;1
[00:00:04.348,175] <inf> main: 3.31;1

When I run the modified version as above, soil_val_mv / 1000.0 is still always zero. There doesn't seem to be any variance when I squeeze the bottom of the PCB where the sensor is (in either the original soil_read_loop or the modified version). The ble firmware sample still reports 100% moisture no matter what I do, though humidity and temperature remain accurate. Am I misinterpreting something or missing something obvious?

Edit: Pin 15 was the culprit all along. I was pressing the continuity probe into the castellation hard enough that it was making contact, but the joint was cold. Hand soldering sucks. Sorry for the bother. I have a working v1.2.0 board now!

@rbaron I am going to run a few more tests and then make an attempt to conformally coat this thing. Do you have any advice on how (or if) I should tune the calibration parameters once I've finished applying the coating?

[rbaron]

Nice! I'm glad to hear you got it working. I will close the issue.

For coating and calibration, I've only ever used acrylic spray coating, and from my manual tests the coating layer is so thin that I saw little change in the capacitance measurement with and without coating, so I didn't bother maintaining two sets of calibration coefficients.

If you want to explore this further, I would compare the soil_read_loop readings from an uncoated board and a coated board with your preferred method. If they differ too much, you'll ideally have to calibrate it. I wrote this and this comments outlining how I did it.

It's up to personal preference, but depending on how much they differ (e.g.: up to 10-20%), I would not bother recalibrating it. Some other effects, like potential drift (#2) and positioning (#104) probably cause more variance in readings than fine tuning by calibration can resolve in my experience.