Write + read error in am2320

[louisvangeldrop]

Target device: NodeMCU ESP32c3-devkit

Description
I'am using the example from folder %MODDABLE%\examples\drivers\sensors\am2320

Steps to Reproduce

1. Build and install the app using this build command: mcconfig -m -d -p esp32/c3_32s_kit
2. I get a Write error at line 88:io.write(cBuf) in debug window:

                Timer.delay(10);
		
		// start conversion
		cBuf[0] = Register.COMMAND_READ;
		cBuf[1] = reg;
		cBuf[2] = 0x02;
         	io.write(cBuf);
		Timer.delay(2);

I have looked at the Adafruit am2320 driver: https://github.com/adafruit/Adafruit_AM2320/blob/master/Adafruit_AM2320.cpp
and changed the delay-duration in Timer.delay(10+100). As a result I don't get the write error.

However I also have to change the Timer.delay(2) into Timer.delay(2+5) to avoid an io.read error.

[louisvangeldrop]

I have changed the "wake" code just like the Adafruit-code into:

               for (let i = 0; i < 3; i++) {
			try {
				io.write(new ArrayBuffer(8));	// write 8 length buffer
				written = true
				break
			}
			catch (e) {
				Timer.delay(100)
			}
		}

That works.

[phoddie]

I guess @mkellner didn't encounter the read/write failures that you are experiencing when he wrote the original driver.

100 ms is a looong time to wait. I quickly checked the data sheet and it seemed to indicate a shorter time should be enough. Does it fail with shorter times (like 10ms)?

Would you like to make a PR to merge this change?

[louisvangeldrop]

I have changed 100 ms in 10 ms and it works.

[louisvangeldrop]

BTW I have no idea how make a PR. Maybe I am too old for it.

[phoddie]

No problem. I've merged your changes. They were incomplete (how written is handled is omitted from the snippet above), so I made a guess. I also did some cleanup on the driver, since it used some older coding patterns.

The full source is on this gist. Could you give that a try? I confirmed it compiles but I don't have the sensor to confirm it actually works. Once you confirm that, I'll commit it. Thanks!

[louisvangeldrop]

Tests running last night and today showed that the 10 ms delay is too short. Sometimes the temperature or humidity value is undefined. With a 100ms delay I do get twice a "catched" exception and with 50 ms three times. Both delays produce correct temperature and humidity values.

[phoddie]

Thanks for giving it a try. The timing / retry behavior seems like a mysterious incantation. Since I don't really understand it, it seems safest to duplicate the Adafruit approach. I've updated the gist to do that. Does that work for you?

[louisvangeldrop]

It works. Thx for the help

[phoddie]

Cool, thanks. Will get that committed.

FWIW – the delays are long enough to be a bit annoying. They exist because the sensor automatically powers down to save power if the sampling interval is high enough. An alternative would be to use a repeating timer to keep the sensor active. There could be an option to disable that, but at least the default behavior would be more-or-less non-blocking.