9-bit 16N degrees celsium is reported as device disconnected

[AndrewKirkovski]

Working with 18B20 sensor in 9 bit mode
In async mode when temperature reaches 32C, library reports -127
31.5 and 32.5 are reported fine
Same happens at 48C and 64C, i would suspect same at 80C and 96C :)
Looks like CRC check works incorrectly

[AndrewKirkovski]

Update.
Actually it is isConversionAvailable that is working wrong. You are comparing LSB to zero to verify there is something there. And LSB === 0 for 16*N celsius

[tomdeboer]

Ah, good catch! Strange that I never noticed that before! Let me look into that tonight.

Meanwhile, feel free to change it to return (scratchPad[TEMP_MSB] | scratchPad[TEMP_LSB]); while you are at and I'll merge it.

[AndrewKirkovski]

(scratchPad[TEMP_MSB] | scratchPad[TEMP_LSB]) will still report wrong value for 0C

[tomdeboer]

Hi Andrew,

I have quite extensively tested what you described, but I'm having varying results reproducing the error you are experiencing. There is definitely some different behaviour at 32 degrees. At that particular temperature, isConversionAvailable gets called multiple times (5 times usually), instead of just once at any other temperature. It does give me a valid 32.00 degrees though, so I'm not sure what you are experiencing.

Can you tell me more about your code and your setup? Are you using parasitic power? Please look into it more if you can. I will do the same another time and hook it up to an oscilloscope.

[AndrewKirkovski]

Initial description is wrong, second comment is precise
Temperature getting works ok, isConnected too.

isConversionAvailable has wrong check. Should check bytes that can't be zero. Both TEMP_MSB and TEMP_LSB are zero at 0C so you can't say use just scratchPad[TEMP_MSB] | scratchPad[TEMP_LSB]

bytes 5(0xFF for DS18B20) or 7(0x10 for DS18B20) are better for this check because they can't be zero when conversion is ready.

Though I guess the best way is to follow specs and check all bytes