Counting of pulses from flowmeter seems unreliable

[SvenRoederer]

During the last event I found that the same size kegs have very different total number of counts. This leads to an unusable prediction of the fill-level.

I assume, this is caused by the counting-code of the RPi-software. As far as I remember it's using "poll"-mode for the pulses. It's quite realistic that the pulse-frequency is to high or the gpio-lib to slow. Currently we are using https://pypi.org/project/RPi.GPIO/

Options to fix this:

• find a better library
• use a dedicated counting-circuit

[MicDiFebo]

Beernary counting should be based on realtime-interrupt-counting like it was on our prototype with an Atmel µC. The pulse frequency of the counting device itselve is not really high - as far as I remember we have 171 pulses per liter. Even if you would leave the tap open for a minute and beer is floating constantly i guess you will only have about 10 liters. That means 1710 pulses/minute or 28,5 Hz - not very high ;-)

[SvenRoederer]

As of the low frequency the effect of a missing pulse is roughly 6ml, only.
But have a reliable counting again a "realtime-interrupt-counting" seems the best option for now. Adding some µC for this which :

• counts pulses on one end
• talks to the RPi on the other end via serial-port
  seems to be enough.

Serial communication should at least have two commands:

• "getcount" - returns the number of counted pulses (unsigned int)
• "resetcount" - resets the counter to 0

Having this installed we can still have the RPi-counting connected on parallel and compare the difference of counted pulses and try to improve the RPi-code

[SvenRoederer]

As discussed on chat, the serial port of the RPi is planned for the serial RFID reader (see PR #6). So a regular serial (rs232 like) connection is not possible. I see following options:

• adding a 2nd serial port of the RPi (USB <--> serial; probably there is some driver "GPIO <---> rs232" around)
• using RS485 to build a serial-bus with RPi, Reader and Counter-chip)
• finding a I2C or SPI counter chip (I had no luck with this)
• adding an SPI / I2C interface to the counting-controller

[SvenRoederer]

During MCH2022 I installed a small Arduino to count the pulses in parallel to the RPi. It's connected to the RPi via USB (cable looped out to the external USB-ports) for power and serial-interface.
Arduino-Code added with 6ab59d9