harvie / rotarydial

Arduino library to read numbers from a rotary telephone dial

Rotary Dial Arduino Library

This library was written for use with the dial pictured above (Dial from Czech telephone Tesla AS-10, green and yellow wire). But it should work with anything capable of producing bursts of pulses.

This implementation is for the North American system, where [1, 9] pulses correspond to the numbers [1, 9], and 0 is represented by 10 pulses.

Connection & Circuit

This library uses interrupts, so you have to connect the dial to pin which supports interrupts (this is different for different arduinos, but most common boards can do interrupts on pins 2 and 3).

Unfortunataly this library is static, so currently you can use it only once at a time (= with one dial).

Internal pullup is used, so all you need is to connect NC terminals of rotary dial to GND and Pin 2 (or whatever pin you choose in the code).

TODO

• Possibility to register callback for processing received numbers
• Make timing constants user configurable
• Allow inverting of logic to use NO instead of NC dials (not sure if there are such)
• Allow more than one dial at the same time
• Provide method to detach interrupt if needed
• Handle dial timeout directly in library? (= send ENTER when user is done typing numbers, currently we handle this in example sketch)

See also

• Wikipedia: Rotary dial http://en.wikipedia.org/wiki/Rotary_dial
• Rotary-Dialer Library by Mark Fickett https://github.com/markfickett/Rotary-Dial
  • Similar library, but does not use interrupts and needs third wire with "enable" signal
• Do you want to make PS2 keyboard using rotary dial as an input? https://github.com/Harvie/ps2dev/
• Brmlab Hackerspace Prague http://brmlab.cz/

rotarydial's Issues

Allow multiple instances

Modify library to allow multiple instances listening on multiple pins/interrupts.

This can be done once this is fixed:

arduino/ArduinoCore-avr#85
arduino/ArduinoCore-avr#58

Incorrect digit can be read from a partial pulse train

Nice library! Very succinct and efficient. Truly a pleasure to read and use.

I suspect there is one failure mode though. If a client delays read()ing out the latest dialed digit, and if the next digit is being dialed when client read()s, then the next digit may have some but not all of its pulses skipped. This will result in a digit that is lower than what's been actually dialed returned next. And the user would have no way to detect this and reject corrupted input.

One potential solution is to track when a new pulse train begins -- even if the previous digit is still available(). If that happens, discard either the previous (complete) digit or the entire new pulse train(s) until the last available digit is read. Lossy! But at least no false/partial digits. Another possibility is to queue digits as they are dialed, and read them, first-in-first-out, form the queue. Still lossy, but not until the queue overflows. Finally, adding/taking into account the third (dial/ready) wire could help, too.

Either solution is probably too much trouble, and -- realistically -- of little value in real life. But I thought I'd mention this to warn users of this failure mode, so that they make sure to read() timely, or else risk getting false digits.