Many microcontrollers and logic devices (programmable or otherwise) are capable of drawing very little current when nothing is happening. If a double-throw switch is wired to such a device, then when the switch is moved to one contact-closure position, it's possible for the controller to sleep, drawing essentially zero current, until the switch is moved to the other position. Even if the switch were to sit between the two positions for an extended period of time, or repeatedly oscillate between one position and "open", the system wouldn't have to care, and could remain in a near-zero current state.
Monitoring a single switch input, by contrast, requires consuming a certain amount of quiescent current whenever the switch is closed; the amount required is related to the speed with which software must notice if the switch is open. This is fine for many scenarios involving mechanical switches (in many cases, one can either arrange to have switches be open most of the time, or will be able to get by polling a switch e.g. once every 100ms or so). Rotary encoders, unfortunately, satisfy neither of those conditions. They're as likely to be on an active spot as an inactive spot, and in many cases one will need to respond immediately to a change if one is to have any hope of maintaining an accurate position count.
If a rotary encoder has just two contacts, there will necessarily be a position where both are active simultaneously; the only way to determine when the encoder leaves that position would be to feed current through those contacts to test their state, likely increasing quiescent current by an order of magnitude beyond what would otherwise be achievable (depending upon required response time). If there were three or four contacts wired as:
** ** ** Out1: -AAAAAA----- Out2: -----AAAAAA- Out3: AAA------AAA ** ** ** ** Out1: -AAAAAA--------- Out2: -----AAAAAA----- Out3: ---------AAAAAA- Out4: AAA----------AAA
then it would be possible for a processor or other circuit monitoring such an encoder to ignore the state of any closed contact until it saw a different contact close. In the first case, the processor could resolve 1/6 cycle when awake, and be assured of waking up when the device moves to any position marked with ** after having been on another such position (typically 1/3 cycle; just under 1/2 cycle worst-case). The second case would allow 1/8-turn resolution when awake, and would wake up when the device moves 1/4-3/8 cycle. The second design would be somewhat easier to interface with a quadrature decoder, since feeding Out1 and Out3 into one RS latch, and Out2 and Out4 into another, would yield clean debounced signals for the decoder circuit.
I've seen rotary encoders with more than two contacts, but they've all used the extra contacts for purposes of yielding a more detailed binary position. Do any companies make encoders which use extra contacts to allow near-zero-quiescent power while maintaining an accurate position count? Alternatively, would there be any other approach which could be used to monitor an encoder (possibly a higher-resolution one) with near-zero quiescent power?