I spent some time investigating whether the frequency sensing could be done by monitoring the strings themselves but concluded to capture time was too slow both because of the inherent time window required for accurate frequency detection and the lags of the filtration needed to remove guitar string harmonics.
Accordingly the Milwaukee design uses tension measurement where the tension is deduced by sensing the deflection of a helical spring pair by a potentiometer. A standard potentiometer with a linear taper and a 1 inch body appears to be accurate enough to support over 256 (eight-bit) levels which appears to provide just adequate tuning accuracy. The bigger problem is the musical scale does not map to equal tension intervals and thus to equal potentiometer rotational increments. So precious resolution is wasted at high notes or tuning accuracy is lost at low notes. As noted in the Theory section, improved resolution for low notes is provided through the use of two springs (which in series produced a low spring constant) one of which is bound (wrapped) with a loop of cord to prevent it from further extension when it is about halfway through the tension range. This increases the spring constant, permitting increased resolution at lower frequencies while effectively decreasing the resolution for higher notes which have greater frequency differences. The complexity of the mapping of tension to frequency is handled by a lookup table in a controller microprocessor.
In order to reduce the torque along the shaft of the potentiometer, pulleys are fit to the potentiometer shafts with the shank of the pulley (1/2" outer diameter) receiving a cord from the springs to turn the potentiometer with change in spring length. The actual pulley flange has a hole drilled through it so that part of the cord wrapped around the shank can pass through the hole and be singed by a wire tie to prevent slippage between the cord and the pulley.