Hardware
Some effort has to be put into place to implement suitable vinyl movement sensing.
This discussion on the ti e2e forum lead to a suitable RF design for the vinyl to clock generation hardware. The following is taken directly from the forum and was proposed by Derek Payne, a TI RF engineer :
In case you are unfamiliar, a mixer is a device which combines two frequencies to produce a third frequency, based on the results of some trigonometric identities. In this case (the most common), the mixer implements addition and subtraction. Typically we refer to a mixer by the RF port (high frequency), the IF port (intermediate frequency) and the LO port (local oscillator frequency). The operational direction of the mixer can vary, such that either RF or IF ports can be the input or output.
I've drawn a diagram representing my first mixer suggestion. To keep matters simple, there are two identical PLL circuits. One generates 1.2GHz from 100kHz; the other generates 1.2GHz + up to 60MHz from 100kHz + up to 5kHz. A mixer would be used to combine the 100kHz reference and the 0-5kHz signal from the tachometer, such that the signal seen at the phase detector is now 100kHz to 105kHz (much smaller variation relative to nominal frequency). Furthermore, now there is only 5kHz * 12000 = 60MHz of variation at the VCO, which can definitely be within the tuning range of the VCO. The first PLL locks to the 100kHz to 105kHz signal and produces 1.2GHz to 1.26GHz at the VCO output. The loop bandwidth of this PLL can be made greater than 5kHz which should be acceptable for tracking the 1-2kHz/s change at the tachometer. Meanwhile, the second PLL is identical to the first, but has a fixed frequency input and output. The resulting VCO outputs are combined at a subsequent mixer operating as a subtractor, so that 1.2GHz is subtracted from the RF port and the 0-60MHz multiplied-up component is all that remains. In practice you would also want antialiasing filters on both the input and output to prevent locking to 100kHz - 0-5kHz, and to eliminate the sum of the two VCOs. Eliminating the VCO sums is easy; eliminating the difference at the mixer can be more tricky, but some mixer designs can more favorably produce a sum or a difference specifically and suppress the other unwanted function.
The drawback of this scheme is that the VCO is operating at very high frequency relative to the PFD frequency, so changes at the VCO take a long time to propagate through the N-divider. If it takes long enough, the PFD may "cycle slip" when the N-divider reverts its phase relative to the R-divider before the VCO can bring the feedback path into alignment again. Cycle slipping slows down the loop considerably, and although a stable loop can recover, it can be difficult to keep up with rapid changes at the reference (even when the loop bandwidth is theoretically high enough). This means you may not track the full 1-2kHz/s rate of change as desired.
As I mentioned, if you break this scheme up into two segments with lower multiplication ratios (e.g. 100 and 120), cycle-slipping is much less of an issue since the PFD can operate at a much higher rate without requiring terrifically high VCO frequency, and the LO can now be at a more ubiquitous frequency than 100kHz (e.g. 10MHz). Another image detailing the cascaded scheme is shown below.
