See project's main page for further caveats, if you are not already familiar with them.
(Comments from WE Payne, around 2011, tweaked...) The OOK wireless modules Sparkfun were selling around 2011... and possibly later, still... top out around 15 to18 kHz when running on 9v with a 5v logic input. The bottom seems to be about 50 Hz. (50 Hz to 16 kHz with 1 Hz resolution is about 14 bits but I believe your overly optimistic to use it like that.) The slow end output looks like the input. At a few kHz there is some lag, more on rising edge than falling, it affects duty cycle. By 15kHz the RX output barely slews to +5v before it turns around and the phase is lagging 180 deg from the TX input. Got scope photos. So we are learning the limitations of these links, and there's a ton of possibilities as long as we respect their limits.
(Comments from WE Payne, around 2011, tweaked...) The "battery monitor" costs nothing, brilliantly simple, the low quiescent low dropout regulator will add $1.50 but saving the first couple of 9v batteries will pay for that. Running out of ideas that fit in without adding chips.
(Comments from WE Payne, around 2011, tweaked...) The OOK (brand... see above) transmitters draw hardly any quiescent current, so it maybe worthwhile to redesign the power supply around a low quiescent current regulator, even though it will become the 2nd most expensive part after the PCB itself. It also requires special decoupling capacitance for stability, ESR has to be low enough but not too low. Go figger. Probably adds $1.50 to BOM, quiescent battery life of 4 days extended to perhaps a few years.
(Comments from WE Payne, around 2011, tweaked...) OOK radio are NOT transparent links. Research into the operating envelope reveals that besides the "squelch noise" for instance at LOW pulse rates (160 Hz) I have seen the RX "double clocking" i.e. the frequency out of RX is twice the frequency into TX. That disturbs me much more than noise out with nothing in. The output impedance is too high to overcome even a modest pull up. The pulse width is not faithfully transmitted, and the phase lags increasingly until it becomes completely inverted near the upper limit. What are the dependences on signal strength? Link frequency, model #, expected unit-to unit variation? These and others beg to be explored and documented as a technical resource to the community.