At the heart of the beacon is the 30m transmitter itself. It uses a 74HC240 octal inverter IC as a power amplifier in push pull configuration. One inverter of the 74HC240 is a VXO using a 10.125MHz crystal in HC6 case (many thanks to Ian Braithwaite G4COL for giving me this nice crystal).
There are many circuits on the internet for 74HC240 transmitter circuits and a good place to start looking is Testes com 74HC240, a page in Portugese but useful to even the non-Portugese speaker, and with many useful links.
In the circuit (left) I have used is a combination of the VXO from one of those links, with the push pull circuit shown on PY2OHH's page. Both the VXO and the push pull amplifier are unlike most of the other 74HC240 circuits on the internet. The push pull amplifier drives a trifilar transformer which is supposed to transform the impedance to near 50 ohms. It is followed by a low pass filter. I was unable to find values on the internet for the 30m band so I had to guess the values, erring on the higher-frequency side (I hope) since in any case it is a low pass filter intended to suppress harmonics, rather than a band pass filter. I do not have the necessary equipment to make an accurate measurement of output power but guess it might be somewhere in the region of 200mW.
A BC109 transistor inverter was necessary to control the inverted control inputs of the 74HC240 since I wanted a "1" in the d0 bit of each byte to signify "carrier on", it seems more intuitive.
Notice in the top left of the circuit diagram that frequency shift is provided by a strange type of varicap! It is simply an ordinary 5mm LED. See my Common Diodes as Varicap Diodes for detailed experiments on this topic. Having played with this idea in the past the LED in this application seemed ideal. Why use an expensive varicap for such a simple application?
In order to validate the idea I performed a simple experiment to chart the frequency shift in this VXO against LED reverse voltage (see right). Whilst I am not entirely convinced of the accuracy of these results, due to drift and instability troubles with the frequency counter, it does reassuringly show a reasonable degree of linearity and plenty of shift for this application. I only want to be able to shift 15Hz (0-15Hz in 16 1Hz steps).
The most significant 4 bits of each control byte (D3-D6, since D7 is unused) are applied to the diode via a simple digital to analogue converter consisting simply of weighted averaging the 4 digital signals using 10K, 20K, 40K and 80K resistors. A further resistor network (hopefully!) reduces the shift range to approximately 0-15Hz as required.