This is the description and circuit of my Diversity MEPT. I have used a 74HC86 as Oscillator/Buffer and 2 x 74HC240 as PA.

Right Click and select View Image for full view.

Due to the need for critical frequency matching I decided that the best approach was identical circuits on a PCB.

Right Click and select View Image for full view.

Sorry for the error. Pin1 goes to zero volts not +V

Run on 5 Volts and using just one gate in each bank of four on the chip and an output of 35mW the 74HC240's are not stressed.

When experimenting with oscillators using these chips I have noted that anything you do on any gate is reflected on all others, (via the power supply architecture on the chip I believe).
With the 74HC240 even enabling unused gates produces a positive move of the output frequency of a Crystal Oscillator which is using the other group of gates. I tried to use this effect to produce FSK of an Oscillator without the use of a Varicap Diode. It works, the snag being the shift is greater than we need for QRSS and the switching spike is huge.
Published designs using the 74HC240 for a single chip CW transmitter which keys the 'Enable' complain of incurable chirp, Hmm.

With the 74HC86 it is an opposite effect. FSK'ing of the Crystal Oscillator produces a mirror of the shift on the other gates.
I asked myself can we make use of this?
Yes, build 2 Crystal Oscillators on the same chip adjust the frequencies so they are close and they pull each other, get them very close and they lock.

Now FSK each oscillator 180 degrees out of phase, the positive shift of the high frequency one reinforces the negative shift of the other. This occurs when they are in a near fully locked state. Adjustment is tricky but once set up they seem to stay. The high oscillator with the positive shift becomes the master, if the frequencies are identical they fully lock, the Master will swamp the Slave, controlling its frequency to the extent of cancelling its shift.

Has this effect been noted before I wonder or is this the 'G3ZJO Lock'

The Oscillators are in fact adjusted to be 10Hz apart in frequency. The Master is set to say, 10.100000Mhz, the FSK shifts it to 10.100005Mhz.
The Slave is set to aprox 10.0999990Mhz the FSK shift is adjusted to be aprox 2.5Hz. When pulling takes place this shift is reinforced to 5Hz, i.e. to 10.100000Mhz.

The Locking and shift reinforcement can be seen in action in the following picture. The two oscillators started in lock, the PCB component side was sprayed with freezer concentrated on the Master Oscillator. The Master (top trace) is seen gradually approaching the Slave as the temperature recovers. At 130 seconds from the left hand side of the display the two carriers are seen beating together, after the next shift they lock, shaky for 45 seconds, then solid.
Note how the FSK Slave shift increases when in this state.

Right Click and select View Image for full view.

Very cold starts can be a problem if the Slave starts a little too high it is swamped requiring a finger to be poked in to pull it low. (A re-set button could be provided.)
It is still early days but at normal ambient temperatures start ups are perfect and 16 Hour periods of lock have been maintained using both breadboarded and the final PCB version.
In striving for minimum bandwidth (10Hz shift) I may be running the system too tight, tests are continuing with both oscillators set for about 5Hz shift prior to locking them. This results in a final shift of around 12.5Hz, it may avoid swamping as the FSK will force them further apart.
08.03.2008. Further I have found another lock, tighter than both previously described, when in this state the negative shift can be turned to zero level. Tests will continue to establish the long term stability of this. Maybe we are getting only the mirror of the positive shift.

When testing on the breadboard I switched off my bench light and the lock went haywire, this lead my the use of Black, Infra Red LED's (ex TV remote controls) in place of the Visible Red LED's normally used as varicaps.

The use of two 74HC240 for the PA may be an overkill, I didn't try using two halves of the same chip as I was aware of the undesirable effect mentioned above. They cost less than .5Euro each after all.

Another extravagance is the use of a LM7805 Regulator on each unit, again at 1.5Euro for a bag of 15 we get super regulation and isolation.

PA PCB Print View.

Osc. PCB Print View.

Finally the PIC Keyer, everyone has their own favourite method of generating FSK we just need two 180deg. outputs.

PIC PCB Component View.

PIC PCB Print View.