AVR Freaks

I suspect using square waves for FSK which then feeds an RF system is not a great choice.

Using a Sin wave would be much better.

This spectrum, from Wiki Square Wave, shows that a square wave has an infinite spectrum of harmonics.

Using two square waves will also generate additional spectral noise dues to the phase discontinuity when you switch back and forth between F1 and F2.

Using a DDS signal source both gives you a sin wave, and a smooth transition, phase wise, when you switch between the two frequencies.

Following the DDS you also need, as mentioned above, a LPF to remove the harmonics of the "stairsteps" generating the sin wave.

Here is the square wave spectrum:

This image is a very crudely generated 40 kHz signal, using a DDS with only a relatively few "stair steps".

Passing the signal through a LPF retains the fundamental frequency and gives a decent appearing sin wave.

Having trouble uploading this image, will try putting it in its own post.

Analog devices, and others, have DDS chips.

An overclocked XmegaE5 would also work well for this.

JC

Surely 110kHz has harmonics at 330k, 550k, ...
And 114kHz has harmonics at 342k, 570k, ...
.
There will be other components during the switch phase. But most of the time, you have either F1 or F2.
Like Kartman, I would expect a bandpass or even a lowpass filter will round the edges of the signal.
.
Synchronising the frequency shift to happen at the completion of the current period. i.e. go from 4.05us half-cycle to 4.39us half-cycle
.
David.