I have now got to the section in my book that
describes various transmitter layouts.
I was going to use direct generation of the CPM
signal using a DDS. The book suggests an alternative
technique of applying the CPM to the PLL.
This would allow the use of a cheaper DDS.
However there are the effects of the PLL filter to
take into account. The books says the phase error
will be acceptable. We shall see. I can try both ideas.
I feel a bit stupid not thinking of this myself.

What I am a bit worried about are the modulation
sidebands of the CPM. The technique I am using will
not allow any form of pulse shaping or filtering.

I had a look at the local DSTAR repeater spectrum
and the sidebands are not as well suppressed as
I thought they might be.