The recent resurgence of interest in BPSK on the LF bands reawakened my unease at using this hard switched mode with high power transmitters.

I recall ages ago using Coherent (10 B/s BPSK)  and hearing audible clicks coming from my PA and / or antenna loading coil.  It all survived, but feels iffy.

And as for the keyclicks that must have been spreading a long way either side of the carrier !

So, I've been looking at using a DDS to generate a Gaussian ramped phase shift, where the phase glides from 0 to 180 or back following a curve.

I've implemented an approximation to the ideal curve by including a 32 step phase ramp, using a Gaussian curve lookup table, into an AD9852  DDS frequency source.   A lonk on the PIC allows hard swited PSK to be selected to allow quick comparisons, or for when sharp edges really are needed, or for faster response

The spectra of both types can be found here http://www.g4jnt.com/RampedBPSK.pdf

At the moment PIC code exists for driving the high resolution AD9852 DDS with its 48 bit frequency setting and 14 bit phase register (although I only use 8 bits of phase from the lookup table).  

The same PIC code could be modified for the simpler to drive AD9850 DDS beloved by many, although the phase register in that device only has 32 values (5 bits), meaning the ideal curve will be further degraded

Andy  G4JNT