Replacing the DAC with a DDS, apart from the difference of interface, requires a bit more than a mechano/electro change.
The problem with using a DDS in band is the alias produced along with the modulated carrier. There are two classical solutions
A tuneable RF bandpass is incorporated to remove the alias from around the carrier. Tricky to tune and track.
Alternate: use a fixed frequency carrier, fixed frequency BPF, then a second synthesiser as an LO to up-convert to the require in band frequency.
There is yet another method, increase the sample rate. We note that as the sample frequency is increased the power density of the alias is reduced until it is lost in the backgound noise. We note that a sample rate of 8192K is sufficient to meet the 70db below carrier requirement for in band spurious.
Now due to the interpolation / decimation required for the stereo encoder we already have a 576K sample rate so logical would be to increase the sample rate to 8196k by further interpolation but this runs up against a speed problem with the DDS.
So use is made of the linear frequency profiles to carry out a linear interpolation. Since the sampling freqeuncy is so high compared withe maximum modulation frequency any interpolation will be linear with very little distortion. So load the interpolation values at a lower rate and let the DSS interpolate to simulate a higher sample rate.