I've experimented with a couple of the cheap Chinese DDS boards, and have had endless problems programming them - either with PICs or Atmel chips. Most of the software that's around the 'net is either written for DDS ICs I've never seen, or just doesn't work! I'd be very interested in your programmer software and details of the "cheap" DDS board you use.
The CARDINAL programmable crystal oscillators look like a good, cheap solution. I wonder if I can get some "blanks" and programme them myself....?
I still favour my cheap 74HC CMOS approach, though. The parts are dirt cheap, widely available, and if you're really anal about frequency accuracy, you can phase lock your PLL reference to a "standard" like the BBC 198kHz (accurate to 14 digits!) or other reference frequency station. Incidentally, I use the 198kHz off-air to control the reference oscillator of my frequency counter, so when it reads 62XX.000000kHz, I know that the rig's bang on frequency!
Incidentally, I've been playing around with a Lulu of a rig on the bench, and 10 Watts carrier / 40 Watts peak is now pretty easy with a "car battery" supply. On a quick test into an inverted-V dipole here in the southern Netherlands, I got reports from Switzerland, Italy, Germany, the UK, Ireland and Spain. Signal strengths were described as "fading a bit" and pretty weak, but generally audible, and achieved S9 in Switzerland on peaks! I'm quite astonished at the coverage attainable by such a simple (and cheap) set-up. I used to just do Medium Wave (often through a ¼-wave "tower-block sloper"), and got some great daytime results on pretty low power, but now I'm hooked on SW and will definitely put on something reasonably powerful over Christmas. I had some big FETs arrive in the post this morning, and my target is 250W carrier / 1kW peak - just to prove that it can be done cheaply.
I'm considering ways to modulate the 250W beast, and PWM looks favourite, but I'm tempted to go for Outphasing - I have an idea for a modulator that generates square waves at 135° to each other (derived from different outputs of a ÷8 CMOS IC). These will provide the references for a couple of 4046 VCOs, which will be frequency modulated in phase opposition to each other by injecting the audio into the loop filters..... This makes the PAs relatively simple - it's just the output bridge that gets tricky. I've built outphased rigs before, but only ever with pairs of 807s or 813s in the outputs!