The other day I realised, looking at the THS4631 (output amplifier) datasheet, that the frequency response with a gain of 75 was going to be awful. I thought about maybe having a quad op-amp, each one multiplying by the 4th root of 75 and feeding into the next, but that was a no-hoper too. Finally, I think I have found a solution thanks to this thread on the ever-useful EEVblog forum:

The amplitude of the AD9850 varies with frequency from 1V pk-pk to 300mV or less. I needed some sort of AGC to correct the amplitude to be stable for all frequencies, and the answer was the AD633, a variable gain amplifier. This will output a 1.4V pk-pk sine wave at all frequencies with a few external components. This will then go into a multiplying amplifier: the ADL5931. This takes the sine wave and multiplies it with the output of a 12-bit DAC, giving an output amplitude of 1mV to 4V pk-pk. Finally, this will be combined with a DC offset from another DAC and amplified by two consecutive amplifiers (one is the THS4631, other TBC) with a gain of 2.5 and 2 in that order.