We want to generate a sine waveform of around 5-10Mhz for medical applications. The board was designed by a team member using AD9833 and 50Mhz clock. Now we realize that AD9833 gives a noise prone sine wave output at frequencys greater than 3MHz and it even fails to construct a valid output at high frequencies.
We do not want to redesign the board. Is it possible to over clock the AD9833? How much distortion will we get? Also are there any componet with similar footprint as AD9833 that can act as a replacement.
There are no alternatives to 10 pin pkg.
Filter the DDS output to reduce all nonharmonic spurs.
You can trying overclocking, but it may be no earthly use, device can block HI Fclk.
Max useful F out will be no more 0,6 x Fclk.
I think redesign will be inevitably with AD9850/51(as best choice)
Overclocking is never a good idea in production -- only for hobbyist for one off use. Even if you find a few part that works on your bench, others units could fail.
On the other hand, if you are willing to filter the output, you can still use the lower clock frequency of 30 MHz. E.g., when generating 5-10 MHz signal with 30 MHz clk, the output also contains the alias band of 25-20 MHz, at close to the same magnitude, plus higher frequency alias products. It would take a high order low-pass (or band-pass) filter to remove all those alias products.
A high-pass filter with pass band to 10 MHz and stop band starting at 20 MHz, say 40 dB down, would remove most "visible" alias products (on an oscilloscope). You can still see them on a spectral analyzer. If your have tougher surious requirements, it will require a more complicated filter. I've done some of these filter designs using the following book:
A. B. Williams, F. j. Taylor, "Electronic Filter Design Handbook", 2006 McGraw Hill, other editions will also do.
Sorry, used wrong clock freq. on previous post. Max. clock of AD9833 is only 25 MHz, so with signals from 5-10 Mhz, first alias appears at 15-20 MHz. The filter is a lot steeper now, transitiioning from passband of 10 MHz to a stop band of 15 MHz. The desired signal will also drop in amplitude due to the (sin x)/x function on sampled signals. It's still possible to design such filter, but it requires higher order and tighter component tolerance.