Hi MrAl,
Does the filter have to be adaptive (related to output frequency)?
spec
Hi there spec and atferrari,
Well the data sheet on the AD9850 (the one i was really referring to) shows a passive seven pole elliptic filter. It actually works with a 125MHz crystal to get up to 40MHz out. I know the sine output is clean up to about 2MHz but i could not test any higher than that.
Yes i heard there were boards being sold with the wrong filters too, but not sure which ones. It also depends on what kind of accuracy and total harmonic distortion you are looking for. I know some frequencies wont be as clean as others but dont remember too much about this. There is an app note somewhere that explains the differences though and why this happens.
The output originates as a DAC output so it should be clean up to a certain frequency and then there will be more bumps as the frequency is tuned up higher. You'd need a good scope to see this. My scopes have limited bandwidth so i can only go so high.
I suppose an adaptive filter would work better but if you could tune it by hand you could set it for a clean output. For example, even a second order bandpass filter set to the required frequency would put out a nice clean sine. How you tune it would probably be with a variable capacitor and that would probably be limited too.
I dont think you would want to use this for audio unless you could verify the output was clean for every frequency you had it set for, but then again you would probably want to use a THD meter for that anyway.
One interesting note here is that i dont think you can get a triangle, if you need that. It's just sine and square. Also no pulse output at less than 50 percent duty cycle. If you buy a programmable frequency generator (under 100.00 USD) you get full programmability which even sometimes means you can program in a pulse train with different width pulses.
So this is partly about what you need and what you can get away with and what you want to spend. For me, i did not need super clean output for every frequency and if it was not clean for one frequency then i would just go to a different frequency and test with that instead. The nice thing was that the frequency could be tuned with something like 0.05 Hz resolution, so if you had set it for 10000000.00Hz you could increase it to 10000000.05Hz with a change of one bit in the setting code word. That's something like 0.0000005 percent resolution which is unheard of in the analog world.