According to most DDS datasheets, they can produce up to half the maximum frequency. for example for AD9850, max clock is 125 MHz and it can produce up to 62.5 MHz. But in higher frequencies, the number of DAC steps is reduced and the wave form will start having steps until it is not really a clean sine wave. Is there any equation ( for example fclock/4 or something ) that shows the maximum frequency in which DDS produces most reliable sine wave?

  • \$\begingroup\$ No, "most reliable" is not an exact expectation. If you can define a criterion (mathematically), then probably an equation can be constructed. \$\endgroup\$ Oct 12, 2013 at 9:15
  • \$\begingroup\$ That's why you need to post-filter the dac to remove the higher frequency energy content and also use a sinc compensation filter to normalize the amplitude. It's impossible to reconstruct a sinewave with only two samples and that's a fact. Anything more than two samples is do-able but the more samples the better. \$\endgroup\$
    – Andy aka
    Oct 12, 2013 at 9:29
  • \$\begingroup\$ @Andyaka I read the article about sinc compensation you kindly provided in another question. It was written by Maxim and pointed MAX265.As I have no experience on CIC ICs, Should I use that IC or I can make any active filter I want? \$\endgroup\$
    – Aug
    Oct 12, 2013 at 10:09
  • \$\begingroup\$ The maxim article was just an explanation of what you need to do to normalize the DDS chips output level as frequency gets higher. Any filter type will do - it doesn't apply to maxim or anyone else's chips. \$\endgroup\$
    – Andy aka
    Oct 12, 2013 at 10:23

1 Answer 1


This picture is crucial to understanding what is happening: -

enter image description here

Analog Devices use the example of a reconstructed 20MHz signal (F\$_{OUT}\$) on the diagram and as you can see, at 20MHz it's amplitude is a wee bit less than what it would be if it were much lower in frequency. By "much lower" I mean something like 1MHz and that would have been represented by a vertical line much closer to the Y axis. Here's a drawing showing just that: -

enter image description here

As you can see the out-of-band artifacts are much lower and the first artifact (at 99MHz) is much further away from the 1MHz fundamental frequency. This makes it: -

  • Easier to filter-off the artifacts because the gap between 1MHz and 99MHz is massive compared to the original AD drawing where F was 20MHz and it's first artifact at 80MHz.
  • The amplitude of the 1MHz sinewave is virtually unity whereas the amplitude of the 20MHz signal has fallen a couple of dB (or so) compared to the amplitude at 1MHz.

Filtering is therefore important; not only do you need to remove the high frequency artifacts in order to get a decent sinewave shape but, you have to compensate for the DAC being an imperfect device that progressively attenuates the wanted signal as you approach the reference clock frequency.

This is normal for all DACs that I'm aware of. A "perfect" DAC would prodice a single impulse of energy at each sample point with the energy level equating to the amplitude and this would give a flat pass-band and not the sinc envelope associated with standard DACs.

  • \$\begingroup\$ So it means the decent maximum frequency that can be obtained is completely dependent on the output low pass filter?.Higher frequencies can be obtained if one can have a good LPF with steep Bode diagram on the target frequency and then compensating well the attenuated signal amplitude . It this conclusion correct? \$\endgroup\$
    – Aug
    Oct 12, 2013 at 13:11
  • 1
    \$\begingroup\$ That conclusion is correct at all frequencies other than 100MHz and multiples. It's also worth noting that you can generate frequencies above 100MHz but you need a band-pass filter to remove the fundamental and tune in to the harmonic you require. It's all there in the picture but of course trying to recover anything close to 50MHz is a real pain (say 40MHz) because there will be an artifact at 60MHz and filtering this out requires complex analogue filtering that moves a notch filter around to kill the unwanted artifact. \$\endgroup\$
    – Andy aka
    Oct 12, 2013 at 13:20

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.