Optimizing Arduino and the AD9851 DDS signal generator
While this article concentrates on the AD9851 device, Analog Devices produces many others that operate in a similar fashion. The devices combine a clock reference with a digital divider and a phase locked loop to provide a sinewave output with very fine resolution. In the case of the AD9851, the digital divider is 32 bits and the clock frequency is typically 125 MHz. This results in a frequency resolution of approximately 30 milli-Hz. A single resistor, Rset controls the current output level and therefore the output voltage level.
The digital nature of the DDS process results in an output signal and many signal images. The images follow a sin(x)/x envelope. This envelope is shown in the Analog Devices website and is shown in Figure 1.
From this figure it is easy to see that as the output frequency increases the first system image and the output frequency move closer together, while at the same time the amplitude of the image increases. The AD9851 DDS board is shown connected to an Arduino in Figure 2. The Arduino is used to set the DDS output frequency. A DC block, the Picotest J2130A (shown in the figure, www.picotest.com) or the P2130A 500Hz - 8GHz Blocker, is used to connect the DDS output to a 50Ω oscilloscope port to view the spectral response. The unfiltered AD9851 output is measured so that we can see the digital spectrum rather than the filtered output.
The output frequency is set to 3 MHz and the unfiltered output spectrum is shown in Figure 3. As expected we can see the 3 MHz fundamental as well as the first and second images at 122 MHz and 128 MHz, respectively. The clock feedthrough is also visible in this measurement at 125 MHz. The images are far from the 3 MHz fundamental making them simple to filter using a low pass filter.
The output frequency is set to 50 MHz and the unfiltered output spectrum is shown in Figure 4. Now we can see the first and second images appear at 75 MHz and 175 MHz respectively, as expected. The first image amplitude is only 2.5 dB lower than the fundamental and also less than an octave from the fundamental, making it difficult to filter.
Further increasing the frequency to 60 MHz the unfiltered output spectrum is shown in Figure 5. Now we can see the first and second images appear at 65 MHz and 185 MHz respectively, again as expected. Now the first image amplitude is only 0.5 dB lower than the fundamental and very close to the fundamental, making it even more difficult to filter.