We have designed a dual channel RF board based on the AD9910. We have manufactured 5 boards using two AD9910's on each board. Three of the five boards work perfectly. The other two boards suffer what appears to be an almost identical issue. One board has a channel with an output that is always a distorted 250MHz waveform (see attached) and is not the 10MHz sine wave we expect. We have various frequencies programmed into the profile registers and the part does not change when attempting to use any of the registers. We can query the registers of the DDS in the bad channel and they read back exactly what we expect.
The attached measurement shows a 10MHz sine wave in the good channel (Channel 2/turquoise) and the other waveform on Channel 1 (blue) is the non responsive channel and shows a distorted 246.4MHz waveform. We have verified all bias voltages are within specification and the 1GHz input clock is 1Vpp differential and looks good. We do not use the internal PLL. We have a second board which displays the same exact behavior except sometimes on power up one channel is bad and the other is good. Then after power cycling it can switch and the other channel is bad. Sometimes after power cycling both channels are good or both can be bad. The outputs always remain consistent after power up. Given there is no power sequencing required as per the specification and we are using identical PCB artwork for every board and 7 of the 10 DDS parts involved work every time we are at a loss as to what is going on. They all have the same code programmed into the parts. We have also carefully inspected every pin under high magnification on all the misbehaving parts and compared all the bias point for every external component to a good channel and see no issues. The parts do not run hot. So do we have bad parts? PLEASE HELP!