We're using an AD9467 on a custom board running into a MicroZed board and are seeing strange results when running the ADC. the design is based on the AD9467_FMC project. Using the test software the IP core delay is successfully set, and all the output pattern tests pass, but when looking at analog signals we start to see strange behavior on the digital outputs... It looks like a couple of bits are flipping simultaneously and causing large amplitude excursions in the digital signal. Looking at all the test patterns on the ILA (at the output of the AD9467 IP block) confirms that the digital lines are functioning correctly, as do the patterns when the ILA is connected directly to the LVDS buffers within the AD9467 IP core. I've attached a plot of the data we received back from DMA for a sine wave as well as an ILA capture showing the bit-flips that seem to be a cause of the issue.
Data from DMA (plotted in matlab)
ILA capture showing the bits that are
It doesn't seem to be an issue with a single LVDS pair because bits 11 and 12 are on two separate LVDS pairs.
Any ideas what could be going on?