I'm using an AD9954 as a linear swept reference for a fixed /50 PLL in a microwave FMCW generator. The FMCW generator sweeps over nearly an octave, so the DDS needs to do the same. Because of the wide operating & loop filter bandwidth, and the factor of 50 in reference to output frequency, DDS spurious must be kept to a minimum to prevent spurious signals in the output. To achieve this, I've been studying the layout & artwork for the 9954 eval. board, as I know it has good spurious performance (at least in comparison to my first board layout attempts!). I have two questions regarding the PCB artwork:
1. I noticed that no pin analog ground connections are made via the paddle, which is of course well connected to the analog ground plane. From a routing standpoint it seemed easier to simply run short tracks from the analog ground pins to the paddle, but I noticed during troubleshooting my board design that adding short haywires from 2 or 3 specific analog ground pins to the nearest ground via reduced the spurious levels by several dB. So despite the paddle being well via'd to ground perhaps this is not the way to go?
2. I see that all of the 0.1 uF bypass capacitors connect directly to the ground plane, as one would expect. However, pin 25 looks a bit odd because a track from the paddle runs over to the ground side of the 0.1 uF cap & ground via as well. Does this track between the paddle & the ground connection for the bypass cap on pin 25 provide some additional spurious signal suppression, & if so could someone explain why it does? Thanks.