I'm starting a design using the ADV7513. The hardware user's guide recommends using LC filters on each voltage rail. This is primarily to achieve very low noise on AVDD and PVDD. The obvious source of the noise is transient current on DVDD. For the LC filters to do their job with the recommended values (10uH inductor and 10uF cap) there needs to be a series resistance over 1 ohm to prevent peaking that would actually make the noise worse around 16KHz. If the series resistance is in fact over 1 ohm, there would be over 100mv of DC drop through the inductor for DVDD. This is terrible voltage regulation. It eats up any margin left to keep within the recommended range for DVDD.
The ADV7513 evaluation board is not yet shipping. It is recommended that the ADV7612-7511 evaluation board be used (the Blackfin version). This uses 0805 10uH inductors. There is no part number shown that I can find. My guess is that the series resistance is significantly greater than 1 ohm for this small a part. Another evaluation board (ADV7511 ADV7343) takes a more conventional approach without LC filters (it does use EMI filters, but these use ferrite beads that work at much higher frequencies). There is a separate regulator for the ADV7511 rail.
I am inclined to go with the approach used in the ADV7511 ADV7343 evaluation board (separate regulator for DVDD, no LC filter). This avoids making what appears to me to be a very marginal design. The series resistance of the inductor will either be too high to get adequate voltage regulation or too low and get severe peaking at 16KHz. It would be possible to put the series resistance in the 10uF cap, but that has other problems. There are other LC values that might work for DVDD, but they do not involve a 10uH inductor with a 10uF cap.
I am assuming around 100ma for DVDD, and something around 25ma for AVDD. These are numbers from the ADV7511. There are hints that DVDD will vary considerably depending on data rates and processing selected. I hope that AVDD and PVDD do not vary nearly as much or have transients anywhere close to DVDD (which might require another separate regulators).
Am I missing something here? I need to complete this design in the next couple of days. It would be very nice to avoid using a separate voltage regulator for DVDD, but without more information I see no other way to proceed. If anyone has an analysis that shows that the LC filter design is the more solid than I have indicated here, I would very much like to see it.