pll come out of lock state after spi read

We have not observed such behavior before. I think you might be writing accidentally to some bit that cause a change in device configuration. Can you be sure your read sequency is implememted properly? Can you check your phy layer really sees the sequence required? 

See that you can write to a register, and read the change you've wrote. I recommend you read one register at the time, and find out which register causes this and report back to us your settings.

In the past we had a couple of people report that they could not read from the device unless they adjusted the device address. However, they found a workaround. We asked these customers to send us a scope screen but never received any - so I am not sure if they had a programming issue of some sort. Here is a sequence they used and it worked for them. Can you try this as well and see if it's something similar? This device ships to a large number of customers, and nobody else ever reported such behavior.

Sequence these customers found to work:

The Address requires a shift of one bit for the Reads to work.  So, To execute a read operation we send the following where “AAAA” is the address we want to read, “0000” through “1000”:

MSb_________________________________LSb

“0000 0000 0000 0000 0000 000A AAA0 1110”

By the way, check out this app note for help on the SPI/I2C IF:

http://e2e.ti.com/support/clocks/m/videos__files/180965/download.aspx

Thanks for the application note..will verify accordingly :)