I am using the AD9910 with the PLL locked at 800 MHz. The input clock is 40 MHz and I am using VCO4. It locks up OK until the AD9910 gets warm. I am measuring 55 to 60 degrees C case temperature with an IR thermometer. It seems a little hot but maybe its OK. It has the thermal vias under the belly pad. The value of CFR3 is 0x2438C128.
Anyway, I measured the PLL loop filter voltage (VCO tuning voltage) on both units on my board. It is low, 200 mV when locked and down to about 100 mV after loss of lock. the data sheet indicates 800MHz should be near the center of the tuning range, so I would have expected the VCO tuning voltage to be around 800 mV give or take. If I cool the part with freeze spray, the VCO voltage picks back up to about 200 mV and the lock bit comes on and I get good output as expected. After it warms up again, the voltage drifts down and it loses lock. Both the lock detect bit and the output look like it's out of lock.
I suppose I could force a different VCO but since this is to be a production board, I would like to know they will all work.
Has anyone experienced this kind of issue before? Any hints?