USRP1, FPGA and ADC clock

I was wonder how the RX and TX data from/to ADC is read by Altera FPGA.
I see from the schematics that FPGA and ADC take the clock from the same
output of AD9513.
From Ad9862 I read that the ADC data is latched using CLKOUT1, but this
output is not connected
to FPGA, so at which moment is the data from ADC sampled by FPGA, with
which clock?
And the same question about latching the DAC data in AD9862 send from
I looked at the clock and it looks distorted and I wanted
to separate the ADC and FPGA clocks.
Is it possible to drive the FPGA and ADC from different clock sources?
different frequencies for FPGA and ADC allowed?
Do the clocks have to be synchronized? If not how much phase shift
ADC clock and FPGA clock is allowed?

The FPGA and ADC are run off of the same clock, directly from the main
oscillator. They need to be the same frequency. Why would you want to
use different clocks?


I use one of my USRP’s for other things than radio (ultrasound imaging)
and I my idea was it would be beneficial to run the FPGA at a higher
frequency than the ADC (n*ADC_CLK). I was writing my own FPGA code for
that and since CLKOUT1 and CLKOUT2 are not wired to FPGA, I just have no
idea how to synchronize the ADC data with the clock.

The reason for separating the ADC clock from the FPGA clock is that the
ADC clock is not clean enough for my experiments.

I also discovered that if I remove R1012 from “master_clock” and connect
it to any other output of AD9513 the USRP runs unstable. I get random
“ADC code jumps” when reading the data from ADC and that depends on the
length of the wire connecting the R1012 to the clock, which looks like
the ADC data is read by FPGA somewhere on the edge.

I didn’t investigate the original FPGA code but If really there is no
synchronization clock between the FPGA and ADC apart from the
master_clock then it means that it is only a coincidence that the USRP
works at all.

In this case whether the data is read correctly depends on the clock
track length from AD9513 to AD9862 and AD9513 to FPGA. Since it is
difficult to estimate the delays in the ADC one doesn’t really know in
which moment the ADC outputs the data in respect to the clock.

Is it true that the phase of the master_clock at FPGA and ADC is
important to get correct readings?

Sorry for a long email but I just try to understand the hardware and
whether I will be able to get my FPGA code to get working.