XRM(2)-DAC-D4/1G User Guide
V2.2 - Mar 8, 2018
Figure 6 : Low frequency clocking scheme
Here the data generation in the FPGA runs at the same speed as the DAC sample clock fed to the DACs, so the
FPGA generates 1 sample per DAC sample clock cycle instead of 4 samples at 0.25 *DAC sample clock cycle.
This removes the need for OSERDES, DDR registers etc and simplifies the interface to the DAC and the data
generation.
DCM alignment is no longer required since all data generation and output uses the same global clock. The
unconstrained delay which results from using the clock-capable input appears as a simple phase offset, which is
irrelevant since the DAC synchronises to the DCLK signal generated synchronously with the data via a toggling
bit; data still changes on each edge of this clock. The DAC DLL Bypass bit is normally set when running in this
mode since the operating speed is typically less than 125 MHz.
Clearly there is some overlap in the clock speed ranges which these two architectures can support which is also
dependent on FPGA speed; the user should choose the one best suited to the application. Code for this style of
operation is not included in the example code.
On Virtex6 and later boards, this restriction does not apply as there is no MMCM used.
2.12 Clocking on Virtex6, Kintex7 and Virtex7
On the ADMXRC cards using Virtex6, Kintex7 and Virtex7 FPGAs, clock distribution is simpler. Only one regional
clock input is used as the clock source, selected to be the clock pair which can drive the clock regions above and
the below the one directly clocked by the selected clock pair. This routed via a BUFMR and the BUFRs
(configured to divide by 2) to clock these additional regions and the OSERDES components. The incoming
clocks (at the half the DAC clock rate) are distributed via BUFIOs to drive the OSERDES components. The
master BUFR clock is also routed through a global clock buffer to provide the clock for the data generation
circuitry in the FPGA fabric.
Page 14
Hardware
xrm-dac-d4-1g-manual_v2_2.pdf