We all know that if we want to get the best performance, we should place input clocks to our FPGAs on clock capable or dedicated clock pins.. When it comes to generating output clocks from an FPGA though, it’s not as simple as just outputting the clock on a IO pin to get the best performance .
To give us the best performance (e.g., lowest jitter, best duty cycle and timing performance) when working with Source Synchronous Outputs (SSO), we need to use a clock forwarding approach.
In a clock forwarding approach, we leverage the IO structure provided by AMD FPGAs, more specifically the Output Double Data Register (ODDR). The ODDR outputs data on both edges of the clock, and as a result, we can set the D1 input high and D2 input low. These are then output on every clock edge, creating an output clock at the frequency of the ODDR clock.
We can demonstrate this is working as expected by creating a simple test bench for the clock output module.
Most importantly, this ensures we keep the clock on global clocking resources, which we can observe by looking at the implementation view.
Zooming in allows you to see the ODDR register as it is described in the RTL.
By using the ODDR for clock forwarding, and the associated data leveraging IO flip flops the clock and data are aligned. As the delay through the ODDR and the IO flip flops is designed to be identical. As a result, the delays will be balanced against each other, providing better timing performance.
If we updated the design to remove the ODDR and simply output the clock and reimplement the design, we will see the ODDR bypassed and the global clock buffer being connected directly to the output pin. This uses local routing which results in more skew and jitter on the output clock signal. It will also contribute to the overall clock performance metrics and impact the overall device timing closure.
This means that the receiving device has to declare a larger clock uncertainty because the jitter is larger. If this is another FPGA, for example, it can make achieving timing closure more challenging.
When it comes to constraining the forward clock, we can use timing wizard to define the forward clock constraint or we can enter it by hand in the XDC file. When using the timing wizard, both create generated clock and forward clock constraints pages and result in the same type of constraint being created.
create_generated_clock -name clk_out1 -source [get_pins design_1_i/clk_output_0/U0/ODDR_inst/C] -divide_by 1 [get_ports clk_out1]
There are also a number of steps we can take to achieve the lowest jitter outside of clock forwarding. The first step is to use the PLL instead of the MMCM. To reduce the jitter on the clock, we should also try to physically and temporally separate it from potential noise sources.
Physically, we can try to isolate the clock pins from other signals (especially single-ended signals) which we can do by creating virtual grounds around the clock output. We can achieve a virtual ground during the design of the circuit board and FPGA by assigning the IO bank pins to drive out a logic low. Those pins are then connected to ground on the PCB design.
If, temporally, we are using the same clock internally then we should be using a different clock phase. The exact phase might vary from design to design, but we can see the impacts of using a particular clock phase by monitoring the VCCint ripple / bounce with the clocks set at different phases.
If you follow these clock forwarding rules, hopefully you’ll be able to increase performance when working with source synchronous outputs.
Workshops and Webinars
If you enjoyed the blog why not take a look at the free webinars, workshops and training courses we have created over the years. Highlights include
Introduction to Vivado learn how to use AMD Vivado
Ultra96, MiniZed & ZU1 three day course looking at HW, SW and Petalinux
Arty Z7-20 Class looking at HW, SW and Petalinux
Mastering MicroBlaze learn how to create MicroBlaze solutions
HLS Hero Workshop learn how to create High Level Synthesis based solutions
Perfecting Petalinux learn how to create and work with petalinux OS
Embedded System Book Do you want to know more about designing embedded systems from scratch? Check out our book on creating embedded systems. This book will walk you through all the stages of requirements, architecture, component selection, schematics, layout, and FPGA / software design. We designed and manufactured the board at the heart of the book! The schematics and layout are available in Altium here Learn more about the board (see previous blogs on Bring up, DDR validation, USB, Sensors) and view the schematics here.
Sponsored by AMD
Comments