RocketIO decoupling

The RIO User Guide (UG024) stipulates that the RIO supply pins need to be filtered using a Ferrite bead and a capacitor. It's very specific about the bead being either of 2 Murata parts and the capacitor being a 0.22uF 0603 device.

The ML300 board however doesn't adhere to the capacitor recommendations and uses a larger device. How important is it to stick rigidly to the User's Guide, does anyone have any relevant information on this?

I'm having problems finding a supplier of the ferrite beads. Similar question: is it really so important that I use either of the 2 parts stated in the UG?

TIA.

Rog.

Reply to
Roger
Loading thread data ...

the

I'd guess that's because they didn't try any other parts, so they can't be sure they work. However, I seem to recall the latest advice was a 1uF cap. Are you sure you have the latest version of the user guide? You should know that above 50MHz, 0402 and 0805 caps (~0.4nH inductance) have slightly less impedance than 0603 ones (~0.5nH), and the uF value makes very little difference. The inductance of the package dominates the impedance. I use 0402 parts because they're smaller, so more space for other goodies!

and

That depends on whether you think you'll need support from Xilinx. If you go your own way, they may well say "come back when you follow the design guidelines" before they help. You can see their pov, there are a lot of designs to support, and an FAE who can support Gbit designs is expensive.

stated

Almost certainly not, provided you find a close-ish equivalent. But I refer you to my previous comment about support.

In a way, if only because you felt the need to ask the questions, I'd suggest you follow what the user guide says as best you can. When you've got it working that way, why not play about with different parts and values to get a feel for what is and isn't important? You could experiment with the ML300 board. I also recall that Xilinx reckon the ML310 board is a better example of what to do.

Cheers, Syms.

Reply to
Symon

There are lots of things that sometimes you "can get away with" on MGT use. What you get away with depends on how fast your MGT is going to run. The decoupling strategy affects the power supply noise and has the effect of creating jitter within the MGT or output signals. In short it limits the maximum data rate you can achieve.

Worth also saying is that some packages of Virtex2-Pro have capacitors built in if that is what you are using.

We have bought both sizes of these parts from DIGI-KEY. There are in the US but delivery to UK is usually 2 days. If you run into issues getting them there email me on our Support email, listed on our website, and I can probably find a few for you.

John Adair Enterpoint Ltd. - Home of MINI-CAN. The low cost Spartan3 CAN Bus Development Board.

formatting link

Reply to
John Adair

Roger,

You are not required to use the 2 Murata ferrite beads that are mentioned in the user guide. What you should use are ones with similiar or better characteristics of low DC resistance (600 ohm @ 100 MHz). TDK for instance has similiar versions and Murata has over versions that we don't mention such as the BLM18EG601SN (0.35 ohm @ DC, 600 ohm at 100 MHz).

The use of the external capacitor is only required for the wirebond (FG) packages. The flipchip (FF) packages have this capacitors inside for Virtex-II Pro and Pro X.

Please don't use the ML300 as an implementation guide for RocketIO it was developed very early in the life cycle of Virtex-II Pro and as a piece of hardware it is very expensive to revise it when new knowledge is learned. In contrast, the ML321, ML323 and ML325 boards are up to date with the user guide.

Ed

Roger wrote:

Reply to
Ed McGettigan

Thank you all for you help. I can make some more progress with my design now.

Rog.

Reply to
Roger

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.