20MHz Clock distribution

Bob,

I think you run a huge risk of slow rise/fall times, signal reflections, and reduced signal swing by driving 24 components from one clock driver output. The device inputs themselves amount to 24x4pf, but the actual split in the transmission line for each of the 3 stacks will present a major load. I would DEFINITELY not recommend trying to drive this from one output. At the bare minimum you will find lots of jitter in the system, more than likely you will find some device inputs will not switch reliably at 20MHz anymore due to signal reflections (you likely would find double switching around every clock edge).

Instead what I recommend is to deploy at least a 1:6 clock driver, and drive each stack with 2 clock signals (4 components each from one clock output). Of course the safest solution would be driving from a 1:24 clock distribution but for 20MHz you might bet away with driving 4 components at the same time. Here a link to a potential device:

http://focus.ti.com/docs/prod/folders/print/cdclvc1106.html

Best regards, Fritz

Hi Bob, you are welcome.

Without calculating the total system budget I am not sure I know the answer. Here is my hunch:

The cable length depends on the type, the loss profile per unit length, and the ambient noise. With a shielded cable on a lab bench you would probably be fine driving several foot of cable. A shielded cable has a higher loss/inch but a much better noise immunity. Your real system will have VCC variation, temperature changes, weak/strong silicon, and clock jitter requirements you need to factor in, and you want to preserve a certain safety margin in your timing. Also, longer cable often imply a ground shift between your TX and RX unit, which will eat up your margin. Maybe a IBIS model simulation would give you a quick answer. Think of the parallel hard-drive cables inside a PC that was used to connect to one or two HDDs. These cables came lengths under 3 foot and they worked fine inside an unshielded PC.

The output impedance of the LVC1106 is 50Ω - so you won't need any output termination. Just be sure to minimize the cable length from the first to the last of the 4 receiver, as the first receiver will see an incoming wave of only 50% signal swing and only the reflected wave will get the signal up to 100% signal swing.

If you need longer cables, you may need to consider going with differential signaling.

Good luck. Fritz