Re: VXS & Signal Distribution

["C. Cuevas" <cuevas@jlab.org>]

Hall D Electronics:

Hi,

VITA 41.0 does not have backplane Payload to switch skew 
specifications.  Remember that the switch slot can be located in any 
physical slot in a star configuration, and any two physical slots for a 
dual star implementation.  For our JLAB 20 slot VXS Wiener/Elma crates, 
I would wager that the largest skew variation is between Payload Port 
17(slot 1) and PPort1(slot 9) for the 'left' side of the crate.  Since 
the backplane is symmetrical, pair trace lengths for Port 2 vs PPort 18 
*should* be the same.  The pair to pair skew is most likely in the Elma 
backplane specification and the folks from Wiener owe me that Elma spec. 
(Elma knows the copper trace lengths)

The protocol specifications:
41.1  InFiniBand
41.2 RapidIO
41.3 Gig Ethernet
41.4 PCI Xpress
41.5 Aurora
will most likely have different skew (max) specifications listed, and I 
believe the 22ns spec listing , but must admit that I do not know how 
they derive that spec.

Talk to you tomorrow,
Chris
~~~~~~~~~

Gerard Visser wrote:
> Hi Ben,
>     I agreee, it can't possibly be 22 ns! I don't know what document 
> this is from, but it is probably RapidIO specific. That's not really 
> applicable here. I would certainly hope/expect that the skew and maybe 
> even propagation delays are specified in ANSI/VITA 41.0-2006, the base 
> physical specification for VXS. I don't have that standard (yet). I 
> hope you guys do - can you take a look at what it says. (Don't send 
> out the whole standard document, that's probably a violation of its 
> copyright terms, I should spend the $25 to buy it. But if you see a 
> skew specification, I presume you could just restate it in an email, 
> will be helpful for the meeting tomorrow.)
>     Thanks,
>
>     Gerard
>
> Benjamin Raydo wrote:
>> Gerard/Ed,
>>
>> The attached image may be of use for tomorrow's discussion (this is 
>> the skew
>> spec referenced by VITA 41.2/VXS). Doesn't say much other than that 
>> we must
>> manually measure pair-to-pair skew on each slot of the backplan if we 
>> are to
>> guarentee a setup relationship between trigger and clock. I think the 
>> large
>> pair->pair skew (22ns max) in the spec comes from fiber links that
>> eventually get translatted into VXS copper pairs (I don't see how such a
>> large skew can develop on our "small" 20 slot backplane alone).
>>
>> Have fun,
>> Ben