Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [802.3ae] PMA Loopback




Burt Christian wrote:

> Hi,
>
> According to the PMA loopback definition in [45.2.2.1.2] and [51.8],
> the PMA shall accept data on the transmit path and return it on the
> receive path.
>
> Another useful PMA loopback mode would be one where the input
> from the PMD service interface would be retimed and returned on
> the transmit path.
>
> Is this mode defined ? If not, why not ?

This mode is not defined due to the complexity involved.

10GE is defined as two asynchronous data paths - one in the
transmit direction and one in the receive direction. The two
paths operate at approximately 10.3125 or 9.95328 for LAN
and WAN mode respectively. The only thing we know about
the frequency of operation (at the PMA) is that Tx and Rx will
be within 200ppm (120ppm for WAN mode).

Why is this relevant? Because looping back the data is
relatively simple. However, in a line loopback mode, the
Tx data can no longer be timed off the Tx clock. To
ensure that all data is looped back it must be timed of the
recovered clock - i.e. loop timed.

When we mention loop timing, Roy Bynum gets offended.
Furthermore it assumes that the recovered clock has the
quality required to generate a Tx clock which meets the
jitter and wander specifications. Beyond this, you would need
to develop a messaging system to control line loopback
from the far end device.. As you can see, the complexity
of the PMA increases significantly to implement this function..

(Not that it isn't possible, its just beyond the scope of the standard).

--
Tim Warland  P. Eng.
Applications Engineer
Quake Technologies   (613)270-8113 ext 2311

If you can dream it, you can do it.