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

Gearbox reality check


The purpose of the "gearbox" you mention is simple and straightforward. However,
it should be noted that the gearbox is implementation specific to the PMA
service interface. The 66:16 gearbox converts 66-bit sub-frames and slices them
up into 16-bit chunks for transport over 16-bit interfaces such as the XBI
proposed as a PMA service interface:

Other 10 GbE PHY implementations may utilize a 66:6 gearbox or no gearbox at all
in the case that the PMA service interface is a 1-bit serial interface.

Best Regards,

Roy Bynum wrote:
> Wesley,
> I am afraid that the use of the term "gear box" belies the complexity of what is going on.  Just to do a reality check.  The so
> called "UniPHY" from which the "WIS" is derived now has four separate data clocking domains: common system @ ~ 10.0, open loop data
> transfer @ ~9.2, 64B/66B block encoding/SONET Payload Envelope (SPE) @ ~ 9.5, and the SONET signaling @ ~9.9.  Because the alignment
> between the SONET frame and the block coding is not on even byte boundaries there will have to be a major buffer to compensate
> between the ~9.5 SPE transfer rate and the ~9.9 SONET signaling rate, perhaps as much as 1.5 the SPE.  This buffer is also necessary
> to allow the sync overlap of difference between the 66 bit coding and 8 bit mapping within the SPE.
> Compare this to the already existing HEC framed WAN compatible PHY interface that Nortel already has working.  It only has three
> clock domains: ~10.0, ~9.5, and ~9.9.  If I understand it correctly it only has a frame buffer to obtain the size of the frames
> before mapping them.  Because the synchronization of the data is derived from the SONET and is on the standard byte boundaries,  a
> complex "gear box" is not needed.
> If Rich Taborek is to be believed there are two separate PHYs now.  One PHY is for the CWDM/Parallel PMDs with legacy 8B10B block
> coding.  Another PHY has 64B/66B block coding for serial PMDs. There is no such thing as a single PHY under current confederations.
> I am beginning to suspect that the "UniPHY" is actually going to "cost" a lot more than simply having three separate PHYs instead of
> only two.  ( This was a question that was put to the group at a previous meeting.)  I also think that it is going to take a lot
> longer to have working interfaces that can get to market if we continue to push for a "UniPHY".    The  "WIS" was a good idea, but
> it is turning out to be so complex that it may delay the adoption process.  There is already a WAN compatible PHY interface that has
> been demonstrated.  I suspect that a CWDM short read PHY based on 8B10B (XAUI) will be very easy to get working.  Without the "WIS",
> the 64B/66B LAN only PHY may be moderately easy to get operational as well.
> As a customer, if I am going to have separate PHYs for short reach and serial, an additional PHY for WAN actually makes would not
> add to the complexity of any decisions that I would be making.  In some respects, having totally separate PHYs for LAN and WAN makes
> it easier to support.   The complexity and training needed for serial LAN only installations is not complicated by the possibility
> of having WAN components.  Only for those installations that need WAN functionality do I need additional training for my support
> personnel.  Only for WAN installations do I need additional functionality in my network management systems, if the PHYs are
> separate.
> Given the closeness to the July meeting, I think that it is time for a reality check.
> Thank you,
> Roy Bynum
Richard Taborek Sr.                 Phone: 408-845-6102       
Chief Technology Officer             Cell: 408-832-3957
nSerial Corporation                   Fax: 408-845-6114
2500-5 Augustine Dr.        mailto:rtaborek@xxxxxxxxxxx
Santa Clara, CA 95054