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

RE: Management PMA/PMD registers for XAUI






Hi Boaz,

Pat will correct me if I am wrong but below she said "the 10GBASE-X PMA is
pretty much just the retimer" and if I turn it around to be "the retimer is
pretty much the 10GBASE-X PMA" I would therefore personally suggest that you
would implement the PMA register in that device.

Bye for now,
   David







Boaz Shahar <boazs@xxxxxxxxxxxx> on 11/04/2001 08:33:32

Sent by:  Boaz Shahar <boazs@xxxxxxxxxxxx>


To:   "'pat_thaler @agilent.com'" <pat_thaler@xxxxxxxxxxx>, Boaz Shahar
      <boazs@xxxxxxxxxxxx>, jgaither@xxxxxxxxxxxxxxx, stds-802-3-hssg@xxxxxxxx
cc:    (David Law/GB/3Com)
Subject:  RE: Management PMA/PMD registers for XAUI





Pat,
Suppose the retimer is there: SInce the retimer does not have MDIO port, and
PMD+RETIMER+DTE-XS is valid system, as you mentioned below, I guess that
using only the MDIO port of the DTE-XS is enough for compliancy, especially
if its all optional, isnt it? (That is, no need to implement 8B/10B PCS and
PMA address spaces for compliancy).
Regards,
Boaz

> -----Original Message-----
> From: pat_thaler@xxxxxxxxxxx [mailto:pat_thaler@xxxxxxxxxxx]
> Sent: Wednesday, April 11, 2001 2:54 AM
> To: boazs@xxxxxxxxxxxx; pat_thaler@xxxxxxxxxxx;
> jgaither@xxxxxxxxxxxxxxx; stds-802-3-hssg@xxxxxxxx
> Subject: RE: Management PMA/PMD registers for XAUI
>
>
> Boaz,
>
> Unless I've missed some change that occurred in D3.0, MDIO is
> an optional
> part of the standard. No sublayer is required to provide an
> MDIO. There is
> some text that recommends or requires access through
> proprietary means to
> some managment bits when an MDIO is not present, but I don't
> think there is
> such text for the PMA lock bit.
>
> Secondly, I don't think that a DTE-XGXS to PMD with nothing
> in between would
> be a valid implementation. The stack with a XAUI interface is:
>
>   RS -> DTE-XGXS -> PHY-XGXS -> PCS -> PMA -> PMD.
>
> When the PMD is 10GBASE-X, then the combination of PHY-XGXS
> and PCS is just
> translating a 4 lane 8B/10B signal to an XGMII signal and
> back into a 4-lane
> 8B/10B signal. It is obviously possible to remove most of the
> PHY-XGXS and
> PCS logic to just pass through the 8B/10B signal.
>
> BUT XAUI and the fiber media have separate jitter budgets and
> there isn't
> enough slack to combine them. One might get away with that in
> a very careful
> implementation, but not if one is going to use the XAUI as a
> compatability
> interface. (If you want to do an implementation without internal
> compatability interfaces and you meet all the specs at the
> MDIO, you can do
> your internal layering any way you want.)
>
> If one is doing a XAUI to 10GBASE-X PMD module, then there
> will be at least
> a retimer in it. The 10GBASE-X PMA is pretty much just the
> retimer. Rich's
> diagram in the presentation you mention doesn't get into that
> detail, but
> the words that are in the draft on the subject do mention the
> presence of
> the retimer.
>
> Pat
>
> -----Original Message-----
> From: Boaz Shahar [mailto:boazs@xxxxxxxxxxxx]
> Sent: Tuesday, April 10, 2001 12:22 PM
> To: 'pat_thaler@xxxxxxxxxxx'; Boaz Shahar; jgaither@xxxxxxxxxxxxxxx;
> stds-802-3-hssg@xxxxxxxx
> Subject: RE: Management PMA/PMD registers for XAUI
>
>
>
> Pat,
> I'm also thinking that the XGXS does not need  a separate
> Lock bit, and all
> its status contain enough info.  The question is if the
> standard requires it
> for compliancy or not.
>
> For example: in a system of DTE-XGXS+PMD (As depicted in Rich
> Taborek prsnt.
> from May 2000, taborek_3_0500 p. 8), where the XGXS
> implements both 10GB-X
> PCS and 10GB-X PMA, the compliancy requirement is to
> implement the address
> spaces of both the PMA/PMD and the 8B/10B PCS, or the
> register space of the
> XGXS is enough for compliancy of the system?
>
> Boaz
>
>
>
>
> > -----Original Message-----
> > From: pat_thaler@xxxxxxxxxxx [mailto:pat_thaler@xxxxxxxxxxx]
> > Sent: Tuesday, April 10, 2001 9:54 PM
> > To: boazs@xxxxxxxxxxxx; pat_thaler@xxxxxxxxxxx;
> > jgaither@xxxxxxxxxxxxxxx; stds-802-3-hssg@xxxxxxxx
> > Subject: RE: Management PMA/PMD registers for XAUI
> >
> >
> > Boaz,
> >
> > There aren't any separate PMA registers. They are PMA/PMD
> > registers and the
> > bits in them that apply to 10GBASE-X largely relate to PMD
> > functionality.
> > The one bit that is specifically PMA is 1.1.2 receive link
> > status which
> > tells whether the PMA is locked. XGXS has lane sync bits and
> > I don't think
> > it needs a separate lock bit.
> >
> > Pat
> >
> > -----Original Message-----
> > From: Boaz Shahar [mailto:boazs@xxxxxxxxxxxx]
> > Sent: Tuesday, April 10, 2001 11:23 AM
> > To: 'THALER,PAT (A-Roseville,ex1)'; Justin Gaither; 802.3ae
> > Subject: RE: Management PMA/PMD registers for XAUI
> >
> >
> > Pat,
> > The XGXS layer implements all the functions that are
> > associated with both
> > the 10GBASE-X PCS and 10GBASE-X PMA defined in clause 48. If
> > in the XGXS
> > case there is no need for the implementation of the PMA
> > registers (And I
> > think that there is NO need for that), why it is needed in
> the case of
> > 10GBASE-X PMA?
> > Boaz
> >
> > > -----Original Message-----
> > > From: THALER,PAT (A-Roseville,ex1) [mailto:pat_thaler@xxxxxxxxxxx]
> > > Sent: Tuesday, April 10, 2001 7:58 PM
> > > To: Justin Gaither; 802.3ae
> > > Subject: RE: Management PMA/PMD registers for XAUI
> > >
> > >
> > >
> > > Justin,
> > >
> > > As currently defined, there is only a single PMA in the
> > stack which is
> > > attached to the PMD. An XGXS doesn't have a separate PMA
> > > sublayer. I think
> > > we should leave it that way. We already fragment the physical
> > > layer into a
> > > lot of sublayers.
> > >
> > > Regards,
> > > Pat
> > >
> > > -----Original Message-----
> > > From: Justin Gaither [mailto:jgaither@xxxxxxxxxxxxxxx]
> > > Sent: Tuesday, April 10, 2001 7:19 AM
> > > To: 802.3ae
> > > Subject: Management PMA/PMD registers for XAUI
> > >
> > >
> > >
> > > When a 10GBase-X PCS is being used as a PHY XGXS, it has a
> > XAUI PMA.
> > > Should this PMA have all of the Management registers
> > > specified (ie. 1.0,
> > > 1.1, 1.2&1.3, 1.4, 1.5, 1.8, 1.9)? Or are these registers for the
> > > PMA/PMD that actually tied to the fiber PMD?
> > >
> > > Regards,
> > >
> > > justin
> > >
> > > --
> > > Justin Gaither                       Phone: 512-306-7292  x529
> > > RocketChips a Division of Xilinx     Fax:   512-306-7293
> > > 500 N. Capital of TX Hwy.
> > > Bldg 3                         email: jgaither@xxxxxxxxxxxxxxx
> > > Austin, TX 78746               WWW:   www.rocketchips.com
> > >
> >
>