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

Re: Clause 45: MDIO Electrical Specifications


I submitted a comment to D2.1 on this issue, included again at the bottom 
of this note.  The comment is that D2.1 does not reflect the ad hoc
decision that was reported in Irvine, namely to adopt JEDEC JESD8-11 
for MDC/MDIO. The suggested remedy is that we simply reference
JESD8-11 (, "Free Standards") as the MDIO interface.

For D2.2 (Friday?) and D3.0 (March meeting) I suggest leaving text as is,
"complete" in some sense, or accepting my comment and resetting to the 
ad hoc recommendation and referencing JESD8-11 directly.  Additional
tweaks can wait for D3.0 comments.

FYI, my recollection of why JESD8-11 is proposed:  

-- In earlier meetings, the consensus was that the Clause 22 electrical 
   interface was no longer practical (voltages too high).

-- In earlier meetings, the consensus was that 10G did not want to 
   spend energy defining new "logic families", that is, non-standard 
   interfaces, for XGMII and MDIO, since there seem to be plenty of
   digital interface standards to reference. (We seem to be spending 
   energy now). An ad hoc was formed to report a recommendation at Irvine.

-- Many comments were made that the new interface should last for 
   many years, not just a solution for 10G only. (since MDIO
   may need to connect to phy's of different generations.  Changing 
   addressing and electrical now, but not again for a long while
   since three generations of MDIO in one box is unthinkable.)  

-- HSTL was repeatedly adopted/affirmed for XGMII interface.  As a practical
   matter, systems with XGMII interface will have a 1.5V supply available.  
   No other P802.3ae interfaces infer a supply voltage.  
   (OK, as a practical, practical matter, it seems that at least for 0.18um
   parts, the "HSTL" XGMII interface can be designed to be powered at 
   either 1.5 or 1.8V, even though standard only says 1.5V.  In non-standard 
   systems using 1.8V "HSTL", perhaps only 1.8V supply would be available,
   but then I would expect those systems to design MDIO to operate at
   either 1.5 or 1.8V, also outside standard.)

-- JESD8-11 provides two approaches for future proofing:

   ** If P802.3ae says nothing about "Normal Range" (Vdd 1.5+/- 0.1V)
      versus "Wide Range" (Vdd 0.9-1.6V), then every Clause 45 MDIO
      must support operation at interface Vdd=1.5V, insuring compatibility.
      Part vendors could choose to support wide range, and system
      implementers could select wide range parts, and operate the interface
      on any supply between 0.9 and 1.6V.

   ** If P802.3ae requires, in addition to JESD8-11, specifically 
     "Wide Range" support, then compatibility is also assured, since every
      Clause 45 MDIO interface would operate with an interface supply between
      0.9 and 1.6V. (i.e. 1.5V+/-0.1V, down to 1.0V+/-0.1V).

-- XGMII spec came from JEDEC (JESD8-6), JESD8-11 was available and seemed to meet
   the consensus goals, it was proposed in ad hoc, and despite prodding from Ed, no 
   alternatives were proposed, recommendation (by default) was reported 
   at Irvine.

From (unapproved) Irvine minutes, I only find:
"Mr. Turner gave a quick ad-hoc MDIO report. A number of individuals from both the
802.3ae and 802.3af committees worked on the new electrical interface specification and
they selected an interface with an expected long life. This interface will be compatible
with lower voltage devices that emerge in the future. They recommended the JEDEC
standard JESD8-11 interface. The ad-hoc work is finished, Ed proposed
these ad-hoc conclusions be added to the next version of the draft standard."

Do I think JESD8-11 is perfect "as is"?  No - while Normal Range 
requires >2mA drive, Wide Range only requires an anemic 100uA (as does table in D2.1).
Though 2.5MHz is slow, perhaps we should still say something about expected loading, 
desired range of edge rates, etc.  But I'm confident part vendors would make good 
guesses here.  For the standard, I'm just suggesting that we start back at 
JESD8-11, and make any *needed* refinements as D3.0 comments.


Edward Turner wrote:
> Rick,
> I support your effort to bring forward these potential issues to the group at
> large.  I also encourage further reflector discussion if people feel there is
> something broken in the draft.  It is important that we make the correct
> decisions when these items come up for vote at the meeting.
> I was not sure if you were at the meeting and wanted to explain some of the
> background thinking that had gone on 'off the reflector' during the ad-hoc and
> at the Irvine meeting.
> I also encourage any of you who are attending the meeting and have points to
> raise on the MDIO electrical interface to come along to the Clause 45 sub track
> so that we can hear your concerns when we discuss the electrical interface
> comments.
> Regards
> Ed

Clause: 45
Page: 45.222
Line: 15
CommentType: DISAPPROVE (Technical)
Resolution of Draft 2.0 Comment 1115 ""adopt[ed] an instance of the
JESD8-11 standard with a VDD of 1.2V""  JESD8-11 (, ""Free
Standards"") was selected by an apathetic ad-hoc.
However, JESD8-11 does not support 1.2v only operation.  The choices,
Normal and Wide range, are mentioned in the title of JESD8-11
2.2.1 Normal Range (1.4 to 1.6V Vdd)
2.2.2 Wide Range (0.9 to 1.6V Vdd)
In either case, the sending and receiving Vdd for this interface must
track within 0.1V (Note 1 in 2.3.1 and 2.3.2)
An MDC/MDIO implmenter could select to support either Normal or Wide
range since interoprability at 1.5V is maintained.  If all connected
parts in a system are wide range, a supply lower than 1.5V nominal
could be used.
When an XGMII (HSTL JESD8-6) is present, 1.5V will be available.
Change 45.4.1 to read
""The electrical characteristics of the MDIO interface are defined
in JESD8-11.  Pin input capacitance is limited to 10pF maximum.""
Retain NOTE, change to read ...""Vdd of 1.5V""
Delete Table 45-41.
Annex 45A, change 1.2V Vdd to 1.5V Vdd throughout.