RE: [10GBASE-CX4] Cable / Connector pinout
I think it would be a mistake to use cables incompatible with a present standard. The XENPAK connector mentioned is a quite different style of connector, and the only advantage of using that pin order for CX4 would be that a board could be built without crossover to take the XENPAK signals to the CX4 connector. However, it seems most unlikely that the signals pesented to the XENPAK connector (whose intended destination is <10 cm. away) would be able to drive any useful length of the proposed CX4 cable, so some signal conditioning would be required in the module anyway. Thus there need be no correspondence between the CX4 pinout and the XENPAK pinout.
The only reasonable alternative to the IBx4 pinout would be one already in use on the XAUI test boxes built by IXIA, which is incompatible with any of the other proposals here, either in lane order or polarity, but is in use on the type of signals proposed for CX4.
Peter Bradshaw
Applications Manager
BitBlitz Communications
-----Original Message-----
From: DOVE,DANIEL J (HP-Roseville,ex1) [mailto:dan.dove@hp.com]
Sent: Tue 12/10/2002 1:58 PM
To: 'pat_thaler@agilent.com'; horie@fla.fujitsu.com; stds-802-3-10GBCX4@ieee.org
Cc:
Subject: RE: [10GBASE-CX4] Cable / Connector pinout
I agree with Pat on all points. Thanks Pat.
Dan
-----Original Message-----
From: pat_thaler@agilent.com [mailto:pat_thaler@agilent.com]
Sent: Tuesday, December 10, 2002 1:26 PM
To: horie@fla.fujitsu.com; stds-802-3-10GBCX4@ieee.org
Subject: RE: [10GBASE-CX4] Cable / Connector pinout
Horie,
The pin out reflects the way the wires run in the cable. For Infiniband
cable, the cable includes a "crossover" so that both ends have the same
connector pin out. For example, S1 on one end of the cable is wired to S16
on the other end so that one ends RX0+ connects to the other ends TX0+.
If one uses the pinout you suggest, the two ends of the cable have to have
different signal assignments (as in Ethernet 10BASE-T MDI and MDI-X
connector pin definitions). For 10BASE-T, this wasn't a big deal as there
are only four pins so it is pretty easy to swap pins where needed with an
MDI/MDI-X switch. For 16 bits and 3+ GigaBaud signal rate, this isn't
convenient and one should stick to having both ends use the same pinout.
Pat
-----Original Message-----
From: Takeshi Horie [mailto:horie@fla.fujitsu.com]
Sent: Tuesday, December 10, 2002 12:31 PM
To: 10GBASE-CX4 (E-mail)
Subject: RE: [10GBASE-CX4] Cable / Connector pinout
Dan,
I thought we dont have to follow IB pinout definition, although IB cable
is used. Do I miss something?
Better way is to have pinout order consistent with XENPAK so that one
SERDES chip can be connected to CX4 and XENPAK only with minor trance
change.
How about the following pin assignment?
IB 4X connector
------------------------
IB Pin Signal
------------------------
S16 TX3-
S15 TX3+
S14 TX2-
S13 TX2+
S12 TX1-
S11 TX1+
S10 TX0-
S9 TX0+
S8 RX3-
S7 RX3+
S6 RX2-
S5 RX2+
S4 RX1-
S3 RX1+
S2 RX0-
S1 RX0+
------------------------
TX*=IBtxO*
RX*=IBtxI
Thanks,
- Horie, Fujitsu Laboratories of America
> -----Original Message-----
> From: owner-stds-802-3-10gbcx4@majordomo.ieee.org
> [mailto:owner-stds-802-3-10gbcx4@majordomo.ieee.org]On Behalf
> Of Howard A. Baumer
> Sent: Tuesday, December 10, 2002 10:11 AM
> To: 10GBASE-CX4 (E-mail)
> Subject: Re: [10GBASE-CX4] Cable / Connector pinout
>
>
>
> Dan,
> I have Vol 1, Rev 1a and Vol 2, Rev 1. I didn't have
> Vol2, Rev 1a.
> After looking at Vol 2, Rev 1a all is fine and the cables match the
> spec.
>
> Thanks,
> Howard
>
>
> "DOVE,DANIEL J (HP-Roseville,ex1)" wrote:
> >
> > Hi Howard,
> >
> > I am looking at the spec and find that the
> > cable signals are consistent with my draft
> > and your stated IB Signal is different.
> >
> > The document I hold, Volume2 rev 1.0a has
> > change bars on those signals as if they were
> > the most recent change.
> >
> > This suggests you *may* be looking at an old
> > draft. (please oh please oh please!!)
> >
> > Can someone more familiar with this matter
> > please reply to this?
> >
> > Thanks,
> >
> > Dan
> >
> > -----Original Message-----
> > From: Howard A. Baumer [mailto:hbaumer@broadcom.com]
> > Sent: Tuesday, December 10, 2002 8:44 AM
> > To: 10GBASE-CX4 (E-mail)
> > Subject: [10GBASE-CX4] Cable / Connector pinout
> >
> > All,
> > I just noticed that the connections for the Cable
> assemblies we have
> > been dealing with do not match the connections in the IB
> specification.
> > Here is what I have determined:
> > IB IB Cable
> > Pin Signal Signal Difference
> > S1 IBtxIp(3) IBtxIp(0) X
> > S2 IBtxIn(3) IBtxIn(0) X
> > S3 IBtxIp(2) IBtxIp(1) X
> > S4 IBtxIn(2) IBtxIn(1) X
> > S5 IBtxIp(1) IBtxIp(2) X
> > S6 IBtxIn(1) IBtxIn(2) X
> > S7 IBtxIp(0) IBtxIp(3) X
> > S8 IBtxIn(0) IBtxIn(3) X
> > S9 IBtxOn(3) IBtxOn(3)
> > S10 IBtxOp(3) IBtxOp(3)
> > S11 IBtxOn(2) IBtxOn(2)
> > S12 IBtxOp(2) IBtxOp(2)
> > S13 IBtxOn(1) IBtxOn(1)
> > S14 IBtxOp(1) IBtxOp(1)
> > S15 IBtxOn(0) IBtxOn(0)
> > S16 IBtxOp(0) IBtxOp(0)
> >
> > The Cables IBtxI lane order is reversed from the IB
> Specification. I am
> > looking at "InfiniBand Architecture Release 1.0 Volume 2 - Physical
> > Specifications" page 160 & 161, table 24.
> >
> > Howard Baumer
>
>