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

RE: ONLY one ref multiplier?: PMA clock reference




Joel, all

On reference clocks in general:

I completely agree that 155.52MHz and 622.08 MHz are good choices for the
WAN PHY. If we don't spec. two, we should probably do as the OIF and spec.
622.08MHz. This obviously implies a x16 CMU in the SerDes. So far, I think
we can find consensus. The differences are limitted to the LAN PHY case. In
the LAN PHY I would like to keep the door open for x66 CMUs with 156.25 MHz
REFCK. The rationale would be that the PCS needs the 156.25 MHz clock for
the 64b and 66b interfaces inside (some PCS implementations will generate
this themselves). The point is, that I want to make it possible (spec. or
leave open) to move part of the 66:16 gearbox solution into the SerDes/PMA.
If this doesn't help PCS and PHY design, I don't care. Again: LAN PHY only,
we already agree on the WAN PHY (I think).

On jitter transfer and FPGAs:

I'm very much in favor for keeping the door open for FPGAs. But I'm not sure
how it relates to this issue. The SerDes transmit clock is generated of the
REFCK. Jitter on the FPGA output is handled by the SerDes input. So any
jitter on FPGA outputs will not be transferred onto the line.

Kind regards,

Henning


-----Original Message-----
From: Joel Goergen [mailto:joel@force10networks.com]
Sent: 26. juni 2000 20:22
To: stds-802-3-hssg@ieee.org
Subject: Re: ONLY one ref multiplier?: PMA clock reference



Henning,

My thoughts are to keep the total jitter as low as possible such that the
available budget would allow fpga solutions to be implemented within the
scope
of the SPI4 for time to market reasons.  The fpga technology has
demonstrated
some incredible leaps over the last few years and looks really promising for
critical time to market applications using 16x622 rates.  In such
applications,
the jitter transfer at 155Mhz may not be acceptable.  Since most
transponders
entering the market appear to support two frequencies, I would be in favor
of
defining either a min/max range, or specifically two frequencies (perfer the
later).

The goal, of course, is to write a standard .... but I perfer to write one
that
does not make new implementations difficult.  If, during the creation of our
text, we can show a benifit in reduced jitter transfer, then we should
define
two clocks.  If we don't have two clocks, I am afraid we lose the
engineering
benifit of future implementations within an fpga environment.

At this point, it would be most benificial if an fpga vendor would discuss
the
possible limitations a 155Mhz clock might have on current and future
releases of
that technology.

Take care
Joel Goergen
----------------------

"Lysdal, Henning" wrote:

> Stuart, Justin, all
>
> I appologize if my answer to the fairly simple question by Justin:
> "Are there any that uses the 155MHz as a reference for OC192?"
> has confused some.
>
> My point was to show technical feasibility, so I'll give it a second try,
> and see if we can get the discussion back on track:
>
> Given the choice between a 155.52MHz reference clock and a 622.08 MHz
> reference clock most of the transceiver vendors (SerDes customers), I
know,
> CHOOSE 155.52MHz.
>
> It might be easier to get good jitter performance with 622.08MHz
> (644.53MHz), but it is also more expensive.
>
> Maybe we should start discussing which parts of the OIF spec. should be
> copied for Ethernet rather than going over the details of which
frequencies
> goes where. I guess we can all agree, we need RXDATA, RX_CLK, TX_DATA and
> TX_CLK. What else do we need to specify?
>
> Regards,
>
> Henning
>
> -----Original Message-----
> From: Stuart Brorson [mailto:sdb@axiowave.com]
> Sent: 23. juni 2000 21:56
> To: stds-802-3-hssg@ieee.org
> Subject: RE: ONLY one ref multiplier?: PMA clock reference
>
> I did not intend to "beat anybody silly".  Rather, I wanted to point out
> that designing 10 Gig circuits with very low jitter is *hard*, and the
> difficulty is compounded by using lower speed reference clocks.  My friend
> from (unnamed Danish company) unwittingly proved my point by announcing
that
> their 10 Gig SONET mux chip used a 155 MHz reference.  Unfortunately for
> him, that chip is known in the industry for having jitter problems.
>
> More importantly, my real point was that the chip vendors (and the
standards
> body by extension) should not restrict the designer to using a lower speed
> (i.e. 155 MHz) clock.  IMHO, designers want to choose either 155 or 622,
and
> that's the choice the 10 Gig SERDES chip designers correctly give them.
>
> In any event, I discovered after posting my initial message that the
> question of one vs. two clocks involved not the question of 155 vs. 622,
but
> rather how to synthesize both 9.95328 GHz  and 10.3125 GHz (WAN vs. LAN)
> with only one reference clock.  Read before you post, I always say!
>
> Just to put my two cents into this latter discussion:  Since LAN line
cards
> and WAN line cards often use different optics -- and are therefore
different
> designs -- I see no reason that one needs to generate 9.95328 GHz and
> 10.3125 GHz off the same crystal.  Different boards can have different
BOMs
> and call out for different oscillators.
>
> Stuart Brorson
> Axiowave Networks
> 100 Nickerson Road
> Marlborough, MA 01752
>
> -----Original Message-----
> From: Jonathan Thatcher [mailto:Jonathan.Thatcher@worldwidepackets.com]
> Sent: Friday, June 23, 2000 3:30 PM
> To: stds-802-3-hssg@ieee.org
> Cc: Geoffrey O. Thompson (E-mail); David Law (E-mail); Robert M. Grow
> (E-mail); Steven Haddock (E-mail); 'henning.lysdal@intel.com'; 'Stuart
> Brorson'; Fred Weniger (E-mail); Joel Goergen (E-mail)
> Subject: RE: ONLY one ref multiplier?: PMA clock reference
> Importance: High
>
> Warning!
>
> Excuse me, but me thinks that you (plural) have broken the "respectful and
> dignified" rule here.
>
> Apologizing in advance, frankly, doesn't carry much weight. Try this
> metaphor: "Sorry sir. But, I am going to beat you silly and take your
> wallet." Isn't a courteous mugger kind of an oxymoron? Perhaps this is the
> foundation of an insanity plea? I mean, you can't exactly plead ignorance,
> can you?
>
> Oh, and jumping on the guy while he is down sounds to me more like a
> "Central Park" incident than a technical discussion.
>
> Finally, running down the street screaming with your wallet in the air and
> cash falling out isn't especially wise either.
>
> It would best if everyone avoid the use of company names in all IEEE
> P802.3ae communications. Period.
>
> jonathan
>
> >-----Original Message-----
> >From: Joel Goergen [mailto:joel@force10networks.com]
> >Sent: Tuesday, June 20, 2000 2:15 PM
> >To: stds-802-3-hssg@ieee.org
> >Subject: Re: ONLY one ref multiplier?: PMA clock reference
> >
> >
> >
> >
> >As a guy who work with Stuart and others on these issues ...I
> >double the "Right
> >On, Brother!"
> >-joel
> >-------------
> >Fred Weniger wrote:
> >
> >> Stuart,
> >>
> >> As a SerDes vendor, I can only say "Right On, Brother!"
> >>
> >>   At 08:57 AM 6/20/00 -0400, Stuart Brorson wrote:
> >>
> >> >Please allow me to make a quick comment about 155 vs. 622
> >MHz clocks here.
> >> >I was involved in OC-192 IO card design at my former
> >employer, Nexabit
> >> >Networks (now Lucent Technologies), and have had some
> >experience in this
> >> >department.
> >> >
> >> >It is not my desire to disparage the fine products of Giga
> >here, so please
> >> >accept my apologies in advance.  However, the 10 Gig SERDES
> >products from
> >> >Giga (i.e. GD16555 and GD16554) had jitter gen problems, even on the
> >> >company-supplied test board.  Amongst other problems,
> >Giga's test board
> >> >incorporated a 155 MHz clock.  Designing a low jitter
> >PLL/SERDES chain is
> >> >not very easy.
> >> >
> >> >It is noteworthy that the OIF has speced a 622 MHz
> >reference clock freq for
> >> >the 10 Gig framer/SERDES.  That means that clueful PLL
> >vendors have every
> >> >reason to design low-jitter 622 MHz clock modules which can
> >be used -- or
> >> >modified for use -- with 10GigE also.
> >> >
> >> >In any event, most vendors with whom I am aware --
> >including Giga -- allow
> >> >the user to select either a 155 or a 622 MHz reference
> >clock.  This allows
> >> >the board designer freedom to choose the design problem he
> >wants to tackle:
> >> >either a lower speed 155 MHz PLL with stringent jitter
> >specs (and a very low
> >> >jitter SERDES), or a higher speed 622 MHz PLL with all the
> >intricacies of RF
> >> >design, but perhaps with an easier jitter (i.e. board
> >noise) problem.
> >> >
> >> >Why not allow two clock frequencies and leave the board
> >designer the freedom
> >> >of choice?
> >> >
> >> >
> >> >Stuart Brorson
> >> >Axiowave Networks
> >> >Marlborough, MA 01752
> >> >
> >> >
> >> >-----Original Message-----
> >> >From: Lysdal, Henning [mailto:henning.lysdal@intel.com]
> >> >Sent: Tuesday, June 20, 2000 5:17 AM
> >> >To: 'Jscquake@aol.com'; rtaborek@nserial.com;
> >stds-802-3-hssg@ieee.org
> >> >Subject: RE: ONLY one ref multiplier?: PMA clock reference
> >> >
> >> >
> >> >
> >> >Justin,
> >> >
> >> >In answer to your question:
> >> >
> >> >My company (formerly Giga) has been shipping OC-192 SerDes
> >since 1997 and
> >> >the majority of our customers use 155.52 MHz reference clock.
> >> >
> >> >Regards,
> >> >
> >> >Henning
> >> >
> >> >
> >> >-----Original Message-----
> >> >From: Jscquake@aol.com [mailto:Jscquake@aol.com]
> >> >Sent: 20. juni 2000 00:18
> >> >To: rtaborek@nserial.com; stds-802-3-hssg@ieee.org
> >> >Subject: ONLY one ref multiplier?: PMA clock reference
> >> >
> >> >
> >> >
> >> >
> >> >Hello Rich,
> >> >
> >> >Your proposal sounds good,i.e. to have only a single clock multiple
> >> >(1/4 division) for the reference clock, but I am not sure
> >if this is wise.
> >> >Using
> >> >a lower rate frequency clock autmatically implies worse
> >jitter performance
> >> >for
> >> >the PLL's. This is not as much of an issue for the WDM case
> >as it is for
> >> >serial
> >> >but every psec (or even sub-ps) counts for the serial
> >versions. So I would
> >> >opt
> >> >to be NOT too restrictive in saying only 155-156Mhz xtal
> >osc are allowed.
> >> >Note that the present community of OC192 people use the
> >higher clock rate
> >> >for the reference. Are there any that uses the 155MHz as a
> >reference for
> >> >OC192? Having said all this ... are there readily available
> >> >644.53125MHz xtal osc.?
> >> >
> >> >Justin
> >> >
> >> >In a message dated 6/16/00 1:03:26 AM Pacific Daylight Time,
> >> >rtaborek@earthlink.net writes:
> >> >
> >> > > Henning,
> >> > >
> >> > >  Sorry about the confusion. I did mention in my note
> >that there would have
> >> >
> >> >to
> >> > > be
> >> > >  two optional clock references specified in the XBI, one
> >for the LAN PHY
> >> >and
> >> > > the
> >> > >  other for the WAN PHY.
> >> > >
> >> > >  What I should have said is that only one clock MULTIPLE
> >be specified. For
> >> > >  example, 161.1328125 MHz is 1/4 of 644.53125 MHz and
> >155.52 MHz is 1/4 of
> >> >
> >> > > 622.08
> >> > >  MHz. One fourth is a good multiple to use. This means that other
> >> >multiples
> >> > >  should not be required anywhere in the standard, even
> >optionally (i.e.
> >> >1/8,
> >> > > 1/2,
> >> > >  1/16, 1/1, etc.)
> >> > >
> >> > >  Best Regards,
> >> > >  Rich
> >> > >
> >> > >  --
> >> > >
> >> > >  "Lysdal, Henning" wrote:
> >> > >  >
> >> > >  > Rich,
> >> > >  >
> >> > >  > I don't see how you can avoid having separate
> >reference clocks for LAN
> >> >and
> >> > >  > WAN (with realistic PLL design).
> >> > >  >
> >> > >  > In the LAN case there are several options
> >> > >  > 156.25 MHz (seems to be prefered among serial folks)
> >> > >  > 161.1328125 MHz
> >> > >  > 644.53125 MHz
> >> > >  >
> >> > >  > In the WAN case the OIF specifies 622.08 MHz. I know
> >of a lot of people
> >> >
> >> > > who
> >> > >  > also like 155.52 MHz
> >> > >  >
> >> > >  > Now the problem is: how do you synthesize 9.95328 GHz
> >and 10.3125 GHz
> >> >from
> >> > >  > the same reference. If you use a 10 kHz reference,
> >it's easy, but you
> >> >will
> >> > >  > most likely have problems with transmit jitter.
> >> > >  >
> >> > >  > So I haven't been discussing the WAN case at all,
> >since I was under the
> >> > >  > impression that WAN PHYs will use existing SONET
> >SerDes using 622.08
> >> >MHz
> >> > >  > refck.
> >> > >  >
> >> > >  > Regards,
> >> > >  >
> >> > >  > Henning
> >> > >  >
> >>
> >> Fred Weniger
> >> Gigabit Product Marketing Manager
> >> 805-388-7571
> >> fax: 805-384-5065
> >> weniger@vitesse.com
> >> VITESSE
> >> SEMICONDUCTOR CORPORATION
> >
> >--
> >Joel Goergen
> >Force10 Networks
> >1440 McCarthy blvd
> >Milpitas, Ca, 95035
> >
> >Email:  joel@force10networks.com
> >Direct: (408) 571-3694
> >Cell:  (612) 670-5930
> >Fax:   (408) 571-3550
> >
> >

--
Joel Goergen
Force10 Networks
1440 McCarthy blvd
Milpitas, Ca, 95035

Email:  joel@force10networks.com
Direct: (408) 571-3694
Cell:  (612) 670-5930
Fax:   (408) 571-3550