RE: Minutes of serial PMD specs telecon 17 Oct 00

[Rohit Mittal <RMittal@xxxxxxx>]

 

 

Rohit,

Thanks for the reply.   (I have fixed the email address for
the serial pmd exploder in this reply).

I did realze that the  RJ was not concentrated at a single frequency; it is a filtered noise source and is
broadband.  I didn't realize that the DJ was also broadband; I thought that this, being deterministic, was applied at a single
frequency.   However, your answer has prompted several new questions.

First, when we apply sinusoidal jitter at a given frequency below 50 MHz, do we apply the broadband DJ and RJ simultaneously in the same test (so that the sinusoidal jitter of a given frequency, broadband RJ, and broadband DJ, are
all applied together)?  Or, is the sinusoidal jitter applied alone in one test (actually, many tests with one for each frequency)
and then the DJ and RJ are applied together in another test?  

Above 50Mhz, you apply all simultaneously. Its true that sinusoidal will be at one frequency, whereas teh others are going to be broadband. That way you get total of 0.7UI jitter applied(.1UI for sin, .38UI for DJ and .22UI for RJ). I don't think you test thoroughly at all points for sine above 50Mhz. Just take fc/2 as the worst case to apply the .1UI for sine jitter 

Second, how is the DJ actually generated?  I ask this because, as I mentioned above, I initially thought the DJ applied had a single frequency (similar to the sinusoidal jitter)?  I presume the DJ signal, even though it is broadband, is deterministic.  

I don't know the exact answer to this. Some of our vendors said that they followed the 802.3z method of using some cable length, plus filter after the pattern generator. Basically DJ arises because of ISI, so you need some way of limiting the bandwidth of the system. That, I believe, is the reason its broadband.

That brings up a good point, why does the FC spec say that the majority of the DJ is concentrated at fc/2. I don't know (any comments appreciated)

Third, is the filter that the DJ and RJ are passed through before being applied to the system a bandpass from frequency
(line rate)/1667 to (line rate)/2 (i.e., approximately 6 MHz to 5 GHz for a 10 Gbit/s line rate or, equivalently, 637 kHz to
531 MHz for Fibre Channel as indicated in Table 3 of MJS-2)?   

Thanks for the information on this.

Regards,

Geoff Garner

Rohit Mittal wrote:

I hope I got your question right.There is a difference between applying sinusoidal jitter and RJ/DJ. When you apply sine jitter you apply jitter at each frequency below 50MHz eg. you add xx amplitude sinewave (f=20Mhz) to the clock which is used to generate the data and see if you can tolerate that jitter. Then, you apply yy amplitude sinewave (f=40Mhz, say) to the clock .... and see if you can tolerate the jitter. (This is similar to Sonet). (Note:xx,yy are given by the 20db/decade sloping graph in the spec).Above 50Mhz, you apply a white noise source. You filter this white noise source. So, basically you are applying noise power. You do not put zz amplitude of white noise at one frequency, say, f=100Mhz. (zz=.22UI, if I remember correctly). Presumabely, if all the RJ and DJ was concentrated at one frequency point, say, f= 100Mhz , you would fail to pass the jitter.So this is the difference between sinusoidal and RJ/DJ

-----Original Message-----
From: Geoffrey Garner [mailto:gmgarner@xxxxxxxxxx]
Sent: Friday, October 20, 2000 1:52 PM
To: Rohit Mittal
Cc: DAWE@xxxxxxxxxx; PIERS (A-England,ex1)'; 'Öhlén Peter'; 'Jonathan Thatcher (WWP)'; 'Petar Pepeljugoski (IBM)'; 'Scott Bradshaw (Roshnee) '; 'Vipul Bhatt (Finisar)'; 'Bob Hughes (Telect)'; DOLFI@xxxxxxxxxx; DAVE; 'Ed Cornejo (Lucent) '; 'Jack Jewell (Picolight)'; 'Jason Yorks (Cielo)'; 'Joey Jarriel (Cielo)'; 'John Dallesasse (Molex) '; 'Marc Verdiell (LightLogic)'; 'Mike Dudek (Cielo)'; 'Paul Kolesar (Lucent)'; 'Robert Coenen (Gtran)'; 'Schelto.Van-Doorn (Infineon)'; 'Steven E Swanson (Corning)'; 'Tad Tszostak (E-mail)'; 'Twhitlow (E-mail)'; 'Peter Stassar (Lucent)'; 'Mike Hackert (Corning)'; Buchheit@xxxxxxxxxx; Steve'; 'Chris Simoneaux ( Picolight)'; 'Tom Hanson (Corning)'; 'John Abbott (Corning)'; 'Krister Fröjdh (Optillion)'; 'David Hinzel (ETA)'; 'Walter Thirion (Jatotech)'; CUNNINGHAM@xxxxxxxxxx; DAVID; E-mail; gmgarner@xxxxxxxxxx
Subject: Re: Minutes of serial PMD specs telecon 17 Oct 00

Rohit,

I am replying to the distribution list and the new serialpmd reflector.  I apologize if
some receive 2 copies of this; I am replying to both because apparently not everyone is signed up for the serial pmd reflector yet.

The jitter measurement lowpass filter cuttoff, at high frequency, for SONET/SDH was originally chosen for SDH to be a frequency sufficiently high that it would include all significant alignment jitter, but also not be so high as to be beyond what test sets are capable of.  They chose this cuttoff (represented as f4 in G.825)
as 10 times the regenerator 3 dB bandwidth for the respective rate.  This was initially done for the lower rates.  As higher rates were specified, the various filter bandwidths generally scaled from the lower rates.  The frequency f4 is 80 MHz for STM-64 (10 Gbit/s), 20 MHz for STM-16 (2.5 Gbit/s), 5 MHz for STM-4 (600 Mbit/s), and 1.3 MHz for STM-1 (155 Mbit/s). Note that the scaling by a factor of 4 is not always exact; sometimes the numbers are rounded.

It is true that in a real network, the jitter present above 80 MHz for STM-64 (or, in general, above f4 for each rate) is expected to be small.  However, it seems it is still necessary to specify this bandwidth for test purposes to guarantee that results are reproducible.  In addition, if broadband jitter is being applied to a piece of equipment as in MJS-2, the results will be impacted by how high in frequency the broadband jitter extends because, for the same power spectral density amplitude, increasing the highest frequency means more jitter is being applied.

Related to this, I had mentioned in the conference call that I recalled a 50 MHz high frequency cuttoff, but needed to track down the document (I first came across the number in an offline discussion).  I believe I have found the source of this, and also some more complete information.  In MJS-2, Tables 3 (page 20) and 5 (page 23), and also in Tables 10, 14, and 30 of FC-PI (Rev 9), sinusoidal jitter applied in the tolerance test is swept to a maximum frequency of 5 MHz.  Since for 10 Gbe the line rate is about a factor of 10 higher, the corresponding high frequency cuttoff would be 50 MHz. However, on looking more closely at these tables, I do see that for deterministic jitter and random jitter the high frequency cuttoff is line rate (fc)/2.
For 10 Gbit/s, this would be 5 GHz which is larger than the 80 MHz by a factor of 62.5.  This could certainly affect the test results (i.e., applying jitter up to 5 GHz versus 80 MHz).  Have I correctly interpreted these high frequency cuttoffs?

You indicated you have more information on this; I would be interested in it.

Thanks.

Regards,

Geoff Garner
 

Rohit Mittal wrote:

 

A few comments/questions below:

a) Is it true that 10GFC has a band pass for jitter generation measurement. Where did that come from. FC doesn't have it, just like GbE (and I presume 10gbE)

b) I'll be curious to know where the upper limit - 80MHz in the case of Oc192 came from. My measurements have shown that the upper limit is pretty arbitary. Even GR253 (the core Sonet spec) says that " the amount of frequency above the low pass cutoff frequency is expected to be very small" pg. 5-87. I don't know if anyone else can confirm that they saw the same thing when they measured jitter at any Sonet rate.

c) I'm not sure if I got action item 001017.3 correct, but the way we (Sonet equipment vendor) generate jitter is exactly the way specified in FC MJS-2. If anyone wants more info, I can gladly supply that

Thanks
Rohit

ps: Thanks a lot for the detailed minutes. It lets me keep upto date.

-----Original Message-----
From: DAWE,PIERS (A-England,ex1) [mailto:piers_dawe@xxxxxxxxxxx]
Sent: Thursday, October 19, 2000 12:08 PM
To: 'Öhlén Peter'; 'Jonathan Thatcher (WWP)'; 'Petar Pepeljugoski
(IBM)'; 'Scott Bradshaw (Roshnee)'; 'Vipul Bhatt (Finisar)'; 'Geoff
Garner (Lucent)'; 'Bob Hughes (Telect)'; DOLFI,DAVE (A-Labs,unix1); 'Ed
Cornejo (Lucent)'; 'Jack Jewell (Picolight)'; 'Jason Yorks (Cielo)';
'Joey Jarriel (Cielo)'; 'John Dallesasse (Molex)'; 'Marc Verdiell
(LightLogic)'; 'Mike Dudek (Cielo)'; 'Paul Kolesar (Lucent)'; 'Robert
Coenen (Gtran)'; 'Rohit Mittal (ONI)'; 'Schelto.Van-Doorn (Infineon)';
'Steven E Swanson (Corning)'; 'Tad Tszostak (E-mail)'; 'Twhitlow
(E-mail)'; 'Peter Stassar (Lucent)'; 'Mike Hackert (Corning)';
'Buchheit, Steve'; 'Chris Simoneaux (Picolight)'; 'Tom Hanson
(Corning)'; 'John Abbott (Corning)'; 'Krister Fröjdh (Optillion)';
'David Hinzel (ETA)'
Cc: 'Walter Thirion (Jatotech)'; CUNNINGHAM,DAVID (A-SanJose,ex1)
Subject: Minutes of serial PMD specs telecon 17 Oct 00

Dear all,

Please find minutes of Tuesday's conference call.

Piers
--------------------------------------------
Present
-------
Krister Fröjdh, Peter Öhlén     Optillion
Geoff Garner                            Lucent
Piers Dawe                              Agilent
Vipul Bhatt                             Finisar
Robert Coenen                           Gtran
Mike Hackert                            Corning
Schelto van Doorn                       Infineon
Steve Buchheit                          LightLogic

Headline summary
----------------
We shared what little we knew about the PMD meeting on 24 October in Austin,
Tx.  We discussed jitter, SONET/SDH vs. Fibre Channel/Ethernet.

New actions
-----------
001017.1        Schelto to confirm the Fibre Channel jitter speaker for 24
Oct in Austin.
001017.2        Schelto to upload MJS2 to his web site.  ??Done??
001017.3        Geoff to find out how SONET sinusoidal jitter is generated
(in testing)  (Here is what FC MJS-2 says: Sinusoidal jitter is generated by
modulating the clock into the BERT with a sine wave of selectable frequency
and amplitude.)

Actions pending
---------------
000926.1        Who? to make presentation on SONET jitter (in Austin?).
Geoff Garner is able to attend on 24 Oct and can speak to this subject,
though will have little time to prepare.  Geoff, Jonathan, are you aligned
on this?

000926.2rev     Peter Öhlén, Scott Bradshaw and/or Rob Coenen to summarise
the developing consensus on measurement for dispersion.  For Austin?

000926.3        Dave Dolfi and Piers Dawe to reconsider link model in
dispersion penalty scenario

001003.3        Scott Bradshaw to develop a presentation on attenuator
management.  Draft looks good, though it could mislead by showing a 15 dB
link loss while the draft standard has a maximum 13 dB including connectors.
Scott will you talk us through your  presentation in Austin?

001003.4        Geoff Garner to try to find why OC-48 and OC-192 receiver
reflectance specs are what they are.

001003.5        Vipul and equalisation team to quantify horizontal eye
closure due to PolMD at 1550 nm.

001003.6        Vipul and equalisation team with help from Steve Swanson and
Paul Kolesar to quantify the fraction of old high-PMD could-be-a-problem SMF
out there.

Austin meeting
--------------
Some suggestions for an agenda are:
        Jitter
        1550 power levels
        Attenuator management (1550 nm specific)
        PolMD issue (handled by Equalisation ad hoc, but Vipul cannot attend
Austin so likely to be held over)
        Dispersion penalties, testing, "golden" components and fibre.  For
draft presentation see Krister's email of 9 Oct "RE: Minutes of serial PMD
specs telecon 3 Oct 00"

There was concern about whether or not the jitter technical experts would be
able to make it.  Before the November meeting is preferable in the hope that
some consensus can be achieved.  Schelto agree to confirm availability of
the FC experts for Austin.

Q for Jonathan: will there be any kind of web site for uploading
presentations?  (Answer: try emailing PDF to David_Law@xxxxxxxx, receive
URL, email URL to group.  To be confirmed.  Fall-back is to email
presentations to this list (Serial PMD reflector wouldn't cope with the
attachments nor should we spam large lists with PowerPoint) so the
laptop-email-enabled can download, and bring some paper handouts.  About how
many people are we expecting? (Answer, 50-60.)

Other
-----
The new IEEE serial PMD reflector is available.  See the IEEE 802.3 main web
page for subscription information, or
http://www.ieee802.org/3/ae/public/adhoc/serial_pmd/reflector.html .

Jitter
------
There was much discussion on details of jitter (e.g. measurement methods,
differences between jitter and wander, patterns, detection bandwidth).  I am
sure this will continue in detail in Austin.  Here follows some detail.
When we have got our facts straight we should send a summary to the serial
PMD reflector.

Fibre Channel is working on a measurement description for jitter.  The
inference is that IEEE can reference this but need actual numbers for jitter
included in the IEEE standard.  Drafts of the Fibre Channel documents are
located on Schelto van Doorn's web page http://www.schelto.com/ .
See FC-PI draft 9 http://www.schelto.com/t11_2/PIEB/Comm_res/00-485v2.pdf
and draft 10 in preparation.  FC-PI is expected to replace the published
FC-PH, aka ANSI X3.230-1994.
See also MJS (Methodology of Jitter Specification) Document, Rev. 10.0 June,
1999 ftp://ftp.t11.org/t11/member/fc/jitter_meth/99-151v2.pdf , and MJS2
ftp://ftp.t11.org/t11/pub/fc/mjs-2/00-245v0.pdf .  But remember, spec
numbers of record are not in these documents.
The minutes of the jitter group are at
http://www.schelto.com/t11_2/Jitter/jitter.htm .

Geoff is trying to find out the history of the SONET jitter specs.  SONET
jitter / SDH specs (G.958, line jitter) were developed for SDH, around
November of 1989.  G.958 contained jitter transfer, tolerance, and
generation specs.  These have more recently been moved to G.783.   In
addition, the consecutive identical digit test was in G.958; this has been
moved to G.957.
(SONET jitter is mainly specified in GR-253-CORE section 5.6 and also
GR-1377-CORE, section 5.6).
SDH/SONET use a 1 dB penalty idea.  Set up for a convenient BER (maybe not
as low as specified) and add sinusoidal jitter.  Increase power to restore
error rate.  Geoff Garner will prepare more detail by Austin.  Compare FC or
Ethernet: no measure of penalty, criterion is that the target low error rate
is still achieved.

Ethernet has no need for jitter transfer specification because every
transmit clock is assumed to be independent.  SONET (ANSI T1.105.09, with
some additional synchronization specifications from T1.101, and SDH (G.813)
extend to low frequency jitter and wander, which could relate to the WAN PHY
but appears not to be an issue.  e.g. 0.15 UI pk-pk over 4-80 MHz (allowed
output jitter specification at a network interface).  SONET clocks can
receive timing from other clocks, e.g., BITS (Building Integrated Timing
Supply) and SSU (Synchronisation Supply Unit) (general requirements for BITS
are given in T1.101 and for BITS and SSUs in G.812).  The BITS or SSUs
receive timing from other clocks in the network of the same or higher
quality.  SONET clocks can also be line-timed, i.e., receive timing from
another location via a SONET facility (i.e., without going through a
co-located BITS/SSU).
SONET or SDH Network Element system clock (i.e., SONET Minimum Clock as
specified in T1.105.09, GR-253-CORE, and G.813 Option 2, SDH equipment clock
as specified in G.813 Option 1), BITS, and SSU PLL bandwidths are < 10 Hz
(depending on the particular clock type, some are much less than 10 Hz) so
for our purposes we can treat them as steady clocks.

All the jitter numbers in Gigabit Ethernet's clause 38 (Physical Medium
Dependent (PMD) sublayer and baseband medium) are jitter output to be
measured in testing.  I looked in clause 36 (Physical Coding Sublayer (PCS)
and Physical Medium Attachment (PMA)) and didn't find a jitter tolerance
jitter to be imposed in testing) measurement.

SONET's jitter measurement bandwidth is thought to be about 4 to 80 MHz for
OC-192 while 10G-FC? has 6-50 MHz - similar.  GbE jitter numbers are much
larger than SONET numbers.  This may be related to there being no upper
limit for the GbE jitter measurement (the lower limit, 637 kHz, is about
1/2000 of Baud rate, in line with the figures above - bit rate/1667 for FC).
A receiver PLL bandwidth might be somewhat larger than this lower limit.
However, the CDR bandwidth cannot be arbitrarily large.  It may be limited
by the number of consecutive bits for which there might be no transitions,
which depends on the line coding.  The maximum number in G.957 is 72.  For
64B66B coding, the number is to be determined - less than 66.
We don't know where the 50 or 80 MHz numbers above come from.
Patterning-related jitter might be wideband, but one might expect the random
jitter (phase noise) to be mostly at low frequencies.  And see FC-PI 9.4.3
for time and frequency domain masks.  See MJS-2 sec8.2 Table 3 - Jitter
tolerance components.

One reason why SONET uses sinusoidal jitter tolerance testing: a pass
implies a pass for mixtures of sinusoidal and random jitter.

Geoff recommended this reference:
"Digital Network Synchronization", by John C. Bellamy, IEEE Communications
Magazine, April, 1995, pp. 70-83.

        - End -