================ Minutes with corrections =================================

         IEEE p1394b Long-haul Low Cost(S100) Task Group Meeting
           Arch Tech Co., Ltd. and Sumitomo Electric Industries
                    Akasaka, Minato Ward, Tokyo, Japan
                              July 8, 1997

The task group would like to thank Arch Tech Co., Ltd. and Sumitomo Electric
Industries for hosting this meeting.  The meeting started at 9:30am in Arch
Tech with introductions and then around 11am the whole group moved to
Sumitomo Electric Industries due to the shortage of meeting room space and
sitting chairs.  The updated minutes of previous meeting were approved as
written.  Chair Taka Fujimori started with the following agenda:

Agenda
 0. Introductions
 1. Approval of Agenda
 2. Review of Minutes
 3. 1394b meeting Seattle, June 10-11, Reports
   3.1 Long-haul High Performance(S800) TG Report: Taka Fujimori
   3.2 1394b Report                              : Taka Fujimori
 4. Review Requirements
 5. Contributions
   5.1 PMD (SI) POF S200                         : Yamazaki-san
   5.2 PMD      HPCF                             : Morikura-san
   5.3 PMD UTP Cat 5 Emission                    : Takizuka-san
   5.4 PHY Coding Intensive Study                : Taka Fujimori
   5.5 Start-up(Beta Mode)                       : Taka Fujimori
 6. Open Issues  
   6.1 GI-POF	
 7. Discussion
 8. Action Items
 9. Meeting Schedules
      August 6 through 8(Honolulu, HI)
      September 2       (location TBD)
10. Adjournment


-----
0. Introductions
Taka Fujimori explained the p1394b working group that was originally
intended to extend the speed of IEEE1394 to gigabit rate.  As the 1394TA
requested 1394b working group to expand the goal to cover long-haul 1394
technology, the 1394b group formed two sub task groups this May:
  Long-haul  Low  Cost(S100)  Task Group
  Long-haul High Speed(S800<) Task Group

The motivation behind long-haul 1394 technology is home network.  VESA/
Home Network and DAVIC home network are working on standardizing 1394
based home network technology.  The first meeting of S100 TG was held this
late May and the group has the schedule:
                         PMD        PHY
   Phase 1 '97/E:       S200       S400
   Phase 2 '98/E:       S400        --
   Note: UTP is only at S100

He also mentioned that in an IEEE meeting each participant is individuals,
not a representative of his/her organization.

1. Approval of Agenda
2. Review of Minutes

3. 1394b meeting Seattle, June 10-11, Reports
  3.1 Long-haul High Performance(S800) TG Report : Taka Fujimori
S800 PMD Task Group made a long sub-PHY model which converts 1394b copper
signals to long-haul signals.  The group agreed that this long sub-PHY
will not transmit Self_ID packet after bus reset, therefore it's no an
independent node.  However caboose self_ID packet discussed in 1394a WG
may be able to address long sub-PHY.

  3.2 1394b Report                              : Taka Fujimori
Alister came up with modified IBM 8B/10B code for short-haul S800<
transmission.  It scrambles both of data and control before 8B/10B coding.
For data, it uses exactly the same code of IBM 8B/10B after scrambling.
For control, Alister introduced a new set of 16 8bit-long codes, that have
longer runlength.  Since control codes of original IBM 8B/10B are 16bit long,
it's apparently better to have shorter 8-bit long control codes.  
Eric presented emission measurement results by actually making a coding
silicon with field programmable gatearray and a spectrum analyzer.
Random data emission results of (modified) IBM 8B/10B and 4B/5B are basically
the same.  However if control codes and/or padding codes are added,
significant spikes appear beyond random data emission profile.  Since only
modified IBM code is scrambled, other codes can't be better theoretically.

Though the atmosphere of p1394b group was almost picking up modified IBM
code, Taka recognized that the use of Alister's long-runlength control code
concentrates energy at lower frequencies, making emission worse.  In fact
slide 11 of Eric's presentation is proving it.  There's a peak around a lower
frequency.  If Alister's control code is not used, this emission peak disappears.
Since 4B/5B data and control code patters look like random, if scrambler
is applied, the 4B/5B code will give us the best emission while keeping PHY
delay the shortest.  Taka is to confirm this analysis.

Draft document 0.03 was delivered by Eric Hannah(editor), available via 
ftp site: http://www.fireflyinc.com.

Regarding beta mode, Colin suggested an active reject code, which stops doubling
PLL speed.

Beta mode arbitration enhancement was also suggested by Mike Teener.  When
talking at slower speed, the bandwidth just waisted by stuffing can be used
for arbitration.  Since gigabit 1394 communication is bidirectional, return
line for grant is also availble.  The request can be granted before finishing
current transmission.

4. Review Requirements
By reviewing the requirements and models presented by Yamazaki-san at the
last meeting.  The group agreed to address the following 3 models.

 Model 1 : Repeater/Wallplate               ||
      LINK PHY(1394-1995)*--------------*Long PHYb+-------------------
              (1394a    )     DS link       ||      Long  POF/HPCF/UTP
              (1394(b)  )                   || wall

 Model 2 : Dongle                       ||                       ||
      LINK PHY(1394-1995)**Long PHYb+---||-----------------------||---
              (1394a    ) DS            ||   Long   POF/HPCF/UTP ||
              (1394(b)  )                wall                     wall 

 Model 3 : Direct Long Distance
      LINK PHY(1394(b))Long PMDb+-------------------------------------
                                             Long         POF/HPCF/UTP

                                                     * means 1394 Connector
                                                     + means MIC for long 

Does long PHYb transmits self_ID? : YES
It's an independent node.   Since 1394-1995 nodes may be attached to this
long PHYb, Long PHYb has to be PHY ping capable to measure propagation 
delay, ie, it transmits Self_ID.

Do models consider splicing points? : YES
However the spec may express the number of splicing points by N, where N=0,1,2,.
There's a trade-off between splicing points N and distance.

Long PHYb may have zero, 1 or more DS link ports and zero, 1 or more long
distance ports.  Therefore the group addresses 3 or more port long distance PHY.

Model 3 addresses devices with direct long distance port.  In future, Consumer
devices may have long distance port directly.  Another example is a front gate
security camera which shoots visitors and brings video image to living room over
POF/HPCF.


5. Contributions
  5.1 PMD (SI) POF S200                         : Yamazaki-san
Borrowing ATM Forum POF(155Mbaud) spec for S100 is straight forward.
But extending the spec to S200 is not very straight forward.

At S100, 650nm LED Launch power margin is 6dB.  By subtracting 6 out
of eye safety limit(-2dBm), you get the minimum launch power of -8dBm.
Both of SI-POF and GI-POF has the same loss of 13dB after 50m travel.
At the receiver, minimum power is now -21dBm, which gives you 4dB margin
for 2 passive splicing points before hitting the minimum photo diode
sensing level of -25dBm.

POF S100:(1250Mbaud):
  Tx LED(650nm)
 -2dBm(upper limit by eye safety)
   |
   |  6dB margin for LED deterioration, power variation, etc.   
 -8dBm...................
        `   .           ^
              `   .     | 13dB loss of 50m POF fiber
                    `   ._______
                           |
                           | 4dB margin for 2 passive splicing points 
                         ------- -25dBm  Rx 

When extending this design to S200(250Mbaud), launch power will now be
degraded by 2dB due to the peaking, which gives you min power of -10dBm.
50m Fiber loss is the same(-13dB).  And photo diode sensitivity becomes
-23dBm at 250Mbaud.  Consequently, degradation at LED and photo diode
eats up all the margin for passive splicing.  This is why passive
splicing at S200 is not very possible.

POF S200:(250Mbaud):
  Tx LED(650nm)
 -4dBm(-2dBm)
   |
   |  6dB margin for LED deterioration, power variation, etc.   
-10dBm...................
        `   .           ^
              `   .     | 13dB loss of 50m POF fiber
                    `   ._______
                                 -23dBm  Rx 
                             no margin for passive splicing

Yamazaki-san then described the case of ATF Forum HPCF(Hard-Polymer Clad
Fiber).  HPCF gives you substantially lesser loss medium compared to POF.
At the transmitter, launch power will be reduced to -20dBm because of the small
core diameter(200um).  However HPCF has much less loss and leaves margin
for 2 passive splicing points before hitting the min photo diode sensitivity
of -26.5dBm.

HPCF S100(1250Mbaud):
  Tx LED(650nm)
-14dBm
   |
   |  6dB margin for LED deterioration, power variation, etc.   
-20dBm...................
        `      .        | 2.5dB loss of 100m HPCF fiber
                  `     ._______
                           |
                           | 4dB margin for 2 passive splicing points 
                         ------- -26.5dBm  Rx 

Extending this designed to S200 is quite similar to the case of POF.  Both
transmitter and receiver are degraded by 2dB each, leaving no margin for
passive splicing.

HPCF S200:(250Mbaud):
  Tx LED(650nm)
-16dBm
   |
   |  6dB margin for LED deterioration, power variation, etc.   
-22dBm...................
        `      .        | 2.5dB loss of 100m HPCF fiber
                  `     ._______
                                 -24.5dBm  Rx 
                             few margin for passive splicing

To define a POF/HPCF spec for S200, Yamazaki-san suggested the number of
splicing points should be defined by n, where n = 0, 1, 2, ...   Because
POF at S200 satisfies 50m length requirement when n = 0.

Another method to solve this issue is to adopt LASER diode.  However LASER
diode requires APC(Active Power Control), which will increase cost.

  5.2 PMD      HPCF                             : Morikura-san
Following the Yamazaki-san's ATM Forum HPCF presentation, Morikura-san
explained 850nm LED case of SI-HPCF and GI-HPCF.  Wavelength of 850nm
is optimal for HPCF in terms of loss(less than 1dB/100m).

He made a comparison between raw POF and raw H-PCF:

                      SI-POF            SI-HPCF  GI-HPCF
Core/Clad          980/1000um               200/230nm
Bandwidth(@100m)      150MHz            140MHz   600MHz
Loss(@100m)        20dB@650nm              0.5dB@850nm  
Min Bend Radius        25mm                    15mm
Operational Temp  -20 to 70 degree C   -20 to 70 degree C

Since HPCF with 850nm LED easies coupling, wideband receiver design,
high speed LED and fiber loss, he concluded that one optical module can
achieve 100m transmission at all S100/200/400 speeds.  He also showed
reasonably opened eye patterns of SI-PCF@100m and GI-PCF@300m transmissions 
at 200Mbps.  Bit error rate characteristics proved that GI-PCF can reach
300m at S200, twice longer than that of SI-PCF.

Yamazaki-san mentioned that HPCF with 850nm LED is virtually loss less.
ATM Forum spec for 650nm has to be revised to specify HPCF.

The material cost of HPCF is four time more that that of glass fiber,
simply because of large diameter(230um vs 125um).  Cost of the 
optical link is between those for POF link and glass fiber link.  Plug
connector for HPCF requires ferrule.

Taka asked the possibility of Giga bit transmission by HPCF.  The answer
was 50m tranmission with 1.3um LASER would be possible.

Taka also mentioned that there's no error rate guide line in 1394 spec.
In other words, 1394 is a kind of assuming virtually no error, which might
not fit telecommunication convention.  His personal guide line is that
no or single error in a movie.  In this case the error rate depends on
video compression efficiency.

The US FDA CDRH eye-safety specification, kindly faxed by Joe Tajnai and Del
Hanson, was delivered by Taka.  He said in the US, both of IEC825-1 and
CDRH are required.


  5.3 PMD UTP Cat 5 Emission                    : Takizuka-san
no presentation was made.

  5.4 PHY Coding Intensive Study                : Taka Fujimori
Taka presented code comparison among Sony 4B5B+NRZI, NEC 4B5B+NRZI,
IBM 8B10B, modified IBM 8B10B and HP 8B10B.  Estimated long to long
PHY delay will be 288ns for 4B5B and 312-392ns for modified IBM 8B10B.
This PHY delay was analized by means of
  NRZI
  Serial/Parallel
  Arbitration/Data Separation
  Worst Delay BTW Recv/Trans Clocks
  FIFO
  Parallel/Serial
  NRZI

4B5B codes are always 40ns shorter than 8B10B codes.

Taka said this work has to be continued, since he needs to further analyze
1394-1995 PHY delay of 144ns and modified IBM 8B10B code.

  5.5 Start-up(Beta Mode)                       : Taka Fujimori

The following beta mode was suggested for 4B5B.

S100: no beta = QI   = 0000_1111(Bus Reset)
S200:     SS1 = JKSR = 11000_10001 11001_00111
S400:     SS2 = JKSS = 11000_10001 11001_11001

S100 long distance PHY doesn't have beta mode, thus transmits bus reset after
power on or connector insertion.  Higher speed long PHYs wait for a while to
check bus reset pattern before it starts beta mode.

Optionally long PHY may goes to S800 or above speeds.  In this case, this
coding method can be extended to SS3=JKSSSR and SS4=JKSSSS or beta mode for
8B10B may be adopted after S800 proposed by Colin.

6. Open Issues  
  6.1 GI-POF	
     Nakamura-san briefly explained a new GI-POF(not a perflorinated GI-POF)
     just announced by Mitsubishi Rayon late last month, which runs at giga
     baud rate.  Hi is willing to attend the next 1394b S100/800 joint meeting
     in Honolulu and give us a presentation.

7. Discussion

Yamazaki-san raised an idea that, by applying a small optical lens at 650nm
LED, one optical link may be able to accommodate both POF and HPCF effectively,
since the lens concentrates optical power and coupling to HPCF will be
improved significantly.

S800 PMD TG has interests in GI-POF and GI-HPCF.  Those medium should be 
presented in the joint S100/800 meeting in Hawaii.  And then we could pick
up appropriate media that may cover S800 and above speeds as well.

4B5B based code seems to have the shortest PHY delay and the best emission.
We have to keep studying 4B5B and modified IBM code and then prepare a
material by Hawaii meeting.

8. Action Items
   1.  PMD SI-POF S200                      -----  Yamazaki-san
   2.  PMD GI-POF                           -----  Nakamura-san
   3a. PMD SI-HPCF                          -----  Morikura-san
   3b. PMD GI-HPCF                          -----  Sogabe-san
       (3a and 3b might be merged.) 
   4.  UTP Cat 5 Emission                   -----  Takizuka-san/Watanabe-san
   5.  PHY Coding Intensive Study follow-up -----  Taka Fujimori
   6.  PHY beta mode follow-up              -----  Taka Fujimori

   The PMD items will be presented at the S100/800 joint meeting in Hololulu
   on Aug 6th.  PHY items will be presented at the S100 TG on the following 
   day.

9. Meeting Schedules
                August 6: S100/800 joint meeting colocated to ANSI in Honolulu, HI
     Morning of August 7: colocated to ANSI in Honolulu, HI
             September 2: NEC likely hosts us in Tamachi, Minato Ward, Tokyo

10. Adjournment

=============================================================================
Attendees:
Taka Fujimori      Sony           +81-3-5488-6353  fujimori@arch.sony.co.jp
Yoshihiro Tanaka   Hitachi MC     +81-462-96-9903  tana@adc.hitachi-mc.co.jp
Susumu Morikura    Matsushita     +81-6-906-4876   morikura@isl.mei.co.jp
Kazuki Sogabe      Sumitomo E     +81-6-466-5539   dw750134@jnet.sei.co.jp
Akihiro Miyachi    Molex Japan    +81-462-65-2327  amiyachi@molex.com
Sumihiro Okawa     Sony           +81-466-30-4050  okawa@sm.sony.co.jp
Atsushi Yoshida    Hitachi MC     +81-423-25-1711  yosidaat@hitachi-mc.co.jp
Michihiko Sakurai  Sony           +81-3-5448-5362  sakurai@sslab.sony.co.jp
Satoshi Takahashi  Mitsubishi Rayon +81-3-3245-8746 takahashi_sa@mrc.co.jp
Hiroshi Takizuka   Sony           +81-466-30-4050  takizuka@sm.sony.co.jp
Kenji Watanabe     Sony           +81-3-5448-5362  nabeken@sslab.sony.co.jp
Takanari Fujimoto  Mitsubishi     +81-45-224-2748  fujimoto@cd.lmt.melco.co.jp
Masashi Oguchi     Mitsubishi     +81-45-224-2748  oguchi@cd.lmt.melco.co.jp
Toshihiro Kageyama Mitsubishi     +81-45-825-5795  kage@yh.mtcs.co.jp
Yoshio Inagakki    Toshiba        +81-425-85-3295  inagaki@cns.hino.toshiba.co.jp
Yoshiki Nakabayashi Intel         +81-298-47-5371  nakabayashiy@ccm.intel.co.jp
Tatsuo Inoue       Arch Tech      +81-3-5545-7813  inoue@ba2.so-net,or.jp
Shuntaro Yamazaki  NEC            +81-44-856-2082  yamazaki@optsys.cl.nec.co.jp
Yoshiaki Takabatake Toshiba       +81-44-549-2282  tkbtk@csl.rdc.toshiba.co.jp
Kaoru Ichikawa     Mitsubishi Rayon +81-3-3245-8750 ichikawa_ka@mrc.co.jp
Kazuki Nakamura    Mitsubishi Rayon +81-8275-3-8514 kazuki-n@mars.dtinet.or.jp

-------------------------------------------------------------------------------