[RE] Minutes of yesterday's meeting

[David V James <dvj@xxxxxxxxxxxx>]

Minutes by David V James
Ad­hoc on Residential Ethernet
  Tuesday 2005Jun14 15:00-17:00 PST
  Alpha Conference Room
  Building 26 (I presume)
  Alpha conference room
  3500 Deer Creek Rd.
  Palo Alto, CA 94304
  650-485-1000 (main lobby)
Or, via teleconference:
  Meeting ID:             1588
  Meeting Name:           IEEE 1588 q/a
  Phone Number:           + 1 949-926-5900
  Toll Free (US Only):    + 1 877-827-6232
  International Callers:  Access Code + 1 949-926-5900

Attending:
Alexei Beliaev           Gibson
Steve Blockman           ??? (1588 member)
Richard Brand            Nortel
George Claseman          Micrel
John Eidson              Aligent (host & 1588 chair)
Felix Feng               Samsung
Geoffrey Garner          Samsung
Jim Haagen-Smit          HP
David James              JGG
Tom Mathey               Northern Data Systems
Michael D. Johas Teener  Broadcom

Note:
If you were attending and have not been listed, or
observe error in the preceding listing, please send
correction instructions to dvj@alum.mit.edu.
Corrections to minutes (created from sketchy notes)
are also encouraged.

1) Rough agenda:
  a) Ask questions from Geoff
  b) Ask questions from DVJ
  c) Interact and explore during (a) and (b)

2) Discussions
Question: How is the slave clock adjusted to track the master?
Answer: The standard defines how the error is determined;
the actual adjustments and PLL techniques are beyond
the scope of the standard.

Question: What about PLLs?
Answer: The industrial community considers these.
PI(phase integration?) loop controllers are commonly used.
The PLL time constants are tradeoffs between convergence
times (small is good) and error elimination (long is good)

Question: What about forwarding between boundary clocks?
Answer: Telecom folks are considering transparent clocks,
that are forwarded through bridge like things.
Discussion: Network bridges might prefer to not directly
forward frames, since its hard to respond quickly and uniformly.
In the 802 domain, it might be better to have ports sync
to each other internally, and multiple hops then involve
cascaded synchronization.

Question: Sampling rates? Isn't 1 second a bit slow?
Answer: The overhead must be small and implementation cheap.
However, there are 1588 proposals for smaller times, such
as 100ms.
Discussion: For 802, the key is to snapshot basics in a
register, then process them in a delayed fashion. Within
this environment, 10ms and possibly 1ms sampling intervals
can be supported without requiring special microprocessors.
The 1588 group is also considering the possibility of short
frames, w/o the spanning tree information, for the higher
rate things.

Question: For computation, binary numbers are simpler.
Is a specific format mandated by 1588?
Answer: 1588 assumes 16-bit epoch, seconds, nanoseconds.
If something else were used, a minimal requirement would
be to convert at the boundaries. The 1588 group is
considering a fractions-of-nanoseconds.

Question: Precedence relationships?
Answer: Explained. The 4-byte ASCII value has precedence,
although not a numerical value. The ASCII coding is
historical.

Question: When are 1588 meetings?
Answer: 1st and 3rd Thursdays, 08:00-09:00 PST
teleconference person to person in July, USA East Coast.
RE folks should probably come and present our ideas,
the earlier the better.

3) Next discussion meeting:
  Wednesday 2005Jun21 14:00-14:00 PST
  Meeting ID:          802373
  Toll Free (US Only): + 1 877-827-6232
  International:       + 1 949-926-5900

Physical tie-in site provided in Palo Alto or San Jose
(to be determined).

4) Next status meetings (teleconference only, no site provided):
  June 21, 0700 PDT   June 21, 1600 CEST  June 21, 2300 JST/KST
  June 28, 1500 PDT   June 29, 2400 CEST  June 29, 0700 JST/KST

DVJ

David V. James
3180 South Ct
Palo Alto, CA 94306
Home: +1.650.494.0926
      +1.650.856.9801
Cell: +1.650.954.6906
Fax:  +1.360.242.5508
Base: dvj@alum.mit.edu