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

[EFM] More On Split Ratios.

Hi All,
	I've been talking with David a little more about the split ratios.
He's having trouble posting to the reflector so I'm fowarding this to the

-----Original Message-----
From: Horne, David M [mailto:david.m.horne@xxxxxxxxx]
Sent: Thursday, July 19, 2001 8:52 PM
To: 'mike.obrien@xxxxxxxxxxxx'
Subject: RE: [EFM] RE: EPON TDMA - Split Ratios.

Hi Mike, my comments are in-line below, prefixed with <DH>. If you'd like to
include others at your company in this discussion please feel free. As long
as we avoid that darn reflector! (I have still not received the reflected
version of the message you sent me early today)

-----Original Message-----
From: mike.obrien@xxxxxxxxxxxx [mailto:mike.obrien@xxxxxxxxxxxx]
Sent: Thursday, July 19, 2001 3:13 PM
To: david.m.horne@xxxxxxxxx
Subject: RE: [EFM] RE: EPON TDMA - Split Ratios.

Hi David,

	Thanks for the response. I was emphasizing the cascaded 2x2 because
that is what is available now, both in fused fiber and PLC. I agree that it
should be possible to build a splitter with almost any split ratio, but in
the interest of simplicity and time to market I favor sticking with what's
there now.

<DH>I'm no expert in the manufacturing processes of FBT or PLC, but I'm not
clear on why a different split ratio would be a time-to-market issue.
Simplistically, for PLC anyway, it is "just" a mask change (famous last
words?!). I could be wrong on this, but I don't believe cascaded sections
are used in the PLC case, i.e it is a single point fanout, all generated
with automated design tools. The single point fan can also be done with
fused fiber since in theory you can twist, melt, and pull any size bundle.
My understanding is that 64 is the practical limit. I understand your point
about off the shelf components, but given the scale EFM will reach, I think
we have some influential scale in requests for deviations from what is
common today.     

	However, I don't really think it is necessary to specify the split
ratio explicitly, other than to say we support up to N-splits at distance x,
M splits at distance y. where M, N, x and y are determined by the optical
power budget. Then hardware would be designed to support the maximum split
ratio. The service provider could always leave a few ports unterminated if
they choose.

<DH>Since there is a business case link to split ratio choice (Gerry
Peasvento pointed this out in a reflector reply today, and I hadn't quite
thought of it in those terms before), this alone seems to almost mandate
what you just said. EFM can't cater to a particular biz case while shutting
out others, or it becomes a special interest group like DOCSIS or FSAN,
rather than a globally applicable IEEE standard. This argues to making EFM a
"split-agnostic" specification**, which in turn argues to variable time
slots size capability. There has to be some range limit of course, and
multiple transceiver classes will be necessary to cover the full range, cost
effectively. This alone could be quite a long discussion which I'd like to
have at some point, especially the variable time slot possibilities.

**<DH>it also argues to making EFM effectively a "relatively dumb pipe
coupler." EFM cannot possibly accommodate all the potential profitability
theories and schemes for QoS, SLA, etc., that service providers have and/or
dream of in the future. Seems the best EFM can do is NOT PREVENT them from
implementing their scheme, i.e. serve as a transparent coupler. No way can
it ENABLE all these schemes simultaneously and keep the simplicity mantra

	I agree that a monitor channel would be useful in any case. This
could be a special port with a lower power output, or possibly one of the
standard port. 	As a final aside - I think the real trick would be to get
someone to
build a splitter with an asymmetric insertion loss. Downstream the power
would be split evenly (or with some pre-determined ratio) between all the
ports, upstream - All the power would be funneled into one (or two) port.
Then the upstream link budget looks like P2P. Any ideas?? 

<DH> Maybe I'm not quite understanding your suggestion, but I think this is
possible with waveguide couplers. COuld you sketch a picture of the signal
paths and wavelengths for the scenario you describe (at your lesiure; no
hurry). You could FAX a hand drawn, or use Visio or powerpoint or whatever
is most convenient. (If we all had fiber-to-the-home we could just call up
our electronic whiteboard! )

-----Original Message-----
From: Horne, David M [mailto:david.m.horne@xxxxxxxxx]
Sent: Thursday, July 19, 2001 12:55 PM
To: 'mike.obrien@xxxxxxxxxxxx'
Subject: RE: [EFM] RE: EPON TDMA

Hi Mike, that is one approach to implementation, but I believe any split in
the range we are talking is do-able, whether fused (FBT) or planar (PLC)
based. It may even be that the group decides (hypothetical example) a 12-way
split, with 11 getting 99% and 1 port for diagnostics @ 1%, is the way to
go, after analyzing many operational and test scenarios. My view is that
volumes today are too low to believe we need to be constrained to currently
available product lines, and that EFM will be such high volumes in
comparison that if the spec ends up citing 12-way as a recommended
configuration, it will be built by many vendors (even before the spec is
finalized). That is the beauty of standards-based product development, and
the volume scale it can achieve. I think we have to believe the high volume
scenario will play out or there is little point in continuing. Any comments
you have would be appreciated. 

PS I'm just sending this directly to you without copying the reflector.
Reflector is just too slow (I still don't have the reflected version of this
message from you yet, and probably won't for another 12-24 hours). If we
continue this dialog we can post the whole thing later, or not.       

--Dave Horne

-----Original Message-----
From: mike.obrien@xxxxxxxxxxxx [mailto:mike.obrien@xxxxxxxxxxxx]
Sent: Thursday, July 19, 2001 10:24 AM
To: david.m.horne@xxxxxxxxx
Subject: RE: [EFM] RE: EPON TDMA

Hi David,
	The split ratio's are powers of two because the splitters are made
by cascading 1x2 (actually 2x2) splitters.

-----Original Message-----
From: Horne, David M [mailto:david.m.horne@xxxxxxxxx]
Sent: Wednesday, July 18, 2001 9:03 PM
To: 'Roy Bynum'; glen.kramer@xxxxxxxxxxxx
Subject: RE: [EFM] RE: EPON TDMA

Where did the often-cited split ratios of 16 and 32 come from, anyway? It
seems an argument can be made for holding off on finalizing the split ratio
choice until the TDMA scheme is better defined and quantified. That way, it
can be better matched (disaggregated and reaggregated) to Ethernet pipes on
the ends of the EFM segment, and be easier to hand off to multiple service
providers in a PoP. End-to-end Ethernet is the goal, after all. 

For example, a split ratio of 8 may be a better match for 100Mbps dedicated
(logical) connections, once the actual TDMA overhead and guard bands are
known. Compared to 16:1, the cost of the extra feeder fiber (cost of 2
fibers split across 16 subs rather than 1 fiber split across 16) may be
offset by the more compatible handoff logistics. For long drop fibers, with
two 8:1's, the drops will be relatively shorter than a single 16-way split
to the same endpoints. This may even offset the extra feeder fiber. But,
there's no reason to stick with powers of 8 (that I can think of off hand)
unless they make something easier. Maybe 10 or 12 or... are attractive after
more details are agreed to.