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

RE: [EFM-P2P][EFM] PMD considerations

Hi Ulf, that is an excellent starting point for defining the differences.
I'll just comment on single vs dual fiber in this reply. I tried to
determine the distance vs cost crossover point a while back, and it is more
difficult than might be apparent because of the many possible deployment
scenarios. It's very difficult to give any one deployment scenario greater
weight than another, since it is impacted by a number of factors, depending
on the situation.

An illustrative example ( the numbers are from memory so they may be wrong,
but the relative comparison still holds):

An ADS (active double star) deployment with 100Mb drops:

1) To a residential neighborhood of 1/4 acre lots; each active node serves
16 homes. Average drop length 170 meters. Direct-bury fiber. Around 70
meters, the dual fiber approach equals the cost of the single fiber
approach, so in this particular case the single fiber is the overall winner.
If the active node serves more customers, single fiber is even more
favorable (average drop length increases while all else remains the same).
If the active node only serves 8 homes, the dual fiber approach is more
favorable (doesn't scale multiplicatively)

2) to a multi-tenant building, e.g. apartment. Since each living unit is
closer together, and expensive direct-bury fiber is not needed, dual fiber
is less expensive when the node serves 16, all else being the same. Also,
extended temperature optics aren't required so there are further cost

To throw another wrench in the works, what if the model in #1 uses
microconduit, and fiber is blown in only when a subscriber signs up? This is
a real deployment model that is being considered, and has certain
attractions compared to other models. A new set of factors are involved so
the resulting costs are different. Or, a more standard conduit arrangement
where the fiber is pulled. In that case a much cheaper multimode can be
pulled initially, and single mode can be pulled at a later date, if needed
(presumably well after the expected break-even point). 

Can we rule out any of the above scenarios as a niche? An individual network
operator with a defined set of deployment constraints could, but we can't in
general, since we seek broad (worldwide) applicability. The funding model
for the network, the payback period expectations, the per-subscriber revenue
expectation, the uptake rate, the initial subscriber count, bandwidth growth
expectations, service growth expectations...these will all impact the
deployment model that makes sense for a given network operator.  

So, how do we proceed? About all we can do is cast our individual votes
based our individual biases, it seems. We can't cover all the bases, since
there are too many. That makes it hard to achieve consensus.
I vote for only supporting single fiber cases because this gives EFM
distinct identity, compared to other Ethernet PMDs. I also feel (but cannot
prove) that deployment scenarios where single-fiber is advantageous will
outnumber dual-fiber cases, when all things are considered.
The disadvantage is that a shallow cost comparison that only considers fixed
costs of equipment, and doesn't consider infrastructure costs (fiber and
splices/connectors), will come out in favor of a dual-fiber approach.    

--Dave Horne

-----Original Message-----
From: Ulf Jönsson F (ERA) [mailto:Ulf.F.Jonsson@xxxxxxxxxxxxxxx]
Sent: Monday, December 17, 2001 12:49 PM
To: stds-802-3-efm-p2p@xxxxxxxxxxxxxxxxxx;
Subject: [EFM-P2P][EFM] PMD considerations

Hi all,

The following text brings up some considerations regarding the EFM optical
PMD from a component perspective. It has been written with great help from
our Ericsson internal experts on the optoelectrical component side. 

For the physical medium, i.e. the O/E-converters and the fiber connecting
them, a few aspects may be high-lighted:
1. Data rate
2. Single or multimode fiber
3. Single or dual fiber

We will discuss these aspects in more detail and will also try to draw a
conclusion. Hopefully this will start a discussion on the reflector that may
make it easier for us to agree on a (few) baseline proposal(s) in March.

1. Data rate
The choice is between 100Mbps and 1000Mbps. Of course one must pay a premium
for a tenfold speed increase, throughout the entire system (A more detailed
cost analysis will be presented at the January interim). Optimizing an O/E
converter design for 100Mbps instead of 1000Mbps means

* inherent improvement of receiver sensitivity.
* lowered demands on output optical power (consequence of above).
* lowered demands on thermal management (both inherent, due to lower speed,
and consequence of above)
* lower crosstalk

All these factors will facilitate the module design, simplify the assembly
and increase the yield, thus substantially lower the costs. The argument for
1000Mbps, that the higher volumes for this product will yield lower cost,
neglect the impact of EFM as a cost driving application itself. This
application should in itself be enough to create sufficient production
volumes. Thus, it does not seem optimal to let the vast majority of
connections where 100Mbps is sufficient pay that cost premium, especially as
a P2P topology allows for relatively easy individual line upgrades. On the
other hand, 1000BASE-X will in a P2P topology be appropriate for premium
subscribers and for aggregate traffic higher up in the access network and it
will of course be appropriate to use in a P2MP network. Hence, we see a need
to include both a 100Mbps PMD and a 1000Mbps PMD in EFM. 

2. Singlemode or multimode fiber
As of now, multimode systems are significatly more low-cost than singlemode
systems. Though this difference will decrease as the singlemode component
volumes increase, a certain difference will always remain, due to the less
stringent geometrical tolerances in a multimode system. For those
applications where multimode systems are appropriate, there is no need to
pay the singlemode premium. What is important is that a large number of
connections require singlemode systems, both due to present distance
limitations and to future upgradeability.

3. Single or dual fiber
O/E converters for a single fiber system are inevitably more expensive than
those for a dual fiber system, due to the higher complexity. Just as
inevitable is the fact that this difference will be more than compensated at
very long link lengths. The question is the cross-over distance, and the
distribution of potential installations below and beyond this cross-over,
respectively. If it is regarded necessary to include both options in the
standard, how can that be made with a minimum of effort? Let us examine the
implications on the basic parameters.

3.1. Power budget
A dual fiber system can, and should, allow for a wide output power range, in
order to achieve high production yields in a low-cost assembly process.
If wavelength separation is used in the single fiber case, the power
specification should be equal for dual and single fiber. The extra
attentuation caused by the splitters are hidden inside the converters, and
just has to be compensated for by extra laser power and increased internal
receiver sensitivity, respectively.

The single wavelength case is more difficult, due to constraints imposed by
the reflection crosstalk. In order not to have completely unrealistic
back-reflection demands, the span of the allowed output power must be
minimized. Otherwise, the transmitted power from a "low-end" module would
drown in the reflected power from a relatively high power module. Assume
e.g. a power span of 10dB, a link budget of 10dB and a required SNR of 10dB.
This implies a total allowed near-end reflection of below -30dB, which is
not easily achieved.

Thus, if the output power range for dual fiber is e.g. -5dBm to -15dBm, the
single fiber version should probably be a part of that, something like
-12dBm to -15dBm.

3.2. Wavelength
For dual fiber systems, the operating wavelegth window can, from a component
perspective, be selected freely within the SM fiber window 1300-1600mn. A
wavelength separated single fiber system of course have strict requirements
regarding this matter. For such a system it is also required to have two
types of transceivers, for each end of the connection. Depending on the
actual implementation of the components for a single wavelength single fiber
system, some wavelength restrictions could be needed, as the splitters might
have a wavelength dependence. 

3.3. Connectors
For dual fiber, several types of standard connectors should be allowed, e.g.
MT-RJ, LC, MU, etc. The requirements on connector performance can be kept
low, to reduce costs, since the desired power budget is easily achieved, and
there is no back-reflection problem.

The same should be valid for single fiber WDM systems, even though the power
budget is a bit harder to meet in this case. Possibly the connector
attenuation must be a bit tighter specified.

For non-WDM single fiber, the crosstalk problem make low reflection
connectors necessary throughout the entire system.

Eight different P2P relevant configurations, each with its own merits and
drawbacks, can be distinguished. These are:

100 Mbps MMF dual fiber
100 Mbps SMF dual fiber
100 Mbps SMF single fiber
100 Mbps SMF single fiber WDM
1000 Mbps MMF dual fiber
1000 Mbps SMF dual fiber
1000 Mbps SMF single fiber
1000 Mbps SMF single fiber WDM

of these three already exist as standards within IEEE 802.3, namely

100 Mbps MMF dual fiber
1000 Mbps MMF dual fiber
1000 Mbps SMF dual fiber

100 Mbps SMF dual fiber is at present not an Ethernet standard. Still,
components exist and are used when needed. ANSI has standardized a PMD for
100Mbps FDDI over SMF (ANSI X3.184-1993). The corresponding FDDI standard
for MMF is used as a reference for Ethernet 100BASE-FX.
The need to incorporate 100 Mbps SMF dual fiber within the Ethernet family
is obvious. Since it also seems to be the most appropriate choice for a
large number of EFM connections, it should be the first choice for an EFM
PMD. This PMD should of course to a large extent be based on the 100BASE-FX,
with the physical media specifications optimized for low-cost components
with sufficient performance.

To give a variety of options, it seems reasonable to also incorporate
100Mbps MMF dual fiber as well as 1000Mbps dual fiber in EFM. As already
being Ethernet standards, this should be possible without much extra work.

Single fiber systems are a bit more complicated, since the requirements are
more closely connected to the actual implementation, and a PMD are more
different from existing standards. One way to go, since the requirements
(with wavelength for the WDM solutions as a possible exception) is within
the dual fiber specification, only tighter specified, would be to use the
dual fiber PMD as a base and have different categories within that. These
could be one or two single fiber options, but also extended temperature and
extended range dual fiber options. Depending on the progress of the work,
the single fiber options can either be tightly defined within the base PMD,
or kept rather open for different manufacturer implemenations. The important
issue is to let the time-schedule be set by the most straightforward, dual
fiber, solution.

Best regards,
Ulf Jönsson & Hans Mickelsson