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

RE: [EFM] Split ratios


Hi David,

> The apparent consensus with splits of 16 and 32 seems
> to have been arrived at devoid of any engineering
> considerations.

I wish it were an engineering consideration, but there are many factors at play.  Here are some examples. 

1.  Logical
Split ratios come in factorials of 2, so off the bat 1:4, 1:8, ... 1:128 becomes a common logical division one uses.

2.  Market Requirement/Carriers
Typically, local carriers are requesting splits for business in the 1:16 range, and to the resident in the 1:32 range.   This is not an absolute, but what I find as a common range.  As an example, I spoke with an ex-ILEC manager last night who said their calculations showed 1:12 business and 1:48 residential was a good target.   Alternatively, RHK published a paper showing 1:4 for FTTH was the most economical.  As you can see, there is quite a range of opinions here.

3.  Optical power budget
Much work was done on optical power budgets in Portland, showing both challenges and technical feasibility of low cost lasers in ONU with 1:16.   Using APDs and FEC, a proposal showed 1:128 feasibility.  It is possible that the split ratio / budgets definitions will fall into "classes" to allow for low-cost/small split an higher-cost/large split implementation. 

4. Distance, Optics
The distance requirement will affect number of splits.  Some carriers may only need to  cover 5 km, some may need to cover 20 km, some more.  The max reach affects the max number of splits.  Also, the cost of the optics is directly related to the number of splits.  More splits = more expensive lasers and detectors.

5. Bandwidth per User
A Gigabit Ethernet PON has 1000 Mbps, and the bandwidth per subscriber node depends on the number of splits.  This bandwidth per node will be different for FTTH, FTTC and FTTB applications.

6.  Plant management
Split topologies can be flexible within EPON.  For example, a 1:16 could be a combination of 5 1:4 splitters, located where the service provider wants.

7.  Technology
Planar waveguide splitters have the most focus right now, because they will be the least expensive in volume (one chip, with single fiber and ribbon fiber alignment to waveguide).  However, there is a low cost 1:4 single fused-fiber technology that is also interesting. 

8.  Current Plant
EPON may be deployed as a new-build or over-build (most likely) or an upgrade.  Depending on how many fibers the carrier has access to (own, lease, etc) will affect the economics of splitting.  In cases where the Local Carrier has enough fiber resources, P2P may be a preferred solution.

> I am just suggesting we not get too attached
> to those numbers

I agree.

Gerry Pesavento

Carlos, yes, I understand all these considerations. That was not my
question. The apparent consensus with splits of 16 and 32 seems to have been
arrived at devoid of any engineering considerations. Those engineering
considerations are exactly what I was suggesting are needed. The final
comment you made of "just don't ask me why" is what I WAS asking.

Use of 16 or 32 for interim discussion purposes is fine; I am just
suggesting we not get too attached to those numbers since other ratios may
prove to be more beneficial for Ethernet PON based on more detailed
considerations. 16 and 32 have not been cast in concrete, correct?

-----Original Message-----
From: Carlos Ribeiro [
Sent: Thursday, July 19, 2001 4:39 AM
To: Horne, David M; 'Roy Bynum';
Subject: [EFM] Split ratios

At 21:02 18/07/01 -0700, Horne, David M wrote:
>Where did the often-cited split ratios of 16 and 32 come from, anyway?

There are two considerations here. First, there is a maximum limit on the
split ratio, that depends on the power budget and attenuation. I believe
that distance is not going to be the major attenuation factor in the access
network situation; the number of points where you have signal degradation
(connectors, splits, etc) is going to be the major factor. The actual
physical construction of the access network is going to be completely
different, specially if we have any hope of providing FTTH service; there
are a lot of practical and operational issues still unsolved.

The other consideration is economical. A true point-to-point network can be
equated to a 1:1 split ratio. OTOH the split ratio in a PON network can
range from 1:8, 1:16, or even 1:64. Bigger split ratios mean less bandwidth
available for each customer, but it also costs less to deploy. Se rhere is
a cost equation, depending on a lot of factors - technical and marketing
included - that will tell what is the best ratio in economical terms. I
don't expect this answer to be known at this point; we will only be able to
find this number with practical knowledge, ooking at the real situations.

As for myself, I feel comfortable with a 1:16 split ration Just don't ask
me why <wink>.

Carlos Ribeiro
CTBC Telecom