[160177] in North American Network Operators' Group
Re: Muni fiber: L1 or L2?
daemon@ATHENA.MIT.EDU (Owen DeLong)
Fri Feb 1 17:06:15 2013
From: Owen DeLong <owen@delong.com>
In-Reply-To: <CAGbD49r0HKhApLybJ8WUAXAgGAy8XasS1LML0Thhpc6Mf=enWg@mail.gmail.com>
Date: Fri, 1 Feb 2013 13:59:54 -0800
To: Jason Baugher <jason@thebaughers.com>
Cc: NANOG <nanog@nanog.org>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
On Feb 1, 2013, at 1:43 PM, Jason Baugher <jason@thebaughers.com> wrote:
> It's still a 23dB loss for each customer from the CO to the ONT.=20
>=20
> I have an OLT that launches at +5dBm. At 1490nm, I should see about a =
.26dB loss per km. My 1x32 splitter is going to add about 16dB more =
loss. Assuming we ignore connector losses, and also assume that the =
customer is 10km away:
>=20
Nope=85 The power going into each fiber out of the splitter is 1/16th =
that of what went into the splitter.
Yes, your total in-line loss is still 10km, but you are forgetting about =
the fact that you lost 15/16th of the power
effectively going to the fiber when you went through the splitter (in =
addition to the splitter loss itself).
So: CO Based splitter:
Each customer gets (IN - 16dB - (10km x .26db))/32
Splitter at 9km:
Each customer gets (IN - (9km x .26dB) -16db)/32-(1km x .26db)
If we use 5dBm as our input, this works out:
CO: (5db - 16db - (10km x .26db) / 32
/32 is effectively -15 db (-3db =3D =BD power, 32 =3D 2^5)
Substituting: (5db - 16db - 2.6db) -15db =3D -28.6db to each customer.
Spitter at 9km: (5db - (9km x .26db) -16db)/32-(1km x .26db)
Substituting: (5db - 2.34db -16db)-15db-.26db =3D -28.08db to each =
customer
So there is a difference, but it seems rather negligible now that I've =
run the numbers.
However, it's entirely possible that I got this wrong somewhere, so I =
invite those more expert than I to review the calculations and tell me =
what I got wrong.
Owen
> CO-based splitter:
> +5dBm - 16dB - (10km x .26dB) =3D -13.6
>=20
> Splitter at 9km:
> +5dBm - (9km x .26dB) - 16dB - (1km x .26dB) =3D -13.6
>=20
>=20
> If someone can explain why this math would be wrong, I'd love to hear =
it and I'd be happy to run it past our vendor to see if they agree.
>=20
>=20
> On Fri, Feb 1, 2013 at 3:16 PM, Owen DeLong <owen@delong.com> wrote:
> Actually, this is an issue=85 I should have seen it.
>=20
>=20
> You have 3 loss components=85 Power out =3D (Power in - loss to =
splitter - splitter loss) / nOut - loss-to-customer
>=20
> So, if the loss to the splitter is 3db and you have 20db (effective =
320db on a 16x split) loss on each customer link, that's
> a radically worse proposition than 20db loss to the splitter and 3db =
loss to each customer (which is effectively 48db
> loss on a 16x split).
>=20
> It's still do-able, but you either need amplifier(s) or very short =
distances between the customer and the MMR.
>=20
> Given this consideration, I think the situation can still be =
addressed.
>=20
> Put the splitters in the B-Box and allow for the possibility that each =
subscriber can be XC to either a splitter or
> an upstream dedicated fiber. The provider side of each splitter would =
be connected to an upstream fiber
> to the MMR.
>=20
> So, each B-Box contains however many splitters are required and each =
splitter is connected upstream to a
> single provider, but you can still have multiple competitive providers =
in the MMR.
>=20
> This setup could support both PON and Ethernet as well as other future =
technologies.
>=20
> Owen
>=20
> On Feb 1, 2013, at 1:04 PM, Jason Baugher <jason@thebaughers.com> =
wrote:
>=20
>> I should clarify: Distance x loss/km + splitter loss. =3D link loss.
>>=20
>>=20
>> On Fri, Feb 1, 2013 at 3:03 PM, Jason Baugher <jason@thebaughers.com> =
wrote:
>> I disagree. Loss is loss, regardless of where the splitter is placed =
in the equation. Distance x loss + splitter insertion loss =3D total =
loss for purposes of link budget calculation.
>>=20
>> The reason to push splitters towards the customer end is financial, =
not technical.
>>=20
>>=20
>> On Fri, Feb 1, 2013 at 2:29 PM, Scott Helms <khelms@zcorum.com> =
wrote:
>> Owen,
>>=20
>> You're basing your math off of some incorrect assumptions about PON. =
I'm
>> actually sympathetic to your goal, but it simply can't work the way =
you're
>> describing it in a PON network. Also, please don't base logic for =
open
>> access on meet me rooms, this works in colo spaces and carrier hotels =
but
>> doesn't in broadband deployments because of economics. If you want =
to
>> champion this worthy goal you've got to accept that economics is a =
huge
>> reason why this hasn't happened in the US and is disappearing where =
it has
>> happened globally.
>>=20
>>=20
>> > Bottom line, you've got OLT -> FIBER(of length n) -> splitter ->
>> > fiber-drops to each house -> ONT.
>> >
>>=20
>> So far you're correct.
>>=20
>>=20
>> >
>> > All I'm proposing is making n really short and making "fiber-drops =
to each
>> > house" really long.
>> > I'm not proposing changing the fundamental architecture. Yes, I =
recognize
>> > this changes the economics and may well make PON less attractive =
than other
>> > alternatives. I don't care. That's not a primary concern. The =
question is
>> > "can PON be made to work in this environment?" It appears to me =
that it can.
>> >
>>=20
>>=20
>> Here is where you're problems start. The issue is that the signal =
*prior
>> to being split* can go 20km if you're splitting it 32 ways (or less) =
or
>> 10km if you're doing a 64 way split. AFTER the splitter you have a =
MAX
>> radius of about 1 mile from the splitter.
>>=20
>> Here is a good document that describes the problem in some detail:
>>=20
>> http://www.ofsoptics.com/press_room/media-pdfs/FTTH-Prism-0909.pdf
>>=20
>>=20
>> Also, here is a proposed spec that would allow for longer runs post
>> splitter with some background on why it can't work in today's GPON
>> deployments.
>>=20
>> =
http://www.ericsson.com/il/res/thecompany/docs/publications/ericsson_revie=
w/2008/3_PON.pdf
>>=20
>> --
>> Scott Helms
>> Vice President of Technology
>> ZCorum
>> (678) 507-5000
>> --------------------------------
>> http://twitter.com/kscotthelms
>> --------------------------------
>>=20
>>=20
>=20
>=20
