| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Dear Jim, You know, the technical business is a
funny thing. When you talk to the applications engineers, you get one
story. When you negotiate the contract, you get quite a different one. What
I’ve heard from my friends in On a different matter, I have heard that
the SCTE should be sending us a liaison regarding their work on a new sort of
PON system called RFOG (radio frequency over glass). This system is completing
the downstream-only video overlay scheme that we started with the 1550 to
1560nm wavelength by adding an analog upstream transport. They are
looking at using the 1590 or 1610 bands, and they will need 20nm of width (most
likely) because these upstream lasers are at the ONTs. If we want to
coexist with this system (admittedly a new request), then we should think twice
about using up the entire spectrum. As it is, if we stick to 1577, then
this system could use 1610 with a nice 20nm guard band. If we spread out
to 1600nm, then where does that leave them? If I can elevate the discussion for a
moment, the evolution of PON has been marked with the tendency to spend our
spectrum like water. When it came time to add new features and upgrades,
we were forced to go back and revise our systems to narrow the windows, add
blocking filter specifications that should have been in there in the first
place, jump through multi-rate TDMA hoops, and so on. This is not a good
path. The 1577nm option is a small step towards being more conservative
in our allocation of spectrum. But I’ll admit that this sort of
argument in not strictly in our scope – it is more a general concept
position… just take it for what it’s worth. Sincerely, Frank E. From: Jim Farmer
[mailto:Jim.Farmer@xxxxxxxxxxxxx] Sorry to have let this lie for a few days
- had to go off and earn a little money. Thanks for the replies. It seems
Pete has established that existing fiber and passives should work up to
1625 nm. We have gotten similar information from a large vendor of
optical fiber and components. He tells me that the oldest long
distance fiber he knows of being deployed in the In addition, please see the note from We are content to let PR(X)30 operate
at 1577 nm, and if the market decides that is all that is needed, so be
it. With due respect, it may be that vendors who are working on 1577 nm
product are doing so because of the perceived initial market for a 10 G
product, which may be a slightly specialized market. But we remain
concerned about two things: The cost of meeting the narrow bandwidth, both in
terms of money and power, and the filter used to separate the downstream RF
overlay from the lower 1577 nm wavelength. This latter would not be an
issue except it is a filtering function that must be done at the ONT, where we
worry about every penny of cost. I've been trying to get solid numbers
from vendors, but I, too, am having trouble finding consensus. The narrow bandwidth (+/- 3 nm) issue
is an OLT issue, and so is not quite of as much concern as is anything at the
ONT, but we still worry about it. I can accommodate the bandwidth by
measuring the wavelength of each laser and determining the temperature at which
the cooler must operate to keep its wavelength centered in the window.
This works, but does cost money and power for the cooler and control circuit, as
well as an extra step in production. While traditional telecom
markets may not be concerned, we feel that emerging deployers, such as
municipalities and cable companies, may not tolerate increased costs. We
don't see how a product with a cooler will be competitive with something that
does not have to be cooled, all else being equal. Regarding the issue of filtering the 1560
nm maximum video wavelength from the 1574 nm minimum data wavelength, I, too,
have not been able to get a quantitative answer. But intuitively,
the narrower you make the transition region, the more costly the filter, and
the more loss it will have. We had one discussion with a company that
buys filters and packages them, and at first they thought this was a non-issue
based on their experience with CWDM filters. But then we pointed out that
we need a lot more rejection at the reject band (probably on the order of
50 dB), and we have to do it at ONT prices. This made them worry.
Right now, with current-generation product, we have a transition region from
1500 to 1550 nm - 50 nm transition - and we are talking about
reducing it to 20 nm or so. I hope someone with filter design experience
can give us guidance. We continue to seek same. In answer to Marek's question, while we
normally talk about 1550 nm for broadcast use, the actual band over which
transmitters are available is 1550 - 1560 nm, corresponding to the actual
amplification band of commonly-available EDFAs. It may be possible to
restrict the actual wavelengths used to 1550 - 1555 nm. (The transmitters
use a cooled DFB laser with good stability, so the occupied band is relatively
narrow, but transmitters are produced at various wavelengths.) I have
already floated this idea by the RFoG reflector, and as usual, initial reaction
was mixed but generally unfavorable. But I think we could ease the
filtering requirements some by doing this. Now if I just knew where the
cost curve flattens out, so I'd know if I was tilting at windmills or
addressing a real concern... Regarding the bending issue, I think the
primary issue occurs in or near the end customer, and this is likely to be new
fiber anyway. We have bend-insensitive fiber available now from all
vendors. This thread seems to have bifurcated last
week due to time zone differences. I think I have all the responses
below, in the correct temporal order. Thanks, jim From The G.652B standardization added the PMD
requirement in October 2000. Prior to the above date, there was not a
standardized method to deterimine PMD. Different manufacturers tested
differently. So, fibers made earlier may have met the standard but they
were not measured. G.652B fibers can be integrated with later fibers
such as G.652D; however, system capabilities are limited to that of the legacy
fiber. Jim Farmer, K4BSE From: Dear Pete, Thank you for your additional
information. I had missed that 2006 amendment on the splitters. On the status of real PON deployments, it
is difficult to get really solid information. Since I work for an
equipment vendor company, I don’t have direct access to what the
operator’s measure on their fibers. All I have to go on is what
they tell me. What they tell me is mixed.
Many companies are deploying PONs which are guaranteed to work out to 1600nm.
Other companies only go out to 1580nm (at least, that is what their
specifications / measurements go to). Yet others are interested in using OTDR
equipment, and so the long wavelengths need to be reserved for that. Actually, I’m glad that we are
having this discussion – hopefully the interested parties who have access
to the data will come forward, and we can hear the first-hand story of what is
really out there. Based on that information, we can make a better
informed decision. Sincerely, -------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- From: Marek Hajduczenia [marek_haj@xxxxxxx] Dear all, Speaking as a member of this group, I
believe Frank has one very strong point which we cannot forget about - at the
beginning of the project we set forth to provide 10G-EPON specs which will be
backward compatible with the existing 1G-EPON deployments. If we are to live up
to that promise, I would expect nothing less than a situation in which I can
replace ONUs and OLTs on both sides of the ODN and run 10G-EPON at the magnificent
data rate of 10.3125 Gbps. As a carrier, I would not want to go out and
requalify my ODN. I would not want to go and change my splitters. As a matter
of fact, I would not want to touch anything apart from ONUs and OLTs. That said, consider how much fiber is
deployed and which does not meet more stringent G.652-B specifications. Some of
carriers who enter the fibre deployment phase only right now will not have any
problem meeting 1625 nm marker (newer fiber and PLC splitters), other will have
tremendous constraints moving past 1580 nm, for whatever technical reason
(ODTR, additional filters, poor fiber or splitters etc. etc.), but mainly due
to the fast that their fiber is simply older and will not qualify (probably)
for more stringent G.652-B specifications. There is a parallel discussion on
the same topic going on at FSAN and I believe (Frank, correct me if I am
mistaken here) this is exactly what the conclusions were. We do not have
however the possibility of defining optional wavelengths for the same PMD, as
some are proposing for FSAN NG-PON. As far as I understood IEEE approach to
defining specifications, we try to boil the available options down to the set
of common features. That means that we need to guarantee support for a more
narrow wavelength range in this case i.e. stick with 1580 nm limit upper, at
least for certain PMDs. Our project was born with a set of
constraints we have to live with, when considering the target deployment
scenarios. Additionally, we have to also look at what is happening on the
market of optical components. I agree with Frank's comment from the last
recirculation here - I have been also tracking progress in several
companies developing optical subassemblies and for now I have only seen products
focusing on 1574 - 1580 nm band, while 1580 - 1600 nm band seems to be
pushed much further in the future. Additionally, companies seem to be unwilling
to invest into development of components the volume of which may be much more
limited. In short, most people I have spoken to so far seem to bet more on
PR(X)30 type devices than 10 and 20 classes. That seems to follow the trend
from the past and if we believe it happens, than application will indeed drive
adequate volume to get the cost down at the target wavelength, despite the fact
that it might be more narrow than standard CWDM band of 20 nm. If we are to
look at the problem at hand through the prism of backward compatibility, then
indeed PR(X)30 type PMDs are in a better position to be deployed, mainly
because of the quantity of compatible ODN. The main problem I see here is that
we are trying to predict what market does at times when I think we all learnt
that market is not predictable. An interesting point to note in favour of
PR(X)30 PMDs: if any of the current GPON favouring carriers were to switch to
EPON technology, PR(X)30 would be exactly the PMD they would need to use. Much
as it seems unlikely, I did receive querries from a few large carriers on such
a possibility. In such a case, PR(X)10 and 20 PMDs could not be used for
obvious reasons. Additionally, when considering harmonization with FSAN/ITU-T
NG-PON systems, PR(X)30 PMDs are yet again the ones which will receive the main
focus, since they are considered for reuse for XG-PON systems. Should that move
forward, this will represent a clear signal to the manufacturers to focus on
these devices. In the long run, the cost of a more narrow band PR(X)30 OLT PMD
may be comparable if not lower than uncooled DML based PR(X)10/20 PMDs. Jim and
Alan are right here - the bigger the volume and more companies in the segment,
the better the prices. Concerning filter cost: I have seen
already presentations on this topic done at our group and at FSAN and
conclusions are contradictory. Some say a decrease to 15/14 nm isolation
between video and data channel should have minimum impact on the filter
cost, others say it is going to hurt us bad. If I recall right, we had only one
presentation on this topic at our group. I am not an expert in this area
myself so I rely on others for data and information. Yet in the view of
conflicting information, I find myself wondering where we really are.
Perhaps people workign with filter design could contribute with more down to
the ground relative cost comparison for 20, 15, 10 nm channel separation ? Just a mental note - aren't RF Video
systems supposed to operate at 1550 nm ? Who would push them to 1560 nm ? Regards Marek From: Pete Anslow [mailto:pja@xxxxxxxxxx] Frank, While not wishing to enter the debate on
what the wavelength choice for 10G EPON should be, your statement “The G.671 standard for couplers specifies the two traditional
windows: 1260~1360 and 1480~1580. So, if we continue to specify that our
ODN is composed of fibers recommended in G.652 and couplers recommended in
G.671 then we really shouldn't go beyond 1580nm.“ is not really correct. G.671 Amendment 1 (03/2006) available
here: http://www.itu.int/rec/T-REC-G.671-200603-I!Amd1/en adds an “Optical branching component
(wavelength non-selective) for PONs” which is specified from 1260 to 1360
nm and 1450 to 1600 nm. For the loss of G.652 type fibre, some
information was presented in: http://www.ieee802.org/3/av/public/2007_03/3av_0703_anslow_1.pdf see slides 5 to 11. The red curve on slide 6 is derived from
measurements of 1549 fibres at 1550 and 1625 nm together with full spectrum
measurements of uncabled fibres. The only room for uncertainty between
these results and your requirements is whether the bends seen in a PON
environment produce a significantly different curve from those seen in the
transport environment. Regards, Pete Anslow Nortel Networks UK Limited, External +44 1279 402540 ESN 742
2540 Fax +44 1279 402543 From: Dear Group, Let's not forget that it is not just
splitters and the raw fiber we need to worry about at 1600nm, but also the
effects of bends. And that is something that is more field operations
dependent. This is the biggest uncertainty that has made my operator
contacts concerned about using wavelengths longer than 1580nm. I will
point out that the data I’m getting from network operators is very
sporadic and somewhat contradictory. But, if there is doubt, then I think
the right way to go is to assume the worst case. There is another thing to consider:
We may consider proprietary specifications all we want, but in the end we
should be more standards driven. If I look at the G.652-A fiber
recommendation, all it tells me is that the loss (both basic and with bends) is
specified at 1550nm at the longest wavelength. Only G.652-B specifies the
losses at 1625nm. The G.671 standard for couplers specifies the two
traditional windows: 1260~1360 and 1480~1580. So, if we continue to
specify that our ODN is composed of fibers recommended in G.652 and couplers
recommended in G.671 then we really shouldn't go beyond 1580nm. The alternative would be to require the
fiber to conform to G.652-B, and then to require the couplers to conform to a
revised G.671. But that sort of defeats the whole idea of backward
compatibility with existing PONs, which did not have such restrictions.
So that is our problem. I realize that the window is somewhat
narrow, but it seems to be the only safe choice at this point. Sincerely, Frank E. From: Jim Farmer [mailto:Jim.Farmer@xxxxxxxxxxxxx] Before submitting a formal comment, we wanted to run this by
the reflector for comment. Regarding the upstream wavelength action
reflected below, from the September meeting in Cost and power dissipation: The use of a tight
wavelength band (e.g., 1577 +/-3 nm) at any wavelength is going to
significantly increase the cost of the laser, both due to the manufacturing
tolerance imposed and by the need for heating the laser. Any DFB laser so
far as we know, has a temperature drift of about 0.1 nm/deg. C. If one
were to provide for a normal indoor temperature range, which can cover 60
degrees, then one will have a 6 nm wavelength shift due to drift. We have
found that there is market demand for wider temperature range OLTs, which can
easily have a wavelength drift of 10 nm over their operating temperature
range. Of course, the answer is to temperature-stabilize the laser, but
this requires an added heater/cooler and control, and the best estimate we can
make right now is that it will add 2-4 watts per PON to the OLT power
dissipation when operated at room temperature. Granted, this is small
compared with the total power dissipated by the OLT, but in an age in which we
are trying to minimize the power draw of all equipment, it is going in the
wrong direction. And it will cost money, not to mention that the laser
will have to be specified to a tight wavelength tolerance (or the wavelength
tuned by the cooler), adding more cost. Concerning the availability of devices, we expect that this
application will drive adequate volume to get the cost down at whatever
wavelength we choose. We’ve been told by one laser manufacturer
that the wavelength specified has little bearing on cost, though of course,
tolerance and volume play a big part in determining the cost. Specification: We question the statement that fiber and
couplers are not fully specified beyond 1580 nm. Some people are using
1610 nm for OTDRs. A check with a couple of major manufacturers of fiber
and couplers indicates that specifications are controlled out to 1625 nm.
See, for example, http://www.corning.com/WorkArea/showcontent.aspx?id=15535, and http://www.ofsoptics.com/resources/AllWaveFLEX136web.pdf. As for couplers, I understand that planar couplers are fine
over this wavelength range, though fused biconic couplers may not be as good. Yet another concern we have with the 1577 nm selection is the
filtering when a 1550 nm video carrier is used. The video optical carrier
can be as high as 1560 nm. For a 1577 nm downstream data carrier, which
can go as low as 1574 nm by specification, the WDM to separate the wavelengths
has a 14 nm transition region. If we use 1590 +/-10 nm for data, the
transition region expands 43%, to 20 nm. This can help reduce ONT cost,
as well as the cost of the WDM at the OLT. For these reasons, we seek reconsideration of the decision to
abandon 1590 nm. We would like to receive comments from the reflector,
then we are planning to submit a formal comment before the deadline. Thanks, Jim
Farmer, K4BSE |