Thread Links | Date Links | ||||
---|---|---|---|---|---|
Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
This is in line with what I interpreted Jim’s
argument earlier. Actually I think option (3) is part of
option (2), the only difference is option (3) favor one implementation using
cooled TX, which significantly increase OLT TX cost and power. While option (2)
is more generic, you can use either TX implementations (cooled, semi-cooled, or
uncooled). From practical perspective, PR(x)30 will possibly
dominate 95% of the actual deployment. And I feel PR10/20 for most of us is a “whatever”.
In light of this, I would rather the specs
to be more generic for PR10/20 to at least possibly cover uncooled TX. So I tends toward we should switch back to
option (2) which is D2.0. The question still- Can anybody elaborate how PR10/20 going to
be deployed, and how it maters? Frank From: Marek
Hajduczenia [mailto:marek_haj@xxxxxxx] Jim, so let me try to wrap it all up because I
must admit I am getting lost in here myself. We have now three options on the
table (1) use wavelength allocation plan as per
D2.1 (all 3 PMDs use 1577 +- 3 nm) (2) use wavelength allocation plan as per
D2.0 (1577 +- 3 nm for PR(X)30 and 1590 +- 10 nm for PR(X)10/20) (3) use something in between which You
propose i.e. 1577 +- 3 nm for PR(X)30 and 1590 +- 3 nm for PR(X)10/20 Did I get this right ? Now, in the course
of discussion we acquired one new option, which has never been on the table. As
much as it seems to make sense, I am wondering what is the added gain of
specifying new 6 nm band for PR(X)10/20 instead of using the same band as for
PR(X)30. Distance between video overlay edge and the edge of digital data
channel is the only things that comes to my mind. Did I miss anything ? Note also that we have a number of
contradictory comments in our pool for this meeting - some calling for option
(2), some calling for option (3). Regarding the cost of the resulting
filters and the impact of separation gap between video channel edge and digital
channel edge (using relative cost values) - can we have reliable estimate of
how much it would cost to have various filter options manufactured using e.g.
20 nm CWDM like type 1580 - 1600 pass band filter for D2.0 PR(X)10/20 ONU as
the base cost ? I have seen a number of presentations on cost filters and
usually channel separations on the order of 15 nm between the edges of adjacent
bands are quoted at 110 - 130% of costs of filter with 20 nm separation between
the edges of adjacent bands. Frank I believe can confirm that. A question then:
are they plain old wrong and the cost hit is much greater ? Do You have a way
to provide solid values for relative cost in such cases ? Thank You for this interesting discussion.
I must say that it has been a while since a topic generated so much
traffic on the reflector. regards Marek From: Jim
Farmer [mailto:Jim.Farmer@xxxxxxxxxxxxx] Thanks for the good comments on this
thread. We would like to add a bit more clarification or our
position before we meet next week. We believe that PR(x)30 can stay at 1577
nm. We would like to see the ability to operate PR(X)10/20 on 1590 +/-3
nm. We would like to say 1590 +/-10 nm, but this is not feasible.
It requires a wider filter bandwidth, which is more difficult, as has been
pointed out. This is one of the reasons why we reduced the proposed
occupied bandwidth of the 1590 nm option to +/-3 nm, even though this requires
a cooled laser at the OLT. We have reviewed the presentations
from May 2007. While we agree in principle with the presentations, it
seems that they do not take into consideration initial wavelength calibration
of the filter, nor do they take into account temperature drift of the
filter. And since the filter in question is at the ONU, it must be very
low in cost. A filter designed for either 1577 nm or 1590 nm must be
wider than the occupied wavelength range of the laser, in order to account for
initial calibration accuracy of the filter, and the temperature drift.
When we added these effects, the transition region of a 1577 nm filter (which
must attenuate the 1550 nm broadcast signal), became unacceptably small.
The transition region for a 1590 nm filter went to 14 nm, which is tight but
might be possible at ONU prices. This is what we show on the slides we
sent to the reflector earlier, and which we seek permission to present in We have been talking to filter experts
about how to make low cost filters that will meet the requirements. One
of the experts we have consulted is our parent company's Dr. Matt Pearson in "He's correct in his comments - 6 nm
is definitely easier than 20 nm (which is why we recommend it in Jim's
proposal!). "He is also correct that both the
filters and lasers are available to meet these specs. (DWDM relies on
that!). Our concern is more related to the costs, where DWDM costs are
outrageous, even CWDM costs are too high for FTTH. So we need noticeably
easier specs than CWDM. In fact, I would argue that we need noticeably
easier specs than today's FTTH..! -- There are so many more expensive
aspects to 10G than 1.25G (DFB, APD, 10G electronics, extra blocking filters,
etc), that if they want any hope of getting optics at a reasonable cost
then they have to compromise something somewhere... "Thin films and DFB lasers can meet
either spec. We believe (certain) PLC technology can also meet
either spec. But some other PLC approaches ... would quite
likely never meet these 10G specs.. So again, it limits the pool of
available suppliers and available technologies that could otherwise help bring
down costs for systems people. "Either way, we will make it work.
We're just trying to make (the cost of the ONU lower)..." Thanks, jim farmer Alan Brown Jim Farmer, K4BSE From: Frank
Effenberger [mailto:feffenberger@xxxxxxxxxx] Victor, I doubt that. >90% of EPON is
deployed in Back to the Mike’s suggestion
– while it is a good idea, it will work if the big concern is the
transmitter specifications. However, the latest comment from Jim
Farmer regards the filters at the ONU receiver. And defining a super-set
of the bands doesn’t help there. Actually, in my opinion, neither the
filters nor the lasers are that big of a deal. I’m not sure where
Jim’s filter data come from, but there are pretty standard thin-film
filter designs that can achieve the sharpness, accuracy, and temperature
stability that we need for 14nm of guard band. Our task force actually
got a model of this back in May of 2007. Actually, one of the
considerations in the difficulty of making these filters is the width of the
pass band, and it is actually easier to make a 6nm width pass band than a 20nm
pass band. Sincerely, Frank E. From: Marek
Hajduczenia [mailto:marek_haj@xxxxxxx] Hi Victor, That is how Mike sees it. That does not
need to be necessarily how things work out in the market. It seems to me that
we are trying to guess which direction the market goes and I think we all agree
that is hardly predictable. Additionally, if I recall right, we are not allowed
to discuss market shares so probably it is better to leave it at this ... Regards Marek From: Victor
Blake [mailto:victorblake@xxxxxxx] To chmine in here
– I’d have to say that to me it sounds like the 1577 is the
exception, not the 1590. -Victor From: Marek
Hajduczenia [mailto:marek_haj@xxxxxxx] Hi Mike, thanks for sharing Your point of view with
us. Please confirm whether I understand You
right. You say that we should go with a wider window and carriers may require
vendors to actually build equipment which complies to a certain part of this
sub-band. In our case, we could hypothetically specify a downstream band
between 1574 and 1600 nm while e.g. a narrow band option between 1574 and 1580
nm could be required by some carriers to remain compliant with their ODN. Is
this what You're trying to relay in Your email ? Please confirm Thank You Marek From: Mike
Dudek [mailto:Mike.Dudek@xxxxxxxx] As an outsider to 10GEPON, but member of
IEEE 802.3 working group I’d like to suggest that the IEEE standard
should be working to provide the best solution for the new future installs of
the IEEE standard while paying attention to the existing infrastructure.
When you come to a point that you are having to drive
the cost of the new standard higher in order to be compatible with existing
infrastructure that may or may not exist in many applications I’d suggest
that the IEEE standard should work for the long term low cost solution, while
making it technically feasible for people with the existing infrastructure to add
additional requirements to make it compatible with their existing
infrastructure. That way you do not burden the long term cost of
new installs. EG if the low cost solution needs a Tx
window of xnm to x+30nm but for compatibility with a non-IEEE standard can only
be xnm +10nm, then the IEEE spec should be xnm to x+30nm and individual vendors
that are using the non-IEEE standard can impose the tighter (subset spec) of
xnm to xnm +10nm. (This obviously only applies if the PAR and
objectives have not made compatibility with the non-IEEE standard a
requirement.). Please note my example is for illustration only the
numbers in it are not meant to apply to this specific question. From: Jim Farmer
[mailto:Jim.Farmer@xxxxxxxxxxxxx] My primary concern is that the 1577 nm
downstream wavelength is inconsistent with use of the 1550 nm broadcast
(auxiliary) wavelength. The problem is that the two wavelengths are too
close together to allow us to build economical filters at the ONU to separate
the two wavelengths. It is a little easier with the 1590 nm wavelength,
though it is still difficult. Originally I wanted to specify the
wavelength band as 1580 - 1600 nm as it was originally. But I found that
when I put in real filter characteristics, I still had an extremely narrow
transition region for the filter. So I accepted that we would have to narrow
the transmit window. I chose +/-3 nm (1587 - 1593 nm) following the
reasoning for PR(X)30. We are adding cost to the laser, but at the OLT,
which is not as cost sensitive as is the ONU. I also had to accept that the auxiliary
wavelength was limited to 1550 - 1555 nm, even though commercial practice is to
use wavelengths up to almost 1560 nm. People may complain about this
restriction, but I think in the end they will live with it. Unfortunately I have not been able to get
quantitative information on the filter complexity - I would like to see filter
vendors jump in with comparative numbers. Some vendors I spoke with gave
me more pessimistic numbers than I used in preparing the slides. So the application is for anyone who wants
to use the 1550 nm broadcast wavelength. This is the only way I see to
possibly make use of 1550 nm overlay practical. And it still demands a
more difficult filter than we demand currently. But presumably advances
in the state-of-the-art will made the filter practical at some point. Thanks, jim Jim
Farmer, K4BSE From: Frank
Chang [mailto:ychang@xxxxxxxxxxx] I just reviewed this thread, and my
interpretation to Jim’s slides is that- 1) The argument is not for PR(X)30 as cooled TX is assumed because of
tight power budget, so narrower 1577nm band considered feasible for PR(X)30. 2) For PR10/20, possibly uncooled optical sources are assumed, so
bring about the argument that larger wavelength band, such as wider 1590nm
band, is only feasible. To satisfy this argument, basically call
for the group to switch back to the wavelength plan originally specified in
D2.0. So actually we are re-visiting the argument the group made during the
baseline stage a year ago. Jim- Can you confirm this is what you are
looking for? As it is clear the PR(X)30 will be assumed
mainstream deployment which requires co-existence with installed 1G version,
can anybody elaborate the scenarios on how PR10/20 going to be deployed? My
question is weather PR10/20 scenarios has to use cooled or semi-cooled optical
source? ] thanks Frank C. From: Frank
Effenberger [mailto:feffenberger@xxxxxxxxxx] To pile onto this thread, I have a
question regarding Jim Farmer’s most recent presentation and
Maurice’s support of it: Did you notice that Jim’s
presentation is asking to change the PR10/20 OLT transmitter wavelength range
to 1587 to 1593nm? (At least, that is how I read it, but I
should say that the exact numbers are not clear.) Perhaps Jim can clarify exactly what he is
asking for… that would be helpful. Sincerely, Frank E> From: Marek
Hajduczenia [mailto:marek_haj@xxxxxxx] Hi Maurice, Just following the
arguments You used in Your email: does that mean that You see PR(X)20 OLT
transmitters as uncooled devices? Are the power levels we are targeting
achievable using uncooled optics? As far as I understand, cooling is necessary
not only to keep the central wavelength within the predefined range but also
assure higher output power level. Can You comment on this? Regards Marek From: Maurice Reintjes
[mailto:maurice.reintjes@xxxxxxxxxxxxx]
We request to make the attached presentation during the 10GEPON meeting
in Thanks, Jim Farmer, K4BSE |