[EFM] RE: Single wavelength, single fiber PMD for P2P

["Anthony D. Anderson" <tonyshouse@xxxxxxxxxxxxx>]

P2P's
 
I reading the e-mails on various wavelength plans, and types of laser,
I'd like to offer my opinion which was discussed to some degree at the 
Raleigh, NC meeting. Many points have been raised, pointing out the
Pros and Cons of each variation
 
I supporting a single wave length 1310nm Tx and 1310nm Rx. Note this
is different from what Vipul was describing in his e-mail relating to single
wave length 1310. I believe he was describing a 1310+ and 1310- for
Tx/Rx. Where Tx & Rx never overlap. Technology exists to differentiate
a reflection from a signal being transmitted at the other end of a link.
For example it can be done with link design and signal/reflections being
maintained below 6db (As described in Vipul's Portland presentation).
Its also possible to differentiate between reflections and a signal being
transmitted over a 10KM link via the LOS (threshold) indicator. FP could
be used at either end of the link, the limiting factor will of course be the
PMD in an ONU where it may be located outdoors. Since we have
spec'd the ONU to operate from -40 to +85C the total range is 125
degrees, as described in Vipul e-mail. It is certainly not necessary
to double the number if we are not defining a 1310+ / 1310- P2P link.
Even with a 125C Temp the delta wave length range is conservatively
only 62.5 nm. (Using 0.5nm/C specificed in Vipul's e-mail, although
I think the wave length drift is closer to 0.4nm/C) We voted on case
temp rating versus ambient for the PMD, the FP laser and IC junction
temp's will be at least 10-20C higher within 10-20 seconds of power
up. In operation, their will be a theta C-A Case to Ambient of 5-30C/
Watt, the smaller the heat spreading area the greater the delta between
Case and Ambient temperature. We also should consider the theta
J-C Junction to Case rise. If it's an IC mounted in a package with a
thermal heat slug, the theta C-A is very low about one degree C/W,
however most of the components that go into a module are ether bare
die mounted to an internal surface, or are in plastic packages where
theta J-C is large. This just means that the active devices operate at
higher Temps then either the Case or Ambient Temperature. I believe
the 0.4nm/C I mentioned is not linear, it decreases as temperature
increases. In short the max wave length drift the FP would be
subjected to depends on the worst case delta nm/C, at 0.5nm/C a
FP centered at 1310nm would drift +/- 31nm or 1279nm to 1341nm.
At 0.4 nm/C it would only drift +/- 25nm or 1285nm to 1335nm. These
are worst case numbers, sample measurments are better. I hope this
resolves the wavelength drift over temperature issue for FP's but I
welcome additional data or feedback. 
 
Another issue raised was the "benefit" of having a single PMD for P2P and
P2MP. Someone said the same thing I said on this issue just the other
day which was "Lets stop trying to design a Swiss army knife." We will
probably need a different PMD for each application. The P2MP PMD's have
different needs and requirements, for one, the laser driver needs a control to
shut it down. P2P links do not have these requirements. But what does
make since is the fact that a true 1310/1310nm Tx/Rx single fiber solution
only requires a single PMD design for both end's of the link. This has several
advantages, only one PMD to keep in inventory, and fewer mistakes during
installation, since there is only one PMD to chose from and only one fiber to
connect to it. Finally, a 1310/1310nm link is much easier to add a 1550 video
overlay, then a 1310/1490nm link.
 
One other point on reflections, the carriers I've talked with say their existing
networks have 20 dB ODN-ORL, thus a 10KM run is no problem. They also
say that new installations will have 30 dB ODN-ORL, which makes hitting a
reach of 20KM much easier. Of course MPN is the actual limitation for
20KM links, and agreeing on a smaller K value is key to getting there.
 
In short for P2P the least expensive solution available today is FP's, using

single mode, single wave length, single fiber networks.
 
Regards,

 

Tony Anderson

 

I've attached some of the e-mail treads from Jack and Vipul on this topic below.
 
In a message dated 2/14/02 3:54:43 PM Pacific Standard Time,
Jack.Jewell@xxxxxxxxxxxxx writes:

 

<BLOCKQUOTE TYPE=CITE style="BORDER-LEFT: #0000ff 2px solid;
MARGIN-LEFT: 5px; MARGIN-RIGHT: 0px; PADDING-LEFT: 5px">Subj: RE:
[EFM-P2P] P2P link design considerations
Date: 2/14/02 3:54:43 PM Pacific Standard Time
From:    Jack.Jewell@xxxxxxxxxxxxx (Jack Jewell)
Sender:    owner-stds-802-3-efm-p2p@majordomo.ieee.org
To:    Vipul_Bhatt@xxxxxxxx ('Vipul_Bhatt@xxxxxxxx')
CC:    stds-802-3-efm-p2p@xxxxxxxxxxxxxxxxxx (P2P reflector)

 

 

 

 

 

Vipul,

 

Thanks for this effort and investigation.  One correction: the 3rd
option of
a dual-wavelength solution with both in the 1310nm band EXCLUDES
Fabry-Perot
lasers because the FP wavelength drift with temperature is so large that

 

just one channel uses up the entire 1260-1360nm band (for the extended
temperature range sought).  The dual-1310 solution is the one best
suited
for VCSELs, and is one I suggested in March, but the current status of
1310nm VCSELs makes it a tough argument (despite large recent progress).
It
can also be met with DFB lasers, but cost issues exist just as they do
at
the 1490nm end of the dual-wavelength (1310/1490) P2P approach.  The
dual-1310nm solution is somewhat similar to the CWDM solution of 10GigE.

 

To summarize, the dual-1310 option will be attractive if: 1) people
accept
that directly-modulated DFBs can be cost effective, OR 2) that 1310nm
VCSELs
will arrive in time, OR 3) if a "DFBs today; VCSELs tomorrow" attitude
prevails.  I agree with the concensus that EFM products will have a long

 

product lifetime and the long-term optimization should weigh heavily on
the
choice of PMD.  As I see it so far, either of the 2-wavelength
approaches
would fill that bill.

 

Jack

 

 

 

-----Original Message-----
From:     Vipul Bhatt [mailto:Vipul_Bhatt@xxxxxxxx]
Sent:    Thursday, February 14, 2002 3:21 PM
To:    P2P reflector
Subject:    [EFM-P2P] P2P link design considerations

 

<< File: EFM P2P Options Comparison.ppt >> Dear colleagues,

 

Yesterday, I met with about 20 members of the 802.3ae (10G
Ethernet), all PMD experts. They were meeting in Santa Rosa to
advance their 10G PMD work. I thank Jonathan Thatcher, the chair of
802.3ae, for arranging it. I gladly grabbed that opportunity because
it was a chance to gain useful advice from the Green Berets of the
PMD Planet! This message summarizes the discussions that took place.
The presentation I gave to trigger the discussions is attached.

 

Motivation: While PMD proposals that address our extended
temperature objective and P2MP objective seem to be enjoying good
consensus, the P2P objective is drawing more than one suggestions on
how to specify the right PMD. Yes, it's important to meet our March
deadline, but it's even more important to do it with well-considered
proposals. The fact that many competent members are unable to
converge on a single P2P proposal indicates the need for more
technical and economic analysis of the problem. Considering the
deadline, we need to work faster. So I decided to go.

 

I began by presenting the table with two options: 1-wavelength P2P
link at 1310 nm, or a 2-wavelength P2P link. In the 2-wavelength
case, I made the initial assumption that the wavelength plan would
be 1310/1490, because that would allow us to have a PMD at the ONU
end that can be the same for both P2P and P2MP. The leveraging of
scale economies seemed like a good idea to me. The table is my
fabrication; any factual errors are mine. The choice of 1490 nm
implied the use of either a DFB laser or a VCSEL.

 

First off, the discussion led to the realization that there may be a
third good option - a two wavelength PMD that uses both wavelengths
in the vicinity of 1310 nm. This will permit the use of Fabry Perot
lasers, the most cost-effective and proven lasers available today.
To reduce the risk of worst case wavelengths being too far from the
zero-dispersion point, some helpful assumptions about temperatures
at the two ends of the link can be made. For example, when one
transmitter is at +85C, the chances are slim that the transmitter at
the other end is at -40C. Thus, when counting wavelength drift with
temperature, we can partially  overlap the temperature ranges of the
two ends. (This idea is promising, but before you can make a votable
proposal out of it for March, you will need to crunch some link
budget spreadsheet numbers. If you want to work on it, let me know
how I can help.) With that issue aside, we moved on to discuss the
two options as presented.

 

The group agreed that longwave VCSELs will not be ready for prime
time soon (prime time meant high volume, low cost, high yields,
proven robustness over temperature, wide acceptance among
customers). Only one member said they will be, by December 2002.
More (8) agreed that by December 2003, they may be.

 

And yet, when asked to choose between one-wavelength and
two-wavelength options, majority (8 to 6) chose the two-wavelength
option.

 

Huh? I asked, how do you explain that? The answer was
thought-provoking: We should define a single-fiber P2P link that is
forward looking. Longwave VCSELs and/or low cost DFB-isolator
combinations will be available in the near future. Products based on
the EFM standard will have a long market life. As for the present
need for P2P deployments, the 2-fiber temperature extended LX will
do just fine. Therefore, we can afford to wait. ("Two-fiber
one-wavelength now, one-fiber two-wavelengths later.") Technical
feasibility can be proven at a later stage in the standards
timeline, but for Draft 1.0, the 2-wavelength option should be
included.

 

I then asked how they ranked the technical risk of option 2,
compared to option 1. Answer: If the risk level of option 1 is 1.0,
then the risk level of option 2 is about 0.5. Reason? Mostly, the
fact that the reliability and performance of a single-wavelength
link depends on sustaining the high return loss of the PMD at the
media interface. But a dust particle just a few microns in size can
produce an air gap big enough to cause reflections which will make
the return loss out of spec. Even if the PMD-media interface is kept
within spec, every potential connector joint in the cable plant,
including the often-used patch cord, is a place where dust particles
can get in and defeat the high return loss assumption. This is not
to say that high return loss implementations are not feasible; this
is only to point out that Ethernet deployments can happen in huge
volumes, and the statistical risk of dust particles causing failure
in even a small percentage of links would be unacceptable.

 

This concludes my report. I hope you find it useful in making an
informed decision. Whatever your preference, please remember that we
can only vote on proposals - no proposals, no votes.

 

Regards,
Vipul

 

vipul_bhatt@xxxxxxxx
408-857-1973

 

_______________________________________________________________

 

In a message dated 2/15/02 1:29:32 AM Pacific Standard Time,
Vipul_Bhatt@xxxxxxxx writes:

 

Subj: RE: [EFM-P2P] P2P link design considerations
Date: 2/15/02 1:29:32 AM Pacific Standard Time
From:    Vipul_Bhatt@xxxxxxxx (Vipul Bhatt)
Sender:    owner-stds-802-3-efm-p2p@majordomo.ieee.org
Reply-to: Vipul_Bhatt@xxxxxxxx <mailto:Vipul_Bhatt@xxxxxxxx> 
To:    stds-802-3-efm-p2p@xxxxxxxxxxxxxxxxxx (P2P reflector)

 

 

 

 

 

Jack,

 

Let me try a thought on you: if we get aggressive with FP specs, it
may be possible to accommodate FP lasers in the 2-wavelength
solution in the 1310 nm band. Here's my argument.

 

We assume that the wavelength drift with temperature for FP lasers
is 0.5 nm/C. The extended temperature range is 125C wide. For two
lasers, one at each end of the link, it may be overly conservative
to allocate a wavelength span capable of handling 250 C temperature
range. This is because when a laser at one end of the link is at its
extreme cold, it is highly unlikely that the laser at the other end
will be at its extreme hot. We can overlap the two temperature
ranges a bit, for a total planning range that is less than the sum
of the two. One can also argue that the OLT end PMD will need to
operate over a much smaller temperature range, say 70C. After all,
switches are normally rated for 0 to 70C, and the OLT end PMD will
be in a switch. (I am assuming that if it is housed in a Remote
Terminal enclosure outdoors, it will be temperature controlled.)
Suppose we work out the details and end up with a total temperature
range of 180C. This translates to 90 nm wavelength range. Suppose we
give back another 20 nm in manufacturing tolerances for laser
wavelength and WDM filter spacing (aggressive, yes). This creates a
goal of 110 nm wavelength span.

 

Now we push on the transmitter specs. Dare I say: 2.5 nm rms
spectral width, k factor of 0.4, rise time of 150 psec. (I can
half-heartedly defend these values if necessary, but for now, I will
keep the message short.)

 

Given all that, it may be possible to span a range of, say 1265 nm
to 1375 nm, while still keeping MPN penalty to a manageable ~1.6 dB.

 

Is my optimism blinding me, or do you think this is a promising line
of thought?

 

Regards,
Vipul

 

vipul_bhatt@xxxxxxxx
408-857-1973

 

=====================

 

>
>
> Vipul,
>
> Thanks for this effort and investigation.  One
> correction: the 3rd option of
> a dual-wavelength solution with both in the 1310nm band
> EXCLUDES Fabry-Perot
> lasers because the FP wavelength drift with temperature
> is so large that
> just one channel uses up the entire 1260-1360nm band (for
> the extended
> temperature range sought).  The dual-1310 solution is the
> one best suited
> for VCSELs, and is one I suggested in March, but the
> current status of
> 1310nm VCSELs makes it a tough argument (despite large
> recent progress).  It
> can also be met with DFB lasers, but cost issues exist
> just as they do at
> the 1490nm end of the dual-wavelength (1310/1490) P2P
> approach.  The
> dual-1310nm solution is somewhat similar to the CWDM
> solution of 10GigE.
>
> To summarize, the dual-1310 option will be attractive if:
> 1) people accept
> that directly-modulated DFBs can be cost effective, OR 2)
> that 1310nm VCSELs
> will arrive in time, OR 3) if a "DFBs today; VCSELs
> tomorrow" attitude
> prevails.  I agree with the concensus that EFM products
> will have a long
> product lifetime and the long-term optimization should
> weigh heavily on the
> choice of PMD.  As I see it so far, either of the
> 2-wavelength approaches
> would fill that bill.
>
> Jack

 

Tony Anderson

 

Dir. PON Technical Applications
ZONU