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

Re: [8023-10GEPON] power budget ad hoc teleconference



 Dear Dr. Hamano:  

Thank you for your constructive criticism.  I cannot agree with your
conclusion however, as I outline below.

With regard to your first point - I think we generally agree.  In fact I
stated that 

"(an unfiltered SOA + PIN) by itself is a slight improvement over an APD
at 10 Gb/s"

I do not suggest using an unfiltered SOA, and I apologize for not making
this more clear.  Any number of alternate filters can improve upon the
situation, but for sake of argument, consider a filter that has no
attenuation in the 10 G band (1304 - 1316 nm) but 3 dB of attenuation
everywhere else.  Such a filter would improve the 10 G sensitivity,
while making the 1.25 G (out of band sensitivity worse.)  Since we have
sensitivity to spare on the side of the 1.25 G (unfiltered) PIN + SOA
side such a filter could improve the net performance.

In the second item I did give a rough, meaning heuristic, or "back of
the envelope" calculation for unfiltered SOA + PIN sensitivities.  I
neglected to go into detail in the interest of space.  My conclusions
can be backed up by more detailed calculations, see for example -
Takashima, et al. paper ThG3, OFC2003; Ruhl and Ayre, PTL, p. 358
(1993), etc...

I should also note the that the dependence of receiver sensitivity on
SOA gain is small.  Recall that since the receiver noise is RIN-limited,
not thermal limited, Q is not a strong function of receiver power.

I will assemble a more detailed response, and distribute it on the email
list before the ad-hoc conference call.  I will be on the conference
call this week, to address any questions that may arise.  If there are
other concerns I will try to prepare a response before the conference
call.  

Considering the alternatives (insufficient optical power to meet link
budget, sensitivity to ONU costs vs OLT costs, and the 1G/10G dual rate
"penalty" and associated "dynamic" TIA for mitigation), I think the SOA
+ PIN approach has significant merit.

With best regards, 

David




David Piehler
Alphion
mobile: +1 732 692 4581

-----Original Message-----
From: Hiroshi HAMANO [mailto:hamano.hiroshi@JP.FUJITSU.COM] 
Sent: Monday, April 30, 2007 10:30 AM
To: STDS-802-3-10GEPON@listserv.ieee.org
Subject: Re: [8023-10GEPON] power budget ad hoc teleconference

Dear Dr. Piehler and Ad hoc members,

I would like to make several comments on your summary about SOA preamps.

in 10G OLT receiver.  To tell the truth, now I feel that non-filtered
SOA preamp. for OLT receiver is not effective and realistic for 10GEPON
upstream.

1. Sensitivity with non-filtered SOA
> 10 Gb/s data presented at the November meeting 
> (3av_0611_spiekman_1.pdf; page 19), show that at SOA with no optical 
> filter between the SOA and the PIN; can improve the sensitivity of a 
> "bare" by 7 dB.  (These results were by no means remarkable (e.g. 8 dB

> SOA noise figure) and similar results can be found in the literature.

In the presentation material referred above (3av_0611_spiekman_1.pdf;
page 19), -27dBm sensitivity is shown as the non-filtered SOA with 80nm
3dB(!!)-bandwidth.
But it shows the sensitivity at the BER of 10-9, which should be 2dB
worse at 10-12.  Regarding 1dB polarization dependent gain, its
sensitivity would be -24dBm, which, I believe, is no better than the APD
receiver sensitivity.

2. Receiver dynamic range
> Now plug in whatever numbers you want for 10 and 1.25 Gb/s PIN 
> sensitivities - I'll use -18 dBm (10G) and -25 dBm (1.25 Gb/s) - The 
> result is a receiver with -32 dBm sensitivity at 1.25 Gb/s and -25 dBm

> at 10 Gb/s.  This by itself is a slight improvement over an APD at 10 
> Gb/s (Frank used -24 dBm in his presentation.)  At 1.25 Gb/s 
> sensitivity this is > 2 dB better than the critical 29.7 dBm 
> requirement for the legacy GE-PON.

I feel somewhat strange in your explanation in the paragraph here.
SOA receiver sensitivity shall be determined by SOA noise, and cannot be
expressed as an additional increment from the PIN receiver sensitivity.
To avoid the degradation caused by PIN receiver noise, SOA gain should
be selected to adjust SOA output power into the perfect error-free range
of the PIN receiver input, 3 or 4 dB away from 10G sensitivity at BER
10-12.  
If you use -18dBm as the 10G sensitivity, PIN receiver input power,
shall be around -12dBm, also considering additional margin for SOA gain
deviation and lifetime degradation.
(I suppose that -27dBm SOA sensitivity in your literature was achieved
on such a similar experiment set-up.)
 
In the PON system application, almost 20-dB dynamic range is required
for OLT receiver to handle the distance variety of ONUs, their output
power fluctuation, and also different ODN configurations.  
As SOA is only a simple fixed-gain block, PIN receiver input also
changes 20dB, up to +8dBm, heavily exceeds the receiver overload. 

I believe that high-power optical source development for 10G ONU
transmitter is anyway indispensable, and as it takes years to develop,
it is significant to show optical component suppliers the development
target, at this time of the moment in Draft.1.
I appreciate any comments or discussions.

Best regards,
Hiroshi Hamano
Fujitsu Labs. Ltd.

---
---------------------------------------------
HIROSHI HAMANO         Network Systems Labs.
FUJITSU Labs. Ltd., Kawasaki, 211-8588 JAPAN
TEL: +81-44-754-2641  FAX: +81-44-754-2640
E-mail: hhlsi@flab.fujitsu.co.jp
---------------------------------------------

%% David Piehler <dpiehler@ALPHION.COM>
%% Re: [8023-10GEPON] power budget ad hoc teleconference %% Tue, 10 Apr
2007 22:55:11 -0400

> I believe that a SOA-PIN combination receiver may have better 
> sensitivity by several dB over an APD receiver for the upstream for 
> the
> B++ situation with 1G/10G coexistence.
> 
> At Robert's request, here is a summary of what I stated on tonight's 
> call with some additional backup information.  In this note, I will 
> make use of "back of the envelope" calculations.  However, I have also

> made more detailed models, and we have experiments under way to verify

> predictions.  Also I assume no FEC in any of the following.
> 
> In addition, to reducing the power requirements on the ONU 
> transmitter, such a receiver would be RIN-noise limited, and not 
> suffer from the "dual rate problem" (see for example 
> 3av_0703_effenberger_4.pdf) - and the "smart TIA" would not need to be
invented.
> 
> The SOA had been previously discounted as a potential upstream 
> receiver since a "noise blocking filter" for the 10G upstream signal 
> (e.g. at 1310 +/- 6 nm) would potentially block the legacy GE-PON 
> upstream signals (1260 - 1360 nm).
> 
> 10 Gb/s data presented at the November meeting 
> (3av_0611_spiekman_1.pdf; page 19), show that at SOA with no optical 
> filter between the SOA and the PIN; can improve the sensitivity of a 
> "bare" by 7 dB.  (These results were by no means remarkable (e.g. 8 dB

> SOA noise figure) and similar results can be found in the literature.
> 
> Roughly speaking a 1.25 Gb/s PIN would see a similar improvement with
an
> unfiltered SOA as a pre-amplifier.   
> 
> Now plug in whatever numbers you want for 10 and 1.25 Gb/s PIN 
> sensitivities - I'll use -18 dBm (10G) and -25 dBm (1.25 Gb/s) - The 
> result is a receiver with -32 dBm sensitivity at 1.25 Gb/s and -25 dBm

> at 10 Gb/s.  This by itself is a slight improvement over an APD at 10 
> Gb/s (Frank used -24 dBm in his presentation.)  At 1.25 Gb/s 
> sensitivity this is > 2 dB better than the critical 29.7 dBm 
> requirement for the legacy GE-PON.
> 
> These results can be improved upon by inserting a "noise blocking 
> filter".  The filter certainly cannot completely block the 1260-1360 
> nm legacy GE-PON signal, but since there is some 1.25 Gb/s sensitivity

> margin to spare, one can attenuate some of the 1.25 Gb/s signal that 
> lies outside of the wavelength spec of the 10 Gb/s signal.
> 
> As a result sensitivities of -30 dBm at 1.25 Gb/s and -28 dBm at 10 
> Gb/s should be possible using a SOA-PIN combination, appropriate 
> optical filter, and a shared TIA, optimized for 10 Gb/s operation.  
> Even better performance is possible with a more tightly wavelength 
> controlled 10 Gb/s transmitter.
> 
> The 100 nm operating bandwidth, should also not be a problem, see for
> example:
> 
> In-service upgrade of an amplified 130-km metro CWDM transmission 
> system using a single LOA with 140-nm bandwidth Iannone, P.; 
> Reichmann, K.; Spiekman, L.
> Optical Fiber Communications Conference, 2003. OFC 2003 Volume , Issue

> , 23-28 March 2003 Page(s): 548 - 550 vol.2
> Digital Object Identifier   10.1109/OFC.2003.1248402
> Summary: We demonstrate a 130-km metro CWDM transmission system using 
> a single LOA with 140-nm bandwidth. An in-service upgrade, for which 
> one of the eight 2.5-Gb/s CWDM channels is replaced with 8/spl 
> times/2.5-Gb/s DWDM channels, results in negligible performance 
> degradation.
> 
> As indicated above, more sophisticated models exist, and we are 
> experimenting with this idea.
> 
> I would appreciate any feedback or critisism.  We hope to share 
> additional data with this forum.
> 
> 
> 
> 
> 
> David Piehler
> Alphion
> 
> -----Original Message-----
> From: Lingle, Jr, Robert (Robert) [mailto:rlingle@OFSOPTICS.COM]
> Sent: Tuesday, April 10, 2007 12:09 AM
> To: STDS-802-3-10GEPON@listserv.ieee.org
> Subject: Re: [8023-10GEPON] power budget ad hoc teleconference
> 
> 
> All,
> 
> The next call(s) will occur on Tuesday April 10 at 7PM EST and 
> Wednesday April 11 at 9AM EST.
> 
> (Because of the US time change in March, I believe that 7PM EST 
> Tuesday is now 8AM Japan time on Wednesday.)
> 
> Toll-free in US	866-263-8899
> Outside US 	816-249-6061
> Conf. Code	7707985015
> 
> The following Orlando presentations, among others, may be relevant to 
> the
> discussions:
> 
> 3av_0703_effenberger_1
> 3av_0703_takizawa_1
> 3av_0703_jiang_1
> 3av_0703_stefanov_2
> 
> In the call, I would like to address the following items. Please 
> address them on the Reflector prior to the call.
> 
> 1. Much attention has been focused on the class B++ CH IL case.  The 
> only proposed power budgets for PX10 and PX20 are those in 
> 3av_0703_effenberger_1, slides 6 and 7. Are Frank's numbers for PX10 
> and PX20 met with components available today? Or would these also 
> require development as in the case of B++? Does anyone recommend any 
> changes to these? If not, should we plan to adopt them at the May
meeting?
> 
> 2. In analyzing the power budgets from effenberger and takizawa for 
> class
> B++, it seems clear (at least to me) that the following components 
> B++must
> be
> developed to meet the power budgets:
> 
> -- US 1300nm requires the development of a NEW DML, perhaps cooled, 
> that can output 3-4 dBm minimum at 1300nm, compared to a more typical 
> 0 to 1 dBm at present.
> -- DS w/ APD requires a NEW EML with +3 dBm minimum output power, 
> compared to a -1 to 0 dBm typical minimum output power at present.
> -- DS w/ PIN can be done with existing EML+EDFA; it MIGHT be done with

> existing SOA with Psat(1dB)=10dBm, but might require development of an

> SOA with higher Psat(1dB)=12 dBm.
> 
> Robert
> 
> -----Original Message-----
> From: Lingle, Jr, Robert (Robert) [mailto:rlingle@OFSOPTICS.COM]
> Sent: Tuesday, April 03, 2007 10:07 AM
> To: STDS-802-3-10GEPON@listserv.ieee.org
> Subject: [8023-10GEPON] power budget ad hoc teleconference
> 
> 
> All,
> 
> The next Power Budget Ad Hoc teleconference will occur on Tuesday 
> April 10 at 7PM EST and Wednesday April 1 tentatively at 9AM EST.  I 
> may have to travel Wednesday and ask your flexibility if I need to 
> push this time to 8:30AM or 9:30AM to accomodate.
> 
> The topics will be a review of power budgets presented at the March 
> Plenary by Frank Effenberger and Wenbin Jiang, as well as implications

> of the presentation by Stefanov on SOA output power issues.
> 
> More details will follow.
> 
> Robert
> 
> Robert Lingle, Jr.
> Fiber Design and Transmission Simulation OFS Corporate R&D, Atlanta
> 404-886-3581 (cell)
> 770-798-5015 (office)
> 
>