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

[8023-10GEPON] 10GEPON ad hoc on high split ratio - Task 3.4/3.5/3.6 - RX technology



Dear all,

 

This is a corrected thread. Sorry for the previous mishap - I honestly do not know how it happened.

 

Maurice (thanks for very detailed comments and suggestions) suggested to divide the Task3 discussion into separate 2 threads. I believe that is a good point.

 

Here are once again my questions and the answers I received so far. I will use tags with names to indicate comments in the main question body.

 

1.        What kind of sensitivity can we expect from commercially available 10G PIN diodes. I saw figures in the presentations given in the 10GEPON TF ranging from -14 to -18, which makes me wonder where the truth lies. This is a very important factor since it impacts our overall link budget in the downstream channel. Are we likely to operate closer to -14 with standard devices rather than -18 ? Does anybody have any experience with that that they could share with us?

<Maurice> Leading edge products today can achieve -20dbm typically, meaning that -18dbm sensitivity over all conditions is probably real. </Maurice>

<Marek> That is indeed good news. Does anybody happen to know what production yields can be achieved with the -18dBm PINs? As always, we should assume rather that high yield production devices are used since hand-picked ones will increase the ONU cost in a significant way. The idea I got from the presentations so far is that we should assume PIN sensitivities closer to -14dBm rather than -18dBm. I would be happy to have more comments on this. </Marek>

2.        
What about development/research in terms of PIN diodes for 10G systems ? In 2-3 years, when the standard is accepted, there may be some progress in terms of low cost PIN diodes for 10G systems. Is anybody aware of such development?

<Maurice> Very, Very good point.  PIN receiver sensitivity has been getting better at the rate of about 0.5 to 1.0 db/year.
This has a little to do with the PIN, as PINs with higher conversion efficiencies are fabricated, but more to do with the TIA, as faster process technologies allow for more bandwidth, higher gain, and lower noise.  By 2010 when the standard is released, you could plan accordingly.  Interim solutions and fall back plans could always use APD-based receivers.

This same thing happened in 2003 with GEPON (802.3ah).  Our 1Gb TIA came out and replaced all APD-based receivers for all 1000BASE-PX10/20 applications, since it could achieve -29dbm sensitivity with +7dbm overload using a decent low-capacitance PIN.
 </Maurice>

 

<Marek> If I read that right, that may mean that for now >=29dB budget can be though for achieve with PINs and we may be forced to standardize APDs on the ONU side, at least for the high split, long reach systems where each dB matters. In the future, when the PIN development for 10G systems matures and newer components are available, we may potentially expect both emergence of 10G PIN diodes with sufficient sensitivity figures to replace APDs and their commercial availability + large stock. Do You think that such devices may also find application in 10G Ethernet hardware?</Marek>

3.        
Are -16/-18 PIN diodes for ONUs economically feasible, meaning - are they sufficiently mature and is there sufficient production capacity to assure reasonable cost? I am asking this because perhaps it is less expensive to use a port-amplifiers at the OLT side and increase the launch power rather than invest into higher sensitivity detectors on the ONU side, while ONU module price should be kept as low as possible?

<Maurice> Again a very very good  point.  It's easier to build a damage resistant TIA with PIN, than a TIA with APD.   APD based receivers only have one good point.  Sensitivity.   Their bad points are:  Complexity,  (High-voltage power supply, gain calibration, and protection), Fragility:  (We have had customers blast our 2.5G TIAs with +3-+6dbm, killing the APD, but not harming the TIA, since the APD went into avalanche breakdown) and Cost due to the HV power supply.
When using a PIN-based receiver, higher power is better.  Look at the 1000BASE-PX10/20 power budgets for guidance. </Maurice>

 

<Marek> In short - avoid APDs at the ONU side if possible at all cost. That would shave off some cost but then we have to compensate with a higher power laser on the OLT side to overcome the loss in the channel. What kind of sensitivity penalty do we have when using PINs instead of APDs on the ONU side? This is a P2P channel so probably we should be able to have some feedback on the numbers.</Marek>

4.        
What about low grade APDs for ONUs ? Any ideas regarding the price comparison for low grade APD versus good quality PIN ? Where is the trade-off point?

<Maurice> No contest:  In volume, the APD adds <Marek>cost</Marek> in overhead circuitry  (Think DC-DC converter, adjustable, current-limited HV power supply)  This does not count the cal time.  APDs themselves tend to be more expensive, as
there are fewer vendors, and thus less competition. </Maurice>

 

<Marek> Right. that more or less confirms what I concluded above. If we want to maintain the target ONU price envelope, probably we want to try to avoid APDs. You give very good reasons for that. </Marek>

5.        
Regarding the OLT RX modules - there is little doubt that we need an APD. What sensitivities can we obtain with good quality APDs in the burst mode ? What kind of burst mode penalty do we observe in the upstream channel at 10 G ? Does anybody have any reasonable data on that ?

<Maurice> I'm not an expert in the impairments with an APD/Burst-TIA, as we don't make those. </Maurice>

6.        
Again, what about progress in terms of APD diode sensitivity ? I heard about some research going on in terms of QD (Quantum Dots) APDs, though it is considered still largely immature and probably it will not become commercially available within the next 2-3 years. I may be mistaken here. Does anybody have any more precise information on this topic ?

Best wishes

Marek Hajduczenia (141238)
(PhD Student - COM RD1)
SIEMENS Networks S.A. - IC
Rua Irmãos Siemens, 1
Ed. 1, Piso 1
Alfragide
2720-093 Amadora
Portugal
* Marek.Hajduczenia@siemens.com
http://marekhaj.easyisp.pl/index.php
(+351.21.416.7472  4+351.21.424.2082

 

image002.gif

GIF image