Optical components

Fujitsu Pushes PON Splits

The Fujitsu Laboratories arm of Fujitsu Ltd. (Tokyo: 6702; London: FUJ; OTC: FJTSY) says it has developed an amplifier array that could increase the number of homes or businesses a Passive Optical Network (PON) serves.

The technology was presented at ECOC last month and was announced in press-release form Thursday.

Fujitsu's scheme uses Semiconductor Optical Amplifiers (SOAs) to boost upstream traffic. Four of these sit in the PON remote node, each one serving 32 incoming lines. That gives the remote node a split ratio of 1:128 as opposed to the more normal 1:32.

The key is that the SOAs activate only when an upstream burst of traffic arrives. That limits the noise the amplifiers add to the line -- noise being a big problem in normal SOA deployments, according to Fujitsu. The company also says it's got a way to fabricate the four amplifiers in one module, which could help save space in the remote node.

Why this matters
The number of potential PON endpoints is increasing, not only because of broadband buildouts but because PONs are being considered for new roles such as mobile backhaul and data-center interconnections. Fujitsu further argues that cloud computing will cause more stuff to get attached to the network in general, which would imply that more PON endpoints could emerge wherever PONs are being built.

One way to scale PONs accordingly would be to increase the split ratio. A 1:128 ratio has been discussed for some time; Fujitsu's proposal just sounds like it could make that concept a little more practical.

For more
Some recent tidbits about the progress of fiber access:

— Craig Matsumoto, West Coast Editor, Light Reading

Duh! 12/5/2012 | 4:51:17 PM
re: Fujitsu Pushes PON Splits

So it looks like the reach extender box has to monitor the downstream to keep track of the bandwidth map, so it can turn on an SOA at the start time of each allocation and turn it off at the stop time.  Not only that,  but it also to keep track of which leg of the 4:1 split is associated with each Alloc_ID, so unless I'm missing something, it has to be provisioned with a mapping from ONT serial_number to splitter port/SOA.  All pretty complex.  One of the justifications for  amplification rather than OEO regeneration was to eliminate the need for the extender device to be deeply involved with the TC (other than an supporting an embedded ONU for element management functions using OMCI).

Last I'd seen, SOAs were pretty pricey, mostly as a function of packaging.  They're cooled devices, and fiber coupling is the principal contributor to the device's noise figure.   Maybe Fujitsu (or their supplier?) found a way to cost reduce these array devices?

Also, as explained in G.984.6 Appendix 1, 128:1 split isn't a problem as long as the OLT receiver is DC coupled with reset between bursts, and a blocking filter is used to remove out-of-band ASE.  So it's not clear whether they're trying to relax one (or both) of those constraints, or to get some additional loss budget in the system.

In any event, I've kinda lost track.  ITU published G.984.6 in 2009.  Has anybody actually deployed significant numbers of GPON reach extenders in the intervening years? 

paolo.franzoi 12/5/2012 | 4:51:16 PM
re: Fujitsu Pushes PON Splits


By similar do you mean one that is all optical then yes - Alphion.

By similar do you mean one that monitors per Duh!'s commentary then no.

I have not seen a lot of extended reach PONs deployed.  I keep reading about people wanting to use them for all these things, but am confused on why they think this is a good idea:

- Verizon Wireless rejected using FiOS for wireless backhaul.  Ever ask them why?

- Business stuff over PON only makes sense in the T-1 replacement market.  Data Centers?  Nope.

Remember doing a larger split means less average bandwidth per user.  Most of the bigger split stuff I saw came out of the 3rd world.  3rd World PON is city only - which again makes me think that PON amplifiers are a small market.

Hate to have to diagnose one of them failing....




Pete Baldwin 12/5/2012 | 4:51:16 PM
re: Fujitsu Pushes PON Splits

I know there have been many 1:128 splitters presented at OFC over the years, and I think I've even seen 1:256 listed in the papers.

I haven't read them all but I doubt Fujitsu's setup here is all that radical. I did find the amplifier array interesting, which is why I wrote it up. Anyone heard of anything similar out there?

Duh! 12/5/2012 | 4:51:14 PM
re: Fujitsu Pushes PON Splits

The original impetus for reach extension came out of BT, when Matt Bross was CTO.  The idea was to eliminate central offices;  they had models that showed OPEX savings from doing that which would just about pay for the CAPEX of the PON deployment.  That plan was shelved shortly after Bross, um, left.  BT convinced most of the other  big operators to examine that model, and that was going on during the G.984.6 standardization.  I haven't heard of it being deployed anywhere in volume.

The other big application was purely tactical.  If the operator has to reach a remote pocket of high value subscribers (think ski resort), reach extension or remote OLT is the way to go. 

There probably are designs which would benefit from high split ratios.  The problem I'd see with that is not so much in sharing bandwidth (particularly in the downstream) as the amount of latency in the upstream due to DBA.

Now that Fujitsu's press release is up on LR, it appears that they were indeed looking to get additional reach along with the high split.  And they are correct that the upstream is ASE limited.  One could argue, though, that a better solution would be to focus on the noise figure of the SOA devices and possibly tightening specs for the ONT lasers to allow for narrower ASE bandwidth filters. 


Sign In