|
I’m hearing a lot about multi-wavelength optical
transport as a way to do node segmentation. How is this different from
existing node segmentation methods? What’s “new” about it?
What’s new is that you can make more room for on-demand and broadband
services, without having to install new fibers between the headend and
the serving area group, or its constituent nodes. That’s enormously
helpful for systems that need more elbow room, but don’t want to deal
with the time and hassle of putting in new glass.
Using multi-wavelength optical transport generally comes into play
sometime after a system has performed all of the logical and physical
node splits it can feasibly do: You’re at one-to-one combining from the
headend to the serving area group, let’s say, and you’ve physically
subdivided the node as much as the current fiber resources allow. So
what you need, in that case, is a way to carry more content on the same
glass. That’s where multi-wavelength optics is a beneficial option.
The big bonus is that it works using the original fiber from the headend
to carry the multiple wavelengths – so you don’t have to buy and install
new fiber.
How is this different from narrowcast DWDM (dense wave
division multiplexing) systems already being deployed?
In a 1310 nm link, a dedicated fiber is required for each
serving area requiring unique content. At an average cost of $20,000 to
$30,000 per mile for new fiber routes, it adds up quickly.
In a 1550 nm link, two transmitters launch content down the same fiber
to one receiver. One of the lasers typically sends broadcast content,
and the other sends narrowcast, or unique content. Two lasers generally
means an increased noise contribution at the receive end. As a result,
the total amount of content that could be carried was limited in order
to keep the analog carrier-to-noise ratio (CNR) high.
By contrast, with a multi-wavelength system, each of the wavelengths can
carry the full broadcast and narrowcast loads. That means each link is
similar to what can be achieved by a dedicated 1310 nm path.
What limitations exist because of the use of multiple
wavelengths?
Most of the limitations are measurable as composite second
order (CSO) degradation. It happens because of non-linearities in the
fiber itself, but a significant contribution also stems from the
wavelength dependence of the optical passives—the mux and demux—used to
combine or separate the wavelengths.
There are three dominant fiber effects: Raman crosstalk, four-wave
mixing (FWM), and dispersion. Dispersion in fibers relates to the change
in the speed of light for different wavelengths. Because our lasers are
directly modulated, their wavelengths are “chirped” – meaning they’re
spread out by the RF modulation. It’s the combination of chirp plus
dispersion that leads to CSO limits. For standard SMF-28 fiber, chirp
plus dispersion limits directly modulated transmitter wavelengths to
around 1335 nm. Longer wavelengths exhibit CSO worse than -65dBc.
FWM is an optical form of mixing that’s analogous to composite triple
beat (CTB) intermodulation products. Any two wavelengths co-propagating
in the same fiber will generate additional wavelengths at sum and
difference (optical) frequencies. The effect is strongest for
wavelengths that have low dispersion, and maximum at the
zero-dispersion-point (ZDP) of the fiber, where the wavelengths stay
in-phase for longer distances. Critical attributes of FWM are the
amplitude and location of the generated products.
Raman crosstalk happens when the energy of one wavelength is transferred
to another, by virtue of the molecular properties of the glass in the
fiber. That transferred energy becomes a distortion on the signals
involved.
What can be done to overcome those limitations?
There are really two separate things to consider here. The first
includes the techniques that limit the individual impairments
separately. The second ensures that the system performance of the
combined effects are controlled.
For the individual impairments, we use specific techniques, such as
power density control, to limit the Raman crosstalk. Then, we restrict
the wavelength to be short enough to limit CSO caused by chirp plus
dispersion, and long enough to limit FWM effects.
On the system level, we choose particular wavelength spacing to control
the location of the FWM products and to cap the Raman gain. We limit the
FWM amplitude and the chirp-plus-dispersion impact with absolute
wavelengths.
Putting those things together is the key to maintaining the proper
balance between individual impacts and system level performance.
How “multi” is multi-wavelength – how many more
“colors” do I get? And, how does this, as a bandwidth expansion option,
translate into a practical, day-to-day sense? What can I tell my
management, in terms of how much more HDTV, simulcast, and
channel-bonding stuff I can carry?
Like all real-world engineering questions, the answer is “it
depends.” As our description of non-linearities and other impairments
suggests, there is a tradeoff between the number of colors that can be
co-propagated and the power in each of the colors. For short link
lengths and low powers, one could carry eight wavelengths without too
much difficulty, or, one could go to longer lengths with fewer
(appropriately chosen) wavelengths.
That said, commercial introductions are settling around six wavelengths
carried over 20 to 25 km links. For the second part of the question, the
solution is really at the physical layer of the network and provides an
increase that scales exactly with the number of wavelengths. The
particulars on numbers of HDTV channels, etc., is improved by virtue of
the reduction in the number of users sharing the wavelength. Bottom
line? For a six wavelength system, the capacity goes up 6x.
To go a little further here – there really is no restriction on the QAM
content of each transmitter, so, depending on how your serving group
size is distributed, you’ll get six times the coverage. As DOCSIS 3.0
protocols and MPEG-4 compression encoding become the norm, the HD
capacity and the channel bonding capacity increase – but the allocation
of those signals to physical wavelengths is independent of the protocol
or encoding schemes.
What does it cost, relative to having to run more
fiber?
As noted earlier, new fiber runs can cost $20k-$30k per mile. Example:
In a run length of 25 km (15.5 miles) the cost per link is just under
$400k. In that scenario, adding five new wavelengths would cost about $2
mil. Of course, some of that would likely be run along a common pathway
-- so let’s be generous and say it comes down to $1 mil.
A comparable multi-wavelength solution would cost about 5% of that
figure, so the ROI really is very attractive. It’s a lot of bang for the
buck.
|
