By Fred Dawson
March 13, 2013 – Aurora Networks’ acquisition of Harmonic, Inc.’s optical transmission business signals there’s a growing sense of urgency among cable operators that they must come up with a way to support a pure IP distribution play in the multiscreen services arena sooner than many anticipated.
Unlike most deals of this nature, the underlying motivation behind the $46-million acquisition wasn’t the desire to take a competitor out of the game or acquire its technology. Instead, says John Dahlquist, vice president of marketing at Aurora, the major benefit for his firm lies in enhancing its ability to meet anticipated demand for a radically new approach to service distribution that replaces analog optics with digital on the local access plant while moving QAMs (quadrature amplitude modulators) out to the hybrid fiber coax (HFC) nodes where the fiber and coax links connect.
“This acquisition allows us to double our node footprint to over 200,000 nodes in place,” Dahlquist says. “We have a major focus on how the distributed optical architecture will evolve to where operators will use baseband digital to the nodes and at that point convert to RF.”
Aurora has seen rising interest in the baseband option since it began touting a soon-to-be-commercial product line in this space known as “Node QAM.” “A number of people within the industry are recognizing that if they really want to supply the throughput downstream and upstream that they believe will be necessary over the next six years, they’ll have to make major changes in the network,” Dahlquist says.
As previously reported, the Node QAM concept entails moving the RF modulation process now employed at headends and hubs to the nodes for some portion or even all channels in the RF spectrum, thereby freeing operators to exploit the benefits of digitally processing, multiplexing and transmitting service streams for new categories, such as IP multiscreen, while greatly reducing the space and power consumption at cable hubs and headends. Optical digital transmission benefits touted by Aurora include more efficient use of existing fiber plant, lowering operational costs and achieving higher quality signal output at the node to enable new bandwidth gains over the RF coaxial link.
The realization that what had been a fairly limited TV Everywhere strategy involving distribution of on-demand content to PCs must expand to encompass live broadcast channels for viewing on virtually any type of IP-connected device has raised major challenges for cable companies. For a while it seemed many operators were content to postpone a move to end-to-end IP distribution by using gateways in the home to transcode and reformat MPEG-2 delivered premium content to connected devices, but consumer demand for ubiquitous access to video entertainment is forcing operators to plan how to reach them without depending on deploying home gateways that will only serve high-end customers.
The ability to deliver a unicast IP stream formatted to the encoding and adaptive bit rate streaming requirements of each device every time someone wants to watch a TV program or movie raises cost and bandwidth issues which operators have long understood will require a break with current approaches to modulating and distributing content. Now that the timeline to accomplishing these changes is shrinking operators must decide on a course of action sooner than later.
“We believe in our Node QAM product, and by purchasing Harmonic we’ve doubled our potential homes for that product,” Dahlquist says. “We’re also looking at node PON and the second-generation of Node PON modules that we’ll be able to repackage and put into Harmonic nodes.”
Node PON architecture involves use of nodes as the housings for the optical line terminals (OLTs) which interface with high-capacity metro backbones to offload and regenerate locally targeted signals that can be passively split in the field to serve up to 64 fiber-connected customers at distances of up to 20 Km. from the OLTs. With capacity to install as many as five OLTs in Aurora’s node chassis, operators can incrementally build out all-optical networks to serve a growing base of business customers as well as new or remote residential areas, eventually reaching up to 320 customer locations per node.
Aurora’s Node PON business has been in operation for several years with a growing base of customers looking to cost-effectively extend their commercial services to larger companies, institutions and other high-capacity bandwidth users who can’t be adequately served over HFC broadband connections. The Node QAM, on the other hand, is a new concept which is meant to greatly expand service capacity on the HFC plant.
Aurora is well positioned to perform the product adjustments that will make it possible to install its next-generation modules in the Harmonic node chassis, Dahlquist says. In the case of Node QAM architecture, the new modules must be designed to plug into strand- and pedestal-mounted node housings with all the interfaces required to tap into the node power supply, performance monitoring system and other OSS resources, including a new remote QAM management system.
“It’s not too big a job, because the form factors (of the Aurora and Harmonic nodes chassis) are similar,” he says, noting that the same engineer who oversaw development of the Harmonic HLN3134 node was a lead player in Aurora’s NC 4000 node development. “There’s still some investigation that needs to be done before we understand all the details of their technology, but in general it’s not like we acquired a technology that’s new to us.”
Work is already well underway and progressing rapidly, says Scott Weinstein, vice president of new business development at Aurora Networks. “With the closing of the acquisition behind us, we are now focused on a swift integration of the product lines into our systems with a goal of minimum customer inconvenience,” Weinstein says. “We have already started the integration process, with many of the former Harmonic employees who will assist Aurora’s efforts to achieve an integration goal of four months joining our team.”
There’s very little overlap in the customer bases served by the Aurora and Harmonic optical platforms, he adds. Owing to Harmonic’s pullback on investment in development of the product line over the past several years, the company had reached a point where they were either going to have to heavily re-invest in the technology or walk away, which means that the existing customer base is ripe for upgrades.
But Dahlquist stresses Aurora’s intention is to work closely with those customers to ensure their interests are served. “We’re transferring the Harmonic products into our production,” he says. “Certainly our focus is to maintain both product lines. If we were to pull either the Harmonic transmission platform or the node we’d have a lot of upset customers.”
Just what the upgrade path will be is up to each operator, of course, but there’s growing consensus on what needs to be accomplished. “The migration path many operators are looking at is to get out to the node digitally and do more distributed architecture over coax so they’re not carrying so much common information in the RF domain,” Dahlquist says.
In other words, if operators are going to support ongoing distribution of premium TV service in the legacy MPEG-2 mode while simulcasting their entire programming portfolio in narrowcast mode over IP, there needs to be a way to accommodate a huge surge in traffic to the node without adding more fiber. And there needs to be segmentation of distribution routes over the coax to a level of granularity that keeps the volume of narrowcast signals within the bounds of RF capacity limits on the coax plant, preferably without having to push fiber deeper into the neighborhoods.
This is what the Node QAM strategy is meant to accomplish. With a shift to digital optics the throughput capacity on the fiber becomes much greater insofar as wavelengths can be packed much closer spectrally in the DWDM (dense wavelength division multiplexing) window than is the case with DWDM spacing for analog optical signals. For example, typical spacing between wavelengths in AM mode is 100 GHz, which allows up to 40 wavelengths on a single fiber, whereas spacing for digital wavelengths is 25 GHz or less, which leaves room for 160 or more wavelengths.
But it’s not just the capacity gains that are important from the optical transport perspective. By operating digitally operators rely on low-cost small form-factor pluggable (SFP) lasers versus high-priced distributed feedback (DFB) lasers used in AM fiber, and they avoid the operations costs associated with maintaining rigorous performance on AM links as plant conditions change.
Moreover, any payload delivered digitally to the node will avoid the losses incurred through traditional RF combining and optoelectronic conversions on AM links, resulting in signal-to-noise output at the node comparable to headend outputs at 40-42 dB with attendant improvements in harmonics and noise floor performance. These gains open the possibility of expanding available bandwidth on the coax by going to modulation above 256 QAM on the channels that are modulated by the Node QAMs.
Equally significant, as described in a white paper presented at last fall’s Cable-Tec Expo by Ron Wolfe, senior director for global product strategy at Aurora, Node QAM architecture offers major space, energy and equipment cost savings at hubs and headends as operators attempt to keep pace with the need for ever more QAM capacity. “Devices such as laser transmitters, optical receivers and QAM modulators all consume space, power and cooling capacity such that many hub sites are strained to capacity,” Wolfe writes.
The fact that RF signal generation can be done by FPGA (Field Programmable Gate Array) chipsets as opposed to dedicated circuits where QAM signals have to be up-converted onto contiguous frequency channels means operators have great flexibility to fill the QAM RF spectrum in whatever ways work best for them over time. By separating the pure modulation function from signal processing and placing the miniaturized QAM modules in the node chassis, operators can use a digital service multiplexer in the hub or headend to feed the right mix of service signals to each service group, Wolfe explains.
This multiplexer, along with synchronizing and sequencing the IP video packets coming in from various sources, applies all the timing functions, encryption control messages and other telemetry and control information to each service stream to deliver the appropriate services to specific service groups in the network, thereby eliminating much of the space- and power-consuming apparatus that now populates headends and hubs. This digital multiplexing and configuration process operates dynamically across the full service payload to support on-the-fly assignment of any mix of digital content to any given QAM channel at any given node, including broadcast, HDTV, SDV, VOD, nPVR, cable IPTV, digital voice and DOCSIS data streams.
Wolfe tabulates the savings to be achieved by comparing power costs and space consumed by equipment in a typical hub serving 96 service groups of 500 subscribers each on a 750 MHz all-digital TV HFC system with the power and space requirements of a Node QAM architecture delivering the full spectrum of services to the same number of service groups. In the centralized model, to serve 50 broadcast channels, 24 SDV outputs, eight VOD channels and eight DOCSIS channels, the hub would employ 3,890 circuit-based QAMs compared to 96 FPGA-based QAM modules for the distributed architecture. Power costs would be reduced from an estimated $32,693 per year to $8,729 per year while rack space, factoring in the 22 rack-unit space required for the Digital Service Multiplexer versus the five 44-RU racks required for the centralized architecture, would be reduced by 90 percent.
While Wolfe notes that new QAMs support more channels per port than the older “pizza box” models and therefore would take up less rack space, the older models are dominnant and so were used in these space calculations. Even with newer equipment the centralized model would consume far more power and space than the distributed model.
The Node QAM modules in development at Aurora Networks are designed to modulate any combination of services consuming anywhere from a single RF channel to a full load of 158 channels, which means the operator can incrementally add services to those initially modulated at the node over time. Because the encoding, encryption, grooming and advertising mechanism are performed at the hub the mechanics of the Node QAM are completely agnostic to the services running through it.
This architecture also allows great flexibility in the use of the optical distribution system, since the headend multiplexing platform can dynamically assign wavelength resources on the fly, Dahlquist notes. Any given wavelength can be employed to deliver a complete or partial lineup of broadcast digital TV channels to multiple nodes while any given wavelength bearing narrowcast content can be shared across multiple nodes, allowing QAMs at each node to modulate only those streams that are targeted to subscribers in the node service area. Alternatively, the full complement of broadcast and narrowcast signals to be supported by node QAMs can be provisioned over a single wavelength to each node.
Such dynamic provisioning flexibility allows the operator to incrementally utilize node QAM resources cost effectively without having to forecast in advance what the eventual mix of streams will be at any given node or on any given QAM, Dahlquist adds. In other words, there’s no need to pre-wire or alter wiring of the RF combining network with every change in the service mix, and the mix can be dynamically shaped and expanded over time on a neighborhood-by-neighborhood basis.
The company had expected to deliver Node QAM components for initial trials in Q1 of this year but had to push the delivery date to Q2 when complications arose in the production process. “We’ve had to go back through an additional board spin,” Dahlquist says. “These are relatively complicated products that we want to be able to produce at very high volumes. It wasn’t a tech design problem but more a matter of manufacturability.”
When Aurora Networks first introduced the Node QAM concept in late 2011 it was a radical idea that few engineers had given much thought to. Now Node QAM has entered a next-generation HFC discussion that includes CableLabs’ recently announced DOCSIS 3.1 standard initiative and the IEEE’s emerging EPoC (Ethernet PON over Coax) standard.
These architectures offer very different approaches to bandwidth expansion over the fiber feeder links. DOCSIS 3.1 and EPoC entail use of higher levels of QAM modulation in conjunction with OFDM (Orthogonal Frequency Division Multiplexing), a technique widely used in cellular, and a new type of forward error correction, which will raise the throughput capacity on a single AM fiber wavelength to 10 gigabits per second.
But EPoC also opens a path to use of digital optics, allowing operators to leverage the legacy DOCSIS QAM-based infrastructure while moving to use of the same plant to support an end-to-end Ethernet architecture. This requires a component known as an Optical-to-Coax Unit (OCU) to perform the RF modulation of the Ethernet signals onto the coax links. With the OCU it becomes easier to dispense with use of AM optics by allowing EPON to transmit digitally to the HFC node and over RF to the home.
For operators who envision capping DOCSIS and MPEG-2 for legacy services while using EPoC to simultaneously deliver an all-IP package, a key benefit of this mode of migration would be the savings that come with low-cost, easy-to-maintain digital optics, or, in other words, some of the same benefits touted for Node QAM. “We’re very interested in EPoC, too, as another way to skin the cat,” Dahlquist says, noting the concept is analogous to the Bit Coax architecture Aurora floated a few years ago. “It all depends on how the cost and volume issues play out.”
The problem with EPoC, he adds, is it may not work out soon enough as a cost-effective approach, given the pace of the standards-setting procedure and the lack of an installed Ethernet CPE base for operators to build on. “Right now cable operators do need to increase capacity, and they have a sunken investment in set-top boxes that work very compatibly with what we’re doing in Node QAM,” he says.