Optics Expert Offers Rejoinder To HFC Digital Return Naysayers

Oleh Sniezko, CTO, Aurora Networks

Oleh Sniezko, CTO, Aurora Networks

June 8, 2012 –Amid an accelerating pace in cable operators’ use of digital optical transmission technology for return links over HFC networks one supplier offering a proprietary analog solution is questioning whether MSOs are making a big mistake in the shift to digital.
Analog fiber optics technology, the breakthrough transmission mode that made hybrid fiber coax networks possible in the first place, remains everyone’s mode of choice on the downstream, notwithstanding some new concepts such as EPoC (Ethernet PON over coax) that foresee an eventual migration to digital in the downstream. But there’s no question the digital return strategy is coming on strong, prompting most suppliers of traditional analog solutions to add digital to their product portfolios.

In fact, Aurora Networks, a leading supplier of digital as well as analog fiber options, earlier this year reported research results that predict annual deployments of digital return links will outpace analog deployments on HFC networks within two years. “We believe we offer the best analog solutions in the industry and continue to do research in that area with new patent applications,” says Oleh Sniezko, CTO at Aurora. “But we are convinced digital offers a superior solution for the return on HFC networks.”

Mani Ramachandran, CEO of InnoTrans, which has been winning business for its approach to improving performance and lowering costs of analog transmission on downstream and upstream optical links, takes strong issue on both counts, asserting that the InnoTrans analog return is not only the superior analog solution but also trumps digital on cost, performance and capacity measures in all but very long-distance scenarios. “There’s a lot of confusion in the market right now, much of it caused by claims for digital that are made in comparisons with outmoded analog technology,” Ramachandran asserts.

In truth, both parties tend to base their comparisons on benchmarks associated with less than best-of-breed competing options. When the performance parameters of the InnoTrans analog solution are directly compared with those of Aurora’s advanced digital return solution, the analysis becomes more difficult.

That’s because, in the case of InnoTrans, the firm has applied some of the clipping mitigation techniques used with its Chromadigm downstream transmission system on its Adaptive Return Transmission (ART) platform in order to overcome key limitations of even the best performing analog distributed feedback lasers. Similarly, Aurora’s digital return solution has outpaced competitors with cost-saving migration capabilities and the ability to digitally convert a broader segment of RF upstream spectrum.

For example, when Ramachandran makes the point that ART outperforms digital at costs comparable to many traditional analog solutions he points to digital solutions that can support digital conversion of 65 or 85 MHz of spectrum modulated at the 256 QAM (quadrature amplitude modulation) level, whereas Aurora’s Universal Digital Return is capable of transmitting 100 MHz of RF spectrum at 1024 QAM and will soon go to 150 MHz.

InnoTrans, with a 40 dB noise power ratio (NPR) over a 20 dB RF signal range within the traditional 5-42 MHz upstream spectrum window, has enough headroom on the ART module to transmit 200 MHz at 256 QAM, albeit at shorter distances than are achieved with digital. In fact, if plant conditions become such as to support 1024 QAM, InnoTrans will be able to do that as well, Ramachandran says.

Sniezko counters that, whatever the return spectrum the Aurora digital return might support today, it can do that on 1024 QAM-modulated channels at the same performance levels it achieves at lower modulation levels. “With our 100 MHz digital link and RF at 1024 QAM the upstream throughput would be 700 megabits per second,” he says. “What we’re saying is our digital link with 10 bit coding can support 1024 QAM at all the dynamic ranges needed for RFoG (RF over Glass), fiber deep and RF.”

While Ramachandran concedes digital may be a better solution for very long-distance uplinks above 80 Km., he notes that by far the lion’s share of situations operators are looking at for digital return upgrades are much shorter and therefore well within the range of exploiting the benefits of InnoTrans’ ART solution. In fact, he adds, with use of erbium doped fiber amplifiers in conjunction with transmissions in the ITU C band (the 1550 nanometer lightwave “window”), operators can look at ART as an alternative even over very long-haul links.

Meanwhile, in the real world of what operators are looking for as they contemplate how best to expand the throughput over their upstream links, the issue has little to do with which solution is best for transmitting over 150 or 200 MHz of RF spectrum. As things stand today, virtually all upstreams use the traditional 5-42 MHz subsplit band with expansion to 5-85 MHz widely viewed as the next step to be taken at some point down the road.

“85 MHz is a more comfortable place for operators to think about for upstream spectrum expansion at this point,” says John Dahlquist, vice president of marketing at Aurora. “Real upgrades beyond that level will occur as operators go from N +5 [in amplifier cascades on the coaxial network] to N +0.”

With the elimination of in-line amplifiers, operators will have up to 2 GHz of RF spectrum to work with on the coax, at which point many high-bandwidth options for the return will be available, not only at the sub-band level but at the mid-split and even at the top frequency tiers. But the pressing issue now is how to get the most bang for the buck as operators look for ways to use more of the existing subsplit band at higher levels of modulation.

As previously reported (July 2011, p. 12), in Aurora’s digital return applications all upstream channels allocated within the 5-42 MHz sub-band are time division multiplexed onto a 2+ Gbps baseband stream employing a digital sampling rate that can be adjusted to support 8-, 10- or 12-bit bytes. The digitized TDM signal is fed into a low-cost SFP (small form factor) plug-in laser for transmission over a single wavelength at distances that can exceed 100 Km without in-line optical amplification (or up to 200 Km with optical amplification) without requiring regeneration of the signal.

The limiting factor on digital transmission capacity in cable has been the sampling frequency of the A/D converter, which must operate at more than twice the top frequency of the sampled RF spectrum. Aurora has tapped the benefits of Moore’s Law to gradually increase the A/D clock speeds, with its current model operating at about 215 MHz and a good chance the speed will go over 300 MHz before long.

Contradicting Ramachandran’s assertion that such gains drive costs to impractical levels, Sniezko says, “We have managed to keep the costs the same as we go to higher clock speeds. As a result we’re able to deliver these levels of performance at lower costs than a good analog link.”

That last point is strongly disputed by Ramachandran, who says the InnnoTrans solution, taking into account the full link expenses, costs much less, assuming the link is transmitting a single RF return. “When you go to digital, you not only replace the transmitter, you have to install a new receiver,” he says. “With our transmitter, you put it in without having to change your receiver.”

But Ramachandran acknowledges the cost equation may swing in favor of digital return in the case of a transmitter/receiver combination capable of transmitting two upstream RF feeds from a split node, as would be the case with a digital return link that’s been installed with capacity to support the equivalent of 5-85 MHz in RF signals. “There’s no price difference for our 100 MHz receiver,” Sniezko says. “You can choose it from day one for a 5-42 MHz return and go to the higher capacity whenever you want.”

At the node, Aurora has introduced what it calls “Universal Digital Return” (UDR) for its nodes in order to significantly lower the cost of migrating to higher digital return capacity for operators who start out supporting a single 5-42 MHz return. As Dahlquist notes, digital return has always made it possible to accommodate a node split with installation of a “2-fer” node module and companion receiver, allowing the full return from both RF segments on the split node to be multiplexed into the 2+ gbps baseband signal for return over a single wavelength.

Now, with UDR, rather than having to change out the entire Digital Return module to support the 2-fer capabilities or an expansion to higher return spectrum frequencies, operators need only change out the low-cost “personalization” module, which configures the return to requirements for data transmission speeds, operational modes and other factors. “UDR has really changed the cost calculations for digital upstream migration,” Dahlquist says.

But, as Ramachandran notes, there’s no getting around the fact that, whichever manufacturer’s digital return an operator may want to install, frequently the legacy nodes used for analog return are not equipped to support the digital return components. In such instances, which don’t include legacy nodes supplied by Aurora, operators must replace the entire node chassis to accommodate digital return.

Consequently, there are no pat answers with respect to the cost question when comparing a digital return option with the InnoTrans ART or, for that matter, other analog return options. Indeed, Ramachandran says, many situations where operators are contemplating implementation of bonded-channel DOCSIS 3.0 capabilities on their return links can be accommodated by resetting existing analog transmitters, thereby eliminating replacement costs that would be incurred with either the InnoTranns or digital return solutions.

“The biggest reason a lot of operators are looking at digital return is they don’t believe they can handle DOCSIS 3.0 with their existing transmitters,” he says. “But if they’ve upgraded their returns in the past few years with DFBs (distributed feedback lasers), they may not need to replace them, even though we’d love to sell them our solution.”

DFBs have far more dynamic range and greater linearity than the Fabry-Perot (FP) lasers that were originally used for return transmissions over HFC optical links. But even though DFBs may be amenable to adjustments that will allow an expansion of the usable spectrum on the 5-42 MHz subsplit and an increase in modulation levels across all upstream channels, those adjustments can be a hassle, Sniezko notes.

And, in light of future spectrum expansion requirements, sticking with installed DFBs may only postpone the time when replacement will be necessary while incurring higher-than-necessary operating costs in the interim. “There are some operations issues you have to deal with that just aren’t there with digital,” Sniezko says.

Frequently, even when DOCSIS 2.0 or 3.0 CMTSs (cable modem termination systems) are already installed, the full range of tools available for maximizing efficiency aren’t implemented on existing upstream modules, including DFB-equipped upstreams, until operators are ready to institute DOCSIS 3.0 channel bonding on those upstreams. Often, because technicians have not gained experience with many of the DOCSIS 2.0 techniques, this requires help from outside sources.

But, whatever the pros and cons any given operator might see in weighing whether to replace existing DFB upstream transmitters, there’s no question about needing to replace FPs when it comes to upgrading to DOCSIS 3.0 on the upstream. This is where the debate over the choices on offer from Aurora and InnoTrans is really joined.

In a nutshell, InnoTrans says it has overcome the main impediments to efficient use of analog return by employing clipping mitigation to a point that operators can use low-cost off-the-shelf DFBs to deliver performance parameters that go far beyond what can be done with more expensive HFC-optimized DFBs. Moreover, the dynamic clipping mitigation technology provides a self-calibrating means by which the ART module maintains the proper settings over time, creating the same stabilization against costly field maintenance that digital solutions offer.

But when it comes to claims and counter claims as to which solution is best for any given situation, operators will have to run the tests and come to their own conclusions, given how far apart some of these claims are. For example, Ramachandran asserts the dynamic range, which he defines as the flexibility in decibels of RF input power one can tolerate and still achieve a given NPR at the output, is much greater for the InnoTrans solution than it is for digital returns.

Moreover, he says, digital return manufacturers are using a misleading approach to defining dynamic range where the dynamic range they’re referring to is the range of optical power that can be tolerated on the transmitter without affecting performance. “They’re not even talking about the range of RF signals handled through their analog-to-digital circuitry,” he says.

Nonsense, says Sniezko. “Three years ago we led the industry’s switch to a new definition of dynamic range, and that’s the definition we’re using today,” he explains.

Where, in the past, the definition pertained to the range of RF input level at which the carrier-to-noise ratio (or the NPR – another way to measure signal strength) remains at a certain level, that definition proved inadequate when it came to assessing performance at higher levels of QAM, he says. That’s because, at higher QAM levels, the bit error rate can be subpar within a certain RF range at either end of the range that’s delivering the desirable CNR level.

“So we now define dynamic range as the range of RF input levels in which the BER stays at a defined level, typically 10−6,” he says. “That captures the sensitivity of higher modulation to the clipping sides of the NPR curve.” Thus a dynamic range spec’d at 12-13 dB for a 40 dB CNR on two fully loaded 37 MHz return paths, as is the case with the “2-fer” transmission on Aurora’s Digital Return, would be larger if the definition were based on the range that supports that CNR without taking BER into account.