Cable Nodes Becoming a Choke Point
Cable companies have been talking about a network architecture called node-plus-zero for several years, and finally some have begun deploying it, with Comcast and Cox leading the way. Node+0 is finally being deployed now because only recently did it become technologically and economically feasible, thanks to the introduction of a new generation of power amplifiers based on gallium nitride (GaN). (See Comcast Goes N+0 in Gigabit Markets and Cox Reveals Next Steps for D3.1, CCAP.)
Cable operators are also considering increasing the amount of spectrum in their networks, expanding from 1GHz to 1.2GHz. Plus, they're evaluating an architectural change referred to as remote PHY. Although all three are distinct technologies, they all place new demands on cable nodes, and for now there's no way to accede to all those demands simultaneously.
Qorvo Inc. is one of the very small handful of suppliers of new GaN power doubler amps that help make all three upgrades possible. Macom has a gallium arsenide (GaAs) amp that also supports 1.2GHz of spectrum. Shenzhen Sanland Technology appears to have a similar part. A barely known startup called One Tree Microdevices claims to have a device that might serve the purpose in its portfolio.
Cable operators, in their hybrid fiber coaxial (HFC) networks, run fiber from their central headends to nodes placed in every neighborhood, and then run coaxial cable from nodes to customers' premises.
The coaxial run starts with amps in the nodes, supplemented by as many as four or five more amplifiers spaced along the line to maintain signal strength on the way to customers. A network with four or five amps per line would be denoted as node+4 or node+5, respectively. Node+0, then, is an architecture in which all amplifiers are removed from the line, leaving only those in the nodes -- commonly four per node.
Because the terminology isn't obscure enough already, Comcast Corp. (Nasdaq: CMCSA, CMCSK) and some other companies also refer to node+0 as "fiber deep." Not only are some companies ditching the cascade of amps on their coax lines, they're moving their nodes closer to their customers, extending the fiber portion of the HFC network "deeper." That naturally makes the coaxial runs shorter.
Migrating to node+0 eliminates the cost of periodically replacing all those amps, and also the expense of maintaining them all, including the cost of powering them. A cable operator might end up adding 20% more nodes, but eliminating 80% of the active devices in its network compensates enough that an operator might be able to cut its overall power bill by as much as 50%.
Amps tend to have 15- to 18-year lives, and the last major wave of amp upgrades occurred between 12 and 15 years ago. At the time, node+0 was not an option because the approach requires amps in the nodes to compensate for removal of the subsequent daisy chain of amps on the line. Not only did they have to be more powerful, they would have had to do it without blowing the power budget for the nodes. Amps that could do all that simply did not exist back then.
During the last wave of amp upgrades, MSOs had another factor to consider, though. The spectrum available in a cable network is dependent on the frequency range that amplifiers installed in the network can support. The latest amps at the time wouldn't support node+0, but they would support higher frequencies up to 1GHz.
So the choice operators had was that they could buy amps that would get them to 870MHz or amps that went to 1GHz. If they bought the 870MHz amps, it would be a problem if they needed to go to 1GHz shortly after, because then they'd be buying and installing yet another round of amps, a waste of money and effort. The alternative was to buy amps that supported 1GHz from the get-go. If the operator needed the additional 130MHz, it was there. If not, the company had to replace its amps anyway. There was no single correct answer; different MSOs had different needs.
Today, cable companies are in a similar situation. Installed amps are starting to age out, and MSOs are going to have to do something about it. This time around, though, the latest amps -- the new GaN models -- can not only support yet another expansion of the frequency range, up to 1.2GHz, but are also capable of compensating for the removal of the daisy chain of amps on the line, enabling node+0.
It is absolutely no coincidence at all that the latest iteration of cable's DOCSIS standard, DOCSIS 3.1, was engineered to support a spectrum expansion to 1.2GHz. (It also anticipates a subsequent step probably to 1.7GHz, and allows for additional expansions beyond that, should they ever become useful.)
GaN amps represent a legitimate technological leap, but they aren't magic. If operators want to do node+0 and add spectrum, that's difficult. If operators want to do those two things and also prepare for remote PHY, they're asking too much of what's available now.
Why? Node+0 requires more powerful amps for the nodes to drive signals the full distance from node to customer premises, and they have to do it without drawing significantly more current. The only way to increase the power while minimizing the current draw is to switch to GaN-based amps, and even then, the switch to GaN amps is not a perfectly ideal solution. GaN cannot reduce the power required; but it will draw less additional power than all alternatives.
Previous amps operate at 11W or so. The new GaN amps draw about 18W, so with four per node that's 28 extra watts. So far that's been acceptable for those operators eager to do node+0.
There's a complication for cable operators that want to do node+0 and simultaneously increase spectrum to 1.2GHz, however.
A node+0 system at 1GHz typically operates at 58 dBm out, but attenuation increases at higher frequencies, so if an MSO wants to do node+0 and go to 1.2GHz, it needs more power.
Comcast is upgrading its amps now and it wants the option to go to 1.2GHz, so it's asking for 64 dBm out, tickling technological limits all up and down the supply chain.
What about adding remote PHY? Remote PHY doesn't put any stress on the amps, but what it does, however, as it moves resources from the headend to the node, is add another increase in power required by the node.
"You have fixed power consumption the node can carry. You're asking for more performance, but you're asking to cut the power down? Well, it's just not going to happen," observed Kellie Chong, director of CATV and Broadband Access products at Qorvo.
What does appear possible is amp manufacturers reducing GaN device power consumption just enough to keep the overall power budget for the node unchanged when remote PHY is factored in.
Qorvo said it is looking at a technique called digital predistortion (DPD), which is commonly used in the wireless industry but is less well known in the cable industry. DPD algorithms predict the non-linear behavior of amplifiers, Chong explained.
"If you can correct for that, then you might be able to save power from the amplifiers so you can have remote PHY and node+0," she said.
The cable industry can expect that in maybe in a year, a year and a half she said.
There are also rumors of some unspecified digital alternative still being worked on in laboratories that might bear fruit in a year or two.
Every MSO has different needs, and all of these considerations will factor into their decisions on when to embark on all of these upgrades.
It turns out that many MSOs who have networks capable of 1GHz still aren't using the extra 130MHz they bought themselves when they upgraded from 870MHz. Toss in the complications associated with also going node+0 or opting for remote PHY, and many operators are apt to wait a year or two for the next generation of improved amps before they embark on their upgrades.
And it's unlikely that those MSOs already doing node+0 will also do remote PHY any time before that.
— Brian Santo, Senior Editor, Components, T&M, Light Reading