There will be a notable technological shift with the optical components too; many of the new ones are based not on silicon or gallium arsenide (GaAs) but on gallium nitride (GaN).
Among the most awaited of the D3.1-enabling components is the system on a chip (SoC) that Broadcom Corp. (Nasdaq: BRCM) is developing for use in cable modem termination systems (CMTS). Broadcom recently told Light Reading that it will be sampling its CMTS chip in coming weeks. Cisco Systems Inc. (Nasdaq: CSCO) is a long-time Broadcom customer.
On the cable modem side, Broadcom and STMicroelectronics NV (NYSE: STM) began sampling chips for DOCSIS 3.1 cable modems early this year. Intel Corp. (Nasdaq: INTC) recently revealed it is sampling its new Puma 7 cable modem chip, fabricated in a state-of-the-art 14 nm process. CableLabs is in the process of evaluating the first wave of D3.1 products. (See The Chips Fall for DOCSIS 3.1 and Is Intel Auditioning for the iPhone 7?)
CMs and CMTSs tend to dominate conversations about DOCSIS, but cable also has a critical need for figurative nuts and bolts: various amplifiers, filters and other components that will be necessary to upgrade existing transmission lines so that they'll be able to support the wider bandwidths that cable operators will evolve to as they phase in D3.1 technology.
Most cable systems now have 870MHz of bandwidth (some systems are still at 750MHz, some others have 1GHz). D3.1 has provisions for expanding to 1.2GHz in the near term, and 1.5GHz or 1.8GHz in the future. MSOs use most of that bandwidth for transmitting content to subscribers, but a sliver is reserved for the return path. That has always been the band from 5MHz to 40MHz, but with the expansion of total bandwidth available, cable operators have the option to expand the upstream to 85MHz, 200MHz, or more. (See Comcast Puts DOCSIS 3.1 Live in the Field.)
Existing lines, depending on how long they are, might require as many as five or six amps per line. Since they tend to be positioned based on the properties of existing components, increasing the bandwidth by more than a third will often require adding an amp to the cascade and re-spacing them all. The technical term for this is "nightmare."
MSOs would prefer to replace existing components with new components that can support the greater bandwidth, which would be possible only if the new components have higher power, good linearity, and remain at least as reliable.
This cannot be easily accomplished with silicon, but it can with GaN.
Among the top contenders for this business are Qorvo Inc. , Macom and Anadigics Inc. (Nasdaq: ANAD), all of whom are racing each other to fill out portfolios of amplifiers, pre-amps, splitters, filters and other components.
Qorvo claims more than 60 products designed specifically for DOCSIS 3.1, including forward path amplifiers that operate from 45MHz to 1.2GHz, and reverse path amplifiers that operate up to 300MHz.
Qorvo is looking to distinguish itself with GaN circuitry built on a silicon carbide substrate, arguing that the approach renders superior results, including better thermal performance compared to GaN on silicon.
The company allows that it is taking what it expects will be a short-term cost hit for opting to go that route, however. At present, GaN on SiC can be done only on 4-inch wafers, while GaN on Si ICs can be created on either 6-inch or 8-inch wafers, making production of the latter more cost-efficient. So for the time being, GaN on SiC will be more expensive than GaN on silicon.
It will be up to the market to determine if the performance improvement is worth the premium.
Macom has built a portfolio of D3.1 push-pull amps and power doublers, in both GaAs and GaN, dividers, combiners and filters. The most recent additions include three families of diplex filters covering the following frequency splits: 42/54, 65/85, 85/105 and 204/258 MHz. All operate from 5 to 1.2GHz.
Anadigics similarly has a portfolio of push-pull amps and power doublers, mostly GaN, for D3.1 networks. The latest additions include a set of line amps in surface mount packages.
All three are exploiting GaN for another cable need. In new builds, many MSOs are installing new architectures that push fiber deeper to the edge, with more nodes all positioned nearer to customers so that no amplifiers are needed -- thus the phrase node+0.
Anadigics, for example, just introduced a fourth member of a new power doubler line that puts out four times the output power of the company's original line of GaN power doublers.
— Brian Santo, Senior Editor, Components, T&M, Light Reading