Faster Down, Faster Up: Can’t Have One Without the Other

The return path, long the poor cousin to downstream in terms of throughput and spectrum allocation, is beginning to demand its due. Consumer behavior is driving the attention, with the explosion in popularity of video-rich, two-way applications now impacting Internet usage patterns.

Service providers are responding. Cablevision grabbed headlines in May by offering up to 101 Mbps downstream with its Optimum Online Ultra tier, but it also increased its top upstream speeds to 15 Mbps. Over the summer, Verizon increased the top upload speeds on all its FiOS Internet offerings to 20 Mbps.

Comcast is moving in that direction. "Next for DOCSIS 3.0 is the completion of downstream channel bonding and the implementation of upstream," Comcast EVP and CTO Tony Werner said last month in CT’s Communications Executive.

That work has begun, with some markets already having bonded several upstream channels.

A bottleneck?

Downstream and upstream are linked. But it’s not just marketing; there’s a technical explanation.

A limited upstream can inhibit one’s ability to keep pace in the downstream speed wars. The reason is transmission control protocol (TCP), which is used to send acknowledgements (ACKs) from one endpoint to another; for instance, from a browser back to a Web server notifying it of any corrupted information. Techniques such as TCP ACK suppression (TAS) help, but the principle holds.

"When you are using TCP, you have to have enough upstream available so that the downstream can reach 100 Mbps," Todd Kessler, ARRIS VP of product management, said.

Average users might not notice a bump in speed using simple applications like Web browsing. But as bears repeating, usage patterns are shifting. Market research firm IDC, estimates that the number of U.S. households capable of generating interactive video will reach 60 percent by 2013 and that over that period monthly totals of user-generated content will grow from 8 GB to 13 GB, representing a compound annual growth in traffic just shy of 15 percent.

That’s believable. A little over a year ago, the maximum size of video uploads allowed on YouTube was 100 MB. Now it’s 2 GB.

"There is a bottleneck out there on the horizon," Robert Howald, Motorola customer systems architect said in a CT-hosted webcast in September that focused on the upstream.

Capacity options

To increase upstream throughput per subscriber, an MSO can bond four 6.4 MHz-wide channels delivering 30 Mbps each to achieve 120 Mbps. This addresses peak rates, however, not average throughput. Increasing a user’s upstream burst rate, taking TCP mechanisms into account, can actually slow down acks for other users.

"Higher peak rates (help) for 5 percent of users," Howald said. "(The) way to make sure (everyone) is treated fairly from a bandwidth perspective is to find a way to add more capacity."

The way to do this is threefold, Howald said. Split nodes, move to a higher modulation profile and use the low end of the spectrum. As for node segmentation, that’s been going on for several years.

"We advocate them continuing to (reduce service group sizes). It helps the downstream and the upstream, as well," Howald said.

The cable industry in the U.S. is also seeing operators make the transition to 64 QAM. Smaller-tier Midcontinent Communications has been one of the early adopters. But Midcontinent co-owner Comcast is also moving ahead. The move is challenging, wrought with physical layer issues, especially noise.

"When you increase the modulation to 64 QAM you immediately suck (away) a 6 to 7 dB margin compared to a 16 QAM network… When you go to do this, some of the things that didn’t matter before, will matter," Howald said.

This raises the premium on basics, such as those advocated by CT Senior Technology Editor and Cisco Technical Team Leader Ron Hranac, who has presented two CT-sponsored webcasts on 64-QAM upstream with Midcontinent Lead Engineer David Haigh.

Converting to 64 QAM also raises some questions about lasers. "Some of the newer F-Ps (Fabry-Perots) will survive for a while, but they create a situation with less margin…(And,) they do limit you in range and SNR (signal to noise ratio). DFBs (distributed feedback lasers) can do a higher modulation and have good SNR," Howald said.

Fabry-Perot lasers, however, don’t support channel bonding over "any reasonable distance," according to a recent white paper published by Aurora Networks. DFBs can tackle four-channel bonding, but only at distances of less than 20 km.

"The load you are putting on the transmitters is quadrupling. You have to reduce the drive to the laser," John Dahlquist, Aurora VP of marketing, said. "By doing that, if you keep the DFB, you will have to do some kind of cascading…The easiest way is to replace the transmitter with a digital (one)."

Aurora has long championed digital on the return path.

"Digital return technology is not impacted by the additional channel loading that is going to take place. We convert the analog signal into the digital signal right at the return path transmitter…Therefore (we have) plenty of capacity," Dahlquist added.

Bottom of the barrel

Then there is the option of harvesting the bottom portion (approximately 5 MHz to 20 Mhz) of the spectrum, historically ignored due to ingress noise.

With that capacity now needed, however, operators can use this spectrum by implementing synchronous code division multiple access (S-CDMA). Introduced in DOCSIS 2.0 and enhanced in DOCSIS 3.0, S-CDMA stretches the QAM symbols out in time by multiplying by orthogonal codes and transmitting in parallel without compromising the data rate or decreasing channel capacity.

"It is very powerful in the low end of the return band where there is impulse noise combined with ingress," Howald said. S-CDMA can increase capacity by 50 percent compared to advanced time division multiple access (ATDMA).

Motorola is testing the technology in the field with two operators, according to Howald. ARRIS recognizes its noise-resistant value, too.

"It uses a slower symbol rate because it is transmitting multiple symbols at the same time," Kessler said. "If it is a very long symbol time and a very narrow noise, the noise is much less likely to impact that set of symbols."

Other options

Although it would take time to implement, U.S. cable operators could broaden their plant by extending the range of the upstream. Currently, the upper boundary is 42 MHz, although DOCSIS 3.0 has actually defined an upstream extension to 85 MHz.

"Europe has the advantage in terms of total available spectrum. (They use) another 23 MHz," Kessler said. "U.S. operators…could extend the plant upstream and go up to at least 65 MHz."

Operators in the U.S. would have to be careful of broadcast television stations. "But with analog TV transmission going away over time, (extension) becomes an option if (they) need more raw megahertz to send data," Kessler said. "This is an option, but a longer term one."

Constraints have inspired other cases of ingenious engineering.

In a paper presented at the SCTE Conference on Emerging Technologies in April, ARRIS CSO Tom Cloonan shared results from tests indicating that after introducing impairments, four 1.6 MHz-wide bonded channels provide twice the throughput of a single 6.4 MHz-wide channel. The gains accrued not simply because of stat muxing but also because of the narrower channels’ ability to bypass noise.

One co-author of that paper, ARRIS Senior Systems Architect Ayham Al-Banna, also has investigated on behalf of MSO engineers the use of orthogonal frequency division multiplexing (OFDM) and WiMAX links to augment the upstream.

What about RFoG?

Then there’s all-fiber. An RF over Glass (RFoG) micronode in the home modulates cable modem signals to the optical carrier for transport to the hub. A burst mode detects the RF level from the cable modem, turning a laser on when it reaches a certain threshold.

"No noise will activate the laser," Ketan Gadkari, director for RF and optical systems for Alloptic, said. "Only a good signal from a cable modem will activate the laser."

Whether that’s true for all micronodes, which often feature always-on lasers, is another question. But Gadkari believes the technology fits. "RFoG provides a clean enough environment where (operators) can use the bottom of the spectrum," he said.

While reduced noise is one of several benefits to going all-fiber, MSOs naturally are keen to ensure that any RFoG or other passive optical networking (PON) solutions they’re considering play nicely with their service bundles, DOCSIS 3.0 and digital video included.

But that’s a discussion for another day. For now, we’ll note that that effective upstream tactics include HFC plant basics, possibly different lasers and transmitters, advanced modulation and a willingness to take a fresh look at the low end of the spectrum.

-Monta Monaco Hernon is a contributor to Communications Technology.