Multicasting Special Report | Improved Encoders Equal More Diginets
It’s not just consumers’ appetite for more diverse, ad-supported programming that is driving multicasting these days. It’s also significantly improved compression technology that has enabled TV stations to fit more channels into their broadcast pipe at a low cost, while still maintaining acceptable video quality.
New software-based MPEG-2 encoders that run on off-the-shelf computer servers are expanding the potential of stations’ ATSC 1.0 channels, with improvements of roughly 30% in efficiency compared to systems from five years ago and over 50% compared to decade-old encoders.
The new MPEG-2 systems are also far cheaper than the dedicated hardware units of the past, with the flexibility to add program streams with the simple purchase of additional software licenses.
“MPEG-2 is very alive again,” says Elke Hungenaert, VP of product management for Synamedia, a London-based firm formed last year when Cisco sold its video encoding business to private equity firm Permira.
Like major competitors Harmonic and Ateme, Synamedia’s MPEG-2 encoding is based on a single software codec core that it licenses and sells either separately, to run on a customer’s own hardware, or bundled with common servers like HP or Cisco.
Stations are taking advantage of the improved MPEG-2 technology, with many now delivering three or more diginets in addition to their primary HD network stream as way to boost revenues.
Robert Lydick, VP of information technology and station operations for Tegna, says the station group tested next-generation Harmonic encoders in 2018 with the aim of launching a new subchannel and made a “sizable investment” in the new technology.
“In coordination with that, we’ve had really positive results,” says Lydick. “We were able to immediately add at least one subchannel, while meeting our contractual obligations” to the major networks.
Due to channel-sharing agreements in the wake of the RF repack, numerous big-market stations today are even squeezing two HD channels along with multiple SD diginets in their 19.4 megabit-per-second payload. That’s a big change from the traditional model, in which the single HD channel generally was delivered at 10 to 14 Mbps and SD diginets got 2 to 3 Mbps each.
ATSC 3.0 proponents hope to take MPEG-2 compression even further. They are counting on powerful new encoders to pave the way for 3.0 by allowing broadcasters to “channel-stack” the existing 1.0 channels of several stations in a market so they can free up at least one “host station” for ATSC 3.0 services.
John Hane, president of Spectrum Co. LLC, the ATSC 3.0 spectrum consortium founded by station group giants Sinclair and Nexstar, notes that unlike the original analog-to-DTV transition broadcasters haven’t been given a companion channel for ATSC 3.0 so they could simulcast two standards at once to consumers. But he says digital video engineers have created a viable alternative with improved MPEG-2 compression.
“How do we maintain quality in 1.0 and service in 1.0, and launch exciting new services and formats in 3.0 at the same time?” says Hane. “Well, we have found that the companion channel of the 2020s is actually going to be the really nifty, powerful, wonderful new encoders. That’s where we’re going to get the capacity that we have to have if we’re going to make this switch.”
A new model for MPEG-2
The software model is a big shift from the dedicated boxes of ATSC 1.0’s early days. Mark Aitken, SVP of advanced technology for Sinclair Broadcast Group, remembers when a dedicated Harmonic hardware encoder that could compress one high-quality HD stream cost $80,000. A few years later, a $50,000 unit could handle one HD and 2 or 3 SDs, which helped open up the market for diginets.
“So, the next time a guy starts knocking on your door and wanting to do a deal, to rent 2.5 megabits from you, or do a revenue share, you could accommodate it,” says Aitken.
Software-based encoders existed then, but weren’t cost-effective because of the expense of the processing power required to run them. That started to change about five years ago due to steady improvements in computer processing power (as per Moore’s Law) combined with declining server prices.
Now Sinclair has committed to MPEG-2 encoding technology from French firm Ateme, which it buys as software licenses on a per-stream basis while providing its own COTS hardware.
“It’s always been about computational power and memory speed and memory capacity,” says Aitken. “But suddenly we’ve reached the cross point where conventional hardware has got the horsepower.”
Today, Sinclair stations typically have one HD stream plus three or four SD diginets operating in a statistical multiplex or “statmux.” That load will soon go to one HD and five diginets as the station group deploys Ateme in a gradual rollout across all of its stations.
When it comes to overall compression efficiency, the result is “like flipping a switch,” he says.
“Suddenly there are huge gains in the respect of the number of streams you can actually encode and the quality of those streams—because suddenly you’re spreading the statmux across more unique streams,” he says.
Sinclair will also use Ateme’s HEVC encoding when it launches next-generation ATSC 3.0 broadcasts in multiple markets in the next year. But given the growth of the diginet business, buying new MPEG-2 encoding would make sense even if ATSC 3.0 didn’t exist, says Aitken.
With the large audience that big groups like Sinclair and Nexstar can deliver after years of consolidation, reaching 60% of the country or more, programmers want to do business and place their channels.
“The reason we’re swapping out encoders is we’ve got money-making opportunities,” says Aitken. “You swap out an encoder, add diginets, and add dollars to the bottom line.”
Aitken wouldn’t give an exact dollar number for how much Ateme’s encoding costs. He says the economics favor software encoders and that the licensing model is “extremely advantageous,” as a license for a single video channel allows for encoding in a variety of formats including MPEG-2, AVC and HEVC as well as streaming formats like DASH and HLS.
“We only charge by the input; we don’t charge by the output,” confirms Ateme Director of Sales for North America Lenny King. “On the output, you can send it a thousand different ways.”
Ateme makes Titan statmux software as well as its own SDI input board, and sells its encoders in a number of configurations including bundled with third-party hardware like Dell servers. A “future-proofed” system capable of handling two HD and two SD inputs includes encoding and statmux software, VMware software, SDI and ASI cards and the server, says King.
Ateme, which already has a number of cable and OTT customers began targeting the U.S. broadcast market around 2016. Sinclair is its major broadcast customer, and Ateme is going after smaller groups and stations through resellers like Comark Digital Services. A recent win is PBS member station WSBE in Rhode Island, which bought a complete ATSC encoding solution from Comark based on Ateme encoders and Titan statmux software that compresses 2 HDs and two SDs into its current channel.
Harmonic, which set the initial bar for HD quality with its MV450 hardware encoder, counts Nexstar, Tegna, Fox and NBC as major customers and still has a large installed base across Sinclair.
The company is now on its sixth-generation of MPEG-2 encoding with the software-based Electra X series. Electra X is 25% more efficient than the previous Electra 8000 and 9000 hardware models, and 30 to 50% better than the early MV400, 450 and 500 generations.
Compression efficiency comes down to three big factors, says Jean Macher, director of broadcast market development for Harmonic: the core codec, the preprocessing and the statmux. Since the core codec is close to reaching the limits of improvement for MPEG-2, Harmonic is focusing on the preprocessing and fine-tuning the statmux.
“That’s how you manage to squeeze more efficiency into the system,” says Macher.
Harmonic sells Electra X in a variety of ways, either as a bundled appliance; as a software appliance to run on customer hardware; and even in a SaaS model running on either public or private cloud. There is also an upgrade path to ATSC 3.0 with Electra X.
Macher thinks the cloud-based encoding model could be particularly attractive for 3.0 launches. “For all the different deployment models, it’s the same software for us,” he says.
Macher says diginets are driving Harmonic’s MPEG-2 encoding business, as a station’s cost of buying a new encoder to add a diginet is usually paid back in less than a year.
“One of the reasons we still see a very good business in the ATSC 1.0 market is because of the success of the diginets,” says Macher. “The return on investment is quick.”
And sometimes local stations don’t even have to bear that cost. Macher says. In some cases, the programmer has bought the encoding platform and given it to the broadcasters as part of their distribution deal.
“We do that today,” says Macher. “Of course, the flip side of that nice story is that is even more ATSC 1.0 channels to find a home for.”
After the repack, Harmonic is currently supporting multiple-HD channel-share lineups for network owned-and-operated stations in New York, Los Angeles and Chicago.
The most common configuration is two 720p HDs plus three 480i SDs in one 6 MHz channel, though there is one station with two 1080i’s and two SDs.
Harmonic also has an O&O channel-share customer in Charlotte doing two 720p’s and five SDs; an independent channel-share in New York doing one 1080i HD, one 720p and five SDs; and an independent channel share in Allentown, Pa. delivering two 720p’s and a whopping seven SDs.
Macher describes such configurations as “aggressive” and doesn’t recommend them for most broadcasters. “It depends on your threshold, as a broadcaster, and your acceptance of video quality,” he says.
For its part, Sinclair has done its homework through extensive testing with encoders from Ateme, Harmonic and Synamedia and “a huge amount of objective picture measurement,” says Aitken.
“At the end of the day, we can do three HDs in a channel, we can do two HDs and four or five SDs in a channel, or we can do one HD and seven SDs in a channel,” he says.
In the seven-plus-one statmux Aitken describes, the average bitrate of the 720p HD would be 7-8 Mbps while the SD would average 1-1.5 Mbps. Ateme’s King says that in tests it has compressed 720p HD as low as 3 Mbps, but says that 7 to 8 Mbps is a more realistic range for HD sports.
“It depends on the complexity of the content,” says Aitken. “It costs a lot more to do football and basketball than to do baseball or soccer. The reason is because of the fine-grain detail of the focused grass in football and the defined woodgrain of a basketball court.”
A new currency for compression
As it prepares for ATSC 3.0, the Fox network is now doing a lot of work in the lab to figure out just how much content can actually fit in ATSC 1.0, says Winston Caldwell, Fox VP of spectrum engineering and advanced engineering. In the near term, it is also trying to put together a channel-share arrangement in Phoenix in order to launch the second ATSC 3.0 station in that market.
The answer to the channel-stacking question can vary widely depending on the time of day and what content is being shown.
“It’s not just whether it’s HD sports, but what kind of sports?” says Caldwell. “It’s different between a drama, a game show, and something like ‘American Idol,’ which has a lot of dazzling, swirling lights in the background. There’s not a lot of movement individually, but there’s a whole lot in the set.”
Fox has already done some channel-sharing configurations in markets where it had to repack stations. Where it has a duopoly, its strategy has been to pair the sports-heavy Fox HD feed with the MyNetwork TV HD feed, which doesn’t have much sports content and doesn’t require as many bits on average. That has worked well, says Caldwell.
Most Fox stations have an HD plus two SDs while many MyNetworkTV stations have an HD plus three SDs. That load is due to get bigger in the third quarter, when Fox-owned stations in 12 major markets plan to add another diginet, Weigel Broadcasting’s Decades.
While Caldwell won’t disclose what Fox’s minimum HD bit rate today is (some affiliates say the network asks for 12 Mbps), he says the network has some “pretty strong philosophies around a minimum bit rate requirement” because that assures a certain level of picture quality. But he concedes that in a statmux world, bit-rate requirements on any given program stream are inefficient and “the preference is to allow the encoders to do their magic.”
“There are some times Fox is not showing football, and we may be showing an ad where nothing has changed for five or ten seconds,” says Caldwell. “So to use that bitrate really hamstrings you when you’re trying to pack up stuff in ATSC 1.0 to use [free capacity] in ATSC 3.0.”
Affiliates are already contacting Fox about how to manage the network’s HD feed in a channel-stacked world, and Fox is exploring different video quality measurement tools from vendors like Video Clarity to help advise them.
The metric it is currently considering is a standard PSNR [peak signal to noise ratio] measurement which can be reported through various systems. It has also looked at SSIM [Structural Similarity] and JND [just-noticeable difference] measurement as well as human perception tests. Caldwell concedes that Fox may need to create something new.
“There’s no metric that gives us confidence, if we set the encoder to that, that we know we’re getting a certain video quality out,” says Caldwell. “We need to come up with a new metric, whether it’s static or dynamic over time.”
One way to handle the challenge would be to just drop the bitrate “and see when people start giving you phone calls,” says Caldwell. But he thinks that “would be an awful, terrible way to do it.”
The first question Macher fields from Harmonic customers is how many HD and SD streams they can fit in ATSC 1.0 with the new encoders. The second is how that knowledge can help them frame new channel-sharing agreements between stations as they prepare for 3.0.
“I think everybody would like to have some type of magic video quality language that you could put in there,” says Macher. “The problem is that there is no universal metric for video quality today that everybody agrees on. There are different tools, there are great tools, but nothing that is what I call a ‘currency.’ And because we don’t have that, bit rate remains the thing that you can measure, and that remains the currency.”
Tegna’s Lydick says a new currency for 1.0 compression is needed to help make 3.0 launches feasible.
“A current metric that we use in a lot of our agreements is the bit rate,” says Lydick. “And that’s a legacy, or a holdover, of encoders that aren’t even in the signal path anymore. So, we really have to come together, both from an affiliate perspective and a vendor perspective, and look at this challenge and come up with a different metric on how we can qualify and quantify this challenge that we have ahead of us.”