between two TV stations is feasible but limiting, according to a report from
KLCS and KJLA. The two stations recently completed a channel-sharing pilot
project, which they presented as a potential “baseline assessment of sharing.”
“Due to limitations of both time and equipment, it was not possible to review
every scenario sharing could entail,” the report stated.
The stations tried out combinations of high- and standard-definition channels,
but not mobile broadcasting with ATSC M/H. Mobile
was mentioned only as a potential transmission format in the report’s
background summary on video encoding.
“Every broadcast entity chooses its format: HD, SD, Mobile Handheld (M/H) and
data,” it said.
Channel-sharing is similar to multicasting in that more than one TV signal is
transmitted in a 6 MHz channel. Several stations currently multicast as many as
six or seven standard-definition signals in a single 6 MHz allocation, though
the video quality tends to be diminished with each additional stream.
Channel-sharing throws in the complication of multicasting from dual facilities
and coordinating the bandwidth payload of the shared content.
The channel-sharing setup
used by KLCS
and KJLA combined a statmux pool with eight, late-model encoders, a standalone
PSIP generator, a broadcast-quality decoder, broadcast monitors, a
quality-of-experience analyzer, stream analyzers and recorders, and several TV
sets and set-top boxes. The encoders were fed by video servers, satellite
receivers and fiber feeds from the two stations. A variety of distribution
amplifiers and routers were used as well.
The report concluded that channel-sharing was possible on both the “physical
level,” referring to transmission, and the “virtual level,” referring to
whether or not receivers could decode the signals properly. Findings were based
on “empirical data.”
“On the virtual level, we found that all the TVs and tuners
tested were able to receive and correctly parse all the
required information,” it said. “This included virtual channel, both major and
minor, ratings, audio configuration, codecs, program titles and descriptions.”
On the physical or transmission level, the stations could transmit two 720p
signals in a single channel, depending on the “relative ‘digital complexity
’ of the video content,” and how the bandwidth is
managed. Two stations can either use dynamic or fixed bitstream allocation.
With dynamic allocation—or statistical multiplexing, aka
statmuxing—the most action-heavy video would dominate its less
visually frenetic neighbor. E.g., fast-action sports such as NASCAR racing would
be more bit-hungry than a singer standing still on a stage in “The Voice.”
KLCS and KJLA managed to squeeze two SD streams in with two 720p HD streams,
“without major impact to the quality of experience of the overall material,”
and said that more were possible. Three HD signals were transmitted, as well as
one HD and seven SD streams. Testing indicated these combinations were
“technically feasible,” depending on what level of picture quality the sharing
stations would accept.
The report did not specify what encoders
were used, only that “a more efficient use of MPEG-2” was found with the
“latest generation of encoders.” And since these newer models also employ MPEG-4
H.264, it was also tested. MPEG-2 is the federally mandated broadcast TV
transmission format that TV set manufacturers must accommodate. MPEG-4 is used
more commonly for DVDs and Internet video.
“We found that some consumer televisions decoded H.264 off-air,” the report
said. “While this data is anecdotal, if the majority of future television sets
can utilize H.264, greater bandwidth efficiency and/or quality of experience may
be possible. H.264 offered a 10 to15 percent greater efficiency in bitrate
As for picture quality
, the stations
used the Differential Mean Opinion Score, or DMOS, based on an algorithm from
the University of Texas used to “approximate and predict the scores people
would report if they saw two pictures side by side.” The scores weighted as
Zero is “no defects;”
0.01 to 1.0 is considered “production quality;”
1.0 to 7.5, “broadcast quality;”
7.5 to 8.5, “annoying;”
8.5 to 10, “unwatchable;”
10 and above, “probably not aligned.”
The stations measured DMOS with test equipment from VideoClarity while
engineers watched the transmissions on professional HD monitors.
“We found that our engineers, knowing what to look for and prompted by the spikes
found by the VideoClarity, were able to see some faults when the DMOS score
exceeded 6.0,” the report said. “However, these faults were only discovered
with the aid of a seamless split or side-by-side comparison. Without the
comparison, the picture appeared to be nearly perfect at 6.0 because the
picture area where the faults were observed were generally soft or blurred in
The testing also revealed the efficacy of newer encoders. Neither station knew
its DMOS going in. KLCS scored between 7.0 and 8.5 with a peak of 9.0 with
10-yaer old encoders. KJLA scored between 6.0 and 7.0 with six-year-old
encoders. The newer gear brought KLCS to a 1.5 and KJLA down to 0.69 with a
high of 1.56.
Another crucial element of the testing involved PSIP
—Program and System Information Protocol. PSIP is the data used
by receivers to find channels. The stations used a “new version of the PSIP
generator” provided by an unidentified vendor.
KLCS transmits on Ch. 41 and appears on virtual Ch. 58 on receivers. KJLA
transmits on Ch. 49 and shows up on Ch. 57.
“The testing of the PSIP generator was designed to determine how televisions
would react to physical 41 signaling both virtual 58 and 57,” the report said.
Some sets found all information on Ch. 58.1 with no prompting; some had to be
tuned to Ch. 41.1 to get the information; some needed to rescan. All, however,
ultimately were able to receive and parse the required information.
The bitrate used for PSIP was set at 40 kbps in consultation with the
manufacturer. The rest went to video and audio. For the tests, audio was
allocated 128 kbps per stereo pair, 64 kbps for dual mono and 384 kbps for 5.1
surround. All SD channels were configured for stereo, with only the HDs
assigned 5.1. Video compression
from a 1080i versus 720p perspective, since TV sets support both. Retransmission
through cable and satellite systems and set-tops was also considered. Given
that “broadcasters have no control over the final display format in the home,” KLCS
and KJLA opted for the most “efficient encoding structure for final
distribution over the air.”
The stations found that 720p had a better DMOS than 1080i had half the bitrate
by comparing broadcasts of the “NewsHour.” A baseline was then established
using 720p at a constant bitrate of 18 Mbps for two clips. The first was
12-minute video of Las Vegas ranging in encoding complexity from “moderate to
difficult.” Average DMOS was 2.16. A second clip from the Ken Burns “National
Parks” documentary yielded a 2.31.
Putting the two in a single 6 MHz channel, transmitting each at 9 Mbps, yielded
a 3.86 and a 3.03 respectively. Statmuxing
the two yielded a 3.07 and a 3.03 respectively with both encoders set at 11-2.5
Throwing in a third HD stream—a clip of underwater footage—brought the scores
to 3.42 (Las Vegas), 4.11 (National Parks) and 3.6 (Underwater), with the
encoders still set at 11-2.5 Mbps.
At the high end, the two stations packed one HD and seven SD signals into one
channel, using “Vegas” and the “NewsHour” for different versions of the test.
The HD signal encoding range remained 11-2.5 Mpbs, while the SD encoders were
set at 2.5-0.85. In this scenario, the highest DMOS for SD was 3.16. “Vegas”
yielded a 4.55, while the “NewsHour” scored 2.81.
All numbers were within the acceptable DMOS viewing range, the report noted. It
concluded by noting that channel-sharing is technically feasible, that
statmuxing is preferable and advanced video compression is desirable.
“Whether two entities can successfully combine their required program streams
and business models can only be answered by the stations interested in
sharing,” it said.
The full report is available here