- Local Manipulation of 19.4 Mbps
with Artifact Issues Solved!
- From the very get-go in the world of digital
television, many engineers have asked why has our nation's Public
Broadcasting Service (PBS) elected to distribute their digital network
content at the Emission Rate of 19.39 Mbps, when ABC, CBS, Fox and NBC
(the networks) have opted to use the Distribution Rate of 45 Mbps?
-
- It’s rather simple. The Emission Rate approach
gives the local station a complete package, compressed and ready to be
fed into a DTV transmitter in an ATSC compliant stream. It is the
simplest, easiest and least expensive way for a station to broadcast
digital television, be it in standard definition (SD) or high definition
(HD). Within the digital PBS network, it is quite possible for program
material, in an MPEG format, to be managed and distributed entirely
within this compressed digital domain, from source to display.
-
- The other networks say their local affiliates
must insert various kinds of events over network feeds, such as supers,
bugs, EAS, etc., and the use of the 19.4 Mbps rate would introduce
objectionable concatenation artifacts attributed to
decoding/re-encoding; arguing this is especially true in HD programming.
Decode/re-encoding causes degradation in picture quality, irrespective
of the bit rate. The greater the compression ratio involved, the greater
the exposure to these kinds of picture degradation when going through
the decoding/re-encoding process.
-
- As digital television gains prominence, the
number of PBS member stations preferring to insert EAS information,
their IDs, etc., seamlessly into network content, rather than do
cutaways (be it in SD or HD), and have the confidence of maintaining the
network quality of service (QoS), is growing. So, if the
decode/re-encoding “step” could be eliminated, it would eliminate the
possibility of these problems. The benefits would be many. Bandwidth
requirements would shrink and, since that’s how common carriers
(satellite, fiber optic, microwave, etc.) charge, the savings there
alone could be substantial.
-
- Dealing with compressed signals and the
manipulation of encoded bit streams (more commonly known as bit
splicing) has been considered by many as something akin to voodoo or
magic. In an effort to find a solution, Edmund A. Williams, Senior
Engineer in PBS's, DTV Strategic Services Group was asked if he knew of
any company that has the technology that addresses this need, to which
he responded: "There have been several players out there, but none to
the extent that a company by the name of AgileVision has."
-
- AgileVision, a Sarnoff Corporation spin-off, say
they have “a solution” that will maintain network QoS pictures while
permitting local content manipulation in the compressed digital domain,
eliminating the need for decode/re-encode. This would eliminate that
“step” spoken of earlier and the associated picture degradation.
-
- AgileVision calls their product, the AGV-1000, a
“Station in a Box”. The box deals with three kinds of inputs:
-
- The AgileVision device will store MPEG-2 with
AC-3 audio. The units are shipped with two 18 GB hard drives, which
provide 16 hours of storage at SD bit rates or 4 hours at HD bit rates.
The different formats may be intermixed on the same system. The current
internal storage may be expanded up to 12 hard drives, e.g. , using 36
GB hard drives would provide approximately 200 hours of storage at SD
bit rates and over 50 hours of storage at HD bit rates, with this
capacity doubling once 73 GB hard drives are available, later this year.
-
- The Station in a Box essentially eliminates the
need to invest in dedicated DTV equipment such as an HD capable Master
Control Switcher, an HD capable Video Server, an HD Logo Inserter, a DTV
data inserter, and Multiplexing equipment, because these features are
built into the system.. If the station wants to postpone the purchase of
an expensive PSIP authoring system, the AGV-1000 also generates the PSIP
tables required to be ATSC compliant and necessary for viewers to locate
and tune to the station.
-
- This is a software driven system that runs on a
multiprocessor supercomputer platform which takes two 19.4 Mbps
transport streams coming into the system, one from the network and the
other from a local source (i.e. the station’s NTSC master control
output). These local signals will have had to be encoded as ATSC
compliant transport streams. Each of these two inputs enter a Transport
Stream De-multiplexer where they are broken down into individual program
feeds, each consisting of one Video Packetized Elementary Stream and at
least one Audio Packetized Elementary Stream.
-
- Those streams enter a preselector that provides
two streams to each splicer. The preselector’s function is similar to
that of a routing switcher. It is controlled by the internal software
command and control system, which AgileVision calls the Configuration
Manager.
-
- The Configuration Manager processes information
from the outside world for use by AgileVision’s AGV-1000. It can
reallocate processors and adjust software components to match the needs
of the input and output transport streams, e.g., if the system is only
dealing with HD content and there is only one program in each transport
stream, there may only be one set of splicer components and only one
logo inserter in use. But this is not always the case.
-
- The Configuration Manager takes control messages
from an automation system, PSIP from an external authoring system, input
from a datacasting system, and EAS data, passing it to the proper
components at the proper time. If there is no input from an external
PSIP authoring system, the Configuration Manager generates the minimum
PSIP tables required for ATSC compliance, inputting it through screen
prompts on a computer work station.
-
- The Configuration Manager controls the feeds from
the two “live” program transport stream inputs (network/local), and from
the fileserver (delayed for later broadcast content), to any of the
splicers. This is where the process of re-mixing the programs for a new
multiple program transport stream begins that will eventually be the
output of the system.
-
- In reality there is more than one splicer for
each program: a video splicer and an audio splicer associated with the
program path. The splicer has two inputs, the current signal that is
on-air and the next signal that will be spliced to the output. The video
splicer responds to signals from the Configuration Manager, making the
switch or splice to the new program source. All of the logic to make a
syntactically correct and visually seamless splice is in the splicer
component. After the video splicer selects the correct point for the
video splice, it feeds information to the audio splicers instructing
them where to splice the audio, maintaining lip sync.
-
- The video then moves to the logo inserter and the
audio to the audio mixer.
-
- In the logo inserter, the picture content is
analyzed by macroblocks (the stuff that MPEG-2 is made of), identified,
and then modified. The macroblocks that will not be modified, but are
referencing the modified macroblocks, are intracoded to remove those
references. The logo is then blended, and the modified macroblocks are
recoded using a patented "smart re-encoding" process.
-
- The video now passes to the rate control
component, which is tightly coupled to the logo inserter and uses its
picture analysis information. The rate control system receives
information from the Configuration Manager and modifies the bit rate of
the video stream in response to the needs of the statistical
multiplexer. Meanwhile, the audio signal has been processed by the audio
mixer where voice over audio and EAS are added. The audio stream passes
next to the statistical multiplexer.
-
- In the statistical multiplexer, all of the
program streams and other data such as PSIP data and opportunistic
datacasting data are multiplexed into a 19.4 Mbps, ATSC compliant
transport stream ready for presentation to the DTV transmitter.
-
- The block diagram shows a File Server which
records programs from the live inputs and plays them back as additional
sources available to the splicers via the pre-selector. The file server
can also accept and originate file transfers via the 1Gbps Ethernet LAN.
Files can be shared with other AgileVision AGV-1000 systems, or moved to
an archive system or other disk-based file servers on the LAN. The
file-transfer/sharing system (LAN) technology is identical to that used
in most offices around the country allowing virtually unlimited options
for networking architecture by using off the shelf hardware and software
components to implement LAN and WAN interconnects.
-
- The computer processing power AgileVision used to
perform functions is rather impressive. “Under the hood” are 16 Motorola
PowerPC® G4 AltiVec™ 400 MHz processors in parallel. One of these
Motorola chips, alone, is roughly twice as fast (for Digital Signal
Processing applications) as an 800 MHz Pentium® III chip. These chips
are connected in such a way so they can perform 51 billion
floating-point calculations per second (51 Gigaflops) with a cross
sectional bandwidth of 1 GB/second, supported by over 1 GB of RAM and
the AgileVision box is expandable from 16 to 28 processor chips,
providing up to 90 Gigaflops!
-
- Very favorable beta testing has demonstrated that
AgileVision is helping PBS in smoothing the transition to digital for
its member stations by providing the tools that reduce the need for
substantial capital outlay while maintaining the highest of broadcast
standards. The AVG-100 shrinks that digital danger zone to a more than
acceptable margin.
-
