Tektronix, leveraging technology acquired when it purchased MixedSignals last year, has developed a means by which the quality of IP streams using the complicated multi-rate segmentation employed with adaptive rate (AR) streaming can be more accurately ascertained. Initially, the firm’s new Sentry digital content monitor ensures quick identification and diagnosis of video and audio quality issues evident in the output from transcoders prior to entering the AR segmentation process, says Steve Liu, vice president of product management and business development at Tektronix. Post-segmentation analysis will become part of the platform next year, he adds.
“The H.264 codec used with adaptive rate streaming is much more complex than MPEG-2, so you need to look at the H.264 output from the transcoder at the same level of detail that we look at MPEG-2 with our traditional Sentry product suite,” Liu says. “With linear programming you have to do it in real time, and you have to look at all the individual bit rates a single program may require for streaming to different devices.”
Rather than monitoring packet loss as a way to measure quality of experience (QoE), the Sentry platform performs deep inspection and analysis by looking at each packet for sub-par performance caused by over compression, frame freezes and macroblocking. Simultaneously with the H.264 monitoring, The platform also examines the accompanying audio programs encoded in the AAC codec predominantly used in adaptive streaming.
Detection of video macroblocking is especially difficult on H.264 streams, Liu notes. Macroblocking occurs when one or more of the image components consisting of blocks of pixels within a given frame or series of frames show up as single color or low-resolution frame segments. Tiling, also known as quilting or pixilation, commonly used as a synonym for macroblocking, is the name given to the extreme case of single color blocks, but it’s also important to be able to detect the low resolution fuzziness that can occur with macroblocking.
“Macroblocking is easier to detect with MPEG2 because the standard uses fixed-size macroblocks,” Liu says. “With H.264, the sizes of the blocks are variable, which makes it extremely difficult to analyze. You need a state machine that looks at the entire GOB (group of blocks) and determines how they’re defined within each frame. Being able to do that really sets us apart.”
Over compression is another important component to watch for in the H.264 transcoding process, The Sentry platform now applies its Perceptual Video Quality (PVQ) analysis in the AR domain, which employs an eMOS (effective Mean Opinion Score) algorithm to detect and quantify picture blockiness and other artifacts caused by over compression, Liu says.
“If you look at the PVQ and, say, comparing CNN to ESPN, you see that CNN is scoring a 4.9 and ESPN is scoring 4.3 you can allocate more bandwidth to improve the quality on the more motion-intensive video stream,” he explains. “So it becomes a bandwidth optimization tool as well as an important way to keep track of QoE issues resulting from over compression.”
But as important as analyzing AR streams out of the transcoder might be there’s more that needs to be done to provide a full accounting of quality assurance for premium content delivered to IP-connected devices. Beyond the transcoding of a given piece of content into multiple bit rates for all the devices served, there’s also a fragmentation process associated with each type of AR technique used with various types of devices that must be monitored and analyzed in order to ensure minimum QoE requirements are met across all devices.
“Initially we’re focusing on AR at the high-burst point, but in 2012 we will introduce an enhancement to Sentry for AR that acts as an emulation client that pulls content from the system and does a full QoE analysis of the user experience,” Liu says. “We’ll be able to look at timing delays as well as quality of video across multiple client instances. We’ll be able to determine how long the delay is between request and delivery of the content, how long a delay there is when you hit pause, how many profile changes occur and what those profiles are during the program.”
AR fragmentation is a process which allows the receiving device to report back to the server how much bandwidth is available, which in turn allows the server to send out fragments of a given duration (which varies from one AR system to the next) at bit rates suited to the bandwidth availability such that no blocking or buffering will occur on the stream. But if the bandwidth availability is persistently limited to a point where the AR system is sending out sub-par quality video, as might happen if an HD stream to a connected TV is impeded by bandwidth restrictions to the point that the signal is arriving at very low resolution, the operator needs to know something is happening to cause an unacceptable QoE.
“If you’re delivering an HD program to the TV set at 500 kbps, that’s not good,” Liu notes. “So you’ll be able to set the system for alerts when your conditions for QoE aren’t met.
“For example, you might want to be alerted if over a given period of time, say, one hour, a certain type of device is receiving signals below 300 kbps for more than 50 percent of the time. It’s a proactive approach that allows you to be alerted to problems in the network or with the AR process that you otherwise wouldn’t know about until customers start calling.”
The new Sentry for AR distribution can be applied to over-the-top on-demand programming as well as live streams that are part of the broadcast lineup, Liu adds. And the H.264 analysis process can be applied in non-AR situations where over compression of IPTV programming can be a serious issue.
A major MSO, unnamed, has already deployed the new AR capabilities, Liu says, noting the deployment serves to illustrate the scalability of the Sentry system. “They’re using Sentry to monitor transcoding output on thousands of streams across all three of the parameters we’ve discussed,” he says.