Case Study: SyncCast Rises to the High-Def Challenge
Not too many years ago, only streaming media fanatics had the patience to sit through buffer-challenged 120x160 videos on the web. Today, with more than 54 million broadband connections and more than 52 million HDTV sets in U.S. homes—and a new generation of set-top boxes (STBs) entering the mainstream—expectations have changed. Stuttering, blocky video just won’t cut it anymore. Last year, when Microsoft prepared to launch Xbox LIVE Video Marketplace to deliver high-quality SD and HD content to Xbox 360 owners, the company turned to a trusted partner, SyncCast, to build a secure, centralized video hosting delivery platform. (In July 2007, SyncCast was acquired by Thomson, but since this project was completed before the acquisition, we'll keep referring to the company as SyncCast.)
The Anaheim, Calif.-based SyncCast is a leading provider of digital media technology; it has extensive experience in application development, system integration, hosting, connectivity, and content distribution services. The company has placed particular emphasis on the development and integration of complex digital rights management (DRM) solutions for the entertainment and broadcast industries. Since the company’s founding in 1999, SyncCast has often partnered with Microsoft to develop digital media distribution solutions incorporating Microsoft’s DRM technologies. So it made sense for Microsoft to turn to its longtime partner for the launch of Xbox LIVE. For SyncCast, encoding 500 hours of high-quality content in the 3 short weeks before a Nov. 22, 2006, hard launch date—and an additional 500 hours in time for the Consumer Electronics Show in January 2007—was the toughest challenge the company had yet faced.
SyncCast had more than 50 Dell PowerEdge 2950 dual-core, dual-processor servers at its Anaheim encoding farm, and it was prepared to handle the volume. The key issues were time and quality control (QC). Lance Ware, founder and CEO of SyncCast before its acquisition, and now chief technology officer for Technicolor’s electronic distribution services group, points out that even without QC, producing high-quality encodes is a time-intensive process.
"We use two-pass encoding, and we substantially tune the encodes with various tweaks," says Ware. "The encoding rate on even those super high-end boxes is greater than four to five times real time, and up to eight times real time for HD. That’s just the processing, excluding QC and setup." To ensure that the encoded content maintained the high image quality that discriminating Xbox gamers would expect on their HD screens, SyncCast turned to Inlet Technologies’ Semaphore QC automation software.
With Semaphore, SyncCast could run newly encoded files through automated software to flag anomalies without staff having to view every minute of every encoded file. "Xbox has a lot of long-form content, including 24-minute episodic and 90-minute theatrical," notes John Bishop, Inlet’s senior vice president of business development and strategy. "SyncCast didn’t want to invest 80 minutes in watching a movie, and then get down to the end and say, ‘Man, there are a bunch of bad artifacts in those last 10 minutes.’ With Semaphore, they know ahead of time, very quickly, that the file is not going to pass any QC. They can weed it out and re-encode with an alternate profile." Semaphore also enabled SyncCast to maintain a relatively light staff. Without efficient automated QC, it would have required lots of eyes to watch the 600 to 1,000 hours of content that SyncCast ingests each month.
The most common problems that Semaphore identified were dropped frames—common with video that fades to black—and bit starvation, particularly with HD encoding. STBs like the Xbox pose additional problems with buffer utilization, that is, how full the encoding buffer is at any point in time. "Set-top boxes handle that differently than a PC," says Bishop. "A PC has a ton of memory, so it doesn’t have to worry about buffer underflows or overflows as much as a set-top box." Loss of audio is another common problem, and Semaphore can flag the timecode where audio drops out. Staff can then quickly determine whether there is an audio problem or just a quiet section of the content, without having to listen to the entire encode.
Semaphore is customizable, allowing users to develop specific quality alerts for quick identification of audio and visual "hot spots" within content. "The most common metric for quality is quantization, which is a great indicator of ‘did it look good or not,’" says Bishop. "Users can set up thresholds that say, ‘Anything above a quantization of 10, alert me, because I know at that point I might want to visually go in there and look at individual frames because they might be too blocky.’" Given SyncCast’s time constraints, the fact that Semaphore could perform the initial QC for 60 minutes of HD content in under 5 minutes was a big plus.
For Xbox, SyncCast encoded with software-only in lieu of hardware acceleration. "At the time, [hardware accelerators] didn’t have the level of tweaks available for VC-1 that we really wanted for the quality," says Ware. "Hardware is awesome for acceleration … but if you want color space calculations, those types of things aren’t always supported by the hardware guys immediately."
Although the Xbox LIVE job was strictly a Windows Media project, SyncCast also encodes in H.264 and other codecs. Technicolor encodes in everything from HD DVD and Blu-ray Disc to MPEG-2 and MPEG-4. As for establishing a standard among today’s codecs, Ware thinks that VC-1 is pulling away from the pack. "I don’t think it’s driven by the codec itself," says Ware. "It’s driven by the fact that Microsoft is really the only one out there with viable DRM that people can license. Adobe has been talking about Flash DRM for quite some time. We’ll see if that comes to fruition, and if the studios and content owners are comfortable with it."
The HD content for Xbox was encoded at 720p, which is recognized as the high bar for online video at this point. And what about the prospect of having to re-encode at even higher resolutions in the future? Because there aren’t going to be better than 1080p sets out for the foreseeable future, adoption of the next generation of codecs may be driven more by cost savings than by improvements in image quality. The question may then be whether content publishers are willing to take the price hit for re-encoding if some better codec comes out that saves them a lot of bits, and which in turn offers big savings in delivery costs.