ISDCF conducted its 4th Plugfest in January to test the functionality of subtitles and accessibility with SMPTE DCP-compliant servers. The previous Plugfest was in November, and was “virtual.” Content was distributed, and vendors reported their results. The January Plugfest was physical. As with other Plugfests and digital-cinema demonstrations, it was held at the Technicolor ITC facilities. ITC stands for Interoperability Test Center, a unique facility having a lab-sized projection booth, complete with a dozen racks of equipment and projecting into two screening rooms.
The July physical Plugfest led to consternation. Nothing seemed to work correctly, but there was also no reference to judge by. The content was under test, server behavior was under test, TI’s Series 1 subtitle generation was under test, and the interfaces to closed caption devices were under test. There were bound to be problems, and it was impossible to tell where the problems came from. While fingers pointed to bad test content, further investigation in the following weeks found nothing to substantiate this. The November “virtual” Plugfest was a chance for the server companies to vouch that their servers could play the content, or to express where improvements in their products were needed. This was a major step forward.
This month’s Plugfest was conducted with the goal of establishing a reference system where everything worked. These tests are an opportunity for mastering houses to test their SMPTE DCP packaging software, which has led to a growing number of test materials. Of particular note has been the material developed by Fox: a “killer reel” that tests every feature possible in the realm of timed text tracks. The first step was to remove any question as to the quality of the test materials. But that required a server that works. Bravely, Doremi and Sony volunteered their servers for this test, and Doremi scored.
With open subtitles and captions, the Doremi server was the only server that could reasonably play the Fox killer reel, failing only one test in which the rendering engine is to fade the timed text, both in an out. Doremi noted that the fade feature had not been implemented in its rendering engine – but the text still appeared on screen, minus the fades. Stunningly, no other server in the two days of setup and testing came close to the performance of the Doremi server. The errors in other products appeared as issues in font sizing, lack of displacement when a background and foreground were both present, actual position on the screen, and correct display of character sets. Even the TI subtitle/caption implementation for its Series 1 projectors presented problems.
On the closed caption front, both Doremi, USL, and WGBH closed caption devices were tested side-by-side. Doremi’s Captiview system only works (today) with the Doremi server. However, the demonstration showed that all captions, both open and closed, regardless of display device, were displayed in time with each other. For closed caption display devices that are driven directly from the server, synchronization is expected. But the effectiveness of CSP/RPL to synchronize external closed caption playout was proven.
The performance of the USL closed caption system was put to test. USL’s device is driven by CSP/RPL, and consistently was late in initiating the display of closed captions when a movie was started at the server. The problem is that, with CSP/RPL, the closed caption system must pull the closed caption file out of the digital cinema server in order to parse and display it. Two problems can arise: the closed caption system may not have the horsepower to quickly load, parse and display the closed caption text, and the server may be late in signaling the availability of the closed caption text to the closed caption system, aggravating the issue. In practice, when a show playlist is governing a show, the closed caption system should have adequate time to load the text, and caption display should be seamless. But the Plugfest tests do not employ show playlists, and testers expect immediacy when “start” is pressed, representing typical operation in most screening rooms. USL may need to revisit its choice of CPU in its closed caption system.
The good news from this month’s Plugfest is that the industry found its reference device in Doremi. Competition, and the desire to not be shamed in another Plugfest, should lead to significant improvements in the other products. Clearly, more Plugfests will be needed. The next one is tentatively scheduled for early April.