Hauppauge Colossus: HD PVR with HDMI Supportby Andrew Van Til on April 14, 2011 5:40 PM EST
Testing and Evaluation
With an original HD PVR and Ceton InfiniTV 4 Digital Cable Tuner (DCT) on hand, it provided us with the opportunity to record the same content using MediaInfo and perform visual inspections to document differences in quality. For the results above, the Colossus and HD PVR were configured in-line using the YPbPr/TOSLINK pass-through from the STB. They were set to use the highest recording profile in SageTV, with SageDCT used to capture the raw TS stream from the InfiniTV 4 as a reference point. The settings used by the various devices are summarized in the following table.
|Device Recording Profile Summary|
|Video Resolution||Colossus||HD PVR||SageDCT (RAW)|
|480i||3985 Kbps email@example.com AVC (Main@L3.2) (CABAC/2 Ref Frames)||4695 Kbps firstname.lastname@example.org AVC (Main@L3.1) (CABAC/4 Ref Frames)||2948 Kbps email@example.com MPEG2 (Main@Main)|
|720p||7924 Kbps firstname.lastname@example.org AVC (High@L4.0) (CABAC/2 Ref Frames)||11.8 Mbps email@example.com AVC (Main@4.0) (CABAC/4 Ref Frames)||10.5 Mbps firstname.lastname@example.org MPEG2 (Main@High)|
|1080i||12.7Mbps email@example.com AVC (High@L4.0) (CABAC/2 Ref Frames)||11.7Mbps firstname.lastname@example.org AVC (High@L4.0) (CABAC/4 Ref Frames)||16.5 Mbps email@example.com MPEG2 (Main@High)|
There are differences in the bitrate and AVC profiles used by the two Hauppauge devices; the Colossus uses a higher rate for 1080i and the HD PVR uses higher bitrates for 720p and 480i. The InfiniTV 4 was a third set of values. With three points of reference and digital source content, there is little noticeable difference between the captured files in real terms. (If you’re interested in seeing clips of the videos, let us know and we’ll see about including download links.
Alongside the comparative testing, the Colossus was also evaluated for the overall quality and “watchability” of recordings captured at native resolution from the STB. We recorded sixty files with a 33% distribution per resolution at random intervals and channels, and then sampled at five to ten points in each file for issues related to audio and video. We looked for A/V synchronization problems, stutter, and dropouts, and the final result is an 8.33% fail rate. Most of the “fail” issues only occurred in specific locations in the files, and while annoying they were not severe enough to abandon playback. However, there were also entire recordings with no audio or significant enough issues with audio drops or video playback that continued viewing was very unpleasant or impossible.
Before moving on it is interesting to note that the maximum bitrate available in ShowBiz and via the encoder’s property page (shown below) is quite a bit higher (~20Mbps) than what is used by SageTV when selecting the best rate (~14 Mbps). It is unlikely that the value of encoding at this high a level would result in a great enough pay off to justify the increased file size given the bitrate of most TV content, but it is good to know that it is possible if the content warrants it.
I love it when my interests in the HTPC space overlap, so having observed the difference in bitrate capabilities and issues with file capture it provided an easy excuse to start writing code against the Colossus and figure out how to drive the device. For anyone who has done any playing with DirectShow, the following screenshot of the capture graph in GraphStudio should be very familiar. It demonstrates the four components necessary to enable file capture from the device with the property pages for the “Capture” and “Encoder” filters (which are what ShowBiz exposes through its GUI) displayed. The file sink—the component at the end that writes the TS packets to disk as they are output from the encoder (in this case a slightly customized version of the “Dump” sample included in the Windows SDK)—is the only filter not provided by the drivers.
Using that as a starting point I wrote a simple application to prototype input selection (it supports component + S/PDIF or HDMI) and filter configuration. Using standard DirectShow interfaces, I was able to reproduce both the capture issues and the inconsistent bitrate values observed with ShowBiz and SageTV, where after setting the device to the maximum bitrate (~20Mbps) through the encoder interface it actually uses ~14Mbps. I have asked Hauppauge to provide documentation for programming against the device in case there is a different, non-obvious, mechanism for configuring the card, but they did not provide it in time for this review. For anyone interested in looking at the source code for the application it is available here.
Anyone who made the jump to the original HD PVR when it was released will remember that the commercial scanning story was a bit troubled. Fortunately, the work that was done to support the AVC TS files created with it carries over to those generated using the Colossus. Applications like ShowAnalyzer are able to scan the files and detect commercials in them.