Multimedia
Blogs
Forum
Affordable HD
Whitepapers
Newsletters
Events
Advertisers
Blogcast
MillimeterMPEG-2 Encoders, Part 2
Aug 17, 2010 12:00 PM, By Jan Ozer
In the first half of this month's article, I looked at the usability and performance of five MPEG-2 encoding tools: Apple Compressor, Cinema Craft Encoder MP, Innobits BitVice, Sorenson Media Squeeze, and Telestream Episode Encoder. In this issue I'll discuss comparative quality and detail what happened when I tried to import the files into DVD Studio Pro.
A funny thing happened on the way to these quality comparisons. As often is the case, when I test products that I haven't used before, I contact the vendors to make sure that I'm testing them fairly. I contacted Cinema Craft, sent the company my test files, and outlined my test procedures. The company reps said that my test files and procedures wouldn't fairly represent their product (Cinema Craft Encoder MP) as used by their typical high-end users and target customers. Accordingly, I agreed to remove the program from this review. This left four encoding tools, which I evaluated using three test files.
| Related Links | ||||||||
|
The first is the 78 seconds of a PBS broadcast produced by Connie Simmons. As detailed in the first installment, the target data rate was 4Mbps total, with a minimum of 2Mbps and maximum of 7Mbps. For consistency, I encoded with a GOP size of 15, with a closed GOP and two B frames (when this option was configurable). When search or other quality-related options were presented, I always chose those that produced the highest possible quality. I'll call this clip the Connie Simmons test clip. You'll note that while it contains very little high motion, there are multiple finely detailed maps and many artistic dissolves, wipes, and other effects, all of which complicate the encoding task.
The second file was assembled from various clips that I downloaded from Stock Footage for Free, which offers a great range of SD and HD clips at an unbeatable price (free). Because I saw very little difference in encoding quality with the Connie Simmons clip encoded at 4Mbps, I produced these test clips at 3Mbps, with a minimum rate of 1.5Mbps and maximum of 6Mbps. All these clips were 29.97fps interlaced source, so I'll call this file my 29.97 test clip. You'll note that there are several clips in this bunch with extremely high motion, in particular the kayaking clip, which proved a bear to compress.
The third test file is composed of 24fps/25fps progressive source that I downloaded from a number of sites on the Web. I was seeking the highest-quality source clips available; some of these were shot in Red format and others in Sony XDCAM EX. Sites that contributed included www.freehillproductions.com/RED and sony.co.uk, with some amazing clips shot stunt airplane shots by Alister Chapman of Ingenious TV in XDCAM EX format, and the Discrete Cosine blog. I converted all files to 23.96fps and produced them at the same 1.5Mbps/3Mbps/6Mbps encoding parameters used for the 29.97 test clip. (View this test file at Vimeo.)
Let me say that these test data rates are freakishly aggressive and lower than the data rate used on the vast majority of DVDs produced in Hollywood, the event world, and other markets. However, at 7.8Mbps, about the highest data rate you should use for DVDs, all encoders look about the same on most test footage. If you're looking to differentiate between the encoders, you have to encode at very low data rates. Interestingly, a couple of the encoding tools actually performed very well, even at the 3Mpbs target, while others started breaking up badly. If you've seen blockiness, mosquitoes, or other compression artifacts in your DVDs, you're going to want to read on, even if you produce at 7Mbps or higher.
In all cases, I fed the full-resolution clip to the encoding tool, which performed both scaling and encoding. To assess quality, I played the clips side by side on my computer, created a DVD in DVD Studio Pro and reviewed the clips on my 36in. HD TV and grabbed the frames shown below. I'll review my findings clip by clip.
Continue the discussion on “Crosstalk” the Millimeter Forum.