Comparing Mainconcept to Expensive, Production Company Encoders

[unclescoob]

All this talk about encoding quality and speed sure makes me hungry.

Well, for more encoding information, that is.

Ok folks so just thought I'd bring this up for a nice, friendly and informative discussion (this is coming from a n00b, of course so please have mercy).

We talk about Procoder, Mainconcept, Tmpeg, and others I can't remember now. But those seem to be the most mentioned. Now, if by chance a similiar discussion is already in existence here, I do apologize. However, I do roam around here enough and haven't seen one so, here's where I'm getting at:

Mainconcept, for example is respected as one of the top dogs in the encoding world. Now, with a price tag of $500.00 (seen it for $450.00 recently somewhere else), what are some features that differentiate Mainconcept from, say a $25,000.00 encoder used by professional movie studios?

[unclescoob]

Apologies if the question is too farfetched. Anyone know of at least TWO examples?

All just for fun and learning folks. Here's hoping this turns out to be a nice discussion and not another one of those posts where 78,000 view it and 0 answer. Smurf, of course I can't wait to read your input. As well as the other "technical" guy. But it's nice to hear from everyone else too.

[kpmedia]

There are quite a few ways in which an encoder like MainConcept Reference differs from hardware used in broadcasting, post houses or authoring facilities. ("Studios" don't generally have any encoding facilities whatsoever.)

Note: I don't recall all the official names of these functions off-hand, so I'm making up the names as I go. I don't have the time right now to pull out reference materials. I would suggest my names are probably close in some way, as I'm just using general descriptions.

1. Multiple concurrent outputs. Hardware appliances are capable of encoding HD and SD streams in tandem. MPEG-2 and/or H.264. Broadcast, mobile, web, etc. It can push out multiple high-quality streams in a single pass.

2. Bandwidth load balancing. Bandwidth is fixed at a maximum rate on a transponder or broadcast frequency, so if you're attempting to push several channels through the same bandwidth space, you'll ideally want to encode a "group" style VBR. CBR results in lower picture quality, so you'll find a "bobblehead" news channel paired with a racing channel, with both encoded VBR. The one channel needs less bitrate, while the other needs more to maintain the same quality. But it's not just two channels, it's often 3-4 on terrestrial broadcast, or a dozen (or two) over satellite.

3. Advanced GOP enhancements. If you fully understand the concept of what a GOP is, and how it works, you'll know that you can gain quality by customizing the GOP on a scene by scene basis. Most standard MPEG-2 encoding software maintains a constant GOP length and I-P-B ratio across the length of the video. An advanced encoder allows for certain segments to have I-P only, for example, while still encoding in a single pass. (Generally speaking, this is when an encoder is part of a more complex editing system, such as Scenarist hardware.)

4. Detection algorithms more advanced. The benefit of dedicated hardware is that it only has to do one task. As such, it can do it fast. There's no layers of OS, layers of drivers, etc., like a computer has. While software can be quite advanced, it's easier to perform advanced/complex operations in dedicated hardware. Software tends to balance speed/performance with quality, while hardware can push for quality and still maintain speed (balance not required).

Parting thoughts...

I don't use any of this, because it's simply not needed for what we do. A well-tuned copy of MainConcept Reference, Sorenson Squeeze, or Adobe Premiere Pro (also MainConcept based) is more than adequate for the work needed by consumers, professional organizations and small studios. Video interests me, so of course I do pay attention to what's available.

[unclescoob]

Quote:

Originally Posted by kpmedia

An advanced encoder allows for certain segments to have I-P only.

There's a tutorial on the Tmpeg encoder, for example, which states that if you're going above 6000kbps, you should not be encoding B frames anyway, as P frames, which take redundant information from other P frames, along with the I frames are enough for a good picture. Would this be considered a general rule with utilizing any encoder?

[lordsmurf]

Quote:

Originally Posted by unclescoob

There's a tutorial on the Tmpeg encoder

Where? I've seen a lot of bad guides for TMPGEnc Plus. Some of them were very clearly written by people who were basing their advice on their own limited understanding of video. Several guides popular amongst anime fans were especially horrible, doing quite a bit of picture damage to the video. Novices have a bad habit of cranking filters up well beyond suggested values.

Quote:

which states that if you're going above 6000kbps, you should not be encoding B frames anyway, as P frames, which take redundant information from other P frames, along with the I frames are enough for a good picture. Would this be considered a general rule with utilizing any encoder?

Assuming 352x480, maybe.
Assuming 720x480, absolutely not.

Keep in mind that 352x uses half the bitrate as needed by 720x.
And my comments below will all reflect a 720x480 (or 720x576 PAL) reference point.

You come across a few plateaus of quality. For a standard DVD-Video, with typical GOP lengths, you approach a leveling of possible quality at about 8Mb/s. Depending on the complexity of content (including the % of noise in-image), you may prefer 12, 15, 20 or 25Mb/s instead. Of course, none of those are legal DVD-Video values, and are considered lower end broadcasting or typical mezzanine specs. As you start to raise bitrates, you allow for the removal of in-between frames. B frames go first, then P frames, leaving you with I-frame only (intra-frame encoding). At this level, MPEG-2 is almost identical to MJPEG or JPEG2000, which are considered only mildly lossy intermediate codecs, as there's no frame-to-frame loss. An IP-based GOP, at MPEG-2 25Mb/s, makes for a great mezzanine spec.

Broadcast specs go all the way to 50Mb/s, and that can include use of long GOPs (IPB with a ton of P and B frames). It really comes back to content, usage, and adherence to agreed-upon specs.

Another consider is transcoding needs. That's where you round off encoding math, and toss out bitrate, in order to reduce file size. This is done for broadcasting, and then many people are familiar with the DVD Shrink program, which transcodes DVDs. You can easily transcode IPB, somewhat transcode IP (but not really), and forget about trying to transcode I-frame only.

Asking "Should I remove B frames?" is like asking "How I drive 50mph?"

If you're in a school zone, no.
If you're on a race track, no.
If you're on a small highway, certainly. (Speed limit is likely 50-60mph.)
If you're in a go cart, no.
If you're in a car or truck, sure.

It's very situation based. The answer can, and does, change to meet the needs of the encoding scenario.

[unclescoob]

Quote:

Originally Posted by lordsmurf

you may prefer 12, 15, 20 or 25Mb/s instead. Of course, none of those are legal DVD-Video values

And that's what sucks.