Can someone explain the difference between and DVD and Program Streams?

[Bilar Crais]

I've looked for more information on this but can't seem to find any. Is one closed GOP and the other open? Which is open? TmpgEnc DVD Author seems equally happy with both, but I've read that open GOP is supposed to be better quality than closed GOP.

[btvbuddy]

Hi, I am not certain on your question. A DVD file has a .VOB extension. It is a mutiplexed MPEG-2 file (audio/video) combined. A Program stream, meaning a MPEG-2 Program Stream is also a multiplexed a/v file. So they are very similar, you can rename a .vob to .mpg and it will work unless its a encrypted VOB - you may also have issues with sound because many VOB's have ac3 audio and MPEG2 program streams typically do not. On my system I used closed GOP's because they are better for video editing, I think that having open-GOP's would be an issue with having frame level accuracy or with seeking to a position within your MPG like with WMP9.

Sorry I did not answer your question but felt some info would be helpful - if I am wrong on any this guys - point it out.

[Bilar Crais]

It was a valiant effort! I tried a sample recording of each last night, and they both have exactly the same IBP structure (open).

[btvbuddy]

Oh do you mean the recording profiles in BTV Qualities? That may make sense because if you build your own quality you can not specifiy have closed-GOPs. I build closed GOP's using a hardware MPEG-2 encoder - checkout www.canopus.com for more info.

[Zogg]

In the 'First DVD with Nero 16 hours' thread, SHS wrote:

DVD Stream is for non editing
Program Stream is for editing

Z

[btvbuddy]

You mentioned the structures are the same,

just as backgroud there are three GOP structure types that I can build

IBP
IBBP
IBBBP

In a I frame, it's a full frame.

<snagged from google - thanks Mr. John McGowan>
In a B frame, Bi-directionally encoded macroblocks contain a backward
and a forward motion vector encoded using variable length codes in the
macroblock header. Each macroblock has a header which amongst other
things identifies the macroblock as intra-coded, predictive coded, or
bidirectionally predictive coded. The bidirectional macroblocks are only
allowed and used in B frames.

In a P frame, it's a Predicted frame and decoded from the previous I or P frame.

So, when you said it's the same did you mean both have only 1 bi-directional frame in the GOP?

I think there is a net tool to analzye mpeg GOP's is that what you used?

[Bilar Crais]

I switched between the two streams, keeping my profile otherwise the same

(4500-9500vbr, 5 p-frames, 2 b-frames) and ran it through VideoReDo, then advanced frame by frame, and they are both IBBPBBPBBPBBPBBP.

I used to have an analyzer; don't remember what it was.

My main point was that I wanted open frame GOP for better quality. VideoReDo is a frame accurate editor, so having closed GOP isn't really important to me.
As it turns out, it appears to be open.

I would imagine the DVD stream has a different packet setting than program stream, though I can't say for sure.

[btvbuddy]

The current quality you have is

(4500-9500vbr, 5 p-frames, 2 b-frames) and ran it through VideoReDo, then advanced frame by frame, and they are both IBBPBBPBBPBBPBBP.

So, 5 frames from prediction plus two b-frames with open GOP.

I thought that quality comes from more I-frames (full frames) the B/P frames are based on motion alogorithms and macroblocks (with some extra space-based calculus deriv's throw-in, grin..)

The more I frames is the key I thought, so if you do closed GOP with IBPIBPIBP would'nt that be higher quality than open GOP with IBBPBBPBBPBBPBBP? Can you decrease or P/B frames and run your test again?

Maybe I got this wrong, anyone has a PHD in MPEG-2 encoding?

It's pretty interesting stuff if you are trying to squeeze out some high quality DVD's. Of course higher quality usually means longer encode times.

[Bilar Crais]

I used to have a link to some definitive, scientific information on such matters, but I can no longer find the Web site. I did a cursory Google search, and it seems that almost all the information I found gives open GOP a slight edge for both quality and compression efficiency. The way I understand it, closed GOPs can not reference the previous GOP for predictive information, whereas open GOPs can.

[Rich A]

There is a difference between DVD stream and Program streams. Mostly muxing rate and other "intrinsic" parameters you can't control.

I believe SS uses "standard" mpeg program stream multiplexing for what it calls it's "Program Stream", and it's DVD Stream is using slightly diffferent intrinsic parameters which would make the final mpeg more compatible with some of the more critical DVD authoring programs. Program streams can be Mpeg-1 or Mpeg-2 and VCD or SVCD or DVD.

The "user" settings available to you have nothing to do with the internal settings used by encoders and multiplexers to create "book" compliant mpegs. The difference is regarding "how" the stream is multiplexed during encoding. And that "How" is changed when you select to encode a VCD, SVCD or DVD mpeg stream. (or multiplex their associated elementary streams)

If you have trouble getting your encoded BTV Mpeg to be accepted by certain DVD authoring programs, you might try the DVD stream. Although any "decent" DVD authoring program should be able to de-multiplex any mpeg-2 that conforms to the DVD Spec, and then RE-Multiplex it back into a DVD proper VOB file, there are some that are very picky about those intrinsic parameters.

You have two sets of parameters that are needed for DVD compliancy. The first set is comprised of USER available parameters. Things like Bit Rates, GOP structure, Frame Size and the like. These of course, must be within certain constraints to make an Mpeg Program Stream user's settings DVD compatible.

The second set of parameters are used internally by the encoding engine. Normally you don't have any control over these. They are such intrinsic things such as the Multiplexing rate during encoding, pac lens and other things.

Actually a DVD Stream IS a program stream. You have Program Streams, Elementary Streams and Transport Streams. A DVD Stream is a program stream that has been encoded using the encoder's internal set of intrinsic parameters that are complient with the DVD book. Or it can also be a program stream that was created by multipliexing two elementary streams (again using the proper internal parameters)

When you are capturing with the final output as an mpeg file, you are not only "encoding" but also doing a real time multiplexng.

Most encoders don't allow users to mess with those internal controls. Rather they just give you profiles called VCD, SVCD, or DVD or a "raw" mpeg. Each one has it's own interenal settings. Just don't forget that BOTH the user available settings and the encoder's internal settings must be compliant to end up with a good VCD, SVCD, or DVD video.

[Bilar Crais]

Originally Posted by Rich A

There is a difference between DVD stream and Program streams. Mostly muxing rate and other "intrinsic" parameters you can't control.

I was kind of hoping you'd chime in, Rich. I knew if anyone could answer these questions, it would be you.

[btvbuddy]

That's informative, so if given the choice when creating a DVD (using TMPEnc or other tool) we should select Open-GOP with IBP or IBBP or IBBBP ? Also, it's possible to create one thats Open-GOP and I frame only is that allowed in a 'book-spec' DVD. Any direction on which would be the best quality?

Thanks

[Rich A]

Originally Posted by btvbuddy

That's informative, so if given the choice when creating a DVD (using TMPEnc or other tool) we should select Open-GOP with IBP or IBBP or IBBBP ? Also, it's possible to create one thats Open-GOP and I frame only is that allowed in a 'book-spec' DVD. Any direction on which would be the best quality?

Thanks

I doubt you should use an I-frame only Mpeg. That would be a huge file. I-frame encoding is "sorta" like just taking each frame and compressing them each, (like a jpeg image) What makes mpeg so effecient for smaller file sizes are the B and P frames within a GOP (group of pictures). That's where the huge savings in file space occur. The "intra" or I-frame has the most information and has no reference to any other frame. It's a full frame .. just like a still photo. So using only them results in much larger files.

The I-only frame option is available for some encoders (mostly software types) where the user's CPU might not be powerfull enough to completely compile the P and B frames for a normal GOP sequence. (it's a lot of encoding) By switching to I-Frame only encoding, the CPU has much less work to do as it's only actually compressing each frame and doing nothing else. I've never tried an I-frame only mpeg for sourcing DVD. Don't think it would work.

For the GOP structure there is a correlation to the I, B and P frames in regard to the quality and amount of compression. I forget what the optimum was but back in my ATI days when using the All In Wonder software mpeg encoder I found subtle differences when changing the GOP structure. In case you are interested, here's a detailed summary of what I use to author my DVDs (with the Haupauge PVR-250)

Stream type: MPEG-2 VBR
Resolution: 720*480
Aspect ratio: 4:3 Generic
Framerate: 29.97
Nom. bitrate: 8500000 Bit/Sec
Peak bitrate: 8368000 Bit/Sec
Avg. bitrate: 5155000 Bit/Sec
VBV buffer size: 112
Constrained param. flag: No
Chroma format: 4:2:0
DCT precision: 8
Pic. structure: Frame
Field topfirst: Yes
DCT type: Field
Quantscale: Linear
Scan type: ZigZag
Frame type: Interlaced
Scene change detection: NOT FOUND
Variable GOP pattern: NOT FOUND
Notes: GOP is 18 pictures. FIRST GOP CLOSED REST OPEN
GOP STRUCTURE IBBPBBPBBPBBPBBPBB

To get that, I have BTV's parameters set for a base Bit rate of 5500 kbits, and a peak of 8500 kbits. Using DVD Stream. Bi-frames set to 2 and P-frames set to 5.

That is what works for me. My files tend to be a little on the large size due to the higher bit rates, but I'm watching these home made DVDs on an 84 inch projection screen so quality is most important.

[Bilar Crais]

I've tried I-Frame only recording with my ATI setup before...not pretty. Even at high bitrates.