How much better is 12 GOP?

[Stephen L. Noe]

@S.Mullen

The "mandate" to digital TV. WTTW is one of the TOP PBS stations and a big supporter of JVC technology top to bottom. There content is first class.


P.S. If my opinion doesn't match yours, then it doesn't match yours....

[Keith Winstein]

Originally Posted by Mark Silva

And boy does it show. PBS used to have some of the best looking broadcasts....now they're the worst. Your lucky to not see any compression artifacts even for a second or two.

I wish the stations would not be killing their content with multiple SD embedded streams with their HD streams. 19.2 is barely enough as it is for fast movement at 1080i and then reduce it further by embedding a virtually unwatched weather channel (NBC affiliate in my town does that) but the ABC signal has no other embedded stuff thank god and their programming always looks great.

I look forward to saturday or sunday night movies shown on ABC 720P.

I agree that PBS HD has lots of MPEG artifacts, but I don't think it's so clearly a result of the bitrates. I can only talk about Boston, but here PBS HD sends about 13.7 Mbps of 1080i60 video. ABC, meanwhile, sends about 16.6 Mbps of 720p60 video.

However, unlike the JVC HD100, neither broadcaster uses the "repeat frames" feature of MPEG-2 to allow film-source pictures to be sent only once (and the receiver to repeat them to perform the 3:2 pulldown). That means that when ABC broadcasts a 24 fps movie or a prime-time show, they're still sending 60 progressive frames per second and wasting lots of bits on needless redundancy. ABC's "real" bitrate for film-source material is about 12 Mbps. (Compared with the JVC HD100, which has a real bitrate of more like 19 Mbps since the HD100 uses the "repeat frames.")

Still, I agree that those 12 Mbps from ABC look pretty nice as you pointed out! So I think a lot of it has to do with the sophistication of the encoder.

[Guy Barwood]

I would like to see JVC promoting this new codec chip more and more into the future. We need comparison samples, something visual for the non technical to understand what its all about.

I have been aware for a long time not all MPG2 encoders are created equal but I know for certain others many do not realise this. First generation DV was pretty much all the same camera to camera (purely in encoding quality), one standard VHS recorder resulted in basically the same result as another but this is not true of MPG2 and other MPG variants like MPG4 etc

The general community need to be educated on this. Not in forums like this where the fact people are able to interact in a relatively technical way leads itself to being more technically aware of the products on the market, but most videographers I know hardly know a KB from a MB. Quite rightly so too, they are more interested in the subject than the equipment specs

[Steve Mullen]

I've been thinking more about GOP length and "quality." I think quality is directly related to two very different types of artifacts:

1) MPEG-2 Motion Blocking Artifacts from rapid motion

2) MPEG-2 Mosquitoe Artifacts from fine detail

Therefore:

The image quality (MPEG-2 mosquito artifacts from fine detail) of inter-frame codecs (HDV) cannot be inferred from data rates unless you also take into account GOP length. With Constant Bit Rate (CBR) encoding, increasing GOP length increases image quality because the P- and B-frames increase codec efficiency at a given data rate. MPEG-2 Motion blocking artifacts, however, will be greater because I-frames are less frequent.

Conversely, the shorter the GOP, the fewer the motion blocking artifacts because of the more frequent the I-frames. The shorter the GOP, the less efficient encoding will be at a given data rate and therefore the more fine detail mosquito artifacts.

[Steve Mullen]

Originally Posted by Keith Winstein

can only talk about Boston, but here PBS HD sends about 13.7 Mbps of 1080i60 video. ABC, meanwhile, sends about 16.6 Mbps of 720p60 video.

Curious if that 13.7 is when another 1 or 2 sub-streams are present.

In LV during the day, PBS HD is unwatchable. But, during Prime Time it gets a "bit" better. Yet the second SD channel continues. Which makes me wonder:

1) Do they rebalance the HD/SD data rates in the evening?
2) Do they run 2 sub-streams during the day?

I heard a lot of talk at NAB that even some CBS stations would very much like to broadcast 720p60 at about 14-15Mbps which easily gives them 3-4Mbps for a sub-steam like weather and news which is a second revenue source. Of course CBS, the network hates this idea -- unless it buys CNN in which case the second stream becomes VERY valuable!

Likewise, NBC would benefit from OTA CNBC in the day with MSNBC at night.

Since DirecTV cuts 1080i off at 1280 there would not be much resolution lost.

72060 really makes far more economic sense than 1080i.

[Mark Silva]

Don't anyone get me wrong, I've had signal breakups on ABC as well, but I think that has more to do with where you and your antenna are located in relation to where the broadcast towers are. All I know is, the super bowl on ABC looked spectacular and relatively artifact free which was a first for my eyes.

One other thing on films being shown on national networks I found interesting. I was capturing the beginning segment of Terminator 3 on my home system with this neat little external USB HDTV Tuner I have. I took just a few seconds of it to work and converted it to uncompressed .mov to look at it on our fcp 5 system. I realized there the movie clip said its source was 24fps!


Anyone here look at the quality of D-VHS?

Thats a 14.1 bitstream at 1080i.

How does it hold up by comparison?

[Keith Winstein]

Originally Posted by Thomas Smet

While at first glance a 12 GOP may seem much better the way I understand the way GOP's work is that 12 GOP's isn't really all that much better than 6 GOP's. It is better but not no 2x better. Am I wrong on this one?

If you think about it the only difference between IBBPBBIBBPBB and IBBPBBPBBPBB is that the 12 GOP version has one less I frame and in it's place a P frame instead. That one frame reduction from I to P doesn't seem like it would free up a whole lot of bits.

Something just occurred to me today on this. The HD100 doesn't just use a 6-frame GOP -- it uses a 6-frame *closed* GOP. In temporal order, the frames in the GOP go BBIBBP. But since the GOP is closed, the first two B pictures can only refer to the (future) I frame -- not to any P frame. They're not really bilinearly predicted at all. So the 6-frame closed GOP is really bbIBBP, where "b" stands for a "bastard B" frame that only has one parent (the future I frame), not two parents like a normal B frame. By contrast, a 12-frame closed GOP would be something like bbIBBPBBIBBP, where only 2 out of the 12 frames are "bastard B" frames.

I could be wrong on this because I haven't checked it out in detail, but I believe this is how it works. So it's not the case that a 12-frame GOP just replaces one I frame with a P frame -- it also replaces two bastard B frames with legitimate B frames that are each the children of two parents.

On ATSC HDTV, of course, they use 15-frame GOPs or 30-frame GOPs or whatever, but the broadcasters also never send *closed* GOPs. The broadcasters have the advantage (unlike the HD100) that every B frame they send can be bilinearly predicted by the most-recent-past and most-proximate-future reference (I or P) frames.

[Steve Mullen]

Originally Posted by Keith Winstein

In temporal order, the frames in the GOP go BBIBBP. But since the GOP is closed, the first two B pictures can only refer to the (future) I frame -- not to any P frame. They're not really bilinearly predicted at all. So the 6-frame closed GOP is really bbIBBP, where "b" stands for a "bastard B" frame that only has one parent (the future I frame), not two parents like a normal B frame. By contrast, a 12-frame closed GOP would be something like bbIBBPBBIBBP, where only 2 out of the 12 frames are "bastard B" frames.

It dependes, I think, on what you mean by "temporal order."

The frames are encoded with the I frame first and then the P frame(s) (because they depend on the I frame) and then the B frames (because they depend on the I and P frames).

So the "encode order" is:

I
_ __ __ P __ __

_ B1 B2 _ B3 B4

The Closed GOP means the final two B frames are encoded only from the past frames. So these are the "bastard" frames.

The "transmission order" is based on the order we need them during decoding.

This should be:

I P B1 B2 B3 B4 I P B1 B2 B3 B4

[Thomas Smet]

talk about a dysfunctional family.

[Carl Hicks]

[QUOTE=


Anyone here look at the quality of D-VHS?

Thats a 14.1 bitstream at 1080i.

How does it hold up by comparison?[/QUOTE]

Hi Mark,

The DVHS HD recordings can be at a bit rate of up to 28 MB/Sec, and they look stunning. We use DVHS machines frequently on demos of our DILA HD projectors, on very large screens.

[Keith Winstein]

Originally Posted by Steve Mullen

It dependes, I think, on what you mean by "temporal order."

The frames are encoded with the I frame first and then the P frame(s) (because they depend on the I frame) and then the B frames (because they depend on the I and P frames).

Hi Steve,

I think that's not quite right. The coded order (or "transmission order") of the frames in an HD100 GOP is IBBPBB. If we put in the temporal reference (which shows the display order), it becomes: I2 B0 B1 P5 B3 B4. So the display order (which I referred to earlier as "temporal order") goes B0 B1 I2 B3 B4 P5.

The Closed GOP means the final two B frames are encoded only from the past frames. So these are the "bastard" frames.

Not quite... The closed_gop flag means that the "first consective B-pictures (if any) immediately following the first coded I frame following the group of pictures header ... have been encoded using only backward prediction or intra coding." (MPEG-2 cl. 6.3.8). It's talking about the first two B-pictures in the GOP (both in coded order and display order), not the last two. And it's saying that they're encoded only from the I frame that comes later (in display order).

[Joe Carney]

One of the issues is, we are in sort of an inbetween time, along with the mandate should come a converson to h.264 which in theory, should alleviate many of he bandwidth problems and allow for sub channels without affecting the main channels quality. We're not there yet, probably won't be until the very latest possible moment allowed by the FCC. The Europeans are actually way ahead of us on this. (use of h.264).