Field Picture Structure

[mp3dom]

Hi to all!
I have a question regarding encoding in field picture structure; in particular its compatibility with standalone dvd players and software dvd player. Across the net I have red that the field structure is not supported in the dvd specs (the $5000 book) but I have for sure commercial titles that are encoded with field structure (also almost all Dolby/dts Trailers are in field structure). Also If I import those field struct in Scenarist, in its preview I see a black field followed by a green field... but it's ok on standalone.
So I'm assuming that at the end the field struct is Ok (excluding the problems with Scenarist). Now the question: Canopus support both field and frame structure but also the Auto which is a mixed combination that change on the fly based on a picture-per-picture analysis and offer the best of the 2 method... but how about compatibility? The same as frame or field? I have noted that with Bitrate Viewer I have wrong bitrate detection with mixed structure... but this doesn't happend with frame or full field structure. MPEG specs says that interlaced can be encoded with variable structure (the same as Canopus do) but what about dvd specs? Thanks!

[Mug Funky]

does scenarist compile the project okay? it's possible it's internal decoder can't handle it, but it's compiler can.

generally if scenarist imports it and compiles it, you're in the clear. however, you can't guarantee playback in all standalones even if it IS compliant (i don't have the book, so wouldn't know).

you should probably suss out if it's worth the extra efficiency... at sane DVD bitrates interlaced footage comes out fine with frame pictures w/ alternate scan, interlaced DCT and ME.

[mp3dom]

Thanks for the reply.
Yes, Scenarist accept the file and compile it. The only problem is in the preview.
I have a movie with high bitrate (5000-7800-8800) but some complex scenes comes with artifacts... problem that with mixed structure go away completely...
I have tested the compiled dvd on a Panasonic DVD player and it plays ok but it's a player that read everything (seamless layer break on dvd+r dl, dvd+r dl etc) so it's not good for testing specs :)
Thanks!

[Mug Funky]

try it on an old Pioneer. but not so old that it catches fire like one did here the other day...

if possible, it's good to build a collection of players that exhibit problems. then any disc you want to distribute can "run the gauntlet" on those.

[foxyshadis]

Now that's an exciting day at the office. :p

[mp3dom]

Thanks! ^_^
Just another question:
I've seen that I could heavily improve the PQ in the complex parts if I totally discard B-frames (GOP of 12 frames, IPPP..P/IP.. structure).
The question is:
Since I have a very good quality in complex parts (that are 4-5 seconds in length each part) is it better to encode only those difficult parts without B-frames or could I encode the entire movie without B-frames? I think that if it's good for long complex parts, then it's almost good for normal parts...or am I wrong?
Thanks!

[Mug Funky]

the deal with b-frames is they're more efficient, but p-frames aren't allowed to predict off them.

this means the more b-frames, the further apart the p-frames are, which means the less helpful the motion compensation will be. and seeing as b-frames are predicted off the (increasingly less accurate) p-frames, they'll end up either taking up way more space or looking worse (or both).

so your first assumption is true - just encode the wacky parts with no b-frames, and leave the rest as-is.

if your authoring app allows segment-based re-encoding (or at least shoving a new part between 2 i-frames), this can save a lot of time. maestro does this, but i suspect it's the only one that does (not sure about scenarist).

HC's autoGOP should be able to do what you want automatically - make sure you get the latest (0.19 has tweaks to the autogopping).

[mp3dom]

Thanks Mug Funky!

[kolak]

GOPs without B frames are not fully compliant with some players:)

[Mug Funky]

seriously? any idea which players?

because that's something i'm likely to do in case of a difficult segment - replace it with either 1 b-frame or no b-frames.

[Trixter]

At sane DVD bitrates, interlaced footage comes out fine with frame pictures w/ alternate scan, interlaced DCT and ME.

What's "ME"?

[Mug Funky]

a philosophical question :)

it's Motion Estimation.

[Trixter]

Thanks -- I thought that's what you meant, although I got confused because motion estimation didn't fit with the rest of your list (alternate scan and interlaced DCTs are both support features specifically for interlaced encoding).

[Mug Funky]

i think it's a matter of the motion estimation working on odd and even lines. so you're not trying to match interlaced blocks against each other (which would make encoding extremely inefficient).

[kolak]

Some cheep ones: Medion, Grundig.
Almost all professional DVDs are encoded with Frame mode and GOPs which contains B frames. It's fine if GOP is only I frame, but not good when there are only I and P frames.

[Mug Funky]

Almost all professional DVDs are encoded with Frame mode and GOPs which contains B frames. It's fine if GOP is only I frame, but not good when there are only I and P frames.

lol. the main difference between "professional" DVD encoding and what's done on this forum is higher quality masters and a pay packet :)

i can't see why players would have a problem with no-bframe encodes (certainly the MPX3000 allows these GOPs even when "DVD compliant" is turned on, and that's pro MPEG-2 hardware), but it's certainly not something i would do on a title that's getting a replication run of 20,000+:D

[kolak]

I also don't know why they have a problems with GOPs without b frames, but this is a fact and quite often a problem.
If you check manual for CCE-Xtream you'll find special option which forces GOPs to contain b frames and this feature came from real world(not from DVD spec).

There are lots of professional encoders which produce not fully (in real world) compliant mpegs. There is a big different between DVD-Video standard on the paper and in the real world.
The more DVDs you've done the more problems you know and than it's easier for you to avoid them.

[Mug Funky]

for sure, so long as you get feedback on them - returns, or even forum posts (check out the madboards some day :))

just that i don't think i've ever put no-b-frame stuff on a DVD. in the most extreme cases 1 b-frame and aggressive scenechange detection was enough to get the quality good again. typically we wont know a quality issue until the encode is done and has been watched by someone. then all that remains is fixing up very small segments and dropping them into the timeline. compliant or not, no-bframe for an entire feature is not at all elegant...

so i don't really know - i just find it hard to believe unless there's bitrate or muxing issues (ie ratecontrol might be thrown off by not having b-frames, and it might not get caught during muxing for similar reasons. sounds like a bit of a stretch to me though). looks like i wont be doing it now at all :)

FWIW i find most playback issues arise because of poor replication. some pressed discs will give worse error scans than CMC-magnetics DVD+R's burnt on average drives, i kid you not. of course, try explaining to a replicator that it's their problem not yours... usually they only believe you when they find dents on their glass master :mad:

of course, it's good to know every little issue that comes up, so thanks :) i'll have to track down a medion or grundig and add it to the pile.