TMPGEnc GOP structure, bit rate, source range filter Qs

[Ronaldus]

As suggested by others on this site, I have been using the TMPGEnc settings from http://dvd-hq.info/Compression.html. If I have my terms correct, this specifies a GOP structure of I=1, P=4, B=1 -- which is IBPBPBPBPB.

I don't know enough (nor am I sure I want to get that deep) to know the quality implications of that GOP structure or which settings to choose. I tried 2 settings of the "Quantization Matrix" (Default and CG/Animation), and I could not notice any difference on an animated segment.

In general I find the encoding quality is good enough (but not fantastic). However, at the very start when the screen fades from black to a bunch of white clouds against a light blue background, I always see some initial blockiness -- more at this point than at any other scene change throughout the segment. I tried both a CBR at 7000 Kbps and 2-pass VBR averaging the same with 8000 max and 6000 min. Again, I could not tell any difference. In fact, I am wondering why I should bother with 2-pass VBR which takes me a 13.5 encoding time ratio instead of CBR which is much faster?

Can people please help me with the following:

#1. Is IBPBPBPBPB the best "all around" GOP structure to use along with the "Default" Quantization Matrix for all types of video? (for those who don't know enough and don't have the time to conducts hours of tests on every video segment to find out which setting is optimal for encoding)

I really don't want to have to struggle and think about these issues if I'm encoding animation or home videos or something off TV. Are there any negatives to always using IBPBPBPBPB? I would love to have a few guidelines from people who have more experience. I have not tried tweaking GOP structure because I'm clueless what to choose. You can't tweak something you don't understand, and there are some things in life I don't want to understand. How many people who use cell phones want to learn how it all works underneath? Probably not many. Yet it seems like if your goal is the best possible video quality output with TMPGEnc, you have to get a degree in MPEG-2 encoding details. I just want to put the input in, let it take however long it needs, and have it give me the best quality output it can. Am I expecting too much? Maybe video tools haven't progressed far enough to figure it all out for me.

I tried different encoding quality settings from the Highest, High, and whatever was next. As before I could barely tell any difference. The only reason I use 2-pass VBR at the Highest quality is I want the best quality and when I can't see any difference I figure I may be missing something and it's better to be safe than sorry.

#2. Are there any clear-cut cases where CBR is always better than 2-pass VBR? I would guess it depends on how much "fast motion" there is in the video, but I'm not a good judge of that. It would be nice if others have found that if the bit rate is for instance 6500 or greater that CBR is as good as 2-pass VBR most of the time. Maybe someone could point me to a well-known "fast motion" video scene that I could capture and do my tests on because when I've encoded fast pans in both CBR and 2-pass VBR, I see almost the same amount of slight blockiness either way. Maybe my eyes are bad.

#3. TMPGEnc seems like a great tool, but so far my impression is all these little tweak settings don't make much difference to the final output unless you are a video encoding expert. Are there any additional guides besides the one I referenced above on how to tweak TMPGEnc without spending lots of time analyzing your video first?

#4. On a tangent related to TMPGEnc, has anyone experience audio distortion on some players if you allow TMPGEnc to encode the audio to MP2? On a Pioneer player it sounds fine, but on a newer Panasonic it almost sounds distorted/clipped at loud spots as if some normalizing were occurring. However, I could not find any normalize audio option in TMPGEnc.

#5. The compression settings guide I mentioned above says to set:
"No half pixel motion" to disabled.

TMPGEnc has the following check box:
"No motion search for still picture part by half pixel"

I'm a little confused here on double-negatives. If no half pixel motion is disabled does that mean I want it enabled and therefore the TMPGEnc check box should be checked? Basically, could someone please tell me if I want this check box checked or not to match the guide (or if the guide should be ignored here)?

#6. Finally, I use Pinnacle Studio 8 to capture to AVI and edit out unwanted segments which I resave to AVI using the same DV Encoder Codec (I believe). TMPGEnc has a "Source Range" filter under "Settings -> Advanced" to remove an unwanted segment from the source AVI. Is there any advantage to using the Source Range filter instead of editing in Pinnacle Studio and resaving as an AVI? My instinct is there is no advantage because Pinnacle does not appear to be re-encoding, but just because I don't notice any quality difference doesn't mean there isn't a subtle loss due to re-encoding going on. Theoretically, which of these options is the best way to edit? Don Pedro had suggested performing some tests on editing and encoding to try to PROVE that Pinnacle was simply copying using the same Codec rather than re-encoding, but I couldn't figure out how to prove it. (I guess I'm not scientific enough.) Has anyone been able to prove this or know of a way to verify when Pinnacle is re-encoding? (Unfortunately Pinnacle says Rendering both when copying and encoding.)

Thanks.

[petar]

You are using a high enough bitrate so any other setting will make a relatively small (if any) difference to the quality. Bitrate is by far the most important thing when quality is concerned. At the bitrate you're using I wouldn't bother with VBR either.

1. The "default" GOP structure should on average be most suitable. Reducing the number of P/B frames per GOP won't necessarily improve quality. You only have so much bitrate to play with. It depends on the video though but then you asked for the "all around" best GOP. Just my opinion.

2. Fast motion normally refers to action scenes in movies. On a home made video there's very little fast motion in that sense. I've converted many home made videos (from DV and VHS) and came to the conclusion that what's refered to "fast motion" is very rarely applicable to what one records with an amateur camera (you tend to move the camera very slowly, don't you, and you don't normally record car chases or people fighting?).

By the way, how did you do the test? If you take a 10 sec clip, you won't see any difference whatsoever between CBR and VBR. You have to measure it on a whole movie. It's of no use to set a VBR of, say, 5000 with 8000 max on a 10 sec clip, especially if the whole clip is a fast motion scene. The encoder will still use 5000 all through the clip. The idea of VBR is to use 8000 on some parts, then 2000 somewhere else, so on average it's 5000 and your MPEG stream is roughly the size you want it to be.

3. See what I mentioned above. With a high enough bitrate it's difficult to notice differences.

4. Never used TMPGEnc for audio encoding. Seems to be a general recommendation everywhere I look. Extract it to WAV and encode it with something else.

5. I'm no expert on this. I just read, same as you, that should be left unchecked for relatively high bitrate.

6. With TMPGEnc you don't waste time saving the edited AVI and you don't need any HD space for it. DV compresses each frame individually so it doesn't have the MPEG problem that you can't cut on each frame. I can't think of any reason why you'd use Pinnacle and not TMPGenc to cut.

So in conclusion, I'd say if you are not trying to squeeze too much video on a DVD and use a relatively high bitrate, you shouldn't be too concerned about tweaking (since you say that you don't want to be bothered with it) especially if you watch your DVDs on a TV set where you'll see far less difference than on your PC monitor.

[bugster]

You keep saying you want the best possible quality bon't have the time or the inclination to test all the differnt tweaks. Well I have bad news for you. The only way to get 'the best possible quality' is to tweak TmpGenc yourself. You can get ideas about what to tweak and why from guides but only you can decide what looks good to you.

If you want advice, heres what I do.

Load the TmpGenc DVD template. Use motion search precision normal, use either CQ mode quality 80 or 2-pass VBR and Avg bitrate set suitable to fit the movie on the DVD. CQ is (relativley) quick and quality excellent but you can't predict the final filesize so use 2-pass VBR when this is an issue but time isn't.

If you don't understand GOP settings, leave them alone. Same with quantize matrix. The TmpGenc defaults are fine for 99% of cases and 99% of people.

Hope this helps.

[Ronaldus]

petar and bugster, thanks for the replies.

As petar said, bit rate is the only thing I can really notice a difference in quality of all the tweaks I've tried, so I'm under the impression the difference most other tweaks make is very small. I was asking about the best general purpose GOP structure to use because I haven't dared mess with that because I don't know what it means. I don't think IBPBPBPBPB is the default TMPGEnc structure, but rather the settings the compression guide from dvd-hq.info recommends.

Should I always use TMPGEnc's default GOP structure (I don't know what it is -- and I don't have TMPGEnc installed on this computer) or should I always use IBPBPBPBPB from the compression guide? Which is better for all-purpose use?

Even though I used the highest quality and 2-pass VBR at roughly 7000 Kbps on my test segment, as I said before I still see blockiness on fade-ins from black scenes. Unfortunately it does not appear (if I can believe my old Pioneer player's bit rate meter) to be using a higher bit rate at these transitions when it could/should.

It does sound like I should leave things at their default because I don't know any better and I don't have the time to tweak and test every video. I just want to know if anyone had some other "all purpose" settings for TMPGEnc or some links to some other tweaking guides.

Finally, I found out my TMPGEnc MP2 audio distortion problem may not be TMPGEnc's fault. I encoded the same audio to AC3 outside of TMPGEnc and had the same audio distortion problem on the Panasonic player, but it sounded perfect on my old Pioneer. When I tried ffmpeggui, the AC3 sounded fine on my Panasonic, but locked up my old Pioneer. I fiddled with the Panasonic's audio processing settings, but it stills distorts a little. It's strange how all pre-recorded discs are distortion-free, but the ones I make seem to have minor problems.

[petar]

Default GOP is IBBPBBPBBPBBPBBPBB. As I mentioned before, if your test clip is very short you won't notice any difference between CBR and VBR. A video needs to have fast and not-so-fast scenes for VBR to make any difference.

Does the blockiness appear on your PC monitor only or on your TV set too? If it's only the PC, I wouldn't worry about it. Watch a commercial DVD on your PC and you'll find plenty of blockiness, especially on dark grey scenes. I own a DVD rentals store and have watched most movies both on my PC and on a TV and they always look better on a TV. But then maybe because we are used to the picture of a TV set. Picture quality is a subjective topic. Everyone's eyes and mind are different.

Bugster, I've got a question regarding CQ. Played with it and seems to give very good quality. But I've never used it to create a DVD because I don't see the purpose. I may be missing something here.
Here's how I reason it. If disc space is not an issue I use CBR. Probably wasting bitrate comparing to CQ but then half empty DVDR is a waste too. If I want to put some 1.5h-2h of video I use VBR because I can get the size I want. With CQ you can't so if I overshoot I need to encode again (or shrink afterwards) so it kind of beats the purpose. And if I undershoot I feel like I'm wasting DVD space. When would one use CQ?

[daamon]

Ronaldus, we meet again...

A few snippets to add to the wealth of info already provided:

However, at the very start when the screen fades from black to a bunch of white clouds against a light blue background, I always see some initial blockiness -- more at this point than at any other scene change throughout the segment.

It's always possible that there are imperfections in the source (i.e. on the DV tape), or in the AVI from the tape. A wild guess (someone confirm or deny the possibility) - maybe the transition in Pinnacle didn't render properly?

I tried ... and 2-pass VBR averaging the same [7000] with 8000 max and 6000 min.

Your min of 6,000 is way too high. I learnt this from bugster when I first started out. Set it to more like 2,000.

Having the min lower means that those scenes that don't need a high bitrate can go as low as 2,000 (or whatever is set), rather than the 6,000 you set, meaning that, because TMPGEnc is working to an average, even higher bitrates can be used where needed. Otherwise, with 6,000 set, these areas needing a higher bitrate are starved and so quality suffers.

#2. Are there any clear-cut cases where CBR is always better than 2-pass VBR?

Yes, CBR is better when the size of the final output file doesn't matter - i.e. it'll fit even at max bitrate.

VBR generally produces smaller filesizes, fairly accurately based on the average bitrate. And so is better if space is tight.

Good luck...

[bugster]

@Ronaldus, TmpGenc's default settings produce close enough to the best achievable quality in most situations for most people. The only settings worth changing are bitrate and motion search precision. Tweaking GOPS and quantize matrix is only for those who want absolute best quality with slightly unusual requirements such as needing to use low bitrates

Bugster, I've got a question regarding CQ. Played with it and seems to give very good quality. But I've never used it to create a DVD because I don't see the purpose. I may be missing something here.
Here's how I reason it. If disc space is not an issue I use CBR. Probably wasting bitrate comparing to CQ but then half empty DVDR is a waste too. If I want to put some 1.5h-2h of video I use VBR because I can get the size I want. With CQ you can't so if I overshoot I need to encode again (or shrink afterwards) so it kind of beats the purpose. And if I undershoot I feel like I'm wasting DVD space. When would one use CQ?

I use CQ mainly with lower quality source that I encode at 1/2 D1 resolution. I know it's not worth the encode time of 2-pass VBR or worth a whole DVD. SO I can encode 2 movies to fit onto one DVD with 1/2 D1 resolution and CQ that end up better than the same playing time with CBR but much faster encoding tha 2-pass VBR. If My encodes do end up a little over sized I simply author and then use DVDshrink on the files before burning.

[petar]

OK I get it. Thanks, bugster

[Ronaldus]

Good information, everyone. Nice to hear from you again, Daamon.

To summarize about GOP structures,
- The often-referred to guide at dvd-hq.info suggests: IBPBPBPBPB
- TMPGEnc's "Standard" default is: IBBPBBPBBPBBPBBPBB

From the comments here, it sounds like the consensus is IBBPBBPBBPBBPBBPBB is better than IBPBPBPBPB, so do people agree I should use the "Standard" TMPGEnc default while using dvd-hq.info's settings for all the other settings?

Has anyone tried both of these GOP structures with all other settings the same to see if there is any noticeable difference?

I may try it myself, but since everyone's eyes and quality opinions are different, I'd like to hear what others have tried. It would be very nice to have 1 .mcf file with settings I can use on every type of video I may try to encode so the only thing I'll have to change each time is bit rate. I think I'll leave motion search precision at the highest quality because I can let the machine work over night, so bit rate is the only thing I need to tweak. As Daamon reminded me, I should set my minimum lower to not starve the parts that need it. However, I don't think I ever want to go below 4000. I remember a discussion about where the max should be, and I think the conclusion was 9400 (including audio of course), but some rare players had problems as low as 8400. This basically means 4000 min and 8000 max for the video, and I set my average at exactly what the Bit Rate Calculator says.

I haven't been able to tell any quality difference in audio (MP2 or AC3) encoded at 192, 224, or 256 Kbps. This means I will probably always use 224.

I remember when I started on this long project to learn how to do this, I looked around this site for a step-by-step no-brain-needed guide how to do this from start to finish, and I could not find any. There were tons of threads discussing various steps, but putting it all together was hard for me. Some threads said there is no start-to-finish guide because there is no definitive "fits-all" process. However, that's exactly what I've been looking for, and it seems to me for those who don't want to be GOP structure experts that there's really no tweaking needed in TMPGEnc. Therefore, there should be a no-thought-required process that gets good enough results for most people as long as you don't take your average bit rate much below 5000-6000 Kbps.

Are there any guides I have missed that start from AVI capture and walk you all the way through to burning using TMPGEnc, TMPGEnc DVD Author, and VirtualDub as the encoding/authoring tools? The capture and burning step have been the easiest, so I never needed a guide. I just want to convince myself that Pinnacle is not re-encoding edited and re-saved AVIs, and so far I haven't been able to do that.

I believe Daamon uses a process very similar to mine. If no such guide exists, I'll be happy to write up one and post it to see if I can get a consensus that the way I'm doing it is a good brain-less way to achieve good results without tweaking. Another important part of my criteria was the tools needed have to be among the cheapest to buy. Pinnacle was free with my camcorder, Nero free with my drive, so I only had to get the TMPGEnc stuff.

It would be great to know if anyone who has used the TMPGEnc tools to create DVDs has then tried encoding the same material in Vegas, DVD Lab, or other more expensive tools -- and how did the results compare? Anyone?

[Ronaldus]

Hmmm. I guess this thread died, so I'm going to assume my posts are too long which scare most people away from reading them. I will re-post my questions in separate, smaller messages. I'm not sure if these subjects, such as GOP structure, are advanced or newbie, though. I feel like everything I discuss is a newbie topic.

[donpedro]

#6. Finally, I use Pinnacle Studio 8 to capture to AVI and edit out unwanted segments which I resave to AVI using the same DV Encoder Codec (I believe). TMPGEnc has a "Source Range" filter under "Settings -> Advanced" to remove an unwanted segment from the source AVI. Is there any advantage to using the Source Range filter instead of editing in Pinnacle Studio and resaving as an AVI? My instinct is there is no advantage because Pinnacle does not appear to be re-encoding, but just because I don't notice any quality difference doesn't mean there isn't a subtle loss due to re-encoding going on. Theoretically, which of these options is the best way to edit? Don Pedro had suggested performing some tests on editing and encoding to try to PROVE that Pinnacle was simply copying using the same Codec rather than re-encoding, but I couldn't figure out how to prove it. (I guess I'm not scientific enough.) Has anyone been able to prove this or know of a way to verify when Pinnacle is re-encoding? (Unfortunately Pinnacle says Rendering both when copying and encoding.)

Unfortunately, I don't know how to prove that it is 100% true If you can't see difference, I wouldn't worry about it too much.

[donpedro]

I'm going to assume my posts are too long which scare most people away from reading them.

You are definitely right there

[daamon]

Nice to hear from you again, Daamon.

I'm usually floating around in the background... Sounds like you've made good progress from when you first started.

If no such guide exists, I'll be happy to write up one and post it to see if I can get a consensus that the way I'm doing it is a good brain-less way to achieve good results without tweaking.

And I'll be happy to review it (to the best of my ability) and feed back constructive comments. email a MS Word doc or something.

One initial comment before you start:

I'm going to assume my posts are too long which scare most people away from reading them.

You are definitely right there