Linear Quantizer scale or not? Any difference at all?

[defense]

Hi, I have been wondering for about a good three weeks now on whether or not using the Linear Quantizer scale function in CCE means anything significant at all. I have been using DVD2DVDR along with CCE to reencode movies that can't fit otherwise. Now, here is the confusing part. In DVD2DVDR if you hit the help tab and scroll down to the bottom, ChatWalker, the programs author, tells you that for progressive films, you set Linear Quantizer Scale to "false" which apparently means not to check it or use it in CCE.

To make a long story short, that is exactly what I have been doing up until this point. But I decided to give bitrateviewer a try since so many people use it. What I noticed is that on nearly every film I needed to reencode, which happens to be Progressive, (NTSC) I noticed that in bitrateviewer, the Quantscale said "linear." So I figured that I would try a few movies with the linear quantsizer scale option checked, and the output on the reencoded video looked the same as with that option NOT checked. My point is that there is NO QUALITY difference or ANY other difference that I can "SEE" between having the Quantizer option checked, making it "linear" or having it unchecked, making it "nonlinear."

Also, if the original movie says "linear" in bitrateviewer, I would think that it would make sense to keep it "linear," but then I am wondering why the author of DVD2DVDR would say to leave that option "false" for NTSC, Progressive films. See why I said this was confusing? lol...If ANYONE can please explain this option to me and let me know what the purpose of it is, and if I should use it or shouldn't use it, I would appreciate it greatly! Thanks

[defense]

<bump>

Anybody know anything about this Linear Quantizer scale function at all?

In MPEG-2 video, a quantization scale factor used for quantizing DCT coefficients is uniquely determined by Quantizer Step Table and quantization scale code. As for Quantizer Step Table, there are two tables specified, Linear table and Non-linear table, and they can be selected according to bit rate or characteristics of input video. Generally, Non-linear Quantizer Step Table is used when coding bit rate is high.

ref: http://w3-mcgav.lab.kdd.co.jp/mpeg/mpcs/faqe.html

By the way, DCT is somewhat similar to a fourier transform. It turns the video from a time domain to a frequency domain. Sounds complex ... basically it allows one to carry out some mathematical calculations faster and with greater precision than could occur otherwise. So basically the quantizer is a way of setting up the data for this transformation.

Conclusion seems to be: For high bit rate use linear; for low bit rate use non-linear. To be honest, I almost always deselect linear in DVD2DVD-R meaning I am doing a non-linear. I have read (but you can find disagreement on this point) that you should generally do a non-linear transformation unless there is a special reason to do otherwise. I never have such a special reason and I am very satisfied with my results.

[defense]

@robw, thanks for your reply. Then you have basically done what I have done all along, which is ALSO what ChatWalker reccomends in his program, DVD2DVDR when you hit the help tab. He basically says that for Progressive films, you should leave the linear quantizer scale set to "false." That is exactly what I had been doing all along, with NO problems, however, I kept reading about bitrateviewer and figured I would try it. I basically just ripped a VERY small portion of the first VOB of a couple of movies to my desktop, and then ran bitrateviewer on them. I noticed some of them said "linear" under Quantizer scale, while others said "nonlinear." The truth is that it seemed after I did enough tests, that I got just AS MANY "linear" results, as I did "non linear" so my question now is that I would be willing to bet that SOME of the movies which I left Linear Quantizer Scale unchecked in, would of shown up "linear" within' bitrateviewer, meaning that if you take that information, apparently, in order to keep the video the same, you should "check" linear quantizer scale within DVD2DVDR. And if you don't, then your encoded .m2v file is going to be "non linear" when the original film is "linear." You see why I am saying this is confusing?


So in summary, I have three quick questions for you. One, I guess it's obvious that it makes NO difference at all if the ORIGINAL film is listed as "linear" with respect to the Quantizer Scale, and the encoded .m2v file comes out "non linear." Right? I mean, I don't understand why the original film should be changed from linear to nonlinear, but there must be a good reason if ChatWalker says to not check it for progressive films. Also, my second question is, you mentioned that you read you generally only use the linear quantizer scale function within CCE for special circumstances, which you've never had. I'm not quite sure what that means, but I do know that you said this:

I almost always deselect linear in DVD2DVD-R meaning I am doing a non-linear.

I think the key word there is "ALMOST." I'm just curious what times you did select Linear Quantizer Scale, and why?

And last but not least, and to me, this is the MOST IMPORTANT question of them all. If a movie is ORIGINALLY Linear and you change it to Non-Linear, as you and I have done MOST of the time, what potential problems, if any, will there be with playback, or otherwise? And vice-versa. If the film is Non-Linear, and you select Linear, what problems could or would you encounter. As I said, I have had ZERO playback or picture issues on ANY of the movies I have made using DVD2DVDR, but I know that I must of made SOME films which were originally Linear, and ended up having them NON Linear in the end. And vice-versa. And if the answer is that it doesn't make a difference, as it appears it doesn't, then why should it matter which option is checked? Thanks again in advance for any more info on those three questions.

My understanding of this issue comes from looking at guides and generally doing some additional studying in books, the web, etc. on how the encoding process actually works. That is, I've spent some time trying to get a deaper understanding of DCT (discrete cosine transform), 3:2 pulldown, drop frames, etc. As I understand the method chosen for quantizing in part depends on how the bit rate is used. For example, many broadcasters choose constant bit rate because they only have a certain bandwidth or capacity. So basically, as I understand it the bit rate directly effects the need for linear or non-linear quantization. As stated in my previous message, the general recommendation is to use linear only for high bit rate. Since we are generally compressing a high capacity DVD9 disc to a smaller capacity DVD5 (for example when using DVD2DVD-R) it seems logical to me to also bias toward using the non-linear quantizer. That is the logic that I have used.

In my previous message I stated that I "generally" use non-linear. By that I mean that I have indeed done some experiments to see if I could detect a difference between linear and non-linear. I must admit that I see none. So all the more reason to stick with a standard procedure which is that I use non-linear unless I am doing VERY little compression and then only if the original film was done with linear quantizing.

I hope this helps. Its not perfect logic but its the way I've been doing things. By the way, here is a good reference for more detailed information on the encoding process

http://131.170.32.110/Course/not%202001/co421/report98/group5/CO421%20-%20Assignment%2...1%20Report.pdf

also

http://www.broadcastpapers.com/sigdis/Snell&WilcoxMPEGVideo-print.htm

[adam]

For what its worth, the CCE manual specifically says that you should always use non-linear except when using mpeg1, in which case linear is required. I've read the same thing elsewhere as well. But I agree with robw that if there are instances where linear would be preferable, they probably won't apply to what the average person would be doing. I am also in the same boat where I too see no difference between the two, so I figure I might as well go with the general consensus.

[defense]

@Robw, thanks again for all of the information. Your more detailed example and relationship between bitrate and Quantizer scale makes more sense now. I will check out those links you have posted. One other thing I wanted to mention, which is another reason I did choose to maintain the "Linear Quantizer Scale" is because there is a little discrepency within' DVD2DVDR. For example, as I mentioned, if you click the help tab, you will see that Chat Walker says to leave the Linear Quantizer option "false" however, one thing I find interesting is that he also says that DVD2DVDR will default some settings automatically for you when you choose the main movies Vob Set. I have noticed that on several occasions, DVD2DVDR will automatically select Linear Quantizer Scale to "true." As you can see, that is a conflict with what is being said to do for progressive films, and what DVD2DVDR is automatically selecting. What do you think about that? I'm just curious, because it is definitely either a bug, or some type of conflict. Although, I don't think it is a bug, because Chat Walker says that the program is supposed to select options for you depending on the film type, etc. Thanks again for all of the information though, it has been extremely helpful. Also, if you get a chance, let me know what you think about the conflict with DVD2DVDR. Thanks

adam said:

For what its worth, the CCE manual specifically says that you should always use non-linear except when using mpeg1, in which case linear is required.

Humm, I remember seeing a post awhile back leading to a guide which said something about "getting the most out of CCE" and If I remember correctly, that post said that to get the MOST out of CCE, you should check or use "LINEAR QUANTIZER SCALE" in CCE.
Anyway, as you agree, there is no "noticeable" difference either way. I'm just trying to now find out the reasoning behind using or "not" using linear quantizer scale. I.E. why you should not keep an original movie that says "LINEAR" linear, because CCE says so. Either way, I think I will continue to stick to what has worked for the majority of movies I have done, and that has been to leave "linear quantizer scale unchecked, or "NON LINEAR."

[defense]

One other thing robw, could you also verify the settings that you use in DVD2DVDR for MOST or ALL of your NTSC films? Please just tell me whether you check these options or not. I was just going through some of the very first posts on the original DVD2DVDR thread on this forum and I saw where Chat Walker said the Upper Field First needs to usually be "deselected." I know he is from Germany, but I'm wondering if that was a typo, or if he really means to leave upper field first "NOT" checked. He said that on either the 4th or 5th page of that thread...right around there. I ALWAYS check "upper field first" because in the help tab, it says to check it or keep it "true" for NTSC films. Now I'm wondering if I have been setting that option correctly. Here are a list of options to either check or uncheck.


Upper Field First: I ALWAYS check this for Progressive films.

Progressive Frames: I ALWAYS check this also for Progressive.

Linear Quantizer Scale: AMOST always "NOT CHECK IT"

Zig Zag Scanning Order: I ALWAYS check this for progressive films.

Please let me know if the options that you check are the same as me. Thanks again

I always set upper field first for progressive since there are no fields anyway.

Yes, I always select progressive for anything that is progressive

Zig-zag scanning order if progressive otherwise not.

One issue that I have not resolved. You should only select force film / driop frame when the frame type is film with a rating of 95% or higher. To do this accurately you need to make a determination using software such as DVD2AVI. I don't usually make this analysis and just assume that my movies are 95% film. So far, no problems but I'm watching this to see if it ever causes a problem. I have also used DoItFast4U - this software actually makes this determination automatically. It seems to run DVD2AVI twice. Once to check and then a second time after it has made the determination. This seems to suggest that you do in fact need to check this carefully. So I am concerned that I am not giving this proper consideration in DVD2DVD-R. On the positive side, my projects are all coming out fantastic so I have no complaints and I guess this means that everything is set properly .... I need to do some more checking / investigating in this area.

[defense]

@robw, Thanks ONCE AGAIN for the helpful info. I guess ChatWalker just had a "typo" when he said to have Upper field first "deselected" for NTSC films. You brought up another very interesting point with the 95% film issue. I too, have just figured that I was probably ok the majority of the time, because of the progressive film type, and I watch ALL of my projects very closely, and carefully as well. To this point I have not noticed anything negative at all, but I'm sure there will be that movie that comes around which is the exception to the rule and which I will have to make sure I get right. I just use DVD2AVI but that DOITFAST4U seems like a good tool. I might have to try it out. Thanks once again for all of your info.

I believe the point Chatwalker was eluding to relates to interlaced film. It is not always the case but USUALLY Region 2 (Europe) movies when interlaced use upperfield first whereas Region 1 it is the opposite.

[adam]

If you do any type of fine tuning to your resizing, to optimize macroblock usage ala FitCD for example, then you should uncheck upper field first in CCE, even with a progressive source. CCE has a bug where it shifts everything up one pixel when the upper field first flag is set which can screw up your optimizations.

Otherwise this setting makes no difference with progressive sources, since there are no fields anyway.

[defense]

adam said:

Otherwise this setting makes no difference with progressive sources, since there are no fields anyway.

What about a lot of NTSC films which are a "hybrid" of both, progressive and interlaced? How do you set upper field first in those instances? Adam, when you say "resizing" do you mean, changing the resolution? I don't do ANYTHING other then leave the movie the way it is. I check the functions I mentioned above, and that is it. Thanks again for the info, and please let me know about the mixed sources.

[adam]

Well if your source is interlaced then yes you need to set the correct field order before encoding. This applies to interlaced and hybrid material, which some NTSC material is. I assume we are talking strictly about DVD here since we are in the DVD to DVDR forum so...unless you are dealing with anime or dvds of tv series, you will almost never have to worry about this. %99 of all theatrically released NTSC DVDs are stored progressively and are telecined during playback. So on the vast majority of theatrically released NTSC dvds you can just use forced film and you will be fine. I think robw already mentioned the test for when and when not to use forced film.

If forced film will not work on your DVD (anything less than %95 film) than you need to first check your field order and set it accordingly. BitrateViewer can check field order but it is sometimes wrong. A more accurate test is to use TMPGenc. Load your source and double click on the deinterlace filter and set it to even-odd field (field). Now scroll through your movie and see if it looks ok. If the field order is incorrect than the playback will be jerky or the picture may appear to jump vertically. If its incorrect than look on the advanced tab and change the field order and recheck until you got it right. Now set this to the same field order in CCE. If you get it wrong your encode will look very bad, it will play a few frames, jump back a few, etc etc... All you have to do is reverse the field order in pulldown if this happens...just use the gui.

Yes by resizing I mean changing the resolution. Sorry, I mostly make SVCDs so its hard to get this step out of my head. If your not resizing at all then forget about what I said, it doesn't apply to you.

[defense]

Thanks again Adam. First, i'm glad you clarified the "resize" option. I started out making DVD's, but I got a Great deal on the DRU500a, and now I'm getting a $50 rebate, so It's even a better deal. Anyway, fortunately, I have not experienced that jitter/jumpy playback issue as of yet. As you said though, MOST theatrical releases can be set to reencode with upper field first because of the film type. I am glad I know all of thise information that you and robw supplied me with, because when it does happen, if it does happen, which it probably, EVENTUALLY will, then I'll know how to handle it. One last thing, you mentioned to use TMPGenc to check the film type, as it is more accurate then bitrateviewer. That is a good thing to know, because I usually use bitrateviewer and DVD2AVI. I have seen discrepencies between the two before, so I wasn't sure But I will use TMPGenc to get the results as accurate as possible. Thanks again for your assistance.