TMPGEnc Bitrate Settings - Any Benefit In Different

[daamon]

Hi,

I've been using 2 pass VBR on DV AVI at:

Min = 2,000
Average = 6,000
Max = 9,600

And muxing with AC3 @ 192kbps iin TMPGEnc DVD Author. I'm happy with the results, but don't know any better, so...

I've been wondering this for a while - As the source is DV AVI (i.e. good / high quality), what would the effect or benefit (if any) be if I upped the average to a higher setting?

If it's a positive benefit, what sort of figure would be a good one to have?

Cheers.

[gioto]

Have a look to my post "Quick moving scenes problems": we are talking about the same.

Bye

[daamon]

Mmm... Kinda useful.

I don't have a problem with fast motion in scenes, I was just curious as to whether upping the average bitrate would yield any worthwhile benefit in the quality of the final video (then burnt to DVD).

At the same time, I'd obviously like to know if there'd be any negative impact(s)... I know the filesize will go up, but they're short enough clips so as not to matter.

Anyone...?

But cheers for replying.

[disturbed1]

Depends on your footage. Sometime you may just be waisting bits. But if you have the space use it.

The only real negative impact of upping bitrate (minus space) is that some set tops have a problem playback high bitrate DVD+/-Rs. The motor/laser isn't good enough to decode the information, or the blank disc isn't high enough quality to be read at that speed.

I try to keep my average below 7500, and use a max of 8000. I have an old player that stumbles when the peak hits above 8000, even on pressed DVDs.

[FulciLives]

Yes I agree that with any home-made DVD you should keep your MAX video bitrate to no more than 8000kbps and if using PCM WAV audio you might even want to use a slightly lower value such as 7800kbps

Using a video MAX bitrate of 9600 is MUCH too high even with MP2 audio or AC-3 audio

- John "FulciLives" Coleman

[daamon]

Hi disturbed1,

For the moment, it's holiday videos, friends weddings etc. that are gonna be no more than 40 mins apiece.

So it seems that, if I wanna do one per disk, I could use the higher bitrate - but I could probably fit 2 per disc, with a lower (or same = 6,000) bitrate and not really affect quality.

In terms of making a difference to quality, would say an average of 7,500 be noticeable over 6,000?

I've got a DVD-RW, so I can check if my stand-alone can handle higher bitrates, and (as is always the mantra) I'll get decent media.

Cheers for the info. Much appreciated...

[daamon]

Hi FulciLives,

Cheers, also, for the info (regarding the average bitrate up to approx 8,000).

Question:

Using a video MAX bitrate of 9600 is MUCH too high even with MP2 audio or AC-3 audio

Can you explain why? I'm not challenging the statement, just want to understand / learn.

Also,

In terms of making a difference to quality, would say an average of 7,500 be noticeable over 6,000?

Any comment on that one too?

Ta.

[FulciLives]

The best you can do is CBR 8000kbps so 2-pass VBR should only be used when the AVG will be lower than 8000kbps but if it is close to being 8000kbps (such as using 7500kbps) then you might as well just go with a CBR encode.

You say these programs are about 40 minutes each so to get 2 on each disc you could use an AVG bitrate of approximately 7200kbps if using 256kbps AC-3 or MP2 audio. If you are using LPCM WAV audio then you are looking at 5900kbps MAYBE 6000kbps but that might be pushing it (if the total is exactly 80 minutes total).

It's been fairly well documented that the total bitrate of a DVD (including video bitrate, audio bitrate and even subtitles etc.) should not exceed approximately 9800kbps but some think it might be as low as 9600kbps

So a CBR of 8000kbps VIDEO and 1536kbps AUDIO (if using LPCM WAV) brings you RIGHT up to that upper limit.

Now if you are using a low bitrate sound format like MP2 or AC-3 then normally you would NOT go above 256kbps for 2.0 channel (maybe 384kbps for MP2 at the most). So in theory you could set the VIDEO bitrate higher than 8000kbps as long as the VIDEO + AUDIO bitrate don't exceed that 9600kbps to 9800kbps limit but that is JUST not a good idea.

Why?

One I think that 8000kbps is sorta meant to be the VIDEO bitrate limit since you CAN'T really go higher if using LPCM audio. Even though that is only (partly) theory when using MPE/AC-3 keep in mind that it is "harder" for any DVD player to read a DVD recordable due to differences between a recordable and a more "standard" factory made DVD.

So never set your bitrate higher than 8000kbps ... otherwise you are asking for "trouble".

- John "FulciLives" Coleman

P.S.
Here is a little trick when doing 2-pass VBR
When you set your AVG and your MAX make sure that you set the MIN so that the AVG is exactly inbetween the MIN and the MAX

Some examples:

If your MAX is 8000kbps and your AVG is 6000kbps then your MIN should be 4000kbps

If your MAX is 8000kbps and your AVG is 5500kbps then your MIN should be 3000kbps

If your MAX is 8000kbps and your AVG is 5000kbps then your MIN should be 2000kbps

Also if you can help it I would never set the MIN to lower than 1000kbps although some say it is safe to go as low as 300kbps and others will tell you that 0kbps is fine for a MIN

[daamon]

Hi FulciLives,

Thanks for all the detailed info, tips and pointers - all good learning.

Cheers.

[SaSi]

Fulcilives,

Why set the min bitrate so that avg is right in the middle between min and max?

There have been quite a few threads with quotes from experienced members describing why and how using a low min bitrate benefits quality.

I don't challenge your statement about setting min = 0. Having a very low bitrate CAN cause problems, but setting min too high just deprives bitrate from scenes needing this.

I have seen that CCE is practically the best encoder in distributing bitrate appropriatelly and the only one able to lower the bitrate in low-entropy scenes below 1Mbps. Mainconcept and Tmpgenc are rather more "conservative" and, in my opinion, this is one of the reasons why CCE excels.

[Jukka]

I thought the DVD minimum was 3000k not 2000k?

Also I have experimented with DVD conversion bitrates and CBR / VBR. From what I can see, if you are converting a slow moving film or a cartoon type (e.g. Lion King type stuff) then CBR is fine, if fast action movies (e.g. T3 type stuff) then VBR is better.

The way I work is I use TMPGEnc to suggest a CBR rate, and then if I want to use VBR go 1500-2000 above and below (subject to the limits). I've not had any problems with this.

Again, trial and error is good practice + knowing what your eye can see. Some of my DVDs look crap on the PC but OK on my 28" Widescreen at 10 feet.

Good luck,

Jukka

[FulciLives]

Fulcilives,

Why set the min bitrate so that avg is right in the middle between min and max?

There have been quite a few threads with quotes from experienced members describing why and how using a low min bitrate benefits quality.

I don't challenge your statement about setting min = 0. Having a very low bitrate CAN cause problems, but setting min too high just deprives bitrate from scenes needing this.

I have seen that CCE is practically the best encoder in distributing bitrate appropriatelly and the only one able to lower the bitrate in low-entropy scenes below 1Mbps. Mainconcept and Tmpgenc are rather more "conservative" and, in my opinion, this is one of the reasons why CCE excels.

When using 2-pass VBR in TMPGEnc and setting your AVG exactly in the middle of the MAX and MIN you are "forcing" TMPGEnc to do ... more or less ... what it does when you use CQ mode except you can accurately determine the final file size with a 2-pass VBR which you cannot do with CQ mode.

I'm not exactly sure why but this does seems to work well with TMPGEnc.

- John "FulciLives" Coleman

[dwiesel]

Not to throw a monkey wrench in the mix, but from my own trials/errors in setting different bitrates, the values of your (I), (P), (B) frames can affect the effectiveness of your bitrate. When I used a high (P) frame count of 14 with 1 (I) frame and no (B) frames, I found increasing the bitrate from 8000 to 9400 yielded a visibly superior MPG. I know (P) frames use a higher bitrate than (B) frames, so this may be the reason. When I used lower (P) frames, upping the bitrate didn't noticably change quality (though after so many comparisons they all start looking the same when the difference is that minimal).

Also, in my MANY back to back tests using different settings, I found the best quality MPG I got was using (I),(P),(B) = 1,14,0 and setting the average and max bitrates to 9400 and min bitrate to 0 (though I'm not sure how much the min bitrate affects the quality). This really surprised me...I thought keeping the (P) frames low would force the use of more (I) frames (since no compression on these) and produce a better MPG. I do know that (I) frames use the highest bitrate...maybe using higher (P) frames more effectively uses the bitrate.

If anyone can explain/expand on this (or feel free to dispute it), I would greatly appreciated it. Thanks (sorry if this has gotten off the thread too much).

[indolikaa]

When using 2-pass VBR in TMPGEnc and setting your AVG exactly in the middle of the MAX and MIN you are "forcing" TMPGEnc to do ... more or less ... what it does when you use CQ mode except you can accurately determine the final file size with a 2-pass VBR which you cannot do with CQ mode.

I once wondered that, and then told myself I was far too newbie to reach that conclusion...

[dafreak]

don't know if this helps, but i've been using this calculator to determine 2-pass vbr bitrates. It also shows what your P, B, and I settings should be.

calculator

I convert a lot of anime from avi to mpg2 and i'm pretty thrilled with the results i get from it.

[dwiesel]

...have used that calculator as well, and like the results...I'm just not sure what the benefit of setting all bitrates to 8000 does, and it's hard to see why the I/P/B frames change the way they do based on video time.

[SaSi]

You are correct. After too many tests things start looking the same

There are no general rules on how different parameters affect the result. Even more when one tries to compare settings among different encoders.

However, sticking to Tmpgenc, I believe that the more B frames you use, the better image quality you get at a given bitrate.

Encoding a typical video at CQ with a max. bitrate of 4800, Tmpgenc will (typically) allocate
50~60kB for an I-frame,
28~34kB for a P-frame and
8~12kB for a B-frame.

Decreasing the P frames per GOP by two and increasing the B-frames appropriately, results in frame sizes that are almost the same as above. This means that the overall bitrate is decreased (for a given Constant Quality).

In this scenario, you make Tmpgenc decide the actual bitrate provided overall quality meets the CQ setting you choose. Therefore, for the same quality you get a smaller file if you use more B frames and decrease P frames.

This, however, may end you up in trouble with compatibility among several players. Although I don't believe that the IPBBPBB... GOP structure is mandatory in DVD (don't really know), I am certain that some DVD Players may want the content in that format to playback properly.

I think, after months of experimentation, that the best and safest approach is to "invest" in motion search rather than modify the GOP structure from the "well-known-default".

Motion search tries to remove redundancy by shifting the current frame's macroblocks across a motion vector and generate a better approximation of the next frame. This is pure calculation and within the strict MPEG format, so it doesn't impose compatibility penalties.

It only costs in time, while you wait for Tmpgenc to encode a movie and it takes a day or so...

[SatStorm]

This objervation on TMPGenc plus 2 Pass VBR (new), I was the first who mention it in this same forum, about 20 months ago... When TMPGenc plus appeared for the first time.
I had a long conversation with Adam for this, he couldn't understand why TMPGenc acted like that, which is natural since he is a dedicated CCE user, which on those subjects is a far better program.

This encoder, TMPGenc, is unique and doesn't follow what the book says.
In practice, we always have better results when the averagate is equal distant from minimum and maximum values. I once explained why TMPGenc works that way, it is a long post I don't have the patience to write it down again. Maybe you gonna find it in a forum search

Later TMPGenc plus versions don't implement something on the 2 pass VBR mode, only on CQ modes. In the latest version for example, CQ mode works in a mothod I call "rise over the minimum", which means that if you set for example 2000 min and 8000 maximum, TMPGenc gonna encode @ 2000 and each time it "thinks" that more bitrate is needed, will rise it up to what you set at quality. The default 65% setting is what really works best.
In that example, 2000min and 8000 max @65% is about rising to 5500kb/s at most, which is OK for typical sources (not home movies).

With TMPGenc you have to study TMPGenc-ology, that's for sure...

[dwiesel]

Thanks SaSi, I bow to your knowledge and experience. My question will also stick to just TMPGenc. Also, if it makes a difference I'm referring to DVD, not VCD.

I find the whole I/P/B frame issue daunting, as there seems to be no definitive rules (as seems common with settings in TMPGenc). You mentioned increasing the B frames to achieve the same quality with a smaller file. I don't doubt this, but how do you assimilate this with other sites suggesting to use no B frames for best quality? Since B frames are bidirectional, isn't there an inherent quality loss in using them since there's more interpolation involved? I would think restricting frames to just P and I would work the best. I have so much to learn, but am trying.

Also, is there a better forum to continue this discussion of I/P/B frames, I have other observations/questions I'd like to share and get feedback on. Thanks.

[dwiesel]

Just wanted to pass along what I NOW find to be the best settings in TMPGenc for achieving (what I think is) the highest quality MPG...BUT, this only pertains to DVD NTSC projects one hour or less...haven't tested with longer segments:

- CQ, 100 quality, max=9400, min=4000, P/B spoilage = 0, leave "enable padding" unchecked

- motion precision is Highest Quality (very slow)

- I/P/B = 1/3/0, max GOP frames = 15

These are just some of the settings...if interested in others, let me know.
For me, another coincidental advantage of keeping B/P frames low is when I crop scenes in TMPGenc Author...you can't manipluate B/P frames, only I frames.

Of course, it's all subjective, but I did many back-to-back comparisons (and my source was laserdisc), and found the above settings gave the highest quality image for an AVI captured in YUV2 colorspace.

I've been thinking about trying CCE for the simple reason that it uses the YUV colorspace, thus keeping my entire process within the YUV realm and out of the RGB realm (i.e. no colorspace conversion)

Thanks.

[FulciLives]

I've been thinking about trying CCE for the simple reason that it uses the YUV colorspace, thus keeping my entire process within the YUV realm and out of the RGB realm (i.e. no colorspace conversion)

BINGO !!!

I capture with the PICVideo MJPEG codec (on the 19 quality setting) and this is a YUV/YUY2 codec so for that very reason I now only use CCE for my encoding.

I still use TMPGEnc here or there but never with my captures anymore.

- John "FulciLives" Coleman

[dwiesel]

Ah, there's nothing like reassurance from a source you trust.

John, do you find that keeping in the same colorspace actually makes a substantial difference in terms of quality (be it sharpness / brightness / color)? I know it's subjective, but was curious what you had to say.

I miss the zombie image!

[FulciLives]

Ah, there's nothing like reassurance from a source you trust.

John, do you find that keeping in the same colorspace actually makes a substantial difference in terms of quality (be it sharpness / brightness / color)? I know it's subjective, but was curious what you had to say.

I miss the zombie image!

When you use TMPGEnc with a YUV/YUY2 capture it never really looks just right if you ask me. I've tired both checking and not checking that dreaded option. You know the one! Output YUV data as Basic YCbCr not CCIR601

Argh !!!

Either way the output looks different with what CCE does which looks overall much better and more true to the source.

So you will see a brightness/contrast difference as well as color difference. For the better. As for sharpness etc. I think CCE looks just as good as TMPGEnc yet it is MUCH faster.

However if you are going to use CCE then brace yourself to learn AviSynth because that works better for CCE than using VirtualDub ... plus VirtualDub uses RGB colorspace on it's frameserved output which is what you are trying to avoid i.e., conversion from YUV/YUY2 to RGB

- John "FulciLives" Coleman

P.S.
It has been about a week now with my Naked Geek bashing Avatar so it may be time to change back to 'ol mummyface as xcool (opps now banned) use to call it.

[dwiesel]

Thanks John...

(Heavy sigh)...after all that testing with TMPGenc, now to have to do it all over again for CCE (which I just downloaded)...it's a good thing I have an obsessive nature.

If it's not too much to ask (and since I value your opinons), can you point me in the right link direction for getting the best settings, or just share some of the basics...I hate to keep reinventing the wheel (and am low on fuel). It may prevent me from festering away in my PC corner all weekend. Many thanks.