View Full Version : Interpreting the comp. test values & b/(p*f) value?
Backflip
14th September 2005, 06:04
I've been encoding quite a bit lately, but have not really taken a whole lot of notice of the numbers. I've read the GordianKnot guide, but I don't think I fully understand. It recommends staying around 0.270 (av. bits / (pixel*frame)) for a 2CD rip, but also not to have a compressibility higher than 80%, yet in some cases if I go for 0.270 b/p*f I get a high percentage value of 90%+. Then I've read on the forums b/p*f isn't the most important thing to consider, it's infact the comp. test value.
Another thing is the GordianKnot guide goes into a little detail about how using bframes of 1 to 2 will make a rip look better at a lower b/p*f value, but what value would be acceptable now?
Here are some of the comp. test numbers I get. Now, on face value which would you choose, and why? This is for a 2CD rip. 107 minutes (25 fps).
1. 54.8% 0.364
2. 74.1% 0.269
3. 77.4% 0.258
4. 62.4% 0.320
5. 70.9% 0.281
I I initially thought just to go with 3, just because the comp. test % is highest without being too high.
I have VHQ 4, MSP 6, Chroma Motion, Chroma Optimizer and Trellis (1,31) enabled.
Thanks :)
jggimi
14th September 2005, 19:17
You can use a b/p*f value as an initial starting point for resolution / bitrate prior to executing your first comp test. It can help you determine, approximately, if you need 1 CD, 2 CDs, etc.
Once you've done your first comp test, you can ignore the specific b/p*f numbers and instead, focus on the percentage values returned from the test.
The comp test percentage compares the b/p*f value at the maximum possible bitrate obtainable (determined by the test) against the b/p*f value of your current bitrate selection.
Backflip
16th September 2005, 07:40
Thanks, that's what I've mainly been focusing on, taking note of the percentage figures, especially as I'm using CQM's. I go for between 70 and 80%. Unless someone can guarantee that those percentages are too high to actually be utilised completely by the codec, then I'll keep using that target :)
Out of the tests I'll be going with 3. 77.4% 0.258 and perhaps throw Q-PEL and AQ in there for good measure :)
Comp test using Sharktooth's eqm_v3ulr_rev31CD matrix, plus AQ & Q-PEL, I get 79.6 0.250[/b] which seems pretty optimal :)
BTW - I'm encoding Powder (1995). A few dark scenes, and rain here and there, but also very bright scenes and environments for a majority of the movie.
Backflip
18th September 2005, 14:56
Are there any problems with that logic^
jggimi
18th September 2005, 15:22
Are there any problems with that logic^I don't know, as I don't understand the post.
What's the definition of CQM? Configuration and Quality Management? Close Quarters Marksmanship? What do you mean?
The only percentages "too high to actually be utilised completely by the codec" are above 100%. 100% is the maximum bitrate obtainable with the current content and filter set. Best practice is to use a lower range. It varies, of course, but most people seem to aim for 60-80%. If I recall correctly, and I may not, AutoGK aims for 75%.
manono
18th September 2005, 22:17
Hehe, Custom Quantization Matrix.
Backflip, there's no way for us to tell if your settings are optimal. Judge with your eyes. In general, the lower the bitrate matrix, the higher the compress test percentage you want. And if you plan on ever playing your CQM encodes on a standalone DVD/MPEG-4 player, I guess you know that compatibility varies with the chipset and matrix used. That is, a lot of standalones don't play custom matrices very well.
jggimi
18th September 2005, 23:22
Thanks for the definition, manono.
Our best practice (60-80%) covers a very wide range of compression. There's more art than science in it, because visual quality is subjective and there are so many varieties of content and filtration, let alone quantization.
Backflip
19th September 2005, 07:24
I personally believe people generally see the same thing in terms of visual defects, but some acknowledge it as a problem and some don't. But we all see it ;) I do use my own eyes to judge, but was somewhat distracted reading in the guide that you should stay within "this" and "that" guideline when it really doesn't matter (it appears), this is why I brought up the subject because I had revisited the guide(s), recently. Mind you the current guide doesn't take into account CQM abilities.
I'm using Xbox Media Center to play movies, which has little to no problems playing CQM encoded video, thankfully.
Thanks for your help :)
vBulletin® v3.8.11, Copyright ©2000-2026, vBulletin Solutions Inc.