Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: "Target Quality" vs. "True VBR"? (Read 1800 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

"Target Quality" vs. "True VBR"?

Hi,
I've used Nero's encoder for at least 15 years and for the last decade with a mac.

I thought mac version of fb2k would evolve faster to the level where you could encode cd rips with Nero, but maybe I've waited too long.
I had used a decade old "ready built" wine version of fb2k, but I finally upgraded from 4 year old Mojave to Monterey and it's naturally not working.
Wine hasn't done much in the past 3 years, so...

What I have always liked about Nero, is that there is a clear choice about quality: "Target Quality".
Which sounds to me, that bit rate can be anything, as long as the quality is maintained.
I just like that. I don't care about bitrates, as long as they are not too high, so the music don't waste any storage space.

So, maybe I should abandon "Nero on a mac" and use something like XLD's qaac?
Where I instantly get concerned it that it is by name a VBR.
Which might mean that it noly focuses on quality with certain limits in bitrate.

Is this the case?
Are these encoders actually the same in real world concerning about the bitrate limits?

Re: "Target Quality" vs. "True VBR"?

Reply #1
Each encoder uses its own logic when it comes to making bitrate decisions in VBR mode. Apple's CoreAudio/qaac has performed better than Nero in listening tests at low bitrates, but when you use medium to high bitrates, both can easily reach transparency. There's really no such thing as "the bitrate can be anything", since the encoder must have some objective internal metric for making bitrate decisions. If you use TVBR 64 with qaac, the bitrate will average out to ~128 kbps over many tracks, but individual tracks may be anywhere from say 96 - 160 kbps. It's always possible to find outliers, but all VBR encoding methods tend to produce similar bitrates when you use the same quality setting across many files.

Re: "Target Quality" vs. "True VBR"?

Reply #2
Just checked about a hundred of old encodings and my bitrates maybe average on 170, usually between 160-180, some over 190 or under 120 kb/s.
Some songs are just simple and maybe the average per song is not interesting. I guess "the quality of compression" comes from the fluctuation of the compression inside one song.

Nero's q0.5 has been fine for me, I haven't really made any comparison with slightly higher settings with maybe my MDR-7506's.
With qaac, XLD shows 165kbps from 78 to 86, maybe those numbers give identical output?
And maybe in that area, the quality is the same than Nero's q0.5?

Re: "Target Quality" vs. "True VBR"?

Reply #3
CoreAudio has a weird quality scale from 0 - 127, but there are actually only 15 quality steps:

Code: [Select]
Q0 - Q4 (0) = ~40 kbps
Q5 - Q13 (9) = ~45 kbps
Q14 - Q22 (18) = ~75 kbps
Q23 - Q31 (27) = ~80 kbps
Q32 - Q40 (36) = ~95 kbps
Q41 - Q49 (45) = ~105 kbps
Q50 - Q58 (54) = ~115 kbps
Q59 - Q68 (64) = ~135 kbps
Q69 - Q77 (73) = ~150 kbps
Q78 - Q86 (82) = ~165 kbps
Q87 - Q95 (91) = ~195 kbps
Q96 - Q104 (100) = ~225 kbps
Q105 - Q113 (109) = ~255 kbps
Q114 - Q122 (118) = ~285 kbps
Q123 - Q127 (125) = ~320 kbps

TVBR 82 will produce close to 160 kbps, but comparing bitrates between Nero and CoreAudio is not very useful; if you want to compare the quality, you should take some familiar tracks from your library or some audio samples from public listening tests and encode them at various quality settings with both Nero and qaac. You can then do some ABX tests to determine the lowest VBR setting for each encoder that produces transparent results to your ears.


Re: "Target Quality" vs. "True VBR"?

Reply #5
I've participated in several Hydrogen Audio ABX tests and I also ABX'ed a lot for myself. As listening tests confirmed, there's no doubt that Apple AAC is clearly the best AAC encoder regarding quality at a specific bitrate. Nero AAC never really got close.

Some of my notes regarding Apple AAC bitrates of batches at least 100 albums converted:

TVBR: Q78 - Q86 = Q82 -> resulting range: 144-187kbps, average: 165kbps

CVBR:
--cvbr 128 kbps -> resulting range: 129-143kbps, average: 133kbps
--cvbr 144 kbps -> resulting range: 147-160kbps, average: 152kbps
--cvbr 160 kbps -> resulting range: 163-180kbps, average: 178kbps

TVBR Q64<<CVBR128<<TVBR Q73<<CVBR144<<TVBR Q82<<CVBR160<<TVBR Q91<<CVBR192

CVBR 144 is under most non-earbud listening circumstances transparent to me and only through concentrated listening seeking for artifacts I can distinguish the AAC track from lossless (mostly on electronic music). On earbuds one can target even a lower bitrate. Note that CVBR 144 is not selectable in foobar2000 Converter window so one has to set this value through Custom encoder setting.

If you were happy with a Nero average of 170 you'll be absolutely happy too with TVBR Q82 and maybe even with CVBR 144.

Re: "Target Quality" vs. "True VBR"?

Reply #6
Thanks all for your insights!

What I'm still wondering is, if NeroAAC has more or less "range" than AppleAAC?
With Q82 Apple's range is 144-187.
How much is it with Nero's q0.5?

Is there any charts on this?

 

Re: "Target Quality" vs. "True VBR"?

Reply #7
Apple default without parameters is around 190 vbr, so not far from nero.
wavpack -b3.63hhcs.5

Re: "Target Quality" vs. "True VBR"?

Reply #8
The VBR range doesn't matter as much as the encoder tuning. Apple beats Nero in listening tests across the board, so you really shouldn't be using Nero at all anymore.

Re: "Target Quality" vs. "True VBR"?

Reply #9
Like already mentioned pretty much... Apple AAC is what you want if your using AAC(AAC-LC) in general. even for a alternative to Apple AAC is to use the FhG from WinAMP in Foobar2000 since it's similar to Apple AAC and, if I recall correctly, encodes faster if that matters to you. basically this listening test at 96kbps from 2011 sums up how close Apple AAC and FhG (WinAMP) are... https://listening-tests.hydrogenaud.io/igorc/aac-96-a/results.html ; with scores like that I doubt most people would notice anything and even if they can it won't be easy to spot because once someone has to really focus to spot tiny things, you already know the overall sound quality is strong.

as for general bit rates... I tend to see it like this on most equipment (like speakers and typical headphones) in that 96kbps (CVBR) is going to be at least 'good enough' and it's very efficient. I could maybe see 128kbps or so as a bit of a safety buffer but much beyond this lossy encoders lose their appeal if you ask me. because while I get most around here have a tendency to obsess over tiny sound differences, in the real world, they are likely quite small to where the average person ain't going to notice it. even if they do, it's not something you will likely notice in general when just sitting back and enjoying ones music.

because I figure if someone does notice the smaller stuff and it bothers them, at that point, if possible, they are better off using a lossless format like FLAC since I think the whole point of lossy audio is to retain a high percentage of the sound quality but at pretty much smallest possible file size that will suit many, if not most, people well.

p.s. in regards to TVBR vs CVBR... CVBR might be a touch safer than the default TVBR on sound quality if one has to choose, but even here we are splitting hairs. but a slight negative with CVBR that favors TVBR overall is it seems CVBR tends to have, on average, bigger file size vs TVBR. so this is a really small thing that many might not care about. but if you do, then you can make a choice here if you would rather have the smaller file size(TVBR) or 'maybe' a touch better sound quality(CVBR). but when using 96kbps, since I feel this is the lowest bit rate I would suggest people use on AAC(AAC-LC), I tend to be on the side of caution with sound quality and opt for CVBR. but higher bit rates (say 128kbps or more) one might as well opt for the slight storage space savings with TVBR.
For music I suggest (using Foobar2000)... MP3 (LAME) @ V5 (130kbps). NOTE: using on AGPTEK-U3 as of Mar 18th 2021. I use 'fatsort' (on Linux) so MP3's are listed in proper order on AGPTEK-U3.

Re: "Target Quality" vs. "True VBR"?

Reply #10
I'd go for 192 for all codecs. Because there is a very long history of usage and tunings. It works
on VBR, CVBR, ABR and CBR as bitrate is high enough.

Sizewise, Its right there in the middle - it feels just right.

Nero, Apple, MPC default to 176 ~ 192.  Lame VBR was originaly tuned around 192 .
If you your desperate for smaller files try towards 160k, If you want to push it in the other direction,
then 224k.
wavpack -b3.63hhcs.5

Re: "Target Quality" vs. "True VBR"?

Reply #11
When Vorbis came out, in my ignorance at the time (and until much later), it's then novel vbr-based "quality" setting led me to think, and I guess most people, including the OP here, that what it did was aim at some sort of target perceptual transparency level (before I even knew to use such fancy technobabble), towards which goal it would dedicate as much or as little bandwidth as strictly required for each part of the audio track. This seemed to mean anything between 0kbps and whatever the top was for the format.
It wasn't until Opus arrived with its -b setting (which, at that point, seemed to like a weird return to "the old handfisted bitrate settings of the past"), that someone corrected me explaining that Vorbis' -q setting was also an abstraction, so to speak, of a set of constraints around bitrate. I was told that this perception of an absolute perceptual quality target didn't in fact exist, and that all that -q# did was an approximation based on tuned bitrate allocations and weights and other techniques.
That sure bursted my bubble. Everyone and his dog back then were spreading my same perception about Vorbis' "quality" based encoding: "Choose a quality level that sounds right to you and stick to it so that any source material you throw at it sounds exactly as good. The encoder will take care of using just the necessary ammount of bits to achieve that target quality. qX will always be qX and the improvements are about improving bandwidth efficiency to reach the same target quality"

To be honest, It's kind of frustrating that this mythical "quality" setting doesn't actually exist. It seemed so neat and... oh well.
Quite removed from the actual reality of lossy codecs, where it's the perceptual efficiency that's improved upon for different bitrate ranges.