Now that you found your lowest transparent setting...

Topic: Now that you found your lowest transparent setting... (Read 54102 times) previous topic - next topic

0 Members and 1 Guest are viewing this topic.

Re: Now that you found your lowest transparent setting...

Reply #75 – 2016-12-11 19:05:46

Quote from: Satellite_6 on 2016-12-11 18:36:13

Lossless, because lossy is not consistently transparent for me at any bit rate.

Since you have golden ears, it would be a shame if you didn't help out with public listening tests.

Re: Now that you found your lowest transparent setting...

Reply #76 – 2016-12-11 20:27:26

I use FLAC whenever possible, at home either (mostly) 16/44.1 (CD version) or 24/48 if that true (not upsampled) version is available (rarely available on web stores or it comes from my own digitization of analog tapes). For mobile 16/48 (16/44.1) FLAC, dither with modified-e-weighted noise shaping. For playback on old minitower which supports only MP3 I use CBR 320 kbps or VBR V0.

I know that is probably overkill regarding threshold of practical audibility, but I do not want to skate on the border of audibility (neither pursue craving for XXX kHz or DSD, however). Rather I prefer formats that are practical for use and offer robust container for audible spectrum. The above is what I have great experience with and does not disappoint me at all on my favourite records. In this sense it is transparent for me.

Re: Now that you found your lowest transparent setting...

Reply #77 – 2016-12-13 04:44:04

Since I've actually done ABX testing and confirmed that for demanding music in my collection 128k is almost transparent, I've elected to go a notch higher and have never felt that I'm hearing anything encoder-related during listening (as opposed to testing).

Re: Now that you found your lowest transparent setting...

Reply #78 – 2016-12-13 08:29:26

I ditched FLAC, since I couldn't hear a damn difference between that and MP3 anyway.

I buy a lot of music on Bandcamp, and their standard download format is V0 MP3, so that's what I've settled on when I rip CDs as well.

Re: Now that you found your lowest transparent setting...

Reply #79 – 2016-12-13 08:43:39

The difference can be very subtle and only on some tracks but I just did this listening test FLAC vs 320 kbps MP3. I did a direct conversion from FLAC (CD 16/44.1) to MP3 using foobar GUI and lame.exe. Not a strict proof (score 11/16), but I still prefer FLAC to MP3 when possible.

foo_abx 2.0.2 report
foobar2000 v1.3.13
2016-12-13 09:41:30

File A: music.mp3
SHA1: f535306b9fb103b219d433c51b56ad0556deb574
File B: 13 Gvendolína.flac
SHA1: 3e457cced230fac33725c499b6ef8af8b9c6f686

Output:
WASAPI (push) : Reproduktory (Sound Blaster Audigy Fx), 24-bit
Crossfading: NO

09:41:30 : Test started.
09:42:14 : 01/01
09:42:25 : 01/02
09:42:30 : 02/03
09:42:39 : 03/04
09:42:42 : 04/05
09:42:46 : 05/06
09:42:52 : 05/07
09:43:03 : 05/08
09:43:06 : 05/09
09:43:11 : 06/10
09:43:13 : 06/11
09:43:16 : 07/12
09:43:24 : 08/13
09:43:26 : 09/14
09:43:29 : 10/15
09:43:32 : 11/16
09:43:32 : Test finished.

----------
Total: 11/16
Probability that you were guessing: 10.5%

-- signature --
4f6fa72dc3412870e27cd438d2e4f69fa00646c7

Re: Now that you found your lowest transparent setting...

Reply #80 – 2016-12-13 15:42:10

5% is the typical threshold for acceptance, but at least you provided something, unlike the rest of the hit-and-run, self-professed golden ears.

Re: Now that you found your lowest transparent setting...

Reply #81 – 2016-12-13 16:11:15

I know that is why I wrote that is not a strict proof, moreover as I wrote I am not that good in abx testing, those scores 11-12 I can achieve on tracks where I suspect a difference typically only after longer break e.g. In the morning and then taking the test once. I am less successful if I were asked to repeat this score e.g. 5 times in a row. Similar results i obtained when I tried some more aggressive noise shapers against dithering without noise shaping (but in noise shaping case i kept only that log, not those files anymore).

But when the tracks are really the same "digitally", i achieve usually 7-9, e.g. flipping a coin. So maybe I can perceive some difference.

Still I consider 320 or V0 as acceptable when space or flac compatibility is an issue. I chose for this test an album/song which I know that has a lot of drums and cymbals and that I remembered feeling much better when listening from CD than from MP3. Large majority of MP3 is completely ok and I cannot distinguish them from original.

Re: Now that you found your lowest transparent setting...

Reply #82 – 2016-12-13 16:25:28

A lot of words to explain inadequate proof.

Quote from: jumpingjackflash5 on 2016-12-13 16:11:15

I am less successful if I were asked to repeat this score e.g. 5 times in a row.

It also appears you need to review the information here:
https://hydrogenaud.io/index.php/topic,16295.0.html

Additional conversation here will be binned for being off-topic.

Re: Now that you found your lowest transparent setting...

Reply #83 – 2016-12-13 16:39:59

Not using Flac or mpc/apple/ogg/opus. Using Winamp AAC instead since after redoing a test with my more demanding music it sounds great at 256kbps VBR. As for people wondering why i switch so much in the past?, My collection is large so testing is hard & being with a 64gb micro sd didn't help when i found the sweet spot was 256k with FHG.

The rest have track's that have artifacts at even 320k or need to be manually set to 350k+ while I've never had that problem with FHG AAC.

Re: Now that you found your lowest transparent setting...

Reply #84 – 2016-12-13 16:46:33

Since you have golden ears, it would be a shame if you didn't help out with public listening tests.

Re: Now that you found your lowest transparent setting...

Reply #85 – 2016-12-13 17:09:17

Quote from: greynol on 2016-12-13 16:46:33

Since you have golden ears, it would be a shame if you didn't help out with public listening tests.

That kinda dawned on me as i was typing that post out, Might try out if another one happens or just do random one's from a few in my collection.

Re: Now that you found your lowest transparent setting...

Reply #86 – 2016-12-13 19:00:01

Quote from: doom-drone on 2016-12-13 17:09:17

Might try out if another one happens or just do random one's from a few in my collection.

Don't be surprised if you fail miserably with ABX tests comparing high quality lossy to lossless despite being absolutely certain that you won't. Nearly everyone does. It's quite normal.

Re: Now that you found your lowest transparent setting...

Reply #87 – 2016-12-13 19:33:47

Quote from: Slipstreem on 2016-12-13 19:00:01

Quote from: doom-drone on 2016-12-13 17:09:17
Might try out if another one happens or just do random one's from a few in my collection.
Don't be surprised if you fail miserably with ABX tests comparing high quality lossy to lossless despite being absolutely certain that you won't. Nearly everyone does. It's quite normal.

Thanks for the heads up.

Re: Now that you found your lowest transparent setting...

Reply #88 – 2017-07-02 20:27:44

I fail ABX'ing at 192k with FHG AAC but i picked 256k for safety reasons and i have 200gb micro SD space anyways.

Re: Now that you found your lowest transparent setting...

Reply #89 – 2017-07-04 14:30:22

Slightly OT:

For people already using lossless at home:

Abx is important as just a scientific experiment but can turn out a waste of time and space on portables and practical living. The main priority is bandwidth conservation while preserving some fidelity .
Theres no need for 'safety' -You already have 100 % insurance from the lossless archive. Of course many don't want very bad quality either so ther needs to be a balance. Why should abx have much do with anything when you already listen lossless at home ? Anyway the poll results show it doesn't affect anything when it comes to portables, perhaps makes things worse, makes people even more worried, paranoid and use excess 'safety' settings ... For users of *only* lossy its a different story.

Consider:

For big sdcards and limited collections / requirements just use lossless and forget transcoding and abx altogether.

If transcoding consider modern (non-mp3) codec at 100k or less by not using 'full transparency' as the *only* criterea.

For older mp3 players try 96k cbr/abr resampled to 32khz.

Re: Now that you found your lowest transparent setting...

Reply #90 – 2017-10-24 11:12:41

After reading these posts i am surprised a fair amount of people use 320kbps MP3 as it seems like a waste of storage space because assuming your device can play AAC your better off, if your the paranoid type, to use 256kbps TVBR(or CVBR) Apple AAC and be done with it as that will still maintain high quality for the paranoid types and will save a fair amount of storage space to over 320kbps MP3. even Apple AAC @ 192kbps average (or so) is a pretty strong quality setting (given the ABX test linked to below from Jan 2012).

for me personally...

-Computer = FLAC or lossy variations

-Portable = Apple AAC @ 128kbps (of which i recently changed from LAME v2 (190kbps average) as sound quality is easily good enough for me and saves plenty of storage space especially when you got a 16GB MicroSD card to work with and was getting a little tight for space it's a way to get about 33% of your storage space freed up which is a lot when talking around 14.5GB or so and encoding stuff @ 128kbps. basically frees up GB's of space with 128kbps vs 190kbps. so unless you got a boatload of storage space it's not worth using higher than 128kbps (for Apple AAC) for portable use. but if i had space to burn, to where i would not even be close to using it up, i 'may' consider using 192kbps or 256kbps for good measure as it seems like no one complains (well next to no one) about sound quality @ 256kbps on Apple AAC. but even with that said, i might still go with 128kbps because since i keep a backup copy on my hard drive i would rather just keep the extra GB's of storage space for other stuff as i am already using enough of it for FLAC already.

I think Apple AAC @ 128kbps is a good setting for most users and still keeps file size minimal (and you can always re-encode from FLAC to a high bit rate if you ever needed to). like a solid balance between sound quality/file size as sound quality scores still maintained a 4.00+ average according to this test from Jan 2012... https://hydrogenaud.io/index.php/topic,102875.msg846373.html#msg846373 ; i figure anything around that 4.00 range (or better) should be good enough in most music for most people especially if those 4.00 range scores are done by people with 'golden ears'(i.e. above most people) which makes it all that more likely that 4.00 range score would be even higher for many people and would basically be transparent or not far from it for many(probably most?) people. also, according to that test... while 192kbps further cleaned up the sound, as it's in the high 4's for the average, it also used 50% more bit rate to do it over 128kbps. that's why i feel 128kbps (or so) is the sweet spot of Apple AAC as it seems to have a nice balance of sound quality/storage space. plus, i suspect after a certain point... you got to be really focusing to notice some artifacts and when your just sitting back enjoying your music you most likely ain't going to be focusing too much on artifacts etc and even if you can notice a little blip here and there it's probably not to the point it detracts from your overall enjoyment of the music.

as for the general poll... i voted for "One step higher quality than N". because while i don't know my exact ABX point, as i never thoroughly tested too much of my music, i would figure 96kbps is close either way and 128kbps is not all that much larger and i feel gives a decent safety margin for me and, like i already mentioned, i feel 128kbps Apple AAC is a fairly safe setting for most people in general as it's not like if someone was hearing music played @ 128kbps will notice anything obviously negative.

NOTE: but i guess it depends on how one defines that 'one step higher than N' because looking at Foobar2000's Apple AAC settings when using that slider bar to adjust Apple AAC quality and assuming i use 96kbps as my measuring stick, then there is actually two bumps to get to the 128kbps setting (i.e. 96 > 112 > 128). but close enough as a 32kbps difference is still close enough to one step

p.s. as for personal ABX testing... i have not spent a lot of time on them but i can say with both Opus/Apple AAC that once i hit the 64kbps setting (with MP3 it was v7 (100kbps average) for equivalent if i recall correctly) it's no longer easy for me to ABX without much effort which is why i feel 96kbps would even be good enough for me but i just play it a bit safer and use 128kbps as i figure it's not worth getting that extra 25% (or so) disc space savings (like from 128kbps to 96kbps) unless your really straining for storage space but i can't really see myself using Apple AAC under 96kbps setting in general as if i had to go under that i would simply buy more storage space especially considering prices of 32-64GB MicroSD cards right now are around $15-25 which would double or quadruple the storage space i have now in my Sansa e250 v1.

@Kasi1560k

Quote

I fail ABX'ing at 192k with FHG AAC but i picked 256k for safety reasons and i have 200gb micro SD space anyways.

According to IgorC, he says there is no AAC encoder better than Apple AAC @ 96kbps or higher and he mentioned this earlier this year and even recently. i just mention him as he seems like one of the more respected members around here.

I think the only advantage of FhG is it encodes files faster and there might be a edge at low bit rates (i.e. under 96kbps).

but with that said... your already using paranoid high bit rates (basically 256kbps or so on up) so i doubt anything i said here really matters.

Re: Now that you found your lowest transparent setting...

Reply #91 – 2017-10-24 14:15:28

Well, a few years ago I checked AAC quaility using Apple and Winamp FhG. My ears are anything but golden which means for instance usually Lame 3.99.5 -V5 mp3 is absolutely fine to me. But I'm pretty sensitive to certain tonal problems which can be present with practical listening. As for that I realized that FhG AAC (Winamp version) was a little bit (not much) better than Apple AAC at that time.
It took a ~170 kbps setting to make the hard tracks become transparent.
But I admit that the differences were very small, and for a bitrate like 96 kbps on average things may be different.

For a practical choice I think you can simply behave like this (it was the opus developer presumably who for the most part made me think like this):

a) ignore killer samples which are not relevant to you or which you can easily treat in a special way
   (this covers for instance harpsichord music which - in case it should be relevant - can be easily identified and used with a high bitrate setting)
b) if you like to have the overwhelming part of your music be encoded transparent or very close to that
   - use a ~128 kbps setting for mp3 (and a good mp3 encoder like Lame 3.99.5)
   - use a ~96 kbps setting for AAC (and a good encoder like Apple AAC)
   - use a ~80 kbps setting for Opus (and the current opus encoder).
   Qualitywise I'd prefer Opus among these codecs.
c) If ypu like to have presumably all of your music transparent except for a)
   - use a ~200 kbps setting for mp3 (and a good encoder like my lame3995o variant)
   - use a ~170 kbps setting for AAC (and a good encoder like Winanp FhG or Apple AAC)
   - use a ~140 kbps setting for Opus (and the current opus encoder).
Details can very a bit according to your mileage. But I think this is an appropriate guidance.

Re: Now that you found your lowest transparent setting...

Reply #92 – 2017-10-24 14:16:48

Sorry, wrong edit.

Re: Now that you found your lowest transparent setting...

Reply #93 – 2017-10-24 14:47:33

Quote from: halb27 on 2017-10-24 14:15:28

For a practical choice I think you can simply behave like this (it was the opus developer presumably who for the most part made me think like this):

a) ignore killer samples which are not relevant to you or which you can easily treat in a special way
(this covers for instance harpsichord music which - in case it should be relevant - can be easily identified and used with a high bitrate setting)
b) if you like to have the overwhelming part of your music be encoded transparent or very close to that
- use a ~128 kbps setting for mp3 (and a good mp3 encoder like Lame 3.99.5)
- use a ~96 kbps setting for AAC (and a good encoder like Apple AAC)
- use a ~80 kbps setting for Opus (and the current opus encoder).
Qualitywise I'd prefer Opus among these codecs.
c) If ypu like to have presumably all of your music transparent except for a)
- use a ~200 kbps setting for mp3 (and a good encoder like my lame3995o variant)
- use a ~170 kbps setting for AAC (and a good encoder like Winanp FhG or Apple AAC)
- use a ~140 kbps setting for Opus (and the current opus encoder).
Details can very a bit according to your mileage. But I think this is an appropriate guidance.

Those seem to be quite realistic guidelines - specially their first half, considering what you said about said Opus developer's recommendations of using higher bitrates for problematic tracks only.

I wonder if they'd ever make into our wiki somehow.

Re: Now that you found your lowest transparent setting...

Reply #94 – 2017-10-24 14:50:24

It's just my two cents, @ThaCrip , but do all your replies have to be so lecture-like, given their sheer size?

Re: Now that you found your lowest transparent setting...

Reply #95 – 2017-10-25 11:11:57

@halb27

Quote

Well, a few years ago I checked AAC quaility using Apple and Winamp FhG. My ears are anything but golden which means for instance usually Lame 3.99.5 -V5 mp3 is absolutely fine to me. But I'm pretty sensitive to certain tonal problems which can be present with practical listening. As for that I realized that FhG AAC (Winamp version) was a little bit (not much) better than Apple AAC at that time.
It took a ~170 kbps setting to make the hard tracks become transparent.
But I admit that the differences were very small, and for a bitrate like 96 kbps on average things may be different.

If the differences were very small (like you talking 5kbps-ish or less?) would that mean it's possible for another person to feel differently? ; like that 'too close to call' sorta thing from a quality standpoint. like where IgorC's claim of "No AAC encoder is better than Apple AAC @ 96kbps and higher" is still correct (especially for an average) even though what you said is not wrong either as your talking more of a specific thing (and you even said it was 'very small') where as i imagine IgorC when he made that claim is looking at the wide data available. just some random thoughts on this little thing.

also, it appears the Winamp FhG encoder has not been updated in a while (the files needed to encode FhG in Foobar2000 are dated Dec 2013) and i would imagine Apple AAC has not been updated, in terms of sound quality adjustments, in years either(?). so assuming that's correct things likely have not changed since you last tested.

also, i see your 170kbps setting for AAC recommended for basically all music being transparent outside of killer samples. but... at least in Foobar2000, with the Apple AAC encoder there is no strict 170kbps setting as it goes from 160kbps to 196kbps. basically when you start from 64kbps it increases by 16 in each quality setting (i.e. 64/80/96/112 etc) with the slider all the way up to 160kbps and after that it jumps by 32kbps with each raise in quality with the quality adjustment slider til it hits the max setting which is 320kbps.

so with that said... would the 160kbps setting qualify for 170kbps in AAC(Apple etc) and similar to LAME v2 (190kbps) comparing to his 200kbps suggestion? ; because correct me if i am wrong, but when your setting is only 8-10kbps away from what he mentioned it seems like the encoder, on more complex music, will likely shoot up the average bit rates to those 170kbps for AAC and 200kbps for MP3 when needed. hence, i can use the 160kbps(Apple AAC) and 190kbps(MP3) settings and would be within his general settings.

as for your general guideline that you got from a Opus developer with the A/B/C stuff... it appears Apple AAC 96kbps is better than LAME v5 according to this listening test from July 2014... http://listening-test.coresv.net/results.htm ; but i get the general point in that 96kbps Apple AAC and LAME v5 is close enough in quality and he used those settings for THE minimum for quality sound on a lot of music.

but anyways, with the whole A/B/C stuff... to comment more specifically on your B and C general guidelines, i can't really fault that as it appears with 'B' the Opus dev was selecting THE lowest setting where a lot of music is transparent or close to it and with 'C' he more or less found a MINIMUM setting for the more paranoid types of people. sorta like sweet spots you could say for two groups of people.

but it makes me wonder... if Apple AAC @ 96kbps scored that well (i.e. 4.40 average (and only three samples went below a 4.00)) in those July 2014 listening tests, how much better would Apple AAC @ 128kbps be? ; because unless there is barely any difference it seems like you would be not too far from perfection(since i imagine that 4.40 average of AAC 96kbps would shoot up a fair amount with Apple AAC 128kbps), outside of some killer samples, in which case 170kbps range would be mostly a waste of bitrate unless your more of the paranoid type(?). but, since i already mentioned, that Opus dev seems to have made his choices around THE minimum quality setting for more of the common person (option B) and then the more paranoid type (option C) then i can't really fault it.

so given that info... that's kinda why i suggested the 128kbps setting as the sweet spot for Apple AAC in general as it gives a minimal file size, while not as minimal as the 96kbps setting, might clean up the sound a bit better (like raise that 4.40 average score a fair amount) without spending all that much more bit rate and in the big picture 128kbps is still a pretty small file size as i don't know exactly where people draw the line of what's considered 'high' or 'low' bit rates, but, as a ball park figure, lets say 128kbps qualifies as either the lowest point of a high bit rate or the highest point of a low bit rate (even if opinions vary i got to be close) and that just shows you it's still a pretty minimal file size and considering it's sound quality/file size i feel it's difficult to justify using higher bit rates, especially outside of the occasional problem sample or those more golden ear-ish types of people. NOTE: i believe read around here not long ago that some Opus dev (it may have been someone else) considered 128kbps 'a high bit rate'(but they might have been referring to Opus specifically as i can't recall off the top of my head right now), but it's probably not surprising since that encoder tends to focus more in "lower bit rates" and has already achieved quality sound at 96kbps(if not 80kbps like you and the Opus dev you mentioned suggests).

also, it makes me wonder how Opus @ 80kbps compares to LAME v5? ; would it be fair to say, given what i can tell in that July 2014 listening test, that Opus @ 80kbps/Apple AAC @ 96kbps/LAME @ 130kbps(v5) are about the same in sound quality?

but with all of that said... i imagine he (the Opus dev you mentioned) formed his opinion on those core things based on what he observed with people in general out there? ; like with more of the common-ish type of person vs the more 'golden ear'-ish/paranoid type of person.

------------------------------------------

@includemeout

Quote

It's just my two cents, @ThaCrip , but do all your replies have to be so lecture-like, given their sheer size?

I don't know, I guess it's just sorta the way i talk on certain subjects etc. I don't mean anything by it. you can just skip reading most of my post if you want to, no offense or anything. like just skimming through my post above will give you the gist of my thoughts.

NOTE: I actually read this entire topic before I made a post as it's kinda cool just to see what others do and how they approach things etc

Poll

For use at home: