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: Most efficient lossy codec for classical music (Read 7789 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Most efficient lossy codec for classical music

Hello everyone,

to give you some context, let me say I am not per se an audiophile, but I think I have very decent musical ears and high quality headphones (V-Moda Crossfade M-100), that I use for listening for many hours a week.

To the best of my knowledge opus is the best lossy codec out there at the time. According to various listening tests it is fairly close to transparency already at vbr 96 (~107 kbps) [http://listening-test.coresv.net/nonblocked_means_all.png].
That is why I use it for storing (and playback) of my music collection on my mobile phone. Depending on the source material I use between 110 and 130 kbps opus 1.1.0 from a lossless source. For almost all genres that "sounds transparent to me".
Now I have read that since opus uses "non-standard bit allocation", which means that it uses a comparably higher bitrate for quiet, calm, "simple" music than pop, rock, hip-hop, reggae, ... for the same quality - an effect that I have noticed myself before reading about it.

So I have come to wonder whether I should use some AAC encoder or something completely different for classical music. Especially my nine symphonies collection of Beethoven, that I got from hdtracks in 24/96. VBR 130 sounds really good to me there, but it does not blow me away like some Rolling Stones, The Who or Doors Remasters do. Might it be the mastering (classical music is recorded at an opera house and not digitally mixed together from indiviual tracks due to resonance phenomena I guess ?!) or is that due to the codec?

Any advice is welcome. Thank you very much!

Markus

Re: Most efficient lossy codec for classical music

Reply #1
When you say it does not blow you away, do you mean the Opus files compared to the source files?
If so, did you try to ABX at least one set of those? Have you compared it to AAC at the same effective bitrate?

Whether something is recorded or mixed together digitally shouldn't matter.
"I hear it when I see it."

Re: Most efficient lossy codec for classical music

Reply #2
Well, I do not have the source files available right now. I would have to redownload them (~15GB) and do quite extensive tests as you suggested. I find it really hard personally to, say, first listen to the source file and then the lossy one and exactly tell the difference. It is just the feeling that I get at times - "Wow, this sounds amazing" audio-quality wise. I know that that sounds foolish, but it's true ;)

Regarding the mixing - I am not an expert -, but shouldn't there be a difference, if there are like 10 microphones in the opera house or it is done track by track in the studio like almost every contemperary artist does it. As I said, I am pretty positive it would, because especially the kettledrum sounds different in those recordings - just like the real thing. I also fancy myself hearing the resonance and "volume" and also the superposition of the various instruments, that make it sound more like it is live.

Re: Most efficient lossy codec for classical music

Reply #3
A thing to note when talking about amazing sound is to keep the ToS in mind.

Of course different recording techniques will produce different sound, but the codec does not care about that. Some signals are just harder to transparently compress than others, regardless of how they were produced.
"I hear it when I see it."

Re: Most efficient lossy codec for classical music

Reply #4
Unless you are very short on storage space, stop worrying about the bitrate. It is just a number of bits that the encoder decided to apply to meet a certain quality target.

Instead, find the quality setting on your encoder of choice that provides audibly transparent results to you.

Re: Most efficient lossy codec for classical music

Reply #5
Hi, I listen a classical music from time to time. Recently I bought a Kathleen Battle's "a Portrait" album. I did a rip of my cd to flac, and I noticed that the average bitrate of the songs do not exceed ~600 kbps. It seems that the spectral consistency isn't so big, and the encoder has decided to use less bits to interpret the music. The same album converted to Ogg Vorbis Q7 which is my favourite level and means ~224 kbps, gives an average bitrate ~180kbps and about 220kbps maximum bitrate. At this level I couldn't find any difference from the reference Flac file.

Re: Most efficient lossy codec for classical music

Reply #6
Unless you are very short on storage space, stop worrying about the bitrate. It is just a number of bits that the encoder decided to apply to meet a certain quality target.

Instead, find the quality setting on your encoder of choice that provides audibly transparent results to you.

If storage space isn't an issue then why not just stick with the lossless originals?

I only use lossy encoded files where storage space is at more of a premium. Much like the OP, I use opus on my phone, though at 96kbps to pack as much essentially transparent (to me) music on the device as I can. If it was all the original FLAC files I wouldn't be able to carry my entire collection with me. Even if I went completely overboard and encoded at 256kbps or higher I couldn't fit my entire collection on my phone. Opus, and Vorbis before it, pushed that effectively transparent bitrate threshold low enough to make it possible.

I don't worry about finding the best encoder for each specific type of music in my collection. I use the one which seems to do the best overall and use that. It's a personal preference but I like having my collection in a uniform codec. Maybe 20% of my collection is orchestral in nature so any storage savings I'd gain using specific codecs for it would be barely noticeable and outweighed by the encumbrance of multiple encoding steps and all the other things that would impact.

I will say that I just ran a simple and dirty test encoding a Brahms track and a Chimaira track using opus and vorbis. Opus encoded the Brahms track at roughly 100kbps where vorbis did it at 80kbps. Conversely opus encoded the Chimaira track at 80kbps and Vorbis did it at 101kbps. It may be that opus has an Achilles heel when it comes to orchestral music, but that's hard to state with such a small sampling.

For the record I used the opus and vorbis implementations which ship with Linux Mint 17.3 for the above tests:

libopus0:amd64                    1.1-0ubuntu1.2
libvorbisenc2:amd64                1.3.4-2~trusty1

Re: Most efficient lossy codec for classical music

Reply #7
If storage space isn't an issue then why not just stick with the lossless originals?

True, given enough storage space and bandwidth, lossless is a good choice. And you'll be 100% sure there is absolutely no degradation in sound quality due to the codec.

I used to keep all of my rips in FLAC, but I've reasoned with myself that since I can't hear any difference between lossless and MP3 at a decent bitrate, there really isn't any reason to keep the lossless files. I have all the CDs in storage, otherwise it would be illegal for me to keep the files. It's not like I can hear a difference between audibly transparent MP3 and audibly transparent Opus/Ogg/AAC anyway. The only reason to move to a new lossy codec would be to save a bit of space, and unlike the alternatives, literally everything supports MP3.

It's not the best, but it's more than good enough.

Re: Most efficient lossy codec for classical music

Reply #8
A thing to note when talking about amazing sound is to keep the ToS in mind.

Since I am fairly inexperienced in audio fidelity stuff, I don't know what the acronym ToS means and google didnt do the job. But I suppose you mean some sort of suggestive psychology. I listen to a 96/24 HDTracks-Master, so so it's go to be great, although I couln't distinguish it from a lossy version in a double blind test.

Of course different recording techniques will produce different sound, but the codec does not care about that. Some signals are just harder to transparently compress than others, regardless of how they were produced.

What I meant was, me not being "blown away" by the lossy version of the HD-Master might be due to the fact, that those symphonies are produced differently than the HD-Recordings I am used to, that are made in the well defined roam of a recording studio.

So many people have rightly suggested I should stick with the lossless version all along, but unfortunately storage is an issue for me, especially on my mobile phone. As I mentioned those nine symphonies alone have around 15 GB (I would need to check).
Now I know that I could downsample those files to 48/16 and there would be no difference. Here [https://xiph.org/~xiphmont/demo/neil-young.html] is an excellent and in-depth article, that explains that no audible information in form of supersonic superposition or else is lost there and the only reason why HDTracks, SACD's, etc. sound better then normal CD's, is that the (re)mastering is done much more carefully in most cases. To some it might seem strange to go for such a high-quality file in the first place and then "downcode" it to 130 kbps, but that is why I do it.
Still those 48kHz/16bit flacs would leave me with ~5GB, if that math is correct, which would be too much for all my music to fit on my phone.

@KozmoNaut This is probably pretty basic, but how do you use quality settings for opus? I use TAudioConverter through wine on a Ubuntu 14.04 LTS system and it only offers VBR, CVBR and CBR. I would much prefer a GUI.

Re: Most efficient lossy codec for classical music

Reply #9
A thing to note when talking about amazing sound is to keep the ToS in mind.

Since I am fairly inexperienced in audio fidelity stuff, I don't know what the acronym ToS means and google didnt do the job. But I suppose you mean some sort of suggestive psychology. I listen to a 96/24 HDTracks-Master, so so it's go to be great, although I couln't distinguish it from a lossy version in a double blind test.

Terms of Service. The thing you agreed to follow when you joined the site.  There is a link to it at the top of this page if you want to review it.

Still those 48kHz/16bit flacs would leave me with ~5GB, if that math is correct, which would be too much for all my music to fit on my phone.

For a mobile phone I'd just use MP3 transcodes (or whatever) and keep the original lossless files on PC.

Re: Most efficient lossy codec for classical music

Reply #10
@saratoga I have to admit, that I did not read every line of the terms and conditions and I am aware that it's bad for every forum, if too many people start posting spam and advertising products.
However I hope it is fairly evident, that I am not trying to plug any products, but asking a genuine question. I just think, that it is better for the reading flow (so more people reply) to not engage in artificially formal speak in order to be as neutral as possible and also it may be helpful to suggest, that I am not a complete nut, who is listening to a bad master on shitty headphones and therefore there wouldn't be any issue with the opus codec at all. Further I believe that people, who spend time reading about audio-codecs are quite aware what different ways there are to spend their money on music.
That doesn't make me not complying with those rules ok and I will try not "violate" any terms and conditions on this site in the future, but I hope you get my point.

Re: Most efficient lossy codec for classical music

Reply #11
It is expected that you read and follow our rules before posting.

Here's a link:
https://hydrogenaud.io/index.php/topic,3974.html

So that you're clear, he's not talking about rule #14; he's talking about rule #8, to which the following post does not conform:
Well, I do not have the source files available right now. I would have to redownload them (~15GB) and do quite extensive tests as you suggested. I find it really hard personally to, say, first listen to the source file and then the lossy one and exactly tell the difference. It is just the feeling that I get at times - "Wow, this sounds amazing" audio-quality wise. I know that that sounds foolish, but it's true ;)

Regarding the mixing - I am not an expert -, but shouldn't there be a difference, if there are like 10 microphones in the opera house or it is done track by track in the studio like almost every contemperary artist does it. As I said, I am pretty positive it would, because especially the kettledrum sounds different in those recordings - just like the real thing. I also fancy myself hearing the resonance and "volume" and also the superposition of the various instruments, that make it sound more like it is live.


Re: Most efficient lossy codec for classical music

Reply #12
@KozmoNaut This is probably pretty basic, but how do you use quality settings for opus? I use TAudioConverter through wine on a Ubuntu 14.04 LTS system and it only offers VBR, CVBR and CBR. I would much prefer a GUI.

With Opus you have to specify a target bitrate.

I was thinking of encoders like Lame and Oggenc, where you specify a quality value that has no direct relation to a specific bitrate.

Re: Most efficient lossy codec for classical music

Reply #13
I used to keep all of my rips in FLAC, but I've reasoned with myself that since I can't hear any difference between lossless and MP3 at a decent bitrate, there really isn't any reason to keep the lossless files.

I needed to go back and re-rip some old CD's and found a few of them now had problems. I could actually see thru my Aja CD.
Glass half full!

Re: Most efficient lossy codec for classical music

Reply #14
Take a look at Nick.C's signature.
BTW, he is the awesome guy behind LossyFLAC.

Re: Most efficient lossy codec for classical music

Reply #15
The best lossy options you have are AAC, Vorbis, Opus, and our good ol' friend MP3. First off, you should preferably stay away from MP3 for complex music such as your choice, not that MP3 is bad, but because there are better (if ever so slightly) options available.

I don't listen to classical/orchestral music, so can't tell if AAC/Vorbis/Opus would be good for you, but I'll share my perception of these codecs with my rock music library.

The best sound I get is from iTunes AAC TVBR encodes. I haven't been able to ABX a 256 Kbps file from lossless FLAC on any track I've tried. To be honest, AAC encodes are transparent to me at 128 Kbps, might even be lower for most tracks, but I noticed a bit dull sound on a few tracks, like lacking deep bass i.e. they sounded a little flat. Don't remember which ones as it was a few years ago, but there weren't more than 10. This is the reason why I stick to iTunes Plus 256 Kbps just as a fail-safe.

Vorbis is very clear and provides detailed sound quality, but I've felt that transients are a bit sharper, like the drums and acoustic songs feel a bit tinny, i.e. overall the song sounded like it had a bit more treble, just a bit. Don't know if its just me, but if it's common, then I believe you might notice this issue with violins too. Another drawback of Vorbis is less compatibility, and even when supported by mobile devices (Androids do), it can drain battery faster.

Opus was designed for low-bitrate applications like encoding voice for an example. It fares very well with music too, but its superiority over other codecs has only been confirmed till <128 Kbps as you can see in this chart on the official website, listening tests have been in accordance with the claims. Notice how all codecs converge at around 128 Kbps. Also notice they have completely left out graph beyond that, might be because Opus is just on par with other codecs, or worse! Personally, I have to say it sounds almost as good (to me) as any other codec, but I suggest you check for yourself before committing. But even if if sounds better than AAC to you (I highly doubt that, might be equally good at best), I suggest you stick with AAC for compatibility reasons.

Hope this helps you with the decision.
Rap is NOT music

Re: Most efficient lossy codec for classical music

Reply #16
Hey Zaki,

thank you! Your experience with vorbis was exactly the kind of subjective descprition I was looking for regarding classical music in opus and aac. At the end of the day it's only a question of a marginally higher bitrate to achieve transparency I suppose, but I am kind of obsessed in these things as the readers maybe can tell.
And the only reason to choose vorbis over aac would be it being royalty free, right?

I am aware that opus was designed for mostly low complexity audio as you say and also real-time transmission via low-latency. It is superior at low-bitrate audio with the margin to the other codecs getting smaller with an increasing bitrate.
But actually there are listening tests ongoing with live results on opus and others at "higher" bitrates.
It seems to still beats it's competitors at ~150 kbps [http://soundexpert.org/encoders-160-kbps], but unfortunately they didn't compare it to the best aac encoders. From how I understand these ratings, opus can be called transparent at this bitrate to a fair degree of certainty.
I was first suprised to find it so low on the list on this test [http://soundexpert.org/encoders-128-kbps], but they actually compare 129 kbps aac to 116 kbps opus.

Compatibility is not an issue for me, since I use VLC on my phone; the only mobile device, that I use.

Re: Most efficient lossy codec for classical music

Reply #17
I needed to go back and re-rip some old CD's and found a few of them now had problems. I could actually see thru my Aja CD.

CD rot is unfortunately a real phenomenon.

However, I expect that I'm very unlikely to lose my rips, as they are 3-2-1 backed up (3 copies, 2 media types, 1 offsite backup). Cloud backup via Google or Dropbox is both cheap and easy ($10/month for 1TB), so there's really no reason not to back up your most important data.

And in the very unlikely event that I lose all three copies, I probably won't bother ripping every single CD again, only the ones I actually listen to. For the inevitable handful of duds, I'm prepared to re-buy.

Re: Most efficient lossy codec for classical music

Reply #18
Cloud backup via Google or Dropbox is both cheap and easy ($10/month for 1TB)
If you bought a 1TB hard drive, it would pay for itself in 5 months.

Re: Most efficient lossy codec for classical music

Reply #19
Cloud backup via Google or Dropbox is both cheap and easy ($10/month for 1TB)
If you bought a 1TB hard drive, it would pay for itself in 5 months.

I have a 500GB external hard drive that I also use for backups. But it's not offsite with its own associated backup and available from anywhere I wish*, which is what you're paying for with Google Drive.

*I could make my external hard drive accessible from the Internet, but that's not something I'm interested in doing, for security reasons.

Re: Most efficient lossy codec for classical music

Reply #20
Re: FLAC. 
Quote
It seems that the spectral consistency isn't so big, and the encoder has decided to use less bits to interpret the music

Flac encoders make no such decisions.  Flac is lossless.  It keeps all the bits in the raw audio (all the bits that contain data) and then compresses them to store in a file.  The compression may be more or less efficient, depending on the type of audio that is encoded.  When the file is uncompressed, all the bits are still there, no more and no less than before.

The bitrate that is reported for Flac is simply the size of the compressed file divided by the length of the audio.  It is in no way related to quality or encoding decisions, only to the effectiveness of the compression.  You can compare to the bitrate of uncompressed (16/44.1 stereo) PCM audio of around 1,411 kbps (give or take some overhead).

Re: Most efficient lossy codec for classical music

Reply #21
This may be of some help - from when I was testing LossyWav against sample Lossy and Lossless codecs for various kinds of music:



From here:
http://www.giantpygmy.net/gpa/index.php?id=a-high-quality-encoding-solution-pc-dap

Also there's a table with bitrates and the tracks tested:
http://www.giantpygmy.net/gpa/data/uploads/images/play/lossy_lossless_bitrates.png

C.
PC = TAK + LossyWAV  ::  Portable = Opus (130)

Re: Most efficient lossy codec for classical music

Reply #22
...
but I'll share my perception of these codecs with my rock music library.
...
The best sound I get is from iTunes AAC TVBR encodes.
...
Vorbis is very clear and provides detailed sound quality, but I've felt that transients are a bit sharper, like the drums and acoustic songs feel a bit tinny,
...
might be because Opus is just on par with other codecs, or worse! Personally, I have to say it sounds almost as good (to me) as any other codec.

This whole post violates TOS 8 with wild abandon. Please present ABX logs and TOS compliant sample encodes so we can confirm your findings.

You note that you can't ABX AAC at 256kbps, then say that AAC at 128kbps is transparent to you as well. If this is the case then encoding at higher than 128kbps is simply satisfying an illogical "feeling" you have and serves no technical purpose, and is therefore inefficient.

The charts tend to end at 128kbps because above that level virtually all modern codecs become so transparent that deriving valid results from listening tests is nearly impossible. This means they all essentially "sound the same" as they are audibly indistinguishable from the lossless sources by the overwhelming majority of the human race.

Encoding for wide compatibility should only matter if you don't intend to keep lossless sources. If compatibility is your main concern and you intend to encode at 256kbps VBR, then mp3 is the logical choice. It's effectively as transparent as any other codec at these bitrates and it's supported by everything and your kitchen sink. No other lossy codec, including AAC can match it's universal support.

When needed, I lossy encode to Vorbis and Opus for devices which support them. These days every device I own will play one or the other. My main reasons for going with them are quality and freedom. When there are freedom respecting choices available which perform on par with the proprietary options, I see no point in supporting platforms which restrict my freedom. I know we're here to discuss the technical merits of the various codecs, but these issues do factor into my software and format decisions.