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: Yet another topic about the horrors of audio compression (Read 1038 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Yet another topic about the horrors of audio compression

I'm not going to pretend I've not been pirating music for the past 25 years or so, for multiple reasons. First, I was poor, then I realized most audio CDs sold in my country are mp3 ... decodes, and now I'm just OK with a subscription which sort of covers my old bad habits. That's not the topic.

Now more than two decades later I'm completely dumbfounded. Lots of MP3 files I've downloaded have a much better dynamic range and no visible compression. To "improve" the audio stored on my drives I've got my hands on literally dozens of official CD albums and most of the albums and tracks are heavily compressed despite being released approximately at the same time as my downloads.

How is it even possible that low quality 128/192/256kbps mp3s are not overprocessed and sound much better than "true" "HQ" lossless audio? I actually hear more instruments and better vocals for weird low-quality mp3 files encoded by God knows whom than the officially released audio.

My only explanation is that these mp3s are advanced versions sent to radio stations which were never released officially. Now, the question is, does that mean that people who listen to official audio, let's say, spotify, apple music, those who still buy CDs, get a worse audio quality than people who managed to grab "pirated" music? The worst part about this situation is that these advanced tracks are never released officially. By my estimates I have over several hundred of such tracks.

Why is this happening at all? Why don't artists and groups release the actual masters and instead they [or someone else] compress music and turn it into a loud incomprehensible mess?

Attached is such a file, "Ciara - Like a Boy". On top: the pirated 224kbps mp3, vs, at the bottom, the official track from an audio CD.

Another example, "7Б - Я любовь" - a Russian song.

Re: Yet another topic about the horrors of audio compression

Reply #1
It's likely the CDs were remastered to "win" the loudness war.

Note that MP3 is NOT dynamic compression.   It's file compression. It is lossy and there can be artifacts (side effects and quality loss) but you don't get dynamic compression.   In fact, the wave shape changes in ways that make some peaks higher and some lower (without affecting the sound of the dynamics).    Then if you measure the crest factor (the peak-to-average ratio) the MP3 will measure more dynamic than the uncompressed original!   Many of the MP3s I've ripped from CD go over 0dB (MP3 can go over 0dB without clipping) and they "show clipping" in Audacity.  (Audacity is actually showing potential clipping.)

Something similar happens with the vinyl cutting and playback process leading many people to claim the vinyl is more dynamic, even when made from the same master.

MP3 actually has MORE dynamic range capability than CDs!


Re: Yet another topic about the horrors of audio compression

Reply #2
It's likely the CDs were remastered to "win" the loudness war.

I perfectly remember this, I'm just confused because it would mean that everything that's released is remastered by default however from my experience there are different versions of audio and end consumers often get the absolute worst in terms of dynamic range and clarity despite paying top dollar.

Note that MP3 is NOT dynamic compression.   It's file compression.

I know the terms. Also, there's no such thing as MP3, there's MPEG-{1,2,2.5} Audio Layer III. ;-)

It is lossy and there can be artifacts (side effects and quality loss) but you don't get dynamic compression.   In fact, the wave shape changes in ways that make some peaks higher and some lower (without affecting the sound of the dynamics).    Then if you measure the crest factor (the peak-to-average ratio) the MP3 will measure more dynamic than the uncompressed original!   Many of the MP3s I've ripped from CD go over 0dB (MP3 can go over 0dB without clipping) and they "show clipping" in Audacity.  (Audacity is actually showing potential clipping.) Something similar happens with the vinyl cutting and playback process leading many people to claim the vinyl is more dynamic, even when made from the same master.

MP3 actually has MORE dynamic range capability than CDs!

This I don't quite understand. Audio compression is meant to remove information about the source which means dynamic range is probably decreased. It seems unbelievable that the MP3 codec uses full 16 bits [of volume range] to encode audio.

Re: Yet another topic about the horrors of audio compression

Reply #3
Your theory about MP3 increasing the dynamic range doesn't sound plausible to me - I've just tested a couple of CDDA tracks and they look exactly the same after being compressed by Lame 3.100 (lame --replaygain-accurate -c -p --preset extreme).

Re: Yet another topic about the horrors of audio compression

Reply #4
MP3 does not increase dynamic range, and such nonsense was never claimed at all.
MP3 just can hold more dynamic range if encoded from higher dynamic range that is not 16 bit CD.

Re: Yet another topic about the horrors of audio compression

Reply #5
Quote
This I don't quite understand. Audio compression is meant to remove information about the source which means dynamic range is probably decreased. It seems unbelievable that the MP3 codec uses full 16 bits [of volume range] to encode audio.
I was surprised too!  But I once did an experiment -
I reduced the volume by about 100dB.   (That takes two steps in Audacity because Amplify only goes to -50dB)    If you save ("export" from Audacity) as a 16-bit WAV you'll get dead silence...  A file full of zeros...    You can try to amplify but there's absolutely nothing there.    (You CAN re-amplify BEFORE you export because Audacity uses floating-point internally)

But if you export the -100dB attenuated file as MP3, re-open the MP3 and then re-amplify (twice again) and you'll have sound.    It's pretty terrible quality but it's there.  

Quote
Your theory about MP3 increasing the dynamic range doesn't sound plausible to me
The "measured: or "apparent" dynamic range, NOT the SOUND of the dynamics.

Like I said, lots of my MP3s (ripped from my CDs) go over 0dB which is impossible on a CD.  

Here's another experiment -
Take your regular file and Amplify by +20dB (you'll have to allow clipping).   At this point Audacity will "show clipping" but since Audacity uses floating-point it's not really clipped, yet.   You can run Amplify again and it should default to about -20dB depending on the original peaks before you amplified.  But DON'T RE-APPLY amplification (attenuation) because WE WANT CLIPPING for this experiment.

Save the boosted file as 16-bit WAV and it will be (badly) clipped.

Re-open the clipped  file and tun Amplify which will default to 0dB indicating that the (clipped) peaks are 0dB,

Now, export the 0dB clipped file as MP3.

Open the clipped MP3 and run Amplify again, just to check the peaks.    Amplify will default to a negative value, indicating that the peaks have been boosted above 0dB.    And if you apply the (negative) Amplify effect the waveform will "look better" but of course it will still sound terrible.

I believe this side-effect of boosted peaks is worse when the peaks are clipped or otherwise artificially limited/compressed.   But the main reason for using a clipped file in the above experiment is to get lots of peaks at the same-known 0dB level so we can easily demonstrate that MP3 boosts some peaks.




Re: Yet another topic about the horrors of audio compression

Reply #6
MP3 does not increase dynamic range, and such nonsense was never claimed at all.
MP3 just can hold more dynamic range if encoded from higher dynamic range that is not 16 bit CD.

That's all highly confusing to me. I thought audio bitness, e.g. 16 bit encodes the range of volume, from 0 to 65535. If you re-encode 24 bit audio into 16 the dynamic range shouldn't change. The range will be more coarse (I don't think any human will hear the difference) but that's it.

It also sounds quite unlikely to me that a 24bit master encoded into mp3 will sound better or have a better dynamic range than 16bit CDDA.

Re: Yet another topic about the horrors of audio compression

Reply #7
long quote

It all sounds interesting but I don't think people who post mp3s on warez websites do anything you've just said. It all sounds kinda complicated, prone to errors and could in theory destroy audio quality. My MP3s on the other hand sound substantially better than CDDA - this is what the topic is about. Not about how you can make an MP3 have a better/wider/higher dynamic range, but how music is distributed.

Re: Yet another topic about the horrors of audio compression

Reply #8
Quote
My MP3s on the other hand sound substantially better than CDDA - this is what the topic is about.
Simply different mastering.  It was ripped from a different CD, or maybe digitized vinyl.  Or possibly, it came from a more dynamic master and a high-resolution original (but I doubt many pirates are making MP3s from high-resolution originals).    

Quote
but how music is distributed.
If you download an MP3 from Amazon it will (almost always) be from the same-current CDs they are selling.  And the MP3 will usually "measure" slightly-more dynamic. 

If you find an older CD from before the loudness wars got out-of-hand it might be better.

Re: Yet another topic about the horrors of audio compression

Reply #9
Quote
Also, there's no such thing as MP3,
Yet somehow I have THOUSANDS of MP3s!!!   :P

...Yeah, I know they are "layer III" but they are literally named ".MP3" and that's the everyday jargon.

Re: Yet another topic about the horrors of audio compression

Reply #10
Back to your original concern...

The streaming version is probably the latest over-compressed version.  :(

The popular streaming services all loudness-normalize so nobody wins the loudness war but of course that doesn't help with the dynamics, except it removes the incentive to over-compress and over-limit new recordings made today (or newly remastered recordings).    

But I still feel like that constantly-loud or constantly-intense style has been built-into a couple of generations of artists and producers and I don't really expect a lot of new highly-dynamic music to be released.   At least not in the popular genres.    I'm old enough to remember analog and I expected the improved dynamic range of digital to encourage more dynamic music production, but the opposite happened!

Re: Yet another topic about the horrors of audio compression

Reply #11
Attached is such a file, "Ciara - Like a Boy". On top: the pirated 224kbps mp3, vs, at the bottom, the official track from an audio CD.

Looking at that image, I would say that compressed version is not over-compressed, Looks good to me. [1st image only]
Stop comparing tracks. Just enjoy your music. But than again this is HydrogenAudio.

 

Re: Yet another topic about the horrors of audio compression

Reply #12
Also, there's no such thing as MP3, there's MPEG-{1,2,2.5} Audio Layer III. ;-)

Did you just literally say that acronyms don't exist?

Re: Yet another topic about the horrors of audio compression

Reply #13
Attached is such a file, "Ciara - Like a Boy". On top: the pirated 224kbps mp3, vs, at the bottom, the official track from an audio CD.

Looking at that image, I would say that compressed version is not over-compressed, Looks good to me. [1st image only]
Stop comparing tracks. Just enjoy your music. But than again this is HydrogenAudio.

The CD version sounds markedly worse. I'm not enjoying it, sorry.

Re: Yet another topic about the horrors of audio compression

Reply #14
Also, there's no such thing as MP3, there's MPEG-{1,2,2.5} Audio Layer III. ;-)

Did you just literally say that acronyms don't exist?

They certainly do :-) MP3 was just something unofficial which has ultimately stuck.  :D

Re: Yet another topic about the horrors of audio compression

Reply #15
This I don't quite understand. Audio compression is meant to remove information about the source which means dynamic range is probably decreased.
Most lossy codecs, including MP3, remove information by quantization. Quantization is a process that reduces the precision of numbers. You may already be familiar with one form of quantization: rounding.

Well, let's say you have the numbers 3.14159 and 2.71828, and you want to write each of them in a small space that only has enough room for three digits. You can round them and you'll get 3.14 and 2.72, so now they fit, but they've changed a little bit. The first number got smaller and the second number got bigger.

The same thing happens with MP3: sometimes the numbers get smaller, sometimes they get bigger.

Re: Yet another topic about the horrors of audio compression

Reply #16
... and if you don't round off the number that represents the amplitude, but a transform of it, then a "rounding up" might make the losst signal "overshoot" the original; and if that happens at peak, the MP3 encoded signal gets a higher peak and measures at higher dynamic range.
That does not mean that the music sounds more dynamic; it takes more than a few too high samples to give that impression. But it can fool an unsophisticated DR meter - and of course, the concept of "dynamic range" was not in the first place developed to test the "degree of lossiness", so don't complain.
Last two months' worth of foobar2000.org ad revenue has been donated to support war refugees from Ukraine: https://www.foobar2000.org/

Re: Yet another topic about the horrors of audio compression

Reply #17
I perfectly remember this, I'm just confused because it would mean that everything that's released is remastered by default however from my experience there are different versions of audio and end consumers often get the absolute worst in terms of dynamic range and clarity despite paying top dollar.

This is absolutely the case. Perversely this is in part why people started buying vinyl again, despite the medium having inferior performance, the vinyl often (but not necessarily) sounds better because it is less compressed. Some of your mp3s might be needle drops. This in turn had added to the mythology of digital music in general being bad.

Pretty much every popular CD release outside classical since about 1995 is loudness war material. There are some rare exceptions like the "Full Dynamic Range" reissues from Earache, and speciality labels like MFSL. In almost all other cases if there was a release in the 80's or early 90's and a later remaster, the earlier release is superior. This includes "high-res" releases from places like HDTracks, which often use the most recent remaster.

Even albums that were already super compressed are not immune, nobody would have accused Mastodon's Remission (2002) of having too much dynamics, but they nevertheless remastered it in 2014 to crush it even further.

Re: Yet another topic about the horrors of audio compression

Reply #18
Perversely this is in part why people started buying vinyl again, despite the medium having inferior performance, the vinyl often (but not necessarily) sounds better because it is less compressed.
I don't think that explains customer behaviour at this large scale.
Also, even when the vinyl measures to higher DR, it is often some weird effect of the measurement. There was a discussion here some years ago, where a sound engineer at YouTube compared a vinyl rips from the very same master (he knew, he had mastered it) and the numbers came out different.

Yes there are vinyl masters with higher dynamic range. Why? Because record companies demand a DR-squashed master for streaming, but they don't bother so much as to dictate that for the vinyl. Dan Swanö explained this very policy as why one of his albums had better masterings as MP3 included in the EnhancedCD's data session, than in the audio part of the CD. More here: https://web.archive.org/web/20140202202836/http://www.metal-fi.com/taking-swano-challenge/
... but I don't think that is enough to explain the sales figures.



Pretty much every popular CD release outside classical since about 1995 is loudness war material. There are some rare exceptions like the "Full Dynamic Range" reissues from Earache, and speciality labels like MFSL. In almost all other cases if there was a release in the 80's or early 90's and a later remaster, the earlier release is superior.
Earache's "Full Dynamic Range" releases are typically from the DAT used to make the original CD: https://web.archive.org/web/20150619080549/http://www.metal-fi.com/interview-digby-dig-pearson-of-earache-records/
Last two months' worth of foobar2000.org ad revenue has been donated to support war refugees from Ukraine: https://www.foobar2000.org/

Re: Yet another topic about the horrors of audio compression

Reply #19
Your theory about MP3 increasing the dynamic range doesn't sound plausible to me - I've just tested a couple of CDDA tracks and they look exactly the same after being compressed by Lame 3.100 (lame --replaygain-accurate -c -p --preset extreme).
Many CDDA tracks are normalized to near 0dBFS so you can't just pick a random track as an illustration. Try the attached example, I attenuated a highly squashed CDDA track to let the mp3 encoded peaks to grow.

Regardless of audibility, many similarly squashed tracks need to playback at reduced volume in order to get rid of clipping caused by intersample over and/or rise of peaks due to lossy compression errors. The magnitude of these kinds of errors are highly dependent on source materials, codecs and encoding settings.
X