Sample of MP3 Artifacts. . .

[FalconEddy]

Yesterday I stated in another thread that I would put up a sample of 320 kbps MP3 artifacts. http://www.audiokarma.org/forums/showthread.php?t=157457

This MP3 file is a very good representation of the original master file, and has plenty of ride, crash, and hi-hat cymbals in the recording.

The artifact sample is a result of reducing a 24 bit - 96kHz original master file, down to a 16 bit 44.1kHz 320kbps MP3 file; at the highest quality encoding selected. Then, phase inverting one of the stereo files and lining it up (as best as possible) with the other file. It's only lined up within a .008 second tolerance because the software won't let me view waveforms from multiple files at less than a 1:1 ratio. But, it's pretty close.

The artifact sample allows you to hear what's left over. A perfectly matched set of phase inverted files would yield no sound at all.

You will note that the artifact file is fairly quiet as compared to the full version MP3. It would be difficult for the vast majority of people to notice the differences between the original 24bit master, and the highest quality encoded 320kbps MP3.

http://FalconEddyStudios.co/mp3/RockDaBoat-24BitMaster.mp3

http://FalconEddyStudios.co/mp3/MP3-320kbps-Artifacts.mp3

. . Falcon

 

[winters860]

Okay. Seems like there's some hash there.

How about if you encoded a 16-bit 44.1 CD instead of the 24-bit master?

 

[FalconEddy]

Okay. Seems like there's some hash there.

How about if you encoded a 16-bit 44.1 CD instead of the 24-bit master?

You'd get practically the same result. Dithering reduces the bit-depth of the original 24 bit waveform, but it's not a reduction that will alter a specific set of frequencies at varying amounts.

Those types of reductions/increases are the result of the MP3 encoding itself. Different algorithms produce different results. Even the same algorithm won't generally produce the exact same resultant file every time. But, you'd need to match up the waveforms to see the minute differences.

Lower rate (128 kbps) encoding is pretty rude, and can be really harsh sounding. Especially on percussion such as cymbals, or even piano solos. 320 kbps does a MUCH better job of smoothing out the harshness, and delivers more a pleasing audio when the highest encoding setting is used.

. . Falcon

 

[sauuuuuce]

If you are only with in .008 of second on phase alignment of the two tracks, isn't that significant enough to not cancel alot of the signals that would line up fine. .008 seconds is enough time for 125Hz to complete an entire cycle. Please educate me Falcon. I should also mention that I haven't listend to the file yet because I am at work.

 

[CChase]

What are you using to do this?

I'm trying to do this in MATLAB currently, no luck so far though in lining things up.

 

[FalconEddy]

If you are only with in .008 of second on phase alignment of the two tracks, isn't that significant enough to not cancel alot of the signals that would line up fine. .008 seconds is enough time for 125Hz to complete an entire cycle. Please educate me Falcon. I should also mention that I haven't listend to the file yet because I am at work.

Sorry, you are correct. My error. I was looking at the total time I had highlighted (for deletion) at the beginning of one of the file sets to get them lined up.

I wrote that number down where I was making calculations to adjust the difference in the number of samples (to make sure they lined up.)

I just went back and double checked them in another software package. The files are actually offset correctly (they have a zero sample difference begin point.) And I can view them at a 24:1 ratio.

. . Falcon

 

[FalconEddy]

What are you using to do this?

I'm trying to do this in MATLAB currently, no luck so far though in lining things up.

WaveLab, Soundforge, and Acid Pro.

Remember, you may need to adjust the levels on one of the sets of tracks in order to null out the other set of tracks.

. . Falcon

 

[Pneumonic]

Yesterday I stated in another thread that I would put up a sample of 320 kbps MP3 artifacts. http://www.audiokarma.org/forums/showthread.php?t=157457

This MP3 file is a very good representation of the original master file, and has plenty of ride, crash, and hi-hat cymbals in the recording.

The artifact sample is a result of reducing a 24 bit - 96kHz original master file, down to a 16 bit 44.1kHz 320kbps MP3 file; at the highest quality encoding selected. Then, phase inverting one of the stereo files and lining it up (as best as possible) with the other file. It's only lined up within a .008 second tolerance because the software won't let me view waveforms from multiple files at less than a 1:1 ratio. But, it's pretty close.

The artifact sample allows you to hear what's left over. A perfectly matched set of phase inverted files would yield no sound at all.

You will note that the artifact file is fairly quiet as compared to the full version MP3. It would be difficult for the vast majority of people to notice the differences between the original 24bit master, and the highest quality encoded 320kbps MP3.

http://FalconEddyStudios.co/mp3/RockDaBoat-24BitMaster.mp3

http://FalconEddyStudios.co/mp3/MP3-320kbps-Artifacts.mp3

. . Falcon

Hi, Falcon.

From my experience listening to a very wide variety of digital music, both lossy and lossless, I find that higher rate lossy (VBR or CBR) files can be quite pleasing to my ears and enjoyable in terms of tone ...... pace, rhythm and timing. I can toe tap along with most said recordings and, in fact, on some inferior recordings, even prefer the lossy version to the lossless one (usually poorly recorded rock albums which are recorded way hot). Perhaps the algorithms used make such recordings easier on the ears?

However, on exceptionally well recorded material (for example, Chesky, Sheffield Labs, Reference Recordings, Dorian, Opus3, Klavier, Audioquest, Proprius, etc) I can quite easily differentiate the lossless version from the lossy one. Typically, the tell tale sign for me when doing such a comparison is in the "breath" of the music; those parts of the soundstage where you are able to hear the layering and instrument placement which manifests itself as "air" around the instruments. Very well recorded lossless music "breathes" to me whereas encoded lossy material lacks such air, subtle texturing, resolution and clarity.

Further, I also hear not so subtle differences in the dynamics of the music, all the way around not just bass, though this is really only the case when I employ a top notch power amplifier that is capable of resolving such differences. I find that far too many amps are incapable of doing so so it doesn't surprise me that many people are missing out on this clue.

Anyhow, thanks for your efforts on this and I look forward to any further information you might provide on this matter.

 

[wiredbecker]

Cool thread. Thanks Ed

 

[hypertone]

I don't have speakers connected so I can't listen, but there is a flaw when comparing a 44.1K mp3 to the original 96K file. The control and the mp3 file should both be at the same sampling rate. Converting sampling rates discards information as well, so it's not an accurate comparison by which to judge the difference between mp3 and wave.

 

[FalconEddy]

I don't have speakers connected so I can't listen, but there is a flaw when comparing a 44.1K mp3 to the original 96K file. The control and the mp3 file should both be at the same sampling rate. Converting sampling rates discards information as well, so it's not an accurate comparison by which to judge the difference between mp3 and wave.

Although I agree with you, up to a point, I said. . .

The artifact sample is a result of reducing a 24 bit - 96kHz original master file, down to a 16 bit 44.1kHz 320kbps MP3 file

I was only noting that the original file was 24 bit @ 96kHz.

However, I didn't include all the steps taken to get there since I wasn't attempting to compare the original recording with the resultant MP3. The original recording was sample rate converted to 44.1kHz before going through the MP3 conversion.

If you think about it, you'll realize that you can't compare the differences (as a resultant file) between two files with different sampling rates. I did not supply the original file for comparison.

So, all conversions are fully accounted for between the original file and the MP3.

. . Falcon

 

[sauuuuuce]

Just got to listen. Very interesting. I have never had much of a problem with higher quality mp3's. Especially for anything top 40 today. i guess for me the ability for me to carry so much music in my pocket outweighs the loss you have demonstrated. I know some don't feel this way. To each his own! nice post falcon...........

 

[FalconEddy]

Just got to listen. Very interesting. I have never had much of a problem with higher quality mp3's. Especially for anything top 40 today. i guess for me the ability for me to carry so much music in my pocket outweighs the loss you have demonstrated. I know some don't feel this way. To each his own! nice post falcon...........

Absolutely! Although I've had some brief resistance to producers/clients 'buying into' the 320 Kbps MP3 file for initial evaluation purposes only, everyone I've dealt with in the past few years is now on board with it. As a matter of fact, I had several clients that took their 320 kbps MP3 files, and encoded them to a wavefile so they could burn a CD.

Now, an even MORE interesting part (rhetorical question) of the 'food for thought' portion of this comparison is which parts of the artifacts file are left over from the wave file, and which ones remained from the MP3?!?!?

Since both waveforms have different levels for mostly any given number of continuous samples, whichever waveform has the higher energy level is audibly represented within the resultant artifacts file.

Cool, huh?

. . Falcon

 

[Pneumonic]

Absolutely! Although I've had some brief resistance to producers/clients 'buying into' the 320 Kbps MP3 file for initial evaluation purposes only, everyone I've dealt with in the past few years is now on board with it. As a matter of fact, I had several clients that took their 320 kbps MP3 files, and encoded them to a wavefile so they could burn a CD.

Now, an even MORE interesting part (rhetorical question) of the 'food for thought' portion of this comparison is which parts of the artifacts file are left over from the wave file, and which ones remained from the MP3?!?!?

Since both waveforms have different levels for mostly any given number of continuous samples, whichever waveform has the higher energy level is audibly represented within the resultant artifacts file.

Cool, huh?

. . Falcon

I am disturbed that some music people would take a MP3 encoded file and turn it into a CD that us customers would purchase. Then again, hearing much of the crappy sounding material that is being released these days it doesn't surprise me that ultimate sound quality is low on the agenda of too many artists and music execs these days.

My experience is that analyzing music on a technical, or engineering level isn't absolute. Far too often music is talked about in terms of specs, of waveform analysis and of complicated Laplace/Fourier transforms when simple listening skills are all that is needed. I fear this is a skill that is sorely lacking in today's music industry. Heck, you don't even need to be able to play an instrument these days to make it big in "popular" music. Check out the billboard lists for proof. Thus it shouldn't shock anyone that these same "artists" don't give two hoots about how they sound .... they don't even know how they are supposed to sound for crying out loud!

Now, don't get me wrong, you need proper engineering and skilled design processes to get the most out of this hobby, but I find far too often simple listening with our ears takes a back seat to all the technical analysis and study when it shouldn't

Edit: I see that the burned MP3's weren't sold but rather for the enjoyment of your clients.