Interesting page about compression artifacts

dB Cooper · « **on:** 20 Feb 2015, 01:42 pm »

Thought some might be interested in this. They took uncompressed audio, compressed audio, and compared the two in a way that lets you isolate and listen to the portion of the audio removed/discarded by the compression process. It's pretty amazing.

I realize that some here would rather eat their own shoes than listen to compressed audio anyway; nevertheless, those of us who do (internet radio for example) might be interested. Not so 'inaudible' after all, perhaps?

Anyway, here's the link:
http://ryanmaguiremusic.com/theghostinthemp3.html

Guy 13 · « **Reply #1 on:** 20 Feb 2015, 02:51 pm »

Quote from: dB Cooper on 20 Feb 2015, 01:42 pm

Thought some might be interested in this. They took uncompressed audio, compressed audio, and compared the two in a way that lets you isolate and listen to the portion of the audio removed/discarded by the compression process. It's pretty amazing.

I realize that some here would rather eat their own shoes than listen to compressed audio anyway; nevertheless, those of us who do (internet radio for example) might be interested. Not so 'inaudible' after all, perhaps?

Anyway, here's the link:
http://ryanmaguiremusic.com/theghostinthemp3.html

Hi dB Cooper,
bizarre and interesting at the same time.
Sound to me like psychedelic music.
Thanks for the link.

Guy 13

Speedskater · « **Reply #2 on:** 20 Feb 2015, 06:00 pm »

If it's "Not so 'inaudible' after all, perhaps?" why do most people most of the time have trouble differentiating 24/96 from well made MP3s?
I think that the word audible was misused.

Speedskater · « **Reply #3 on:** 20 Feb 2015, 07:04 pm »

Upon rereading the page, it appears that he is a musician/composer and is attempting to make new music out of the inaudible parts of a song that MP3 removes. Then on the technical side he gets into trouble by not pointing out that the subtraction samples have been amplified by maybe 50 or 60 dBs. He also complains about the musical samples used in the early MP3 design stages without suggestions for future developments.

dB Cooper · « **Reply #4 on:** 20 Feb 2015, 08:01 pm »

Thanks for spotting that, Speedskater. In my enthusiasm to share the article I missed it. The digital era version of the old hi-fi shop trick of turning up the system you want to "move" a couple dB more than the other system, perhaps.

I have seen a page that does something similar, showing what was removed from an image by JPEG compression.

Steve · « **Reply #5 on:** 8 Mar 2015, 06:32 pm »

It may or not be perceptible, depending upon the music. I know for personal testing over the years with different individuals, that when playing a cd of solo cello, "muddiness" can be detected with only an approximately 32udb (32 millionth of a db) alteration at 20hz. That would approximately translate to -90 to -100db down at 20hz, and even more db down at 60-100hz or so, the frequency of the cello's "muddiness".

It is very interesting article nontheless

Cheers
Steve

Speedskater · « **Reply #6 on:** 9 Mar 2015, 03:16 am »

I'm glad that the of us don't live in Steve's world.

*Scotty* · « **Reply #7 on:** 9 Mar 2015, 03:23 am »

Sure we do. We listen to varying degrees of mud everyday. Imagine what it would be like to listen to recorded music that sounded more like the real thing? All it would take it less or no compression when ever it was possible.
Scotty

Steve · « **Reply #8 on:** 9 Mar 2015, 08:07 pm »

Yep, agreed, we all do as various recordings are different, some slightly lean and some full. My purpose was just to test how sensitive our perception can be, for my knowledge bank. When I listen at the Peoria Civic center, the sound is quite full imo.

The 0,02 ohm variations in the zobel are to obtain the most natural sound I can. It is surprising to me how many selections can sound so good, considering all the venues and recording techniques.

Cheers
Steve

*Scotty* · « **Reply #9 on:** 9 Mar 2015, 08:50 pm »

Steve, where would that zobel be located?
Scotty

Quiet Earth · « **Reply #10 on:** 9 Mar 2015, 09:33 pm »

3rd moon of Jupiter.

Steve · « **Reply #11 on:** 9 Mar 2015, 11:09 pm »

LOL quiet.

Scotty, across the full range driver in my test speakers.

Cheers
Steve

*Scotty* · « **Reply #12 on:** 10 Mar 2015, 12:43 am »

Well, without the zobel they're certainly going to be lacking something, not to mention that they will tend to sound like a driver reproducing sound instead of a hole in space with music emanating from it.
Scotty

Steve · « **Reply #13 on:** 10 Mar 2015, 03:42 am »

Quote from: *Scotty* on 10 Mar 2015, 12:43 am

Well, without the zobel they're certainly going to be lacking something, not to mention that they will tend to sound like a driver reproducing sound instead of a hole in space with music emanating from it.
Scotty

Beautiful and accurate post Scotty.

Cheers
Steve

Speedskater · « **Reply #14 on:** 10 Mar 2015, 05:21 pm »

How did the subject of this thread change so much? It started with a musician/composer making new music out of what a data compression program tosses in the dumpster and the amplifying it by 50 or 60dB. Now it's drifted to volume compression and graphic equalizers.

Quiet Earth · « **Reply #15 on:** 10 Mar 2015, 06:13 pm »

I think because the title and the subject matter are not the same thing. It's a confusing post to begin with.

bwaslo · « **Reply #16 on:** 16 Mar 2015, 09:11 pm »

To hear similar test, at actual signal levels (or boosted as you want to):
http://libinst.com/diffmaker_example_files.htm

You'll need to install Audio Diffmaker (free) first though, http://libinst.com/Audio%20DiffMaker.htm

There are some other interesting comparisons there, green pens, ceramic caps, etc.

*Scotty* · « **Reply #17 on:** 17 Mar 2015, 02:08 am »

Audio DiffMaker can probably provide conclusive empirical evidence, substantial enough to convince those who think that cables, powercords, tweaks, etc don't actually make an audible difference, assuming of course, that they are willing to examine evidence which refutes a staunchly held position.
I wouldn't be surprised if it required a stereo mike in some cases to aid in capturing the information necessary to detect differences in stereo imaging.
Also, I suspect the program will detect differences if the microphone location is altered between "takes" during the recording process. This will change the comb filter effects that the mike is subject to thus generating a difference signal even if nothing else changed.
It should also be able detect if ABX switchers mask differences that are otherwise audible when the ABX switcher is not in the system.
Obviously the program is so sensitive that it will be very easily to inadvertently cause a difference to be detected due to procedural errors that may occur during testing. A good thing actually, if the difference is still there when you have taken stringent measures eliminate uncontrolled variables then you know you might be on to something.
Scotty

JerryM · « **Reply #18 on:** 17 Mar 2015, 02:49 am »

Great points, all, Scotty.

Steve · « **Reply #19 on:** 22 Mar 2015, 04:04 am »

For public consumption. On AudioKarma there was an article "Green" presented some interesting data. Channel separation was only approximately -60db, with differences measured as high as 1.1db between Green and non treated CDs, if I remember correctly. Harmonic differences up to 2db were also detected in and around -90db, particularly in one harmonic. However, Audio Diffmaker failed to register those differences, completely missed the differences. I am not advocating one should treat cds, but I would say that when I checked the data presented in the article, the Audio Diffmaker was lacking. I would be careful putting alot of weight on such programs.

I thought the article was interesting.
Steve