[Armaegis]

Here's a good post that I bookmarked a while ago regarding lossy compression...
 
http://www.head-fi.org/forums/f15/lossy-audio-codecs-comparison-huge-amount-pics-itunes-update-p-7-a-225356/
 

 

[plonter]

Quote:

Here's a good post that I bookmarked a while ago regarding lossy compression...
 
http://www.head-fi.org/forums/f15/lossy-audio-codecs-comparison-huge-amount-pics-itunes-update-p-7-a-225356/
 

thanks for the link. i was surprised by the OGG results,i thought it supposed to be superior to mp3,but it seems that even at 500kbs there are still some parts missing.   do you think it is because it was an old version of the codec?

 

[aimlink]

Quote:

Radioactive is right.  If you have two files with purely a 1khz tone, they will sound exactly the same, you can't have one the has more "depth" than the other.  If you were playing a 1khz on a speaker and recorded it from different places, then you wouldn't have just a pure 1khz tone.  You'd have the echo and the effect of the room on the tone.  Moving it farther away gives a different effect on the sound.

You took the words right out of my mouth.  I was reading along to see if anyone would introduce this point.  The acoustics of a room has a lot to do with the overall effect.  It may be a pure 1K tone, but not from only one source.  The walls would make for other sources.

 

[khaos974]

Besides the varying decay rates between different frequencies, the placement of your microphones with affect the amount of sound it pick up from the primary sound source and the secondary reflections as well as their timings. This is also why loudspeakers measurements are done in anechoic chambers.

 

[jcx]

Quote:

thanks for the link. i was surprised by the OGG results,i thought it supposed to be superior to mp3,but it seems that even at 500kbs there are still some parts missing.   do you think it is because it was an old version of the codec?

you either didn't read far enough or didn't appreciate the explainations of what losey perceptual codecs are trying to do and why the pretty spectographs are not useful in judging the perceptual coding success - in fact as pointed out by some later posters the "defects" visible in the spectographs can be evidence of efficient, proper operation of good psychoacoustic coding
 
perceptual coding for bitrate compression is intentionally throwing away information that the ear/brain system doesn't pay attention too - look up frequency and temporal masking - good codecs can throw away 3/4 of the Shannon-Hartley "channel information capacity" - and trained testers can't ABX the diference from full bitrate except on rare hard to encode samples
 

 

[plonter]

Quote:

you either didn't read far enough or didn't appreciate the explainations of what losey perceptual codecs are trying to do and why the pretty spectographs are not useful in judging the perceptual coding success - in fact as pointed out by some later posters the "defects" visible in the spectographs can be evidence of efficient, proper operation of good psychoacoustic coding
 
perceptual coding for bitrate compression is intentionally throwing away information that the ear/brain system doesn't pay attention too - look up frequency and temporal masking - good codecs can throw away 3/4 of the Shannon-Hartley "channel information capacity" - and trained testers can't ABX the diference from full bitrate except on rare hard to encode samples
 

I did read the explanations you are talking about, what the op intented to show in this thread is what codec get as closest to the source with less bitrate as possible. as you can see,the mp3 has succeeded in getting closer to the source with less bitrate...thus the mp3 is better. 500kbs ogg drains the battery much faster than 320 mp3,so basically you pay more and get less.  
i personally use only lossless.
the fact that the human hearing can't detect certain frequencies doesn't mean that they are not there!   it is enough for me that 1 person out of a million will be able to hear the difference between lossy and lossless in order to use only lossless.   lossy codecs are  a solution for those don't have the space and battery life required (in portables) in order to use lossless..no more.  in all other cases, sticking to the source is the preffered way to go.

 

[Armaegis]

Most of the time, I am content with 192k bitrate. On some songs/genres I can tell where there are encoding losses, but there isn't an appreciable difference between 192 and 320 for me on those particular songs. It's either go all the way to lossless, or be happy with where I'm at. I choose the latter, for simplicity and size.

 

[Skylab]

Lossy audio codecs for music are going to be as obsolete as 8-track tapes very soon.  They were useful in the era of expensive disc storage and limited broadband.  In the era of 1 TB drives for under $100, and ubiquitous crazy-fast broadband, the rationale for lossy coding is largely passed.  It's not benign, regardless of how transparent it may be.  And it isn't needed.  The only reason it's not completely gone is that the recording industry lawyers don't want you to have lossless.  If they didn't care, it would already be well on its way to dominating the PC-audio world.

 

[Armaegis]

The portable world (laptops, daps) still have some space limitation, but I agree in that a few more years that will become a non-issue.

 

[etiolate]

Quote:

Ok, to clarify my scenario, let's say that when recording B was made, the mike was placed much farther away from the source in order to give the sense of depth. Still a 1000 hz tone, only differing in microphone placement.

assume a perfect microphone for scientific purposes, first off.  secondly, many responses in this thread are bogus. 
there's going to be no "spectral change" if all you are transmitting is a perfect tone in a perfect open medium. 
if you were to neglect room timbre and the like, this whole depth thing is close to non-existant.  this is because the space creates overtones and, thus, "spectral change" if the mic was repositioned.
 
one aspect that was not addressed here is that this probably should be a stereo recording if you want to be able to hear depth in the first place (listening to music or just a tone).  if this was a stereo setup (using a stereo mic like the following picture perhaps), then regardless of the bit-rate, you'd be able to hear where (or at least, hear a change in) the tone was being transmitted as, for instance, the transmitter was revolved around the room. this is due to your brain being awesome about detecting phase shifts and delays (same thing with perfect tone).
 

 
the overtones that would inevitably exist in a musical pitch (ie, from an instrument, voice, etc) that would give your brain more clues as to where the transmitter is exactly and, perhaps, a relative distance (to where it was located previously).  thus, a lossless file would be better considering this aspect.
 
but, as for the real question, i'm going to (like many before me) venture into a realm that i'm not 100% on now (encoding/decoding).  it is true that mp3 attenuates the highs to save bits, but if you're just transmitting a tone, perhaps the attenuation won't be necessary.  see, by you choosing a perfect tone, you chose basically the easiest waveform to digitize and work with in electronic circuity (save DC).  so, maybe the mp3 encoder could be smart enough to be able to replicate any tone, given nyquist is obeyed, when converted to analog.

 

[dvw]

I don't think the frequency will change in relationship to the distance of the mic unless you're talking about Doppler's effect. The perception of depth I think comes from reflection of the sound. I mean the same sound is transmitted in multi direction and they arrived in your ears at different time and that create a space sensation. So reducing the volume does not give a sensation that the sound source is moving farther away. Another way is that if you move the balance, it makes the sound on the left or right louder but it does not move the source from left to right.
 
I am not an expert on this. This is just my speculation.

 

[DrSpiv]

Quote:

Radioactive is right.  If you have two files with purely a 1khz tone, they will sound exactly the same, you can't have one the has more "depth" than the other.  If you were playing a 1khz on a speaker and recorded it from different places, then you wouldn't have just a pure 1khz tone.  You'd have the echo and the effect of the room on the tone.  Moving it farther away gives a different effect on the sound.

 
Right on. And this goes for lossy or lossless compression. Keep in mind lossless doesn't mean high quality, it denotes the manner that data is converted and stored. You could sample a song at 50kb/s and use a lossless compressor, it would sound terrible and be missing most of the original data.
 
The difference is in what is done with the data each time the file is compressed: in a lossless engine the bits are preserved (for the sample rate selected). You could re-compress that 50kb/s file 1000 times and it would sound exactly like the first time you did it, because the file would be identical (assuming no errors in the method used).
 
However, if you took a 320kb mp3 and kept running it through various encoders, the quality would degrade with each pass. Why?
 
A lossless engine looks for patterns in bits and makes a reference table; if you keep seeing the same set of 50 bits over and over again, you store that as an 8bit (or whatever) segment in the table that you can keep pulling from and thus shorten up the space required.
 
A lossy engine looks for sections where they can remodel the way the waveform or data is being represented first, hopefully in a manner that is imperceptible, but uses less data to represent.
 
this is why a color .gif file can look so much worse than a .jpg (.gif is lossless, .jpg is lossy). There is more color information in a .jpg, but the .gif has a very definable way of storing it. If you expanded your .gif data table size (thus increasing the file size) you'd be allotted more color information (this is pretty much what a bitmap, raw, or targa is). Jpeg gets away with it by storing color hue separately of brightness because our eyes are much more sensitive to how light or dark a color is than what color it actually is. Each pixel gets lightness information, but color is stored in 2x2 or 4x4 segments and interpreted at decompression.
 
Make sense? I might be going a little too in depth on that, but it's a complicated subject. The short it that the mp3 will clip out data in busy spots or cut range to try to match a bit rate. If your source is a perfect repeating waveform (1000hz tone, or even a 20khz tone) the two will almost certainly do a fantastic job of compressing this without any perceivable loss. If it has a lot of other stuff mixed in (natural sound from a recording studio?) then the mp3 will make sacrifices to match its selected bit rate.
 
Starting from a 'pure' source, and same bit rate for both engines? I'd choose the mp3 for a one shot encoding. But we cheat: 320kb mp3 vs a 1000+kb flac, and in this case we usually start from a matched source (someone already made a digital data set on a CD), so we might as well just copy exactly what's on the CD so as not to tinker with the resulting bits at all.

 

[DrSpiv]

 
Quote:

I don't think the frequency will change in relationship to the distance of the mic unless you're talking about Doppler's effect.

Doppler is only if you are changing distance in relation to the microphone (moving toward or away from). I mean unless we're talking about spanning large enough parts of the universe that hubble's law comes into contention.
 
But distance from the mic can mean the volume is low enough that sound pressure does not carry enough momentum to overcome the microphone's static friction, so softer things given out as resonant or echoed frequencies from the studio room can be lost if the source (or reflected surface) is of sufficient distance away. Perceived sound pressure degrades with distance, just like light.