Audio channel question

[maxhondur]

Hi,

Suppose my dvd has a regular stereo track and 5.1 channel audio track. Is a 5.1ch, 448kbps ac3, downsampled to two channels, equivalent to a 448kbps ac3 stereo file, in terms of audio quality? I'm guessing that the two can't be directly compared, but surely, the former would have a higher quality audio than a 2 channel 192kbps ac3.. and of course, the stereo 448kbps ac3 is higher quality than the 192kbps file. Do we have any tools to quantitatively say what factors play into better audio quality, when it comes to looking across files with a different number of channels?

[Skelsgard]

Audio = number_of_channels + sample_rate + bit_depth + bit_rate.
Ex: 6-ch 48kHz 16bits 448kbps. The SR, BD and BR are the parameters that define the quality of an audio track. The number of channels are not important in a mathematical way, but in a sensorial way. So leaving the ch_number aside, and 2 of the other parameters being equal, the higher value in the 3rd parameters would represent a higher quality track.
I.e. "48kHz/16bits" is of less quality than "48kHz/24bits".
44.1kHz/16bits is lesser than 48kHz/16bits. (Not that the average human ear can tell the difference in this particular cases).

5.1ch, 448kbps ac3, downsampled to two channels
It would depend on the encoding bitrate u choose for the downmixed 2-ch file. If the bitrate is also 448kbps (as the "448kbps ac3 stereo file") and the "448kbps ac3 stereo file" was downmixed with the same matrix, then the files should be exactly equal in quality.

Now, in WAV (lossless, raw format) every channel has asigned an amount of bits, so the more channels, the bigger the file. 1Hz/1bit/1-ch uses 1 bit of space for every second. Therefore 48.000Hz/16bits/2-ch would use 48000 x 16 x 2 = 1536000 bits/sec--> /1000 = 1536 Kbits/sec --> /8 = 192 KBytes/sec.
1 min = 192 KBytes/sec x 60 sec = 11520 KBytes.
And 48.000/16bits/6-ch = 48.000 x 16 x 6 /1000 /8 = 576 KBytes/sec. 1 min = 576 x 60 = 34560 KBytes.

But in compression, the KB/sec are set for the entire amount of channels. Therefore when u assign 448kbps to 2ch tracks, each channel will have 448/2 = 224 kbps to use, whereas with a 6-ch track, each channel will only have 448/6 = 74 kbps, wich is less then for 2-ch.
Imagine u have a 6-ch track where FR and FL are identical to R and L in a 2-ch track. FR = R and FL = L. And u compress both tracks at 448 kbps. FR and FL get only 74 kbps to be adjusted to while R and L get 224 kbps each. That means that a 448 kbps 2-ch track will be mathematically of higher quality since it had to throw away less info than the 448kbps 6-ch track to achieve the bitrate per channel asigned to each one.

A 192 kbps 2-ch (96 kbps per channel) would be similar in quality to a 512 kbps 6-ch (96 kbps per channel), mathematically speaking. In 5.1 tracks, AC3 uses (I think) a 5 1/3 channel bitrate asignment (1 for each discrete channel (x5 channels) + 1/3 for the LFE (X1 LFE) --> [96 x 5] + [96 x 1/3] = 512 kbps). For 448 kbps it would 80 kbps/channel ([80 x 5] + [80 x 1/3] = 448).

Hope this helps to clarify.:)

[maxhondur]

Thanks for the reply.. I think that clears things up a bit. So if I'm getting this correctly:

If I have a 448kbps 5.1ch ac3 file, and play it back with foobar to my (2-channel) headphones, the quality will be roughly like playing the 448kbps 2.0ch ac3 file that was offered on the same dvd. However, the assumption is that all 6 channels were used in an efficient manner, given that a fixed number of bits is allotted to each channel. If any of the channels were not used, then this would seriously degrade the quality of the music.

Is this train of thought incorrect? It's been a week, and I'm still scratching my head. :)

[Skelsgard]

If I have a 448kbps 5.1ch ac3 file, and play it back with foobar to my (2-channel) headphones, the quality will be roughly like playing the 448kbps 2.0ch ac3 file that was offered on the same dvd.

No. When dealing with compressed audio, always think about "bitrate-per-channel". That will basically determine the quality.
The more space u have to put things into, the less u have to throw away. Always mathematically speaking.
Remember: the 2-ch has 224kbps/ch while the 6-ch has 80kbps/ch. And the downmixing of the 6-ch must be the same as used on the 2-ch file.
But above ~96kbps-per-channel almost all encoders (formats) sound the same. So they will sound pretty much the same (the 2ch and the downmixed-6ch).


If any of the channels were not used, then this would seriously degrade the quality of the music.
Not at all. Remember: BITRATE PER CHANNEL. If a channel exist but doesn´t have any info (let´s say a mute channel), then the bitrate asigned for that channel just won´t be used (cause there´s nothing to encode in that channel).

There are other parameters like channel coupling that alter the relation bitrate/quality, but that´s a little more advanced and imply technical issues wich go beyond my knowledge... I would be talking out of my ass if i thought myself capable of explaining them to you. :)

[maxhondur]

Hi,

I just got back from vacation.. sorry if this post is too old to reply to. Here's my question, I guess :)

It was my thought, that in a 448kbps ac3 file, there is also an issue of well-defined separation and how it can change the quality of a file. For example, if a two channel 448kbps ac3 attempts to mush together and capture 5 singers with its alloted bitrate all at once, a well-mastered 5-channel ac3 will capture each individual singer(ideally, anyways) per channel, which could be much different if not better than the former.

Is this line of thinking also false (or is this what you meant by channel coupling)? :) Again, I know that I'm a little blindsided on this topic, and I really appreciate the input.

[Skelsgard]

a well-mastered 5-channel ac3 will capture each individual singer(ideally, anyways) per channel, which could be much different if not better than the former.
If each voice is encoded into a separate channel, then they WILL be reproduce ONLY in the respective channel (it will be separate). AC3 5.1 is Discrete (that means each channel really exists), also DTS 5.1. AC3 2.0 (DPL II matrix-encoded) is Discrete as it carries 2 discrete separate channels, but is not discrete as 5.1 channels since these do not truly exist as separate channels but they´re decoded thru a matrix.
Channel coupling means that u couple 2 channels to be encoded together ("together" does not mean that they are merge together to 1 (one) channel). Here some info: http://wiki.hydrogenaudio.org/index.php?title=Channel_coupling