[Febs]

Quote:

Originally Posted by ObiHuang /img/forum/go_quote.gif On a sidenote, the people who say that audio and visual are completely unrelated could stand to take maybe a physics course. I feel many don't realize what sampling rate and bit depth mean. Each sample (1/44100 of a second) has to be represented in 16 bits. It's not some complicated artistry. These 16 bits are a bunch of numbers. Music is a just a lot of discreet numbers that express the amplitude and frequency of a tone. How do you anti-picture people react to that? Music and MATH related!? Blasphemy! Compression is just even more math. Look up "Modified discrete cosine transformation."

No-one has argued that audio and visual or completely unrelated. No-one has argued that music and math are unrelated. This paragraph has no bearing whatsoever on the topic of this thread or the arguments set forth by the people who advocate evaluation of codecs by listening rather than by use of spectrographs.

 

[liquidfireboy]

i think this thread needs explanation from a programmer of lossy codecs.

 

[Aerosushi]

Quote:

Originally Posted by liquidfireboy /img/forum/go_quote.gif i think this thread needs explanation from a programmer of lossy codecs.

An explanation was provided in my last post. Again, I strongly urge those who believe this to be a fair and objective codec test take their case to HydrogenAudio. There are several more DSP gurus there who would be happy to help.

@ ObiHuang - my apologies if this sounds harsh, but as Febs eluded, your post was close to meaningless drivel masquerading as science and has no bearing on the topic at hand. You simply fail to understand the relevance of the word "perceptual" in the term "perceptual coding" -> the point at which physics and the human auditory and visual systems intersect.

I would recommend a course in discrete time systems (sampling theorem, FIR & IIR filters, Fourier series and transforms including FFTs, DCTs and other fun algorithms, 1D & 2D perceptual coding). This type of course has been a staple of university engineering programs for at least 25 years. They are typically available in 3rd or 4th year. Correct me if I'm wrong, but I'm assuming you haven't taken one yet. I'm sure one is available at UCLA.

 

[liquidfireboy]

Quote:

Originally Posted by Aerosushi /img/forum/go_quote.gif An explanation was provided in my last post. Again, I strongly urge those who believe this to be a fair and objective codec test take their case to HydrogenAudio. There are several more DSP gurus there who would be happy to help. @ ObiHuang - my apologies if this sounds harsh, but as Febs eluded, your post was close to meaningless drivel masquerading as science and has no bearing on the topic at hand. You simply fail to understand the relevance of the word "perceptual" in the term "perceptual coding" -> the point at which physics and the human auditory and visual systems intersect. I would recommend a course in discrete time systems (sampling theorem, FIR & IIR filters, Fourier series and transforms including FFTs, DCTs and other fun algorithms, 1D & 2D perceptual coding). This type of course has been a staple of university engineering programs for at least 25 years. They are typically available in 3rd or 4th year. Correct me if I'm wrong, but I'm assuming you haven't taken one yet. I'm sure one is available at UCLA.

the idea is...I do visit HA and I understand what Febs and you are trying to say.

what I'm trying to say is, a programmer's words might carry more weight. Either way, I don't really care if other people want to use Blade.

 

[Aerosushi]

Quote:

Originally Posted by liquidfireboy /img/forum/go_quote.gif the idea is...I do visit HA and I understand what Febs and you are trying to say. what I'm trying to say is, a programmer's words might carry more weight. Either way, I don't really care if other people want to use Blade.

I understand completely where you're coming from. It's just unfortunate that some people's misinformed opinions will only be swayed by someone who develops music codecs and is part of the HydrogenAudio community. IMHO, it doesn't take a PhD in nuclear fission to determine the nasty health effects of exposure to radiation.

I've opted not to bring up my own relevant professional background on the topic at hand. I'm simply not into pulpit grandstanding. Sadly, I feel that these misinformed opinions will only be swayed through a conscious effort on their part to seek out the truth...

http://www.hydrogenaudio.org/forums/...p?showforum=30

http://www.hydrogenaudio.org/forums/...p?showforum=40

 

[Otaku-Cyber]

this is deep.....

stuff

 

[xinoxide]

actually, im going to look around even more.

 

[Earlephant]

Interesting thread, great graphs. I'm not sure what all the arguing was about but generally - to a newbie - it looked like lots of people were enjoying themselves. My take on 11 pages was: that you can only hear what you like by listening. The graphs were about fidelity to source. Different concepts equally valid.

 

[zip22]

equally valid, but one applies to what you hear, the other does not.

 

[DSlayerZX]

holy.... thx for the test result, know I know NOT to encode in ogg files and wav pro will do just fine compare to the high bit rate mp3s

 

[donunus]

Quote:

Originally Posted by Febs /img/forum/go_quote.gif No-one has argued that audio and visual or completely unrelated. No-one has argued that music and math are unrelated. This paragraph has no bearing whatsoever on the topic of this thread or the arguments set forth by the people who advocate evaluation of codecs by listening rather than by use of spectrographs.

The picture and the sound are related a little, especially with the high frequency extension but the thing here is that some of the better looking pictures ex. blade compared to lame doesnt tell the whole story. When an mp3 file shows a good graph that can look identical to a lossless/uncompressed original file, it's taking a hit somewhere else(since it does lose the majority of its data size)

In this case compare blade and lames sound quality, then tell me if you still care about these pictures.

If you want great pictures without sound quality loss combined then don't lose any data at all... use lossless

 

[donunus]

Quote:

Originally Posted by DSlayerZX /img/forum/go_quote.gif holy.... thx for the test result, know I know NOT to encode in ogg files and wav pro will do just fine compare to the high bit rate mp3s

It's because of statements like these that show us that this thread should be locked.
People are getting sidetracked by pictures without knowing the sonic qualities of the codecs themselves.
You should take these pictures with a grain of salt. The pictures show that there is missing data in spots of the spectrum for different files, some looking better than others. But... How does one know that this missing data is the reason why one codec sounds good and more identical to the original? You simply don't! ex. ogg really looks bad even at the highest bitrate but majority knows that ogg is a great sounding lossy format so what gives?

Think of it this way, one codec can get a truly great frequency response out of pink noise for example but that does not tell us how bad the codec is at transient response/time domain problems etc because pink noise is pretty much a constant sound. Therefore, if one wants to back up sound with pictures, its gonna take more than just this

 

[00birdy]

I'm a bit confused why you say a 320kbps VBR would be better than 320kbps CBR.

 

[Achmedisdead]

Quote:

I'm a bit confused why you say a 320kbps VBR would be better than 320kbps CBR.

 
Technically, there is no such thing as a 320kbps VBR. The "V" stands for "variable", which means the bitrate will vary according to the complexity of the music. There should not be a sonic difference between them, but the VBR file is a much more efficient one space-wise.

 

[00birdy]

Yeah that's why I didn't know why he mentioned 320VBR ...this is a old post so who knows maybe it was different back then.  When I convert with VBR most songs end up at around 290kbps.