Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[3D Audio]

RedBus wrote on Fri, 03 June 2005 00:44

The Myth is basically that DAWs sound different because of the way their summing is implemented. The CD proves that given the same input files and the same fader levels, you get the same data (bit-for-bit accurate) out of most DAWs.  If you still hear a difference, it is because of other parts of your system.

The part we never figured out (and possibly never will) is why people felt they could clearly hear differences in bit-identical files. I did the math once and not one single value of one 24-bit word among 3 million words (I think it was) was different at all. Yet people still thought they could hear the difference.

I did not read the whole thread so pardon me if this comment is out of line.

[Ronny]

3D Audio wrote on Mon, 26 September 2005 20:53

RedBus wrote on Fri, 03 June 2005 00:44 The Myth is basically that DAWs sound different because of the way their summing is implemented. The CD proves that given the same input files and the same fader levels, you get the same data (bit-for-bit accurate) out of most DAWs.  If you still hear a difference, it is because of other parts of your system. The part we never figured out (and possibly never will) is why people felt they could clearly hear differences in bit-identical files. I did the math once and not one single value of one 24-bit word among 3 million words (I think it was) was different at all. Yet people still thought they could hear the difference. I did not read the whole thread so pardon me if this comment is out of line.

Very easily explained. The mind tends to fill in blanks when there are none. Our ears are great, but our short term audio memory isn't so hot and we know very little about the audio cortex and how other senses, preconceived thought patterns and mood affect the sense of hearing, or at least the information derived from it. The design of the ear is a great device, but it still has the brain, that has to coherently assemble the information into meaningful data or the lack of data and that's where the problem lies. It's harder to train your ears not to hear differences when there are none, than it takes to train the ears to hear differences when there are some. You have all kinds of non audio factors that enter the equation, one is the egotistical notion of not admitting hearing anything, when there is nothing to hear. I call it golden ear syndrome.

In response to RedBus' statement, if you still hear differences, it's likely that it has more to do with your own system, than differences in the audio one, IMHO.

[danlavry]

RedBus wrote on Fri, 03 June 2005 00:44

The Myth is basically that DAWs sound different because of the way their summing is implemented. The CD proves that given the same input files and the same fader levels, you get the same data (bit-for-bit accurate) out of most DAWs.  If you still hear a difference, it is because of other parts of your system. If we start to discuss the implementation in the digital world of the flaws of a particular analog circuit, then the sky is the limit and people will have a right to their opinion if they say the digital version doesn't sound correct, etc. Implementation in digital of the non-linear processes that sometimes goes on in the analog world (e.g. opto-coupler in a compressor) does bring us back to the trade-offs, good and bad, that people make. One cardinal matematical rule about such implementation is that you have to deal with the higher frequencies generated in the process. Usually this means upsampling the incoming data and performing the whole simulation at a higher rate, then filtering it back down to the original rate.  The quality of this conversion dictates the accuracy of the model, assuming that your model was mathematically close to the analog process to start with. But, one shortcut you can take is to not do the upsampling. That guarantees you get aliased audio when you do your analog simulation, but it may end up being an entirely new effect that you want to keep. I'm sure that this wouldn't work for all audio sources you throw at such an algorithm (assuming you were going for a model of a smooth opto-compressor instead of a grungy distorted comp), but it opens up new paths. RedBus.

I do not wish to argue with "new path" but let me point out that alaising takes frequencies and moves them to frequency locations that are "highly non musical" in the sense of thier relationship to the intended (non aliased) sound.

A very simple example: Take a 5KHz sound with harmonics at 10KHz, 15KHz, 20KHz, 25KHz and 30KHz. Say you ampled at 44.1LHz so Nyquist is at 22.050KHz. The 25KHz is aliased to 19.1Khz and the 30KHz is now at 14.1KHz.

With an analog system limited to say 22KHz the sound you can hear is made of 5, 10, 15 and 20KHz. Very harmonic sound.

With the aliasing you have 5,10,14.1, 19.1 and 20KHz. I would say that the aliasing is non musicaly related to the energy.

Of course, when having many instruments together, we have a more complex picture, but I think the concept is the same. I have listened to some non oversampling AD's with the aliasing protection (filters) intentionaly removed. It is really bad.

Regards
Dan Lavry
www.lavryengineering.com  
 

[RedBus]

I only mentioned the aliasing as being an effect in the context of those users seeking a "grunge" sound effect for Guitar or Keyboard parts. There is no doubt that aliasing is horrible for ordinary audio signals,  but the non-musical result can be utilized by some people and may be appropriate to some forms of composition (I'd prefer not to call it music!)

Regards,
RedBus.

[Johnny B]

Dan just said,

"With an analog system limited to say 22KHz the sound you can hear is made of 5, 10, 15 and 20KHz. Very harmonic sound.

"With the aliasing you have 5,10,14.1, 19.1 and 20KHz. I would say that the aliasing is non musicaly related to the energy."


By golly, Dan, I think you are on to something there. So, this begs the question...What if you redesign the entire process so the harmonics  are more even...rather than odd?...

That's always been one of the things cited to explain why people love their analogue gear...the pleasant distortion is more musical because it works well with what people find harmonically pleasing...Now wonder if digital sytems could be adjusted or redesigned so that harmonics where more even than odd...

I'm just doing a little green light thinking here...not at the red light stage...(That's the stage where people say "No, that won't work because...." )...The green light stage is where you just dream about things considered impossible to do as being possible...some people call it "Imagineering."

Anyway, let's think about that aliasing thing that Dan mentioned as giving us the nasty odd harmonics instead of the pleasant sounding good even harmonics...Now if any of this were true, what would you do to go about fixing it? How would you nail down that alaising bit to get it to behave itself in a more pleasant way?  

The exercise here is not whether you are right or wrong...The exercise is merely to spur some ideas, perhaps some new, fresh ideas, that might be worth pursuing later on...

[Jon Hodgson]

Johnny B wrote on Sun, 02 October 2005 08:01

Dan just said, "With an analog system limited to say 22KHz the sound you can hear is made of 5, 10, 15 and 20KHz. Very harmonic sound. "With the aliasing you have 5,10,14.1, 19.1 and 20KHz. I would say that the aliasing is non musicaly related to the energy." By golly, Dan, I think you are on to something there. So, this begs the question...What if you redesign the entire process so the harmonics  are more even...rather than odd?... That's always been one of the things cited to explain why people love their analogue gear...the pleasant distortion is more musical because it works well with what people find harmonically pleasing...Now wonder if digital sytems could be adjusted or redesigned so that harmonics where more even than odd... I'm just doing a little green light thinking here...not at the red light stage...(That's the stage where people say "No, that won't work because...." )...The green light stage is where you just dream about things considered impossible to do as being possible...some people call it "Imagineering." Anyway, let's think about that aliasing thing that Dan mentioned as giving us the nasty odd harmonics instead of the pleasant sounding good even harmonics...Now if any of this were true, what would you do to go about fixing it? How would you nail down that alaising bit to get it to behave itself in a more pleasant way?   The exercise here is not whether you are right or wrong...The exercise is merely to spur some ideas, perhaps some new, fresh ideas, that might be worth pursuing later on...

There is nothing new in what Dan just said, it is basic sampling theorem.

Preventing those alias components is what the input filter on an ADC is all about.

[Johnny B]

Yeah, but he's talking about those harmonic aspects, isn't he?  I could be wrong, I often am.  

[Jon Hodgson]

Johnny B wrote on Sun, 02 October 2005 19:08

Yeah, but he's talking about those harmonic aspects, isn't he?  I could be wrong, I often am.

He's saying that if you don't bandlimit your signal before you sample it then the harmonics will fold back into non-harmonic frequencies, and sound awful.

Nothing to see here folks, move along please...

[Johnny B]

Ok, I knew that....but when we look at the phenom more closely it may have some other possibilities....ie sometimes you can exploit bad behavior to your advantage...maybe I'm not making myself as clear to you as perhaps I could if I had more time...Nevermind, don't give it a second thought

[bobkatz]

RedBus wrote on Thu, 29 September 2005 15:41

I only mentioned the aliasing as being an effect in the context of those users seeking a "grunge" sound effect for Guitar or Keyboard parts. There is no doubt that aliasing is horrible for ordinary audio signals,  but the non-musical result can be utilized by some people and may be appropriate to some forms of composition (I'd prefer not to call it music!) Regards, RedBus.

Aliasing is the worst-sounding kind of "grunge" you can imagine. Well, then again, some of the sampled cymbals you get out of synthesizers sound very aliased.

Prior to digital recording, most of the analog distortion that we heard was very harmonically related, shall we say "pleasant sounding grunge". The problem with aliasing is it is very foreign and ugly to the ear. Create your grunge the old fashioned way, please  

BK

[Johnny B]

bobkatz wrote on Sun, 02 October 2005 23:08

Aliasing is the worst-sounding kind of "grunge" you can imagine. Well, then again, some of the sampled cymbals you get out of synthesizers sound very aliased. Prior to digital recording, most of the analog distortion that we heard was very harmonically related, shall we say "pleasant sounding grunge". The problem with aliasing is it is very foreign and ugly to the ear. Create your grunge the old fashioned way, please   BK

Yes, that's right on the money.

[jimmyjazz]

johnny, you cannot change the alias frequencies.  They are inextricably linked to the sampling rate and the frequencies which are being sampled but which should have been filtered out.

[C-J]

Johnny B wrote on Sun, 02 October 2005 10:01

Dan just said, "With an analog system limited to say 22KHz the sound you can hear is made of 5, 10, 15 and 20KHz. Very harmonic sound. "With the aliasing you have 5,10,14.1, 19.1 and 20KHz. I would say that the aliasing is non musicaly related to the energy." By golly, Dan, I think you are on to something there. So, this begs the question...What if you redesign the entire process so the harmonics  are more even...rather than odd?...

Dan is talking about harmonic overtones of the 5k when he mentions 10k (even), 15k (odd) and 20k (even). The aliased 14.1 and 19.1 are neither even, nor odd harmonics, but non-harmonics.

Quote:

That's always been one of the things cited to explain why people love their analogue gear...the pleasant distortion is more musical because it works well with what people find harmonically pleasing...Now wonder if digital sytems could be adjusted or redesigned so that harmonics where more even than odd...

Well,,, if your heads on straight, with today's equipment you record at 24-bits, and leave 'nuff headroom to completely avoid all unpleasant digital distortion.

C.J.

[Johnny B]

Hmmm, who here is not conversant with the ins and outs of the harmonic overtone series?...both in the perfect form and the impefect well-tempered form..?.

I suppose I could look up the formulas for the 15th billionth time or see if I could dig up my excel spreadsheets which take them out to the trillionth harmonic...but I'm far too lazy to do that right now.

I can say, the calcs are not that difficult.

[Jon Hodgson]

Johnny B wrote on Fri, 07 October 2005 06:14

Hmmm, who here is not conversant with the ins and outs of the harmonic overtone series?...both in the perfect form and the impefect well-tempered form..?. I suppose I could look up the formulas for the 15th billionth time or see if I could dig up my excel spreadsheets which take them out to the trillionth harmonic...but I'm far too lazy to do that right now. I can say, the calcs are not that difficult.

And your point is??

Anyway it sounds like you are confusing harmonics with scales, not the same thing. Harmmonics are integer multiples of the fundamental, very easy calculations.