why bother with 24 bit

[Cazzbar]

Quality posts Glen. Is it in anyway possible you (or anybody) could explain in basic terms how you can accurately reproduce a wave by sampling it at (just over) twice its frequency, basically twice per wave cycle? I've looked at a few math functions associated with the Nyquist Frequency and ermm... it's a bit heavy going!

(by the way I've been completely laughed out of town by all the computer geeks around here stating the principles of the Nyquist Frequency, they're convinced the more times you sample a wave the closer to the original you're going to get, without any idea of enough is enough! so it just shows how far and deep the misconception is!)

 

[SouthSIDE Glen]

Quality posts Glen. Is it in anyway possible you (or anybody) could explain in basic terms how you can accurately reproduce a wave by sampling it at (just over) twice its frequency, basically twice per wave cycle? I've looked at a few math functions associated with the Nyquist Frequency and ermm... it's a bit heavy going!

Man, to say it's "heavy going" is understating it quite a bit .

Personally I find information theory (which is what we're diving into with this stuff) to be even harder to understand than quantum theory. I'm not sure I can give such an explanation justice, since I only barely grasp the high concepts of it myself. But I'll give it a shot at the broad brushstroke level. If there's someone out there that knows this stuff better than I do - and there HAS to be! - please feel free to refine or downright throw out and correct any bad strokes I may make here...

As I understand it, the rough de-sampled wave information coming out of the DAC is run through a series of what are essentially narrow Q bandpass filters. These filters basically break the rough (unprocessed) wave into a harmonic series of "wavelets". Then some math functions rooted deep in heavy information theory (all that math that's over my and most other folk's heads) are applied to these wavelets and they're summed back together. What comes out of those equations when the waves are summed back together is a reconstruction of the original pre-sampled waveform.

Now, I may have some order of operation mixed up there; I'm just not sure if the equations are applied to the filtered wavelets or if the equations are part of the filtering that creates the wavelets, or something slightly different than either of those. But *I think* the basic building blocks mentioned and the basic process are more or less correct - as I understand it, anyway.

Where ol' Dr. Nyquist comes in is in the revelation that these info theory manipulations and equations (based on the work of a whole lot of other folks other than just Dr. Nyquist) require samples at rate only twice that of the original wave frequency in order to have all the information they need to reconstruct a clone of the original wave.

I still find it quite magical myself. But according to those who - unlike me - didn't schedule their college algebra and calculus classes at 8:30am, and therefore didn't sleep right through them, the magic is built right into the math; all one has to do is churn the numbers through to watch it happen.

G.

 

[TerraMortim]

I'm tired of arguing about it. It doesn't matter in the long run, as we're working on music not open heart surgery. Nobody is going to care about any of this shit once the music is out in the public's hands, so it's kind of dumb for us to be arguing about it. There is an audible difference to some people, regardless of what causes it, and what this guy says and the book over there says, doesn't change the way things sound in practice. Higher samplerates are smoother with software compressors and reverbs I find, and some software synthesizers have an unbelievable difference depending on the sample rate. It's night in day in some circumstances, and some it doesn't make much of a difference (most rock stuff, guitar bass most drum situations, 44.1 is about as good as you will get it to sound. Maybe a little extra air in the cymbals would be added at higher rates, but rock is always by nature a little bit lower-fi.). If you can't hear it, then fair enough, but just because you can't hear something doesn't mean it doesn't exist. The difference is there. Believe me, I've compared. Always could use more computing power.

 

[TerraMortim]

hmm. Dude, seriously. Chill out. I don't honestly care if I'm wrong or right. I could care less if the phase of the moon caused a 95 year old circus midget to get an erection, and in turn causes changes in humidity locally, causing me to feel more comfortible, which causes me to create a hit record that will cause me to buy my own small country, and retire drinking beers on the beach for the rest of my life. Kidding aside, I actually have researched a bit about it. I don't recall the Nyquist therum saying some of the things that you were saying (mainly the bit about it perfectly recreating the wave), then again, I don't know everything about it, beyond enough to get the concept of it, because generally in practice it's pretty boring, and doesn't make a bit of difference beyond a basic understanding of it, unless you are planning on developing your own hardware/software, which due to the fact that I have no skills, or desire to do so, won't be too terribly high on my to do list.

That very well may be about the converters. So what? If 96k sounds better (which as far as what I do...it does almost every time) I'll use it. However, there are some differences with some very specific types of effects, primarily compressors and ESPECIALLY reverbs. I'm glad that you admit that there is a difference in tests, and you know why it comes back so mixed? It's all subjective. All of it. Some people go for really open recordings, some people go for really aggressive walls of sounds...some people like distortions running into distortions and then thrown through a fuzzbox for good measure, and some people are mortally afraid to distort anything. I find that this whole sample rate debate is the new analog vs. digital that was plaguing the audio community with endless dick waving nerdfests, when really...NEITHER IS BETTER. Use what works for you, but don't deny that it makes no difference. For some things, it totally does, just the same that running an analog tape at a different speed (no I am not saying this causes the same effect) will give you very slight differences in sound for certain recordings. Some stuff doesn't make a bit of difference, just as in my analogy that much of rock, really makes little difference when you go up in sample rate. Audio sounds different using different sample rates and bit depths, no matter what causes it, and it's all about what works for what you want to do. I've used 44.1 for different purposes when it was perfectly sufficiant for it, and likewise, I've used others for other purposes, where 44.1 created a sound I didn't like.

No one is going to listen to a song allong with a sheet saying what settings and techniques were involved in creating it, unless it's another audio engineer, and with that said, most of the recordings that people in the industry idolize, had they been in the session, they would scoff at that/those technique(s) and most likely laugh the guys right out the studio door. Who are you making the music for? Other industry people? Chances are most people who hear it won't be, and won't give a shit if it was recorded with a tattoo gun etching waveforms into a rusted sawblade, by a trained monkey.

And what in the world are you talking about with my last statement? hahaha. I should wish for less computing power in order to understand audio? How very... scenester of you. lol You'd get allong with all the emo indie rockers out there, or black metallers. Who the hell WOULDN'T want more power in their setups? Your response to that sentance littereally made not a bit of sense. How did you get anything out of that statement other than, "I'd love to have more computing power"? How does the desire to always increase the ammount of power available to someone have anything to do with anything other than. "Gee, lets get more power as time goes on."

But yeah man, don't take shit so seriously. Chill. Physics aren't my life, music is. While I do know some of the physics of it all, it gets to a point where it's really just a bunch of nerds with pocket protectors bitching at each other and arguing about who has the thickest rimmed birth control glasses. Just make music and have fun!

Science H. Logic, can't we all just get alllawng?

*unzips my pants..* "My science is bigger than your science!"...

 

[SouthSIDE Glen]

First of all, Terra, you're responding to a post that I deleted about an hour before you responded to it, because I didn't feel like arguing with you. I had that post up for all of about 5 minutes; you must have caught it while it was still up. But to come back an hour after it was deleted and still respond to it with that rant is kinda ridiculous, don'tcha think? (Not to mention a continuation of what seems to be poor attention to detail.) Take a deep breath, bud. Relax.

Do what works for you. If you have a converter that sounds better at sample rate X than it does at Y, then by all means use sample rate X. Just don't expect to be able to go around spreading misinformation as to WHY it sounds better there without someone who happens to know better calling you out on it. There's already enough bullshit myths flying around the Internet.

G.

 

[Cazzbar]

To me this thread is about understanding how something works, anyone can do and use what they like if it makes them feel good, but if a friend said they're getting a new 5000 dollar quad core computer with 4 gig of memory and 28 terabytes of disk space solely because the salesman said it really speeds up sending and receiving emails, you'd probably question it.

 

[thavva]

There's no drop in quality. The 24-bit files are left alone, no change there. As for the 16-bit files, they are either converted to 24-bit or they arent. If they arent, no change there. If they are, all that conversion does is stick 8 zeros onto the bottom of the value; no actual change in quality there, either.

Wht you might want to double-check, though, is to make sure the addition of 16-bit files to the project doesn't somehow cause any settings to change in the project itself. For example, make sure it's still default set to save your mixes at 24-bit and that the default setting has not accidently changed to 16-bit or something like that. If so, no big deal, just change it back.

G.

Thanks alot G. and Danny. That clears up things. Now I can concentrate on my work and not think so much. Thanks again guys.

 

[Cazzbar]

I'm personally very grateful for Glen shining a clear light on this area, I can now happily sample at 44.1 knowing I'm recording perfectly all the frequencies I can physically hear (dependent on mics, preamps and convertors etc...) Previously I would be thinking double the sample rate and double the clarity. But if I do decide that 96Khz gives me a 'livelier' sound then the reasons will be far more mystical than just thinking I've given my DAC more dots to join up, therefore it must be better.

 

[noisewreck]

I actually wish recording systems would move away from this whole nonsense of PCM encoding. Then we could all happily use 1bit.

 

[SouthSIDE Glen]

But if I do decide that 96Khz gives me a 'livelier' sound then the reasons will be far more mystical than just thinking I've given my DAC more dots to join up, therefore it must be better.

Thaks for the kind words, Cazz .

This is a tough subject to deal with because of the very fact that the real mechanisms used are not naturally intuitive. As anyone who tries looking this stuff up can tell us (like you did), it all looks like a bunch of gobbledygook in the form of sophisticated algebraic equations. The sample rate-as-resolution idea, OTOH, is far more intuitive, and can be laid out pretty easily in one fairly simple picture. While that makes it an attractive idea, it doesn't make it an entirely correct one, unfortunately.

I highlighted that quote on top because I did want to make clear and reiterate that there are those that argue that 44.1kHz is not necessarily the be-all and end-all of sample rates. There are other factors involved that can argue for a higher rate. These usually fall into one of two general categories:

1) There are practical physical and technical design considerations that make strictly adhering to Nyquist frequencies difficult to cleanly accomplish. Noisewreck and others hit on a couple of these a few posts back.

2) There is a use for some frequencies in the 20kHz-32kHz range. Even though 99.5% of us may not directly be able to hear much of anything above 16kHz, frequencies and harmonics in the 20-32kHz range do contribute a delicate "air" to the audible sound.

Point 1 is pretty much a given. Point 2 is a subject of debate to this day. There are those of us who think that a sample rate halfway between 44.1 and 88.2kHz would be ideal because it would address both of those issues without providing sample rate overkill that unnecessarily eats up both processing bandwidth and information quality.

That, however was not the point of the direction this discussion had taken, as you rightly understood. The qustion was, would an increased sample rate above 44.1kHz alone increase the "resolution" of a 20kHz signal?

And the answer is no; increasing the sample rate alone will increase the functional bandwidth - it will increase the frequency response, but it won't increase the quality of frequencies already covered.

G.

 

[noisewreck]

So, in layman's terms you can have a large pizza cut in 10 slices or 16 slices... you still end up with the same amount of pizza. If you want more pizza, get an Xtra Large.

Pass the pepper please!

 

[NL5]

I highlighted that quote on top because I did want to make clear and reiterate that there are those that argue that 44.1kHz is not necessarily the be-all and end-all of sample rates. There are other factors involved that can argue for a higher rate.

Just wanted to quote this. The thread got pretty long, and I think some people missed some stuff.

44.1 is all that is needed to reproduce the audio band. BUT, due to the inherent shortcomings in the converters, there ARE reasons to go above 44.1. Everything above 60khz or so isn't gaining you much, and above 96 may be hurting your audio. If you don't understand this, go Read Dan's article and the thread again.........