This debate is tired, but Noah's post really made me laugh.

.....seriously though, if you can't make a good sounding digital recording, no matter how hard you try, you really shouldn't blame the tools.

math means not physical. you can use math to describe whatever you want, but digital actually alters the signal by doing it. not caring about fidelity is a moot point in a discussion about sound quality. i've listened to a lot of both, and one is better than the other. purely subjective. hmm. are you not saying you prefer digital? or have you avoided stating your preference?

-noah

i'm saying that i like music, and i like to listen to great music, and for me, it having been recorded digitally is not a dealbreaker. as far as a preference, in a vaccuum (metaphorically, obviously), solely on the basis of sound "quality" or "character", sure i'd pick an all-analog setup, the most brilliant of boutique gear that costs a zillion dollars, designed and built to maintain the best elements of ancient gear, modern gear, future gear, whatever. that's not the least bit realistic nor practical for me. i am not a zillionaire.

in the context of the original question though, i think a statement like "all digital always sucks" or whatever it was you were saying, i think that doesn't really do much to answer the question. it's like, "what's the difference between the taste of mushrooms and pepperoni" and the answer is "mushrooms suck". talk about texture, or flavor, or something. i dunno.

i like digital for its clarity. i like shitty analog for it's charm. digital lacks character, or rather it's noteworthy character is more an absence of character. shitty analog is probably better than marginal-quality digital in that it has a nice lo-fi charm to it. lo-fi digital generally sounds either dry or harsh, depending on whether it's clean or clipped. if your goal is to have every instrument be distinct and separate, or you wanna have a bunch of tracks for cheap, i'd pick cheap digital over cheap analog. if you're looking for an old-school sound, go with cheap analog.

this is my attempt at addressing the original question in the context i think is appropriate. considering that in its nature and presentation, the question suggests the person posing it is a novice when it comes to recording. so i assume there's not a giant budget involved.

if i had little-to-no money, and i already owned a computer, i would illegally download a pc-based multitrack audio program, and buy a cheap dynamic mic and get a used mixer or something (i used to use my porta one as a mic preamp when i first got a digital setup. oh yeah, it sounded *great* ). but this would be great for composing multitrack music, like building songs with several tracks, to put songs together, to build them, to hear what the interaction between instruments sounds like. or to make a really cheap demo.

if i had thousands of dollars, i'd buy a sweet analog setup and spend a bank on reels of tape. if i had enough to buy an adat, i'd buy a reel-to-reel deck instead because i hate the sound of alesis adat based on my experience working with it. i do like the sound of a nice analog recording, even a good lo-fi analog recording, something you might make with a 1/4" 8 track machine like the fostex ones i've seen on ebay when looking at what kinda 8 and 16 track tape machines are out there. i've make a bunch of recordings using a tascam porta one that i'm comfortable listening to. i've also made recordings on a pc that have more clarity, and especially a more easily-controlled low end. but yeah, they lack the non-quantifiable "character" of analog tape.

so there, this is my preference, give or take.

as far as the difference in sound between digital and analog, i'd like to point out that i think the room acoustics, sound of the instruments being recorded, quality of microphones being used, placement of microphones, quality of gear throughout the signal path, refinement of the engineer's ears... all of these things mean far more than whether your storage medium is digital or analog, and that's a very important thing to consider when asking this question about which medium is better: is everything else lined up to the point where digital vs analog is that relevant? and digital vs analog has a lot to do with quality of D/A converters, too. let's not pretend that all digital sounds equally "worse" than all analog. the highest-quality digital crushes the lowest-quality analog, in terms of "sound" or fidelity. anyone who denies that is a fool in my book.

and doude. "math". here's one definition of it, from my ever-loved dictionary.com :

That science, or class of sciences, which treats of the exact relations existing between quantities or magnitudes, and of the methods by which, in accordance with these relations, quantities sought are deducible from other quantities known or supposed; the science of spatial and quantitative relations.

if you think math isn't a key player in analog recording, i'd like to ask you to talk for a moment about room modes, harmonics, sabins, decibels, frequency response, phase, electricity, magnetism... i gotta stop there, because i assume you're trying to make a point with your use of the word "math" that i'm not understanding, but i know we could all come up with plenty of other examples of how music, recording, acoustics, all of it is very much in bed with "math"...

cheers.

my name is noah, too. and i say:
do the best you can with whatever you can get our hands on. Just focus on making/ capturing/recording music in a way that is expressive, compelling and moving. use it all. use whatever you have access to. The more you listen, the more you'll hear. You'll hear differences in the way analog and digital recordings sound to you. Play with really basic digital devices like samplers, play with basic analog 4 track cassette recorders. play with casios and play with steinways. (By the way, no matter how you record a steinway, it will never sound like a kawai, noah. And a Kawai will never sound like a Casio. Everything, it seems, insists on having it's own sound.)
Try to remember that sound is fluid. Every wonderful sounding record sounds completely different from every other wonderful sounding record. That's what makes them special. And respect your own opinions about all of this. Everyone has their own opinions. Just try not to post yours on forums as if they are some kind of righteous factual truth, cause that's just annoying.

Another thing to point out in this debate is that digital improves all of the time. A lot of people get a bad impression of digital because they're recording at 16/44.1, or possibly at 24-bit and forgetting to dither it when going down to 16-bit.
And then there's 24/96kHz. Have you HEARD that? The faster the sampling rates, and then more complex the digital word (the bit depth) the more accurate recordings will be. I'm willing to bet that in 15 years analog machines will be VERY hard to find as digital will be so good that it would be a waste of time to keep messing with analog machines.
And I'm a big fan of analog too. Even though a computer itself is an amazing thing, a big ol' 24-track machine is an INCREDIBLE piece of machinery. How can a machine with moving parts be so damned accurate in speed and everything?? Very cool indeed, but I still think that digital taking over is inevitable.
Another thing I do like about analog though, is that the entire signal path is very fluid. Everything going to tape is analogous to what comes in, as far as the materials permit. Like a guitar through an amp un-hindered by digital stomp boxes, you're literally pushing electrons down the cable like something hydrolic , except not with water but electricity.
Still, eventually I think sampling rates will exceed the rate at which magnetic particles can align themselves to electronic impulses.

sndo wrote:And then there's 24/96kHz. Have you HEARD that? The faster the sampling rates, and then more complex the digital word (the bit depth) the more accurate recordings will be.

This is pure misinformation. On high end converters, the difference between 44 khz and 96 khz is minimal. It is prosumer converters that are showing larger differences between 44 khz and 96 khz, as they use the higher sample rates to push the poor filtering artifacts outside of the human hearing range. This does not happen with our Lavry and Weiss converters, where 44 khz sounds remarkably similar to 96 khz due to proper filtering techniques.

And about more bits equalling better recording quality: are you aware that the best preamps have a dynamic range of approximately 90 dB at peak performance? This is about 30 dB less than the 120 dB of dynamic range found in a decent 24 bit converter. Adding more bits to the converter will only result in recording more of the noise floor beneath the preamp. Unless you plan on recording your peaks at -40 dBfs, 32 bit converters would be absolutely ridiculous.

out of curiosity, could you explain some of this a little more? i don't really know so much about digital, but it sounds like you do. there's two things i'm not so clear on.

first, how could more than doubling the sample rate not make any appreciable difference? maybe most people can't easily identify the difference between an 8KHz triangle wave and a 12KHz sinewave (or maybe they can), but hypothetically, if my music included such infomation, wouldn't a higher sampling frequency be much better equipped to try and capture it, and then present it more closely intact? for things like duplicating the exact timbre of the cymbals on a drumkit, isn't the quality of sampling these higher frequencies very important? and doesn't more-than-doubling the sample rate do a noticably better job, even on the best of gear?

and as far as bit depth, isn't it useful to have more than 16 bits, so that the information recorded near the noise floor has better resolution to it? i once had someone explain the significance of bit depth to me, and he was saying how, sure you might have a big dynamic range with 16 bits, but the quietest stuff is pretty much useless. like, to think of it as, if you cranked it up really, really loud, your quietest passages would be so clearly audible, but they'd have such little resolution to them that they'd sound like utter crap. because the way digital makes "loud" or "quiet" is by having more 1's or more 0's in those bits... but the quietest of your quiet, the bottom end of that dynamic range, you've only got, say, the last two bits of actual data that you're working with, the other 14 bits are all zeroes all the time... something like that.

i dunno, the way i had it explained to me once was that the point of adding more bits isn't to get more dynamic range, like lotsa people would like to think it is, but rather to make the lower area of the existing dynamic range more useful. cause the bottom area of your dynamic range should pretty much be ignored as useless, in the digital world. i may have totally misunderstood what he was getting at, or the guy explaining it may have been totally wrong in the first place.

if you could clear up either or both of those issues, that would be awesome!

I'll try to answer some of this. I'm certainly no expert but I've done a bit of research on this stuff....

toomanyhelicopters wrote:first, how could more than doubling the sample rate not make any appreciable difference? maybe most people can't easily identify the difference between an 8KHz triangle wave and a 12KHz sinewave (or maybe they can), but hypothetically, if my music included such infomation, wouldn't a higher sampling frequency be much better equipped to try and capture it, and then present it more closely intact? for things like duplicating the exact timbre of the cymbals on a drumkit, isn't the quality of sampling these higher frequencies very important? and doesn't more-than-doubling the sample rate do a noticably better job, even on the best of gear?

There are two things wrapped up in this question: capture/reproduction and perception of high frequencies. The Nyquist theorem says that for a given frequency, a sample rate of more than twice the frequency will allow you to perfectly reconstruct the wave form. In practice, there are problems with reconstruction filters and such (like mparker said), which may push this a bit higher than *just* twice the frequency, but in theory you should be able to reconstruct a 20kHz wave perfectly with a 40.00001kHz sampe rate. Having 192kHz won't allow you to reproduce it any more accurately. This is a very non-intuitive aspect of digital audio that most people don't grasp immediately.

How high of frequencies we need to capture/reproduce depends on a lot of things. How high can humans hear? Probably 20-25kHz is plenty, although there is some research that suggests higher frequencies are perceived in some way. Another thing to consider is that there is almost always some other component in the chain which limits high freq response in addition to sample rate (mics, speakers...). However, some digital processing, pitch shifting for example, works much better with higher sample rates.

toomanyhelicopters wrote:and as far as bit depth, isn't it useful to have more than 16 bits, so that the information recorded near the noise floor has better resolution to it? i once had someone explain the significance of bit depth to me, and he was saying how, sure you might have a big dynamic range with 16 bits, but the quietest stuff is pretty much useless. like, to think of it as, if you cranked it up really, really loud, your quietest passages would be so clearly audible, but they'd have such little resolution to them that they'd sound like utter crap. because the way digital makes "loud" or "quiet" is by having more 1's or more 0's in those bits... but the quietest of your quiet, the bottom end of that dynamic range, you've only got, say, the last two bits of actual data that you're working with, the other 14 bits are all zeroes all the time... something like that.

Well, what your friend says does make sense. If you record something at a very low level you basically loose bits. But it really is analogous to what happens in the analog domain. Say you record something digitally at a low level and then use software to add gain. That software *should* be adding dither noise to the other bits and what you get is a lower signal to noise ratio. The same thing happens in analog when you record something soft and then try to increase the volume...you hear tape hiss. For practical purposes I think this is more of an issue of gain staging I think in both domains.

For final mixes, and most modern recordings, hardly anyone has enough dynamic range to use up all 16bits, but I'm not sure this is really a justification for saying 16 bit is enough. For multitracking, I personally see a lot of advantage for 24 bit, because you have so much dynamic range that you can record at say -10dBfs and not worry about it. Also, almost any processing you do to the digital audio will benefit from a higher bit depth.

Sorry if that was a bit rambling and nonsensical...I'll try to clarify if you want.

i still need help here. i've heard about the whole Nyquist frequency thing, but i still can't wrap my brain around it. with only two samples, it seems utterly impossible to think you could recreate exactly a waveform as it was before you sampled it. for example, what if the frequency you're talking about is 10K and your sampling rate is 20K. and what if your two sample points fall exacly at the node of the waveform? so both samples would be 0? also, with only two samples, how could you possibly know the shape of the waveform? if you're talking only about sinewaves (since complex waveforms are really just collections of various sinewaves all thrown on top of each other in varying amounts) then maybe. but any waveform more complex than a sinewave (say a 12KHz wave that's somewhere between sawtooth and square in its shape)is going to be made up of all kinda overtones that are way outside the range of your sampling frequency, no?

is there somewhere online anybody could recommend that clears this up in a way that makes sense to me? cause all i can picture right now is anybody who tries to explain this to me getting really frustrated with my inability to get it.

http://www2.cs.ust.hk/faculty/layers/co ... quist.html

That page has some 'pretty pictures' to help illustrate sampling theory. When describing a waveform mathematically, the wave period is the time for the wave to rise from null to it's highest positive level, fall back to null, sink to the lowest negative level and return to null (or the reverse). That entire motion makes a single cycle. To accurately sample this wave, we need at least one point on the positive part and one on the negative and thus two samples per cycle.

Of course, the highest frequency allowed will only be represented as a sine wave. Your concern about the two samples of this frequency falling on the nulls is valid, and the technique of oversampling aims to alleviate this problem - the odds of 128 samples all falling on the null is more remote than 1 or 2 times oversampling. Other problems exist for frequencies at or near 1/2fs, mostly induced by the lack of perfect brick wall low-pass filters. Any sort of aliasing introduced here become very prominent in the lower part of the audible spectrum, and thus filters often intrude a little into the upper reaches of the sampling range.

A particular theoretical problem concerns the representation of square waves as the more accurate reconstruction of these requires an increasing number of harmonics. Take a 10 khz square wave in a 48khz sampling rate as an example. The number of harmonics available is severely limited and thus the resultant signal will resemble a sine wave more than a square wave. However, this mirrors the commonly held ideas about human auditory perception, but lots of disagreement and contradictory research cloud this model. Many people might not discern between a 10khz sine and square wave, some might, and others might find them too annoying to even bother listening to at all.