Recording at 44.1 or 48

[in general the higher the sampling rate, the better.]

More thoughts on higher sampling rates..........

Analog tape recordings (to me) sound superior to digital recordings. How can that be? Analog sound is a continuous recording. Not approximations from slices of sound. It makes sense that having more slices of sound will resemble a full continous flow of analog tape. An argument here: each sample introduces artifact sound as well, and with thousands of slices, the presence of artifact sound might be discernable. This is just a thought. Maybe it's an argument for using less slices of sound per second. Less artifacts. Less analog sound as well.

Checked NYC Mastering houses: Places that people pay for music production. Either they are fools paying for voodoo, or people looking to make great sounding records. Or both....

Pro Mastering House
Trutone, Inc.
http://www.trutonemastering.com/faq.html

Sampling Rates:
44.1K, 48K, 88.2K, 96K – 192K  (and everything in between), in general the higher the sampling rate, the better.

Music Lab Studios
http://www.musiclabnyc.com/equip.html

Multi Track:

Tascam ATR-80 2" 24 track analog recorder, 36 channel MOTU 1296 24 bit 96khz I/O hard disk systemMixdown:
Sony PCM 2600 Dat, Alesis Masterlink 24 bit 96khz recorder, Power 100 w/ Audiomedia II & editing & mastering software.

http://www.tweakheadz.com/16_vs_24_bit_audio.htm

Lets talk about sample rate and the Nyquist Theory.  This theory is that the actual upper threshold of a piece of digital audio will top out at half the sample rate.  So if you are recording at 44.1, the highest frequencies generated will be around 22kHz.  That is 2khz higher than the typical human with excellent hearing can hear.  Now we get into the real voodoo.  Audiophiles have claimed since the beginning of digital audio that vinyl records on an analog system sound better than digital audio.  Indeed, you can find evidence that analog recording and playback equipment can be measured up to 50khz, over twice our threshold of hearing.  Here's the great mystery. The theory is that audio energy, even though we don't hear it, exists as has an effect on the lower frequencies we do hear.   Back to the Nyquist theory, a 96khz sample rate will translate into potential audio output at 48khz, not too far from the finest analog sound reproduction.  This leads one to surmise that the same principle is at work.  The audio is improved in a threshold we cannot perceive and it makes what we can hear "better".  Like I said, it's voodoo. 

So should you record at a high sample rate?  Its going to depend on who you ask.  Some people say "It's all going to end up as 44.1 any way" when the cd is burned.

 

Others will tell you that when an audio interface processes and mixes sounds at 96 kHz the result is better and remains better even after the final conversion to 44.1.

A parachute only works when it's open. Open doors open minds.

Brewer

[wrong]

More thoughts on higher sampling rates..........

Analog tape recordings (to me) sound superior to digital recordings. How can that be? Analog sound is a continuous recording. Not approximations from slices of sound. It makes sense that having more slices of sound will resemble a full continous flow of analog tape. An argument here: each sample introduces artifact sound as well, and with thousands of slices, the presence of artifact sound might be discernable. This is just a thought. Maybe it's an argument for using less slices of sound per second. Less artifacts. Less analog sound as well.

It's not an argument at all, and it doesn't make sense, because it's simply wrong. Not just a bit wrong, but fundamentally  wrong. What you hear from a digital recording is also a continuously changing waveform - it's not 'broken up in bits' at all. The individual samples represent not hard output, but RATES OF CHANGE of the continuous output, and that's a very different thing altogether. This myth about digital outputs has been perpetuated for years, and it's a real mischief. You can't actually have a waveform that jumps instantaneously from one point to another - it's not physically possible. So what the D-A converter does isn't to output loads of bits, but to steer a current (which is manifest as a voltage change) at the output, and believe me, this is continuous. This is why a rate conversion upwards makes no difference - all that is happening is that the samples that are added are interpolated and added at the correct points in the moving waveform that's already there - nothing gets added at all, except a load of redundant samples.

The theory is that audio energy, even though we don't hear it, exists as has an effect on the lower frequencies we do hear.  

How can this possibly be a theory? It's not based on any theoretical understanding of how hearing actually works at all. Therefore at best it is a misplaced supposition. But actually, it's pure BS. If we don't hear something that is a high frequency air vibration, then it can't make a difference to us - QED. If it makes a difference, then we heard it. And you can forget all the arguments about bone conduction responses going higher - this is only true if the higher frequencies are injected directly into the skull, and even then, presbycusis will make damn sure that you don't hear anything from this source unless it is at unfeasibly loud levels - which in air would make the sounds very directional, and unlikely to be picked up anyway. As humans get older, cochleal  damage increases, because the world is a noisy place. And this damage starts at high frequencies (where the finest and most easily damaged ear fibres are), and gradually progresses downwards. It is only exceptional adults that can perceive relatively low levels of even 17kHz tone, never mind any higher than this.

And all of that is before we even consider that most loudspeakers won't even make an attempt at reproducing these frequencies at all, never mind accurately.

A parachute only works when it's open. Open doors open minds.

No, open doors don't inherently open minds at all. In this case, this could only work if you looked through all of the open doors, saw what was there, and assessed the evidence carefully in the light of the laws of physics before making any sort of decision about it. And that's what education does for you. You don't close your mind to anything - but you do learn that there are distinct limits to certain aspects of the performance of anything, and the people that tell you that this isn't so are generally trying to hype you into buying something.

Now in this case, I'd say that you haven't looked through many of the doors at all, and have just selected the first thing that came along that suited you - regardless of its pedigree. I already suggested on page one of this thread what the possible mechanisms for creating an audible difference by changes in sample rate were - and these do not relate to extensions of hearing beyond 20kHz. If faulty equipment after A-D conversion (like dodgy anti-alias filters) produce in-band artefacts, then there is a chance that you will hear them. But that does not negate anything that Shannon, Weaver, Nyquist or any of hundreds of people who have replicated the human hearing findings have said at all. It's only in-band responses that make a difference.

Others will tell you that when an audio interface processes and mixes sounds at 96 kHz the result is better and remains better even after the final conversion to 44.1.

These people are very seriously  deluding themselves.

[Others will tell you that when an audio interface processes and mixes sounds at 96 kHz the result is better and remains better even after the final conversion to 44.1.]

More thoughts on higher sampling rates..........

 

Others will tell you that when an audio interface processes and mixes sounds at 96 kHz the result is better and remains better even after the final conversion to 44.1.

Brewer

if you would point us towards an even casually (shoot out in a studio) configured 'blind' study (let alone a rigorously configured peer reviewed double blind study) that supports this contention I'm sure most readers of this forum would appreciate it

personally I've never found any published info that exceeds the level of marketing hype

until you get into resynthesis you will never 'add' anything to information from the original analog to digital conversion.  The best you can hope for is that digital manipulation of the original conversion does not add noise, subtract information

and pretty much by definition resynthesis will not, can not, 'improve', objectively, the original conversion . . . it creates something different that has some mathematically defined relationship to original. 

While there is credible evidence that humans can be expected to be able to detect differences between digitizing a source @ 8bit22K and 20bit44K there is scant evidence (from peer reviewed double blind statistically rendered studies redundantly executed at more then one location) that statistically significant numbers of humans can reliably detect differences among a source digitized @ 20b44K and anything 'higher'

Nor is espousing superiority of analog systems a rationally functional path to supporting need for higher (greater then 20b44.1k) bit &/or sampling rates . . . it is not merely matter of apples & oranges but matter of noise and distortion that analog imposes on transduction of pressure waves into and out of electric representation . . .  very generally speaking, among gear that reliably pretends to be semi-pro or pro, the weak leak(s) in audio cards is/are analog portions of the circuit . .. not surprisingly these remain the most expensive elements

without introduction of noise reduction filters 2 in analog tape puts you in neighborhood of 60 dB S/N . .. nor do noise reduction filters (Dolby & DBX, etc.) increase information

anyway main point here is not merely to beat a horse that has been skeletal for more then a decade but to indicate fact that if you can simply point us in direction of any credible research merely opening dialog concerning your contentions, I for one, would be grateful . . . I have far more hours invested in analog recording then I do in digital . . . if I could in good conscious [probably intended conscience here] position my self as an 'analog guru' and charge accordingly I could probably be convinced to cut you in for a royalty slice! (acknowledgment of individual who opened my eyes!)  Short of some reliable research your opinions are purely that . . . and tend to be refuted by basic physics: of pressure waves, of pressure wave to electric circuit transduction, of digital to electric circuit transduction; & by even elementary understanding of psycho-acoustics

(problem with idea (and it's very attractive) of higher out of range frequencies effecting lower human detectable ones is that it requires some mechanism that is coherent with both biology and physics, or it remains in the realm of magic . . . which by definition places phenomenon squarely in the 'unique' realm . . . i.e. my opinions about what you hear are as equally valid as your opinions about what you hear, even if I don't even pretend to have physical contact with source of what you 'hear' . . . and I chose to believe that you can not hear difference between same source captured at 20b44k and 24b96k . . . and in this magical realm there is absolutely (ABSOLUTELY squared) no way you can  (there for a moot point that can be settled only via cultural, physical dominance . . .) lacking such a mechanism the only influence can be that higher frequencies will affect lower in-band frequencies directly . .. in which point they would be as accurately converted as any other pressure wave . . . there is evidence that frequencies humans can not hear do exert some influence on ones we can hear, but as odd as it might seem what we hear is still absolutely dependent on what we 'can' hear . . . if XbYYkHz accurately represents what we can hear, increasing the bit & sample rate will not, except magically, increase what we can hear.  And it is reliable evidence that individuals can detect S/N ratios in excess of 20b and potential frequencies that exceed what can be represented by 44.1 kHz sampling rates that is in very short supply.)

Hi. Oretez. Thanks for your insights.

Short of some reliable research your opinions are purely that . . . and tend to be refuted by basic physics: of pressure waves, of pressure wave to electric circuit transduction, of digital to electric circuit transduction; & by even elementary understanding of psycho-acoustics

Maybe its' my sound card, but when I record at 96kHz 32 bit the sound is cleaner and has more alive nature than sound recorded at 44kHz 32bit.

Fact: analogue tape is continuous sound
Fact: slices of sound at X Hz are each: 1/x per sec
Fact: 96000 slices of sound, each slice = 1/96000 sec
Fact: Telefunken V76 mic pre emphasizes even order harmonic sound.

Opinion: Recording at 9632 w/ Telefunken V76 possibly greatly enhances even ordered harmonic frequencies: an observation based on fact.

BTW your reply is great and very interesting. I do believe recording at 96kHz 32 bit captures an essence of sound and after extensive editing / processing / when the file is downsampled to 44.1kHz16 bit, the overall nature of sound retains greatly the presence of the earlier higher sound. My sound card: (EchoAudio / Mona). It's not just me, but musicians I work with all notice this too. Call it the New York City effect. Telefunken Voodoo. Whatever.

My analogue decks: Revox PR99 MK1 / Teac 3340S / Teac 3340

Next point: how many of you work with a Telefunken V78 mic pre? This makes a tremendous difference in quality of sound captures. Established artists use a SM57 mic w/ a Telefunken V76 - it makes the SM57 a lion vs a lamb. Possibly I hear what I hear because the Telefunken V76 is in the audio chain. The Beatles heard it. They took as many Telefunken units / mics back to EMI London from EMI Germany as possible. Abbey Road Studios was full of Telefunken gear.

[not]

Fact: slices of sound at X Hz are each: 1/x per sec
Fact: 96000 slices of sound, each slice = 1/96000 sec

You should try reading what I wrote. Those are not  facts. You cannot do this. Sampling and slicing are not the same thing at all. You simply cannot 'slice' audio like that.

I know that the truth is a little inconvenient to your way of thinking, but that doesn't alter the real facts in the slightest.

BTW your reply is great and very interesting.

You should try reading it more carefully - he's not actually agreeing with anything you said at all. Which isn't surprising, really.

I'm still trying to work out how anyone can record at 32 bit depth?

as long as you avoid messiness of a mechanical, electrical transducer for ambient pressure wave 32bit float recording is a breeze

I hesitate to jump back into these discussions, perhaps because I 'think' I've said this stuff before, over and over . .. well maybe . . . memory is elusive (as well as illusive)

I like (really, really like) Telefunken V72's & V76's . .. but pairing them, as front end for digital recording does nothing to support contention that 96/192 kHz sampling rates are superior

the standard, un-modded, V76 had a pretty severe band-pass @ 40 Hz & 15 kHz.  There was a relatively rare 'm' version whose specs I don't recall at the moment that I think had variable EQ supporting 'up to' 20 kHz.  These were also designed to run on 220 v and required some fairly careful, high priced adjustments to run effectively (i.e. not introduce power supply switching issues) on US mains

They were also constant gain amps that will introduce artifacts at the noisefloor . . . the higher the bitrate the more noticeable the artifacts . . . V76 + any but top of line (and juries out on those) digital reverb (in series) becomes a FX device not a 'transparent' mic pre (which is not necessarily a 'bad' thing, but obviates need for 192 kHz sampling rate . . . generally speaking you don't need high sample rates to get accurate detail on noise!  (again kinda by definition))

Using a V76 & analog tape as your bench mark for superior recording provides no logical pathway to asserting the primacy of 24b/192 kHz digital recording, one might need hi-powered, on steroids CPU to model a v72/76 effectively but it does not require 24b/192 kHz to capture it accurately . . . again I've seen no credible studies that humans can reliably differentiate between 20b44.1kHz and anything with higher numbers . . . nor am I challenging individual contentions . . . merely indicating that those fall in the realm of the magical.  And my magical anecdotal observation has never found a single singer that has ever, under any circumstances, detected the difference between 20b44kHz and 24b96kHz. 

The only area where debate concerning 96 + kHz sampling leaves area of marketing hype (by baby steps) has to do with so called 'rolloff' filtering as Nyquist frequency is approached.  And it might be 'cheaper' to build effective converters based on 96 kHz . . . and if industry moves in directions it tends to, it is not unlikely that eventually we'll forced into using 96 kHz to avoid poorly implemented on the fly downsampling issues . . . but these become marketing, manufacturing issues.

as long as you avoid messiness of a mechanical, electrical transducer for ambient pressure wave 32bit float recording is a breeze

How so?  As far as I am aware, there's not one ADC, operating at 32 bit depth, on the market. Again, TTBOMK, such an animal is technically impossible - but I stand to be corrected on this.

You should try reading what I wrote. Those are not  facts. You cannot do this. Sampling and slicing are not the same thing at all. You simply cannot 'slice' audio like that.

Confused here: 96kHz = 96000 samples of sound per second? Not 96000 slices of sound? Please explain. Thanks.

Using a V76 & analog tape as your bench mark for superior recording

Hi. I use the V76 for digital recording. I use my analog decks for 2 track mixdowns if/when people want them. I noticed listening to analog tracks made years ago, against digital recordings made with CEP, AA2.1, AA1.5, AA2.0 - analog sound is strikingly lifelike (these are recording made before I had the V76) - much more than digital sound. Adding Telefunken V76 into the audio recording chain (going to digital) gives tracks a lifelike/analog sound.

It's interesting that a V76 has no effect on superior sound regarding sampling frequency used. Amazing.

Telefunken V76 known emphasis for even-ordered harmonics is legendary. If the V76 selects even-ordered harmonic sound, won't these be frequencies a soundcard has to deal with? How certain frequencies interact with soundcard sampling frequencies when music is recorded seems to me to be something worth study - by scientists / musicians. Has anyone done this with a V76?

Again, maybe the reason I (and musicians I record/play tracks for) believe my echoaudio / Mona soundcard sounds better at 9632 than 4432 or 4832 is because of the way this card was manufactured?

BTW as long as you are here - why does the AA 2.0 not let me record at 9632?

as long as you avoid messiness of a mechanical, electrical transducer for ambient pressure wave 32bit float recording is a breeze

How so?  As far as I am aware, there's not one ADC, operating at 32 bit depth, on the market. Again, TTBOMK, such an animal is technically impossible - but I stand to be corrected on this.

All 32-bit floating point recording is 24-bit, directly converted from the integer output of the A-D converter, with an empty exponent added. To all intents and purposes, it's 24-bit.

You should try reading what I wrote. Those are not  facts. You cannot do this. Sampling and slicing are not the same thing at all. You simply cannot 'slice' audio like that.

Confused here: 96kHz = 96000 samples of sound per second? Not 96000 slices of sound? Please explain. Thanks.

The difference between a slice and a sample is fundamental to all arguments about digitised sound. Since soundwaves are continuous, digitising has to find a way to represent that continuous sound - no slices involved.

What happens is that at fixed points along the continuous waveform, a sample of its instantaneous level is taken. Any one sample on its own means nothing in particular, but when considered in conjunction with the samples either side of it, they represent a way of coding the rate of change of a signal - so if it doesn't get very much louder between two samples, that represents a low rate of change. And the idea is to reproduce that continuously varying waveform at the output. And, there's nothing missing at all between the samples; the waveform is still changing continuously, just as it was on the way in. All the samples are doing is coding information about the rate of change of a continuous waveform, in order to be able to reproduce it as such. It's never been any other way, and neither can it be. And that seems to be what you haven't grasped about this!

Absolutely the best way to see how individual samples can change the rate at which the output varies is to zoom right in at sample level on a signal in Audition, so that you can pick the samples up with your mouse. If you move one up and down, you will find that it alters the rate of change of the waveform, not just in the place that you move it, but for a number of samples either side as well. But you will note that however far you zoom in, that waveform is still there, and continuous between the samples - because that's exactly how it is in real life.

So the concept of 'slicing' doesn't actually enter into this - an individual slice would only represent itself, but that can't be said of an individual sample - they can only work in conjunction with the ones around them to recreate the continuous waveform at the output. Nyquist knew this perfectly well - and that's why he said that ultimately, the highest frequency that could be represented in a system is determined by the maximum rate of change of the output between any two samples - hence the maximum reproducible frequency being half of the sample rate.

The problem comes if you try to exceed this rate, and that's when aliasing occurs. Unfortunately, to enable a flat response up to the cut-off frequency, an anti-aliasing filter will inevitably introduce artefacts. Nowadays, there are some neat ways around the problem of reproduced artefacts*, and converters sound nowhere near as bad as they used to - and it was this early failure to anti-alias at the top of the audible frequency range that got 44.1k sampling a worse name than it actually deserved - it wasn't the sampling that was the issue, but the appalling filters. Firms like Apogee have got this licked, though - a 24-bit 44.1k signal recorded and reproduced on their kit (or any other that uses similar technology) sounds a damn sight better than anything recorded at a higher sample rate on an indifferent system. Nyquist wasn't wrong - it just took a long time for anybody to come up with a system decent enough to prove it!

Previously, all sorts of attempts were made to get around the problem of audible phase changes at HF introduced by anti-aliasing filters, and there were loads of different designs - some a lot better than others. But ultimately the problem remained - if you want a super-fast fall-off outside the passband, then you are going to get phase anomalies and ripples within the passband you want - and they will  be audible.

* Typically, you can overclock the output stage of the D-A converter, and by a bit of numerical trickery, get away with a much simpler filter safely out of the audio band, leaving the resulting audio sounding miles cleaner than it would from an old brick-wall filter system that wasn't overclocked. And  with no aliasing.

as long as you avoid messiness of a mechanical, electrical transducer for ambient pressure wave 32bit float recording is a breeze

How so?  As far as I am aware, there's not one ADC, operating at 32 bit depth, on the market. Again, TTBOMK, such an animal is technically impossible - but I stand to be corrected on this.

I was being cute . . . haven't read what looked like Steve's response to this . . . I am unaware of any hardware that takes a 'real' world pressure wave and digitizes it @ any floating point resolution . . .  if I remember correctly straightforward thermodynamics precludes any real world solutions for making maximum use of 24 bit integer conversion, so resolutions higher then 24 bit integer are pretty useless

but there is nothing that precludes the 32bit float render of a 'virtual' process . . . that stuff can be manipulated purely as numbers . . . how useful that might be is a different debate

Nowadays, there are some neat ways around the problem of reproduced artefacts*, and converters sound nowhere near as bad as they used to - and it was this early failure to anti-alias at the top of the audible frequency range that got 44.1k sampling a worse name than it actually deserved - it wasn't the sampling that was the issue, but the appalling filters. 

ok. So it's a filter problem for sound clarity. Then why do people choose between 44.1kHz and 48kHz? Video works at 48kHz - that's a good reason for 48kHz. What about audio? The answer seems to be to have a great AD/DA converter.

Searching Google for Apogee AD/DA converters: Apogee ROSETTA 800 8-Channel AD/DA Converter 24-bit/192kHz

192kHz vs 44.1kHz!!!!!!!! Why 192kHz if their converters are so great - and it's the filter (not the sampling rate) that makes great sound? Is Apogee just playing everyone for a fool? Or do they know something that you don't? I'd like to see people here try the Rosetta, and then talk about what's best for recording.


BTTW No one here has a clue for why I get noise when I try to rec at 9632?
http://audiomastersforum.net/amforum/index.php/topic,6364.0.html

Using a V76 & analog tape as your bench mark for superior recording

Hi. I use the V76 for digital recording. [snip]

It's interesting that a V76 has no effect on superior sound regarding sampling frequency used. Amazing.

No one I know contends that V72/76 are not perfectly respectable mic pre's.  Unless you do some fairly expensive mod's they are limited, they are not a Swiss Army Knife transparent mic pre.  (power supply, impedance (for modern mics), slope of the band pass, upper limit of the filter, curve of constant gain, etc.) 

But those are separate issues

If, on one end of the dialectic curve, an unobtainable goal is to represent an original sound source accurately (theoretically the primary reason to use high bit, high sample rate A/D) then the V72 is less accurate then 24b44.1kHz conversion . . . it introduces a higher noise floor and has a lower coherent high frequency limit.  Since using it as the source can be digitized more accurately then it can represent the original, increasing the 'accuracy' of the V72 simply enhances its artifacts at the expense of original source

A v76 does not 'select' frequencies.  The circuits do emphasize, this is distortion of original source.  Due to a number of interrelated factors we, 21st century humans, tend to 'hear' (in our minds) this distortion as inherently musical.  There are biological elements to this but majority of it is cultural, hence 'fashion' not physics

No would should contend that your recordings, via a V76, do not sound better, to your brain, then ones made without it.  I'm not even contending that hi-bit, hi-sample rate recording does not sound 'better' to your brain.  For other readers of this forum I'm just pointing out that there is no scientific basis for this interpretation.  And there are serious holes in the type of things you are trotting out for argument.  Those arguments seem to raise a question as to how thorough or accurate your understanding of recording, including analog recording, is.  But many of your assumptions are not uncommon.

and, again primarily for other forum readers, I tend to avoid any & all vendors who try to sell the position that any bit of gear (hard or soft) is panacea for recording,  any recording gear is far down the list from skill of practitioner (musician) & content.  If you accept 'room' as gear, that is more important then all the rest of the gear combined.  The mechanical transducer (i.e. microphone) is far more important then all other gear that follows it (in the signal path) . . . rest of stuff pretty much follows the same pattern.  As I said I'm quite fond of V76's . . . but it is doubtful I'd ever carry one around for field recording . . . whereas another 'vintage' mic-pre, API3124, I have no problem dropping into a four space ATA rack, particularly if that rack is all I can carry.  And it has been demonstrated endlessly that you can paint the outside of a cardboard box and get all the potential marketing value from any boutique: 'can't record without this' device.  I've seen 1176's go from 'bee's knees' to door stops and back . .. if I live long enough I expect to see them as door stops again. (& I like 1176's) In the 80's I picked up more then a dozen Altec 4 channel PA's for little more then 'I'll take that off your hands, save you the trip to the dump' . . .  It cost $1100 to bring them up to speed (though at that point the API opamps that were added only cost $.34 ea. . . . cost was in the labor) and after that they were a damn good $1100 four channel mic pre & @ $1500 you were getting screwed.

Anyway main point is that if you want to 'learn' how good 24b192kHz recording might be, for anything other then V76 self noise, I would not tend to route an independent source through a V76