Topic: Why record at 24/48 versus 24/44.1?

[alpine85]

(Sorry if this has been covered before)

I just started doing 24-bit recording since I upgraded to the R-44.  I've been recording at 24/44 and have been happy with the results, but I've noticed a LOT of tapers record at 24/48 and then resample down to 44.1.  Is it worth the extra step for that small amount of extra sampling frequency?  I can understand going 24/96 and then downsampling, but why 48?

[OFOTD]

Just like CD's are 16/44.1 for that particular format other newer formats start at 24/48.    If CD's were gone then the 44.1 sampling freq would probably not exist.


ETA: Correction

[Kevin Straker]

I use 24/48 because it is the preferred format for the image software I use. DVD-Audiophile uses 24/48 or 24/96 to create the .iso file.

[stevetoney]

^ Right.  I think 48 is the standard sampling frequency for the audio portion on any video-based software system.

[alpine85]

Yeah it makes sense to record at 48 or 96 kHz if you're using for video, dvd-audio, or for archiving purposes.

But what if you're final medium is going to be 16/44.1?  Is the quality of a something recorded at 48 and then downsampled to 44.1 going to be any better than if it was just recorded at 44.1 in the first place?  Is it worse?  I guess it would be easy enough to do some experiments and see.  Probably would depend on the method used for downsampling too.

I was just curious about peoples reasoning behind doing 24/48...

[stevetoney]

Yeah it makes sense to record at 48 or 96 kHz if you're using for video, dvd-audio, or for archiving purposes.

But what if you're final medium is going to be 16/44.1?  Is the quality of a something recorded at 48 and then downsampled to 44.1 going to be any better than if it was just recorded at 44.1 in the first place?  Is it worse?  I guess it would be easy enough to do some experiments and see.  Probably would depend on the method used for downsampling too.

I was just curious about peoples reasoning behind doing 24/48...

It's doubtful you'll hear a difference between 48 and 44.1.  I don't know for myself, but I've read that people can't hear any difference between 48 and 96.  It's the bit depth that provides that main difference in sound quality. 

As far as the standard of 48 vs. 44.1, I don't have strong feelings one way or another.  However, my personal reasoning for choosing 48 is that it's compatible with the video standard, it seems logical because it's divisible by 96 and 192, and if I were ever gonna need to burn to CDR, I'd have to downsample anyway, so what's the difference between downsampling from 24/48 (which is a theoretically higher quality than 44.1) or 24/44.1.  The end result is a downsample either way.  That's been my logic anyways.

[OFOTD]

I was just curious about peoples reasoning behind doing 24/48...

Well the easiest way to explain it is to be as future proof as possible.    If you have the ability to record at higher bit/sampling rates as well as have the ability to store the larger files then why not.

Why do it?  Because you can.  Say in xx years a new format comes high fidelity format comes out, you'll want the best rate possible to start with.   Nothing is wrong or inherently bad about doing 24/44.1 but you have the means to record higher so why not.  

For me personally I record at 24/48 unless the band or venue has an above average PA then i'll do 24/96.  Also if it is a recording that is personally important to me then i'll also go with 24/96.

In the end do what you have the capabilities and patience to record/archive at.  

[datbrad]

Yeah it makes sense to record at 48 or 96 kHz if you're using for video, dvd-audio, or for archiving purposes.

But what if you're final medium is going to be 16/44.1?  Is the quality of a something recorded at 48 and then downsampled to 44.1 going to be any better than if it was just recorded at 44.1 in the first place?  Is it worse?  I guess it would be easy enough to do some experiments and see.  Probably would depend on the method used for downsampling too.

I was just curious about peoples reasoning behind doing 24/48...

It's doubtful you'll hear a difference between 48 and 44.1.  I don't know for myself, but I've read that people can't hear any difference between 48 and 96.  It's the bit depth that provides that main difference in sound quality. 

As far as the standard of 48 vs. 44.1, I don't have strong feelings one way or another.  However, my personal reasoning for choosing 48 is that it's compatible with the video standard, it seems logical because it's divisible by 96 and 192, and if I were ever gonna need to burn to CDR, I'd have to downsample anyway, so what's the difference between downsampling from 24/48 (which is a theoretically higher quality than 44.1) or 24/44.1.  The end result is a downsample either way.  That's been my logic anyways.

Adding to what Steve said, there is one more strong reason I have heard for using at least 48khz, or anything higher, and it has more to do with the analog portion of the recorder directly in front of the A/D, not the actual fact that it's a higher sampling rate. PCM recorders have to have an anti-alaising filter in the audio path before the A/D, to prevent any sound above half the sampling rate to hit the A/D. This characteristic is independant of the bit depth.

Since there is no such thing as a true frequency dependent brick wall filter, the attenuation has to start a couple thousand khz before that frequency, so that frequencies above the limit will be completely attenuated. This means that while technically, the frequency response of 44.1 recordings should extend up to 22.05 khz, the attenuation of the filter starts rolling at 20 khz, so the 22.05 top range that should be recorded is not recorded.

At 48 khz, same deal, only the high range goes to 22 khz, attenuating the high frequencies so they are completely blocked at 24 khz. So, at 48khz and higher sampling rates, you ease the job of the anti alaising filter, and at the same time capture higher frequencies. You can always drop down to 44.1, but you can't go the other way later.

Good Luck !!

[DSatz]

There is no good reason at all to record at 48 kHz if your eventual delivery medium will be at 44.1 kHz; in fact it makes the eventual product very slightly worse if you do so, because any sampling rate conversion adds noise. If on the other hand your eventual delivery medium will be at 48 kHz then it makes good sense to record originally at 48 kHz for precisely the same reason.

If you are just recording for your own pleasure and you can set the conditions of playback to your own liking, then of course you are free to choose any sampling rate that you prefer. But there is no good reason to think that a significant difference in sound quality could exist between two rates as close together as 44.1 and 48 kHz--at least, not on account of the sampling rate alone.

Of course if there's any real advantage to a higher sampling rate in terms of preserving higher frequencies, you give that up completely when you resample at a lower rate, so again there is no sonic advantage to starting out at a higher rate than necessary for eventual delivery; it's simply a waste of storage space and processing time. Nonetheless some people do it because it makes them feel better, I guess.

--best regards

[OFOTD]

it's simply a waste of storage space and processing time. Nonetheless some people do it because it makes them feel better, I guess.

Well a 1TB drive is under $100 and my Core i7 machine will process a file extremely fast.   So its not waste of either for me.  Storage space and processing speed will only continue to get bigger, cheaper and faster.

[DSatz]

OFOTD, there's always some value of X for which you get no further audible benefit from choosing a sampling rate any higher than X kHz. I'm not defending any particular value of X here; I'm only saying that a sampling rate higher than the highest rate that gives an audible benefit means (kind of obviously) that disk space and processing time are being used to no audible benefit.

If someone wants to spend N times as much money on storage media as they would have to, with no audible benefit, that's their business, I suppose. But when someone is new at this and they're asking these basic-information-type questions, I think we should tell them the truth: The audio hobby has always been one in which some people choose to spend a lot of extra money for benefits which aren't always tangible.

So unfortunately you can't just look at what choices other people are making and assume that "there must be a good reason for it" in terms of sound quality--just as a person can't simply look at the audio equipment that's on the market and assume that in all cases "there must be a good reason" in terms of sound quality for certain features or approaches that the equipment embodies. There is a fair amount of audio equipment that is made the way it is just to be different--or to appeal to beliefs that the designers of the equipment are perfectly well aware are myths. If markets were rational, we wouldn't be in the mess we're currently in.

--best regards

[alpine85]

There is no good reason at all to record at 48 kHz if your eventual delivery medium will be at 44.1 kHz; in fact it makes the eventual product very slightly worse if you do so, because any sampling rate conversion adds noise.

First off, thanks for all the responses, everyone.   Lots of good info here (much of it conflicting - hehe)  and many factors I hadn't even considered.

DSatz - are you saying that a recording that is converted from 96kHz > 44.1 would be noisier than one recorded at 44.1?  Would that noise outweigh the benefit of recording at the higher frequency?  Or is there no benefit at all of recording at 96 if you're going to downsample anyway?

[alpine85]

(snip)

...if I were ever gonna need to burn to CDR, I'd have to downsample anyway, so what's the difference between downsampling from 24/48 (which is a theoretically higher quality than 44.1) or 24/44.1.  The end result is a downsample either way.  That's been my logic anyways.

Well it would be 2 processing steps instead of one, right?

24/48 > 16/44  bit depth conversion + sample rate conversion (2 different processes)
24/44 > 16/44  bit depth conversion only

[DSatz]

alpine85, you asked:

> are you saying that a recording that is converted from 96kHz > 44.1 would be noisier than one recorded at 44.1?

Any conversion adds noise, but whether you will hear it will depends in part on the noise level in the original signal. Usually with live recording there is already enough noise that the difference won't be audible, although it is real. That's why I said "very slightly worse."


> Would that noise outweigh the benefit of recording at the higher frequency?

Whatever benefit there would be from recording at the higher frequency would be lost AND there would be added noise, so it's not one penalty versus the other--it's one penalty PLUS the other.


> Or is there no benefit at all of recording at 96 if you're going to downsample anyway?

The only benefit would be that you would have an original 96 kHz recording, if that is of value to you for some other reason (as it has been suggested, you might have another use later on for the same recording at a higher sampling frequency). But the 44.1 kHz "end product" wouldn't benefit sonically from having been recorded originally at 96 kHz.

When I talk about "sonic benefit" I mean "a more accurate or precise rendering of the original signal." Whenever any processing alters the sound of a recording, there can be opinions on both sides as to which one sounds better; I can't predict which one anyone will prefer.

I'm just saying that, for example, a 96 kHz recording could (wouldn't necessarily, but could under some circumstances) preserve the phase relationships among very high-frequency tones a little better than a 44.1 kHz recording, and that advantage (whether or not it is audible, which is a whole other question) would be entirely lost by converting to 44.1--so you could get the same level of phase linearity by starting out at 44.1, without the added noise of the conversion on top of the loss of phase linearity.

[Jhurlbs81]

I'm just saying that, for example, a 96 kHz recording could (wouldn't necessarily, but could under some circumstances) preserve the phase relationships among very high-frequency tones a little better than a 44.1 kHz recording

DSatz-

could you explain this in a little more detail?  I'm curious what you mean and always enjoy reading your posts..

thanks
jesse

[boojum]

David, thank you for laying this out in clear and succinct detail.  I am on another board which is mostly pro, GS, and many if not most of the folks there doing studio work say they do it at 44.1.  I have suspected there would be losses in resampling which would offset any gains from sampling at the higher rates and that this was why studio work was being done at 44.1.  And because I believe I can learn more from someone who is in the business of getting paid for good recordings than from someone who dabbles in recording as a hobby I have followed what the pros indicate is best practice.  As a result I do everything at 24/44.1.  Had this not come up here with your lucid explanations I would have said nothing as it is so popular here to oversample that it is almost heresy not to.

Cheers     

[notlance]

"Hi, my name is Paul, and I sample at 44.1 kHz."

I feel much better now.

This whole discussion takes on an "angels on the head of a pin" quality.  There are much bigger problems in recording such as where do I place the mic, the loud A/C in the room, the buzzing EXIT sign, the jerk that keeps thumping the mic stand, or ....  The only benefit of discussing 44.1 vs 48 is understanding why it does not make much difference.

If you are trying to make a living at this, the rule seems to be 44.1 for the music business, 48 for video/film.

[datbrad]

If the final delivery medium goal is limited to Redbook CD standard, and never will there ever be a need or desire to create a DVD (audio or video disc) to take advantage of the higher rates offered by those formats, or listen to the higher resolution files on a media server at better than CD sampling rates, then I agree that there is no reason to record beyond 44.1.

Problem is, many industry observers feel that in 10 years, physical media like CDs are likely going to be considered a quaint relic, and wave files on media servers will be the norm. So when that time comes, instead of listening to your now 10 year old recordings at 24/48 or 24/96, you will only have a 24/44.1 in your archive.

[MSTaper]

While I completely understand the explanations and will likely go 24/44.1 when I tape tonight, I have to ask, if this is the norm, why have recorders that record 24/96 or even 24/192? In what situations would this higher sample rate be beneficial? Commercial video applications?

MSTaper

[Mike R.]

While I completely understand the explanations and will likely go 24/44.1 when I tape tonight, I have to ask, if this is the norm, why have recorders that record 24/96 or even 24/192? In what situations would this higher sample rate be beneficial? Commercial video applications?

MSTaper

Marketing.

[BC]

While I completely understand the explanations and will likely go 24/44.1 when I tape tonight, I have to ask, if this is the norm, why have recorders that record 24/96 or even 24/192? In what situations would this higher sample rate be beneficial? Commercial video applications?

MSTaper

I would think high res sampling could be beneficial in a recording situation with excellent acoustics (I'm thinking pro recording session in a studio, concert hall and the like), as opposed to a concert in a noisy bar or ampitheater. There are some classical studios allowing you to download 24/96 or even 24/192 studio-master quality files these days.

[DSatz]

Jesse Hurlburt, I was referring to the possibility that the anti-aliasing and -imaging filters in a (say) 96-kHz recording system might have lower group delay at (say) 10 kHz and above than the corresponding filters in a 44.1 kHz system. Filters are an inescapable part of any digital audio recording system, and a higher sampling frequency allows them to be better and/or to affect the audible portion of a signal less.

But the relative importance of this as compared with, say, room acoustics and microphone placement shouldn't be overstated. It generally takes special test signals to be able to hear the advantages of the better filtering, and special playback systems besides--loudspeakers generally have horrible phase and distortion problems at the highest audio frequencies, so most often electrostatic headphones have to be used if the difference is to be detectable at all--and even then, most listeners hear no difference, since our ears are rather insensitive to phase distortion at such high frequencies. So we are really talking about a subtlety, although I have every respect for those who wish to preserve such subtleties.

As far as I'm concerned the filtering advantage is the only possible audible advantage of a high sampling rate (i.e. higher than "2 x 20 kHz plus a little") even though in practice, it is far more of a measurable than an audible phenomenon. Preserving frequencies higher than we can hear certainly doesn't benefit the frequencies that we can hear; that's been investigated every few years since at least the 1930s and the results have always been either negative or non-reproducible. And the whole notion that a higher sampling rate "fills in the space between the dots better" is just ... sad (and more to the point, incorrect).

--best regards

[Will_S]

While I completely understand the explanations and will likely go 24/44.1 when I tape tonight, I have to ask, if this is the norm, why have recorders that record 24/96 or even 24/192? In what situations would this higher sample rate be beneficial? Commercial video applications?

MSTaper

Marketing.

Mainly that.  But as DSatz mentions, potentially some theoretical advantages that are unlikely to be heard in practice, but why not do it if you have the space.

Also, higher sampling rates are indisputably handy if you want to record bats. For real, I know a several field biologists who need to be able to record frequencies well over 20kHz, although of course they can't hear the ultrasonic part of the signal without processing it first.

[DSatz]

BC, you wrote:

> I would think high res sampling could be beneficial in a recording situation with excellent acoustics (I'm thinking pro recording session in a studio, concert hall and the like), as opposed to a concert in a noisy bar or ampitheater.

Well, "resolution" isn't really a term with a single, agreed-upon meaning in audio, and here I think that you may be confusing sampling rate with bit depth (the number of active bits per sample). The number of active bits per sample determines the dynamic range capability of a recording system. The best available A/D converters at present have a little over 21 bits of resolution under ideal conditions, so their output data is generally stored in 24-bit samples.

The reality of all live recording--even in a very quiet studio--is that there is always an acoustical "noise floor" and an electrical "noise floor," and there is also a maximum sound pressure level; the dynamic range of the recording is what falls in between these two limits. During the heyday of the vinyl LP a 65 to 70 dB dynamic range was considered outstanding; that is about equivalent to only 11 or 12 bits. The earliest commercially available professional "16-bit" digital audio recording systems had only 14 bits of active data per sample, which already blew any analog tape recorder out of the water, even with advanced noise reduction and high-output tape.

What I'm getting at is that very, very few live recordings fully utilize a 16-bit dynamic range, which is absolutely enormous. When 24-bit recording was introduced, the rationale was that it was strictly for high-end professional use so that a 16-bit range could be maintained throughout a complex studio production, e.g. multi-track mixdown and signal processing, which would all occur at 24 bit resolution until the final mix which would be 16-bit.

That's about as far as the influence of rational, engineering-driven planning went. Everything else after that was the market in action. I can sell my toy better if it has this or that specification which is a "professional" specification--and if it uses 24 bits to convey (say) samples that are really only accurate to 15 bits, then I can still peddle it as a 24-bit recording device, and someone will buy it, take it home and congratulate himself on having it.

People like us use 24-bit recording for security--we can set our record levels a little low, not worry about the peaks, then normalize the levels when we dither down to 16, making ourselves look like level-setting geniuses. And for certain laboratory uses, or systems which drive the loudspeakers to extremely high SPLs, a wider dynamic range can be important. But the range between the noise floor of people's domestic playback environments and the loudest that their playback systems can reproduce is generally quite a bit (so to speak) less than a 16-bit range.

Meanwhile the entire mass market has gone diametrically away from high quality audio in favor of portability. I don't personally limit myself to recording at the audio quality level of an mp3, or the incredibly bad sound of most radio stations, or videos on the Internet, but I do have to notice that about 98% of what people listen to these days is that.

[Chuck]

Yeah, this is the raw, blunt truth of digital recording. Thank you, DSatz, for stating it so well in easy to understand language.

44.1kHz, 16 bit, stereo, PCM (CD quality) audio was engineered to deliver full bandwidth and dynamic range of recorded music for humans to enjoy.

I think there is plenty of room to debate the stereo aspect of the CD spec, because two channels only approximates a 3-D sound field. But, the bit rate and sample rate cover all of the measureable aspects of audio reproduction.

[notlance]

Problem is, many industry observers feel that in 10 years, physical media like CDs are likely going to be considered a quaint relic, and wave files on media servers will be the norm. So when that time comes, instead of listening to your now 10 year old recordings at 24/48 or 24/96, you will only have a 24/44.1 in your archive.

I agree that CDs will become quaint in a few years; a few years after that midia servers will be also obsolete.  Nevertheless, I'll be happy to listen to my 44.1 kHz files because a properly converted 44.1 kHz file contains ALL the information contained in a 20 Hz - 20 kHz analog signal.  If a 44.1 kHz sample has ALL the information, then a 48 kHz, 96 kHz, 192 kHz, or infinite kHz sample cannot have more information.

Notice I said a properly converted sample.  There certainly are problems that can arise during the A/D process.  However, if my recorder has audible problems converting at 44.1 kHz, I don't think all those problems will suddenly disappear if I set it to sample at 48 kHz or higher.  Fortunately, my recorder seems to do fine at 44.1 kHz, so that's where I leave it.

I have had clients that insist on 96 kHz.  In that case I don't argue sampling theory with them; I pull up the sampling rate menu item and set it to 96 kHz.  I do find it interesting when they are happy to accept a CD copy of the final mix.

[DSatz]

Um, at the risk of seeming like (or actually being) a blabbermouth/know-it-all...

Chuck wrote:

> 44.1kHz, 16 bit, stereo, PCM (CD quality) audio was engineered to deliver full bandwidth and dynamic range of recorded music for humans to enjoy.

Well, the truth is that when these parameters were chosen, they were slightly beyond the reach of commonly available technology; then as professional PCM recorders and the digital audio CD were brought into production and use, they were achieved first tentatively, and then more definitely. The technology has progressed considerably further since then, and if today's technology could have been available then, we can't know whether the same decisions would have been made. In some ways I kind of doubt it.

This was years before DAT or DASH recorders were invented, so for Sony (and later JVC), the natural recording medium to accompany their PCM audio processors was segmented video: 3/4" U-Matic or in some cases, the industrial versions of the 1/2" Beta and VHS formats. Thus the choice of sampling rate was tied by design to the field rates of the PAL and NTSC video systems. It was also limited by the available A/D converters of that time, which imposed a tradeoff: The higher the sampling rate, the more noise and distortion they caused. As it was, those converters had to be adjusted carefully for lowest noise at each recording session (part of how I made my living in the early-to-mid 1980s).

Those were requirements imposed by the need to sell a complete package to the studios, record labels and other recording clients. The late Dr. Thomas Stockham's "Soundstream" system, on the other hand, used computer data recorders and was thus free to use a somewhat higher sampling rate (50 kHz) and correspondingly simpler, better-sounding filters. Shannon's Sampling Theorem states that any sampling rate is adequate (and thus, equally good) as long as it is more than twice the highest frequency present in the signal--but it doesn't deal at all with the problems of limiting the signal to that frequency range.

Since those problems are most severe when a filter has to cut off very sharply, I can well understand the wish for somewhat higher sampling rates, now that converters and recorders are available to handle them, and now that the characteristics of video recorders aren't a factor any more. For example, 64 kHz would give enough of a margin that the filters wouldn't need to be very steep.

But I fear that nothing will ever satisfy the people who basically don't understand how digital audio works, and who therefore keep ASSuming that higher and higher sampling rates will bring about a closer and closer approximation to the input signal.

[Chuck]

Thanks for the clarification. As always, DSatz, you are a wealth of accurate information on this stuff... 

[darby]

I merely tape shows as a hobby and unfortunately have no real formal training, or professional experience
so in light of the last few posts would someone care to explain the significance (or insignificance) of this chart:
it appears to me the higher the bandwidth, the closer to realistic it is

[DSatz]

A Dirac pulse is a theoretical signal that has infinite bandwidth, but human hearing is real and doesn't. The filters that these impulses are being put through certainly don't, because they're filters--their whole purpose is to limit signal bandwidth. And band-limited systems can't have perfect impulse response.

But listening tests have shown that a recording system doesn't need perfect impulse response in order to be audibly transparent; rather, there are thresholds of group delay at various frequencies beyond which no further audible improvement occurs when group delay is reduced further.

For decades people gladly tolerated the dreadful impulse and phase response of vinyl LPs and analog tape--yet when digital comes along, which is light years better where it counts (between/among channels), they all of a sudden grow virgin bat ears and become allergic? I don't buy it.

That understood, let me raise the "counter-question" of how different those scope traces would look from one another (if that's what they really are, rather than MatLab graphs as I suspect) if they were band-limited to the human hearing range? That would be the first clue to whether any audible differences could possibly be present or not.

But those impulse drawings can only help answer that question if you know the time scale of the impulses, so that you can infer their spectral content. Omitting an explicit time scale from an illustration of this kind is a major lapse of fairness. Another thing I'm sorry to see is the so-called "analog system"--the only thing with impulse response anything like that would be a short length of "analog" wire. Even 30 ips tape is nowhere near that clean.

--best regards

[Jammin72]

If your goal is a CD I see no reason to record at 48 or 96.  Personally I like to archive and listen on DVD and 24/48 or 24/96 is the format for that particular medium.  It's interesting that when we're talking about recreating live sound that everyone wants to limit or draw lines around the range of human hearing. That doesn't happen in a live environment, frequencies much higher and lower than the ones we perceive though our ears interact with and affect the sounds that we do indeed hear. Why wouldn't we want to attempt to include those in trying to capture a truly live sound? I have a feeling that transducers on either side of the equation will (and have) step up and be able to include that information.

I don't know if it's the processors in my humble stereo system or simply one set of genuine placebo earmuffs but the 24/48 files sound better than their 16/44.1 counterparts on my playback system.

For someone asking the question the way you're phrasing it however, record at 44.1 there's no reason to use 48 if your ultimate goal is a CD. 24 bit gives you the headroom that has been mentioned and also provides for larger chunks of data for DAW work which I've found to be beneficial.

Always love the voice of reason in a DSatz visited thread.

[Jammin72]

You should check out some of Tannoy's super tweeters.  They have them as tack on units as well as integrated into some less expensive speakers.  Those combined with a concentric driver do wonders for image creation and cohesion IMO.

[DSatz]

Let me try to give a viewpoint from a non-hobbyist perspective. The audibility or audible effect of "supersonics" (not a term that's well accepted by audio engineers, incidentally) is controversial only in the sense that evolution is controversial: The evidence has been in for a long time and nearly everyone who has studied it agrees about it, but there is a determined opposition that keeps trying to come up with evidence to the contrary. That evidence keeps being shot down by other, later studies but there is often a period of time in which some tantalizing experimental result that might just indicate that there's some truth to the theory is still in play.

Creationists know that if they can find one feature of living organisms which scientists can't explain how it evolved, they'll win some converts. So they keep switching from complex feature to complex feature (bird flight, the human eye, etc.) and then after a while, researchers come up with evidence for the effect of natural selection in the development of that feature. Whereupon the creationists change to a different feature. Can they sustain their movement this way? Maybe, though they don't do themselves any real honor.

There is very good engineering reason to be concerned with the way systems handle signals at frequencies well above the range of human hearing. When such signals are present, they must not dirty up the audible stuff. And systems which limit bandwidth on purpose must do so in ways that preserve the sonic integrity of the in-band material. But there is no credible evidence, despite insistent claims and attempts to prove otherwise, that human beings can hear beyond about 20 kHz at anywhere near reasonable sound pressure levels--or that the presence or absence of "sound" above the human hearing range otherwise affects the perceived quality of audible sound, as long as it's not interfering (e.g. if you add a lot of 30 kHz, some amps go into audible slew-rate limiting).

In situations like this the only thing that can be said in the end is that the vast majority of serious audio engineers don't believe it, based on available evidence.

[DSatz]

mshilarious, I'm not making an ad hominem comparison; I'm comparing the degree of substance behind the two alleged controversies, and saying that in the audio engineering community this isn't generally considered to be one at all. Rather, the limits of the human auditory range are generally considered to be elementary, well proven and well settled--even the Oohashi study that you refer to cites sources to that effect across a range of 55 years in its first sentence.

It's just as in biology, where evolution isn't generally considered controversial. Creationists are a tiny minority among scientists, but almost any creationist will sincerely believe that a much larger number of scientists agrees with him.

That is a very understandable human attitude, I find. Few of us can say "I believe in something which the vast majority of knowledgeable people utterly reject on the basis of overwhelming evidence," without varnishing it with such additional layers of goop as, "Genius is always belittled, and truth must struggle to come to light in a sea of darkness ..."

I'm fond of people who have unusual and non-conforming ideas, yet in my circle of audio engineering acquaintances I can think of only one person who definitely believes in the audibility of sound above 20 kHz, and he was the director of sales and marketing at Earthworks (Eric Blackmer). At least his commercial activity was consistent with his ideals!

I've got the published report of the Oohashi study right here; I anted up and paid the download fee to the American Physiological Society five years ago and read the whole thing. It does NOT claim that people can hear above the 22 or 26 kHz cutoffs that were used in the experiments; in fact the authors continually use the word "inaudible" to describe such energy. They ran EEGs and PET scans which did seem to show significant neurological responses, but not auditory ones, and the authors were at a complete loss to explain the effect. No one has come up with any explanation since then, either. (Nor to my knowledge have any results such as this been duplicated by any other experimenters.)

What is most notable to me is what was not said, however. In published studies that are much more straightforward than this one, whenever you filter out the high frequencies from a recording using all-pass filters so as to control the phase below the cutoff point and keep it comparable for A versus B, you get a rapidly diminishing ability for listeners to determine whether the highs have been cut off or not as you get to about 12 kHz. By the time you raise the cutoff to 16 kHz there are very few people--including trained engineers and musicians--who can reliably tell you whether higher-frequency energy is present or not.

In this study, the participants listened to music that was filtered at 22 or 26 kHz--yet they readily assigned adjectives describing different sound quality to the "full-range sound" versus the "high-cut sound." This strongly suggests that the sound in the audible range wasn't the same, but I see no sign in the published report that the experimenters did anything to verify this one way or the other. And it's not that hard to do.

Enough is enough about this. I didn't volunteer to take on all comers; I'm just trying to say that if and when sampling rates above 44.1 kHz have any audible advantage, it's not because higher audio frequencies are being conveyed. There aren't any higher audio frequencies--because higher frequencies are not audio frequencies.

--best regards

[datbrad]

If you're into listening tests, perform the same procedures with your own gear but use audio instead of noise . . .

I don't know what relevance it has today, but I recall vividly back at the beginning of '90s when I first started using DAT recorders, I experimented with recording at 44.1 and also at 48, this at a time when the DAT tape was the only digital medium to listen to since the CDR for economical consumer use was 7 years away at that point.

Many people, both tapers and audiophiles with so called "golded ears", as well as casual listeners, could hear the difference between 44.1 and 48 on my home system. I did both microphone and SBD recording experiments with one set of a concert made at 44.1, and the other at 48, and on playback the 48 typically seemed to have cleaner highs, more "open" at the top end, if that makes any sense.

I would say that from '93 on, I always ran 48 on DAT as a result of my own admittedly unscientific experiments.

[boojum]

If you're into listening tests, perform the same procedures with your own gear but use audio instead of noise . . .

<snip>

Many people, both tapers and audiophiles with so called "golded ears", as well as casual listeners, could hear the difference between 44.1 and 48 on my home system. I did both microphone and SBD recording experiments with one set of a concert made at 44.1, and the other at 48, and on playback the 48 typically seemed to have cleaner highs, more "open" at the top end, if that makes any sense.

<snip>

Was that a double blind test?

[datbrad]

If you're into listening tests, perform the same procedures with your own gear but use audio instead of noise . . .

<snip>

Many people, both tapers and audiophiles with so called "golded ears", as well as casual listeners, could hear the difference between 44.1 and 48 on my home system. I did both microphone and SBD recording experiments with one set of a concert made at 44.1, and the other at 48, and on playback the 48 typically seemed to have cleaner highs, more "open" at the top end, if that makes any sense.

<snip>

Was that a double blind test?

No, it was only a single blind test. I knew which sources were which when I played them, only the people listening did not know there was a difference in recording sampling rates between the samples played back to them, or what difference they were asked to listen for. Some folks did not hear a difference, but most did, and only then did I tell them what the deal was.

[boojum]

I am not arguing that your upgraded converters were not better, but I am always impressed with how much more often the proposition is proven rather than disproven in non double blind tests.  It seems to demonstrate that bias has a way of creeping in whether or not we believe it does.  For that reason I will always remain skeptical of non double blind tests.

I believe that people who test and want unbiased results use double blind tests to insure that the results are unbiased.  Double blind removes all doubt.   

[Will_S]

Well in my case the only bias was the name I gave the files; "file1" and "file2".  Those names should have been selected by someone who didn't know which source file was which, but oh well.  Also, my test didn't prove that the new converter sounded better, just that it sounded different.

Well, no, that's the problem.  There may well have been an audible difference, but you told folks they were listening to two different sources and they "heard" a difference, some preferring one and some the other.  You likely would have gotten a similar result if you posted two copies of the same file under different names, some people will always "hear" a difference.