Ya know, I looked around on the internet for opinions on what people thought. It's pretty much dependant upon the type of music, the source, and how well tuned the listener is to the source.

Just like a master acoustic guitar player can feel subtle differences between one guitar and the next, whereas someone with little guitar playing experience would say all the guitars feel the same.

Anyway, not to bring up and old debate, or argue anything at all, but I do hear a difference between them with my acoustic guitar as the source.

48khz: mids are very hard, top end was muffled.

96khz: Immediate noticeable difference from 48khz. Cleaner highs, mids are softer and smoother, bass is tighter.

192khz: Top end very airy, the "metallic" tone of the steel strings comes through. Mids have the same sound as 96khz. The sound of the flatpick hitting the strings is more noticeable than 96khz, left hand finger noise across the frets and strings is also more noticeable than 96khz.

I hesitate to say 192 is *better* than 96. Just different. 96khz is definately better than 48khz, not even a close comparison.

What happens after they are converted to 16/44khz for CD, I don't know. All I wanted to say is that I hear a difference.

I would not hear a difference if I was listening to music or instruments I am not tuned to, say for example like a flute or piano.

It depends on the converters as well. My Multiface does sound better on the input at higher sampling rates. And a lot of the higher quality gained can be retained after downsampling with a good resampling algorithm.

Yes...on something with inferior converters, the opposite might occur...mistakes are multiplied at the higher sample rates, and the listener confidently states, "I hear no difference, in fact, 44.1 sounds better!"

I too am an acoustic musician, and I hear the difference on acoustic guitar and even, sometimes, vocals. I suspect cello and violin players, among others, would also note differences. In this case it is not about Nyquist so much as it is about snapshots per time period.

Using a movie anology: we know that 16 fps is to jittery; we see immediate improvement at 24fps, so much so, that the illusion of smooth movement is created.

But even though the illusion of smooth movement is accomplished, we know that we we can drop a frame here and there and slip subliminal messages in between image frames , and the brain reads them! That means that the brain KNOWS that there is missing information in between frames, or it couldn't "read" subliminal stuff when it is put in there.

So if we increase the FPS to 48, what would happen? Perhaps certain motions would be accentuated or more natural...perhaps we wouldn't notice unless we looked at a 24, then a 48. Perhaps, raised on 24 as we are, we would choose the 24 over the 48 quality because it is more comfortable to us?

I believe that audio could be similar...that our brains/ears are capable of feeling the smoothness of more bites of sound per time period, so that, yes, with accurate converters, you will experience somthing different at 192 than you do at 44.1.

That said, most folks are really used to hearing 44.1/16, and I myself have expressed surprise at certain guitar tracks I have recorded at 96...wondering at the crispness of the harmonics. This might account for blind testing where the question, "Which sounds better" is answered with a lot of folks picking the lower sample rate. ;-)

Best,

Mark

I can also hear the differences on my Fireface, when I switch to a higher sample rate it does indeed sound better.

Still the point being made about not using 'higher sample rates than necessary' is that it's not the sample rate that matters here, it's about the analog filters in the converter design. And as Living Sounds said in his post, even after downsampling a lot of the quality was retained, so the better soundquality is not due to the higher sample rate, offcourse even when downsampling there are filters only now they're digital, so even then quality will never be the same as the original 96kHz file.

I'd say use whatever sounds best to you.

So can you recommend me tracking on 24/192 and then downsample it (by POW-R) to 24/44,1 and work with it? 24/192 seems to be pretty CPU and RAM killer...

If your converter sounds better at a higher sample rate and your system can cope with the load then offcourse go with what sounds best.

You can keep a backup of the higher resolution material and maybe remix it later. Or maybe you;ll get an app which handles downsampling even better. then what's available now.

I have yet to hear any discernable difference between 96 and 192.
Most processors are not yet geared to deal with 192 accurately.

The advantage of higher sampling rates have next to nothing to do with extended audio bandwidth. The real, practical advantage is far more effective and analogue-like high end EQ adjustments, more accurate and faster responding dynamics, and much more accurate peak level metering.

But, I agree that the converters play a far more important role here, rathen than the mechanics of delivering 192 rates.

As the samples are reconstructed to an analog signal prior to your hearing them, the analogy with movie frames is completely invalid. One does not hear "more bites [sic] of sound per time period." I'm not claiming there's no difference in sound however with different sampling rates.

-Naren

Perhaps not completely invalid. Or rather, the first part is quite valid, but sort of runs into trouble as we extend it.

The anology goes, in an obvious way, only so far as the ADC...but...now let's "print" a film to VIDEO TAPE...if you try it with 16fps, guess what you are going to see on the video tape? If you do it at 24fps, what are you going to see? In this way the analogy extends somewhat successfully. The video tape can be understood as the an analog interpretation of a series of snapshots. Not perfect, I understand, but in the sense that frames of something are successfully delivered...one can stretch the analogy a bit?

Now, suppose you take 16fps and digitize it and burn a DVD? And now 24fps with subliminals, and now 46 fps...hmmm....better have good DACs on the other end to convert. LOL...again, we see something analogous to our audio experience.

Best,

Mark

I think the resolution of the waveform after it's converted from analog to digital has to be a significant factor. I've never believed that extended frequency response at 96 and 192khz has much to do with the perceived differences in sound that some of us hear like myself.

If we take a 10khz analog waveform and then convert that to digital under the following sample rates, the resolution is what's important I believe:

For each second that passes by with a 10khz tone:

96khz = 9.6 samples per second.

192khz = 19.2 samples per second.


The ADC is now required to "draw" a smooth waveform with the available samples or "plots" like on graphpaper. Obviously, the waveform will look smoother and have a better curve with 19.2 dots versus only 9.6 dots.

That's how I perceive the differences. But to hear any differences as I mentioned you will need to be really tuned into your instrument or source. I don't even believe you have to have super quality monitoring equipment either, since we are not talking about extended frequency response. In fact, I am able to hear the differences on a cheap pair of Optima headphones, and also on my Altec Lansing computer speakers.


ps.

The famed Neumann U47's frequency dropped off at 16khz. Who said higher frequency rates are needed for legendary sound anyway?

As many may allready know, EQing the topend ITB at higher samplerates is much more pleasent.
Higher SR does give you a better horizontal resolution, so placement of sources will be finer.

Exactly, my movie analogy holds here. If you were filming yourself drawing the dots, you would notice the motion of your hand drawing each dot would be much smoother.

I could care less about extended frequency response (thus my Nyquist Comment); I have come to experience a lot of today's mixes as too bright, anyway (and I have been an "offender" myself). A gentle rolloff around 18K makes stuff much less fatiguing, IMHO, and I will be remixing a lot of my stuff this way. Man, some of my favorite 70's mixes probably don't break 16K...LOL.

"It" is all about resolution; my movie anology uses resolution of MOTION as an analog to resolution of sound waves, is all.

Best,

Mark

The movie analogy does not apply at all. By the time you hear it, there are no discete steps involved. In movies, the connection of the discrete 'samples' is done by your eyes and brain, in audio, it's done by digital and analog filters.

To get a better understanding of this, with fewer broken analogies, you should read, re-read, and read again some of the technical explanations for how digital audio works. I'd recommend Nika Aldrich's book "Digital Audio Explained for the audio engineer". A key point is that, due to the filtering being used, the samples are not simply connected dot-to-dot.

For audio in the 20-20k range, the audible quality at the various sample rates will depend completely on the quality of the converters (and surrounding analog omponents) being used, and won't necessarily get better as the sample rate is increased.

Could just be my 50 year old ears.....heh ..but it seems the higher the sample rate, the cleaner, and "thinner" the signal gets.......YMMV

There are those -- including the legendary and erudite high end converter designer Dan Lavry -- who will tell you that, per the implications of the Nyquist theory that the MOST accurate sampling for the "conventional" audio range (20-20kHz) is, indeed in the 40-50 kHz SR range. Those folks will tell you that the science dictates that -- even if one is to extend the range to be covered above the limits of the human hearing range to, say, 30 kHz (and there is NO science indicating humans CAN hear that high) that the most accurate SR would then be in the upper 60-70kHz range.

Those folks -- who base their position on solid, verifiable science -- will tell you that HIGHER sample rates actually induce GREATER inaccuracies.

And, of course, they might suggest that, if you really do hear a difference (and who can tell you what you think you hear, eh? At least, not without proper Blindfold testing) it is this increased distortion that you're hearing.

As we all know, distortion is often perceived as an "improvement" in sound. (Cfr the controversy over the "benefits" of using an external "master clock" with a single AD interface.)

So, if you like what you hear... and you don't mind the overhead... why not?


That said, I always urge people to try to verify what they THINK they hear with properly administered Blindfold testing. Otherwise, one is fighting against a number of psychological factors. You may hear what you want to hear. You may hear what you're afraid you'll hear. But until you remove expectation and distraction from your evaluation, you'll never really know for sure.


BTW... on the conversion thing... I used to have a fairly anti-SRC (sample rate conversion) attitude, particularly when moving from a SR that is not an even multiple of the target rate. BUT I have been recently convinced that today's BEST SRCs are actually quite good at delivering such downsampling with minimal, even negligible 'damage' to the intended signal. But there is a definite range in SRC quality. I would NOT use the SRC included with my DAW (Sonar -- or for that matter the one in Cubase SX, both of which seem to have a very similar and none too auspicious quality as reflected in these tests: http://src.infinitewave.ca/ [can't weigh in on their methodology/accuracy, mind you!] )

And that's just not true!

As Dan Lavry pointed out, the frequencies till 20khz are captured equally well with a 44.1/48 kHz converter than with a 96 kHz converter. Everything the 44.1 converter is not able to capture, any nuissances, are above the sample rate!!! A higher sample rate is not needed to capture to 20khz.

Offcourse you can discuss when processing and mixing happens at a higher sample rate if the outcome will be better or if you might need unhearable frequencies, but there's not much of a point in this if you don't have a mic which can capture these frequencies. Those exist, but still in small numbers...

I am giving up defenind or explaining analogies. For you that hate movie analogies, here's an SOS article that uses one...

http://www.soundonsound.com/sos/aug0...ostscience.htm

Part one of the "Lost Art of Sampling". Enjoy!

After all is said and done, it seems to me the right compromise of power, disk space, and sound for recording on my dual-core AMD and humble RME FF800 is 44.1/24...so that is what I am doing from now on. (but will occassionally sneak in the 96K/24 acoustic guitar tracks!)

Best,

Mark

That was a great read. I saw it a few weeks ago, good stuff!

I do not know the subtlties of the science behind it but I did some double blind tests with my band and wife, recording 10 tracks of drums with a Fireface all at 24 bit. The verdict was that there was a noticeable difference between 44.1 and 48, less but still noticable between 48 and 96 and a very very subtle difference between 96 & 192. When asked what "pleased" the listener, all 5 picked either 48 or 96. Interesting. We decided to do the record at 96.

Agree. In my thread starting post, I stressed that 192 khz on the acoustic guitar was not necessarily better, just different than 96 khz.

Just an addendum to my post; I've noticed this difference between 192 and 96 over the past week after the culmination of numerous sessions, this isn't from just a 1/2 hour A/B test, but a session here, a session there, almost everytime at 192 khz the guitar takes on this "sparkly" and intimate nature.

I don't think I'd want to do 192 with several instruments, as the result may be too intimate, it's not always pleasant hearing every finger and pick noise in an acoustic band...

I guess you could argue that you could perceive 2nd, 3rd and later level harmonics at that level which perceptually is the usual way you will be able to tell the difference between the two. Bearing in mind this will be at 20kHz, 30kHz etc. I was wondering if your mother dallied with a fruitbat? heh (Please do not take the former as a serious suggestion it is merely a little humour used for emphasis). With very fine equipment and very young ears you should be able to perceive the difference between a sawtooth and and a sine as the second harmonic is just within the normal range of hearing. If you listened to a single cycle of the graphs you describe, I doubt anybody would be able to resolve anything other than an audible click.

As has been stated before, the main advantages are in accurately resolving transients, high frequency EQ adjustments and accurately charting peak and average levels. This produces a more accurate mix that can then be dithered to retail levels for a more professional result.

Reg

FWIW here's a little history of sample rates.

The first developers of high fidelity digital audio did a bunch of testing and came to the conclusion that the ideal sample rate was somewhere between 50kHz. and 60 kHz. At the time slightly less than 14 bits was possible.

As the price came down, the SMPTE and AES sponsored research into what minimum digital audio format would be required to achieve the transparency required for professional audio applications. The result of this study was the 48kHz. 20 bit format that became the audio for video standard. This led to mass produced converter chips that were optimized for that format. Only some of the highest-end outboard converters have ever really been optimized for 44.1 so it shouldn't be surprising many common devices, especially low priced ones don't perform all that well at 44.

As for performance above 48k, there's still plenty of science supporting frequencies up to 60 kHz. so 88.2 and 96k are not all that extravagant however listening tests have found that ultra high quality 24kHz. low-pass filters applied to 96kHz. audio are virtually inaudible suggesting filter quality rather than bandwidth is the real issue.

In the real world of less than ideal filters, many people have observed differences in performance at higher sample rates with particular converters. Depending on the device in question, this can be extremely subtle or pretty obvious. Likewise many D to A converters perform better at higher sample rates which adds to the confusion.

The bottom line is to use whatever sounds best rather than what is theoretically required. This is one of those arguments where everybody is right until they start generalizing about gear other than whatever they happen to be using.

Hi, Mark, the movie analogy in the article is not extended beyond the ADC and what happens in the digital domain. The analogy is used in the context of capturing the higher audible frequencies in the digital domain, and says nothing about the number of "frames" of sound that your ears perceive. The article does say:

"Then, when you play back the sound, the stored loudness values are played back in the correct order, reconstituted by the DAC as a variable waveform..."

"Variable waveform" is the relevant phrase here.

-Naren

Could be, but I just do not know, as we are all "theorizing" why there's a difference in tone.

As I thought though, the discussion has evolved from specific instances, as in my ears being able to perceive slight alterations in tone from my guitar, to generalizing differences in overall music.

My disclaimer was that I would not be able to perceive a tonal difference between 96 and 192 for other instruments like a piano or a bass or a trumpet, because I am not tuned into their tonal qualities.

I'll go with what BoB said.

Inaccuracies-- There are mistakes in computers with there 1 an 0 per 100 line.

That's why jet airplanes have back up systems. Because your getting into a Billionth of a sec almost.

Filters--- from what I learned, that's what gave digital it's cold sound in the beginning.

Harmonics---That's what gives sound it's complexity.

an last--I'm not as smart as you people. Good reading though.

Not to argue with Mr. Lavry...but these arguments don't do justice to the underlying physics.

Suppose I create a simple 1kHz square wave with my analog synth. No big deal. Now, according to Mr. Fourier, this simple square wave is accurately represented by an infinite Fourier series of sine waves. Infinite Fourier series of sine waves means lot's of high-frequency content, even in my simple 1kHz square wave.

The more of this high-frequency content I can record, the more accurately my square wave can be recreated. So, higher sampling rates are necessary for something as simple as an analog synth square wave.

It's not necessarily about the conventional audio range or human hearing limitations, but rather the shape of the waveform being recorded. Unless you're recording only sine waves, it's important to understand the relationship between Nyquist and Fourier. For some reason, this relationship is usually left out of the discussions about sampling rates...

Dave

Well there are some things that do bear out what Bob has to say. Not name dropping , but I am very lucky to be asked on a reasonably regular basis by a very good pro audio designer for my opinion of his work.

The conversation normally goes something along the lines of, well I've had it on my test bed and I've sent it off to blah, we concur that nothing really starts to break down until 60-70kHz and we're getting good results down to (cycle rates where you could send messages to the US nuclear Submarine fleet if you believe books like 'The Hunt for Red Oktober'). Needless to say all of his kit performs excellently anyway in the 20:20 range that is typically regarded as important.

Does this attention to detail yield a sonic benefit. I send the kit to people with far, far better ears than mine and they pretty much exclusively say 'Yes'. However, whether this attention to detail simply means that results over the generally accepted audible range are just that much better or that these 'out of theoretical range' issues make a genuine difference, I can't really say.

I can say absolutely that the gent concerned gets far more worked up about the fruits of his labours being passed into conversion designs that limit out the fruits of all that tinkering. Then again he has in the past called an Audient desk "a college unit", described a number of sample recordings for testing into Logic through some very decent converters as "having lost of alot" and kept me on the phone for a hour, (a time I hasten to add that was an education I really would not have missed for the world), discussing his concerns at an EQ response graph in an area I know my poor addled ears couldn't hear if the good Lord himself was booming down the mike. heh

I hate it when I buy a new CD and I can tell they didnt mixed it at 192. It just ruins the album for me

oh wait....no...I never say thattutt

Did anyone ever ponder that hard drive and AD converter companies might be cahoots, conspiring to sell more product by "dumbing down" the digital output when you sample at a lower rate? When running in 96k+ mode, you're getting the regular quality of the unit, and at 48k and less, maybe they've introduced some awful-sounding filter?

You begin to believe that a higher SR is better, thus you have you buy bigger, faster, and more hard drives.

Just a conspiracy theory to liven up the discussion (because this topic needs more drama...)

I think what you're forgetting in the point of diminishing returns. Sure, it would nice to capture EVERY harmonic of your 1khz square wave, but it's not possible nor necessary.

A true square wave (which doesn't exist in the real world) has an infinite number of harmonics which make it look square. In order to record the square wave, you need to capture all the harmonics that lie in the audible range. Outside of this range, it's just wasted hard drive space.

You can't conclude that a higher sampling rate is necessary to capture a simple square wave simply because the square contains harmonics. These harmonics extend infinitely, and at some point you've got to say enough is enough and LPF the signal at the Nyquist freq.

The shape of the wave is really not of interest. If the harmonics which give a waveform it's space are outside the audible spectrum, they have no affect on the listener. Do you want your sounds to sound good or look good?