One concept that is fairly easy to explain is smoothing filters. After a signal has been converted from digital to analog, it's a series of stairsteps. The smoothing filter smooths out the stairsteps to create a continuous waveform, but more importantly, re-create the original waveform.
 
Here's an analogy. A baseball player hits a fly ball to center field and you're in right field. Plot the height of the ball every 10 ms. This produces a series of points that correspond to the height. If you connect these points with a line, you will describe the ball's trajectory. Now, you might think "so if the points are closer together, the arc will be described more accurately." And that's true, BUT like a baseball audio doesn't move in a series of straight lines from point to point, and a smoothing filter recreates a smoothed curve based on the points. If you compare that smoothed curve to the ball's actual trajectory, you'll find the correlation is for all practical purposes identical.
 
The smoothing fllter is an important part of the "sound" of digital audio. Many people think the reason why DSD sounds more "analog" (whatever that means, LOL) is because the sampling rate is so high - a minimum of 2.8 MHz, with 5.6 MHz also being common - you can basically filter out the clock by twisting a couple wires together and hanging them across the output. Well maybe not quite, but you get the idea.
 
Those who think sample rates need to be much higher would make the argument in the baseball analogy of "Well, what happens if the ball hits a fly in between the two points where it's measured? Wouldn't that knock it off course somewhat?" The counter-argument would be that even at 44.1 kHz, the distance between the two points would be sufficiently close to take that into account.

Hmmm- I guess I am still confused.  I can understand that the frequencies beyond a certain level may be insignificant or not matter at all.  To me, resolution is different than the frequency range of the recording.  Resolution to me, means how many spaces are there between samples.  And it seems that the less space, the higher the resolution.  But maybe after a certain resolution is achieved, the human brain cannot perceived any higher resolution.
 
For instance if a movies is 1000 frames per second or 2000 frames per second, can we tell the difference?  It is 1,000,000 frames per second versus 2,000,000 frames per second, can we tell.  It does seem to make sense that after a certain amount of frames (or samples) per second, the human brain cannot tell the difference.
 
 

drewfx1
Craig (Anderton) had a long thread a while back on this (that got similarly sidetracked like this one!) after he found that some synths/effects sounded better when run at a higher rate.
 
Basically some, but not all, Digital Signal Processing (DSP) benefits from being done at higher sampling rates. Anything that produces overtones higher than one half the sampling rate will alias and sound somewhere between "ugly" and "less smooth" at a lower rate. 
 
So if a synth or effect that would benefit from this doesn't oversample internally (as many of us would argue it should), then running Sonar at a higher rate can improve things. In my case, everything I use that would benefit from oversampling happens to already do so internally, so there's no benefit to me. 
 
I would suspect that older stuff - written when CPU power was less abundant - would be more likely to not oversample internally. 

What sample rate(s) do you use, Drew? Just out of curiosity.
 
And (if you know) does everything that comes standard with X3 oversample internally? That's most of what I use, personally.

drewfx1
Again, there is nothing lost between the samples. Everything between the samples is already stored in the stream of samples. It has nothing to do with the brain - the sampling theorem mathematically proves that nothing is lost between the samples. There is no "guestimating" what's between the samples - the actual waveform at every point between the samples is stored in the samples. Since the entire waveform can be, and is, reconstructed from the samples, there is no space between the samples. You can't get any higher resolution because you already have everything - except for the high frequencies that are filtered out at the beginning of the process.
 
The problem is that it is simply not remotely intuitive how the part between the samples is stored, so people just assume it is lost. But it isn't - it's right there in the samples you have.

Might some kind of diagram that points out what is what as far as sample rate and bit depth in relation to the peaks, troughs, wavelengths, etc. of a sound wave be helpful here?...
 
(Not saying that I have such a diagram, just spitballing the suggestion.)

Jeff Evans

Anderton
And I STILL think DSD sounds better than CDs...but in the immortal words of Herman Cain, "I don't have facts to back me up."

In that situation you are comparing a 44.1K sampling rate to 192K and yes that certainly could sound different. (A bit unfair don't you think)  But how would a comparison go between a playback 96K sample rate and 192K then. I wonder if there is any audible difference.

 
I have mixed feelings about PCM vs. DSD because there are too many variables. At a listening test at AES among multiple formats, DSD sounded the "smoothest" or "most analog" to me. Although I'm not a cork-sniffer audio dude I could hear a difference. But that's the word - difference. I'm not sure that translates to "better" just because I subjectively found the sound more pleasing. Also, the technologies are so different - one bit with a multi-megahertz clock is very different from lots of bits with a kilohertz clock. 
 

I think the whole topic of this thread is a bit moot and uninteresting to say the least.  There are far better things one could talk about eg how to get better mixes! And that has nothing to do with sampling rate!

No, but the topic affects how our decisions will be made in the present in terms of recording. Pioneer is offering DJ tracks at 192 kHz. I've heard that in Japan, several companies offer music at 192 kHz...and interface manufacturers who do 192 kHz are telling stores that any interface that does less is destined to be a doorstop. There's a vocal minority, albeit a growing one, in the audio community that swears 192 kHz sounds better than 96 kHz. (I haven't seen a UFO, but if a career pilot sees one and it's picked up on radar, I'm going to ascribe at least some credibility to the pilot and the radar rather than flat out say UFOs can't exist.)
 
So, if we.re going to be expected to make masters available at 192 kHz, we need to know that now. A lot of people on these forums make a living from what they do with SONAR. Some people have already transitioned to 96/24 not because they think it sounds better, but because that's what Apple wants. When I was doing sample libraries back in the days when 44.1/16 was king, I recorded at 24-bit resolution and was very glad I did when sample libraries pretty much insisted that material be 24-bit.
 
In terms of art higher sample rates may not be that relevant but in terms of commerce, they are.

Jeff Evans
I think the whole topic of this thread is a bit moot and uninteresting to say the least.  There are far better things one could talk about eg how to get better mixes! And that has nothing to do with sampling rate!

 
People are led to believe that things like this are a big deal, so we have the occasional thread discussing it in detail.  
 

I have read some very interesting books by people such as Bob Katz and others and most agree that a higher sampling rate is better but we only really need to go to about 60K in order to really get the improvement.  So the closest thing to that we have today is either 88.2K or 96K.  Once you start going much higher you are getting into that very subtle hi fi territory where tons of expense will yield only a very small or even inaudible result.

 
I believe that those discussions were not taking oversampling into account. 
 
With the oversampling converters that are in overwhelming use today the only questions end up coming down to whether ultrasonics are of any value (and for all the noise about them, there isn't much in the way of proof for some reason), and whether the digital filters in the converters can do their job transparently at 44.1/48kHz.

My reference to sampling rates being around 60Khz from people such as Bob Katz are referring to mainly converter rates coming in and going out, recording high quality analog sources such as quality mics/pres etc on very nice instruments.  It was not talking about internal sampling rates for things such as VST's or effects processing.  I think we now know that having higher processing rates for those things are a good thing.
 
In terms of commercial I would have thought millions of low res files being sold being compared to only a handful of very high quality formats is pretty obvious.
 
I don't think the industry is going to start hassling us for 192KHz masters anytime soon.  Making money too is not based on what sample rate you use either.  It is very much about something else.  So even if we do have to this later I am sure we will have time.  And the obvious thing in upgrading all our hardware to be able to do it is another matter as well.
 
I am just realising what is required for an all 96K system end to end and for me that means working with digital mixers and they too have to be able to handle that end to end. Not that many can  do it. Many stop at 48 Khz.

deswind
Question for Drew:
I am starting to understand, I think.
You indicated that there is no space between the samples.
Assuming frequency range is not an issue, at what sample rate is there an audible space between the samples?

There is never any audible space between the samples because the DAC reconstructs a smooth analog curve that passes through each sample point. 

mike_mccue

 
Yes, this ^. 
 
If you aren't sure if you understand sampling, or even if you think you do, please watch all of that video.

drewfx1

mike_mccue

 
Yes, this ^. 
 
If you aren't sure if you understand sampling, or even if you think you do, please watch all of that video.

 
I watch it, once a year, in 720P, because the explanations make more sense when you watch it in HD.

Also beluga whales can hear past 100,000 hz.  Why limit one's market?  Who knows what the future holds!   (And some scientists may need to record at high sample rates for testing with them!)
Also, it is conceivable that the high harmonics may affect sound in the hearing range or may even affect us subliminally.
As I said, I just do not see a stopping point of a sampling rate that lasts for thousands of years.  It just goes against human nature.
However, Drew made a lot of good points.  But will we really stop at a certain rate?  I doubt it, notwithstanding the scientific arguments.  In fact, it appears that we are already using sample rates that go against the scientific logic.

Jeff Evans
I don't think the industry is going to start hassling us for 192KHz masters anytime soon.

 
Always the optimist  
 
Actually in this thread Milton said that some orchestras insist on recording at 192 or even 384 kHz, and higher sample rate recordings are already being offered. We also have to take into account that the record industry is desperate to find something to increase revenue, and will want to be able to sell/hype re-done versions of recordings in a higher-resolution format.
 
Of course, the joke will be if these higher-resolution formats are remasters of existing material without the insane amounts of compression. Then people will listen to the re-mastered versions and say "Wow, this hi-res stuff sounds sooo much better!"

One thing I would love to see is that we no longer use terms like "High Quality" or "High Definition". They have no meaning when applied to audio.  Even Hi Fi has lost its original meaning if in fact it ever had one. . 
 
We even misapply the term Stereo.  
 
 

Higher sample rates have nothing to do with headroom whatsoever.  Whilst in a floating point digital domain, one has virtually infinite headroom due to the range of values which can be represented in floating point.
 
Oversampling in the DAC's interpolation filter gives better SNR, and greater bit depths give better dynamic range which also relates to (lower) noise floor levels.  Sampling rates only change the frequency range which can be sampled and reproduced.

Something else to consider with higher sampling rates, I didn't see in this thread (or overlooked): Accurate clocking. The higher the sample rates, the more accurate your clocking needs to be, since jitter and component issues may become more of a problem. This is why you see these oven-baked atomic clocks offered up from Antelope, for example.
 
If you're sampling at 192Khz with a low-quality clock, you may stand to benefit from halving the sampling rate and improving transient materials, for example. It could be a noticeable improvement.