Music Manic wrote:
Well, having a higher sample rate using effects and synths in the box makes them have more depth and clarity.

That comes squarely down to piss poor engineering of the software kind.

There are a LOT of complete muppets out there writing (Particularly audio plugin) DSP code without really understanding this stuff. There ARE processes in DSP that can produce harmonics (Anything non linear springs to mind), the sample rate AT THAT POINT must be sufficient to satisfy the sampling theorem, but that has little to nothing to do with the input or output rates.

It is called a multi rate system and is entirely standard in DSP software.

Further there are places where using an excessively high samplerate can cause serious problems, the coefficient precision needed for a low frequency IIR filter increases as the filter cutoff becomes a smaller fraction of the samplerate to give just one example.

Now with faster CPUs and more storage space etc we aren't restricted so much.

That's right. And that's why I record mostly at 88.2kHz. It is neatly divisible by 44.1. Storage is cheap. I can hear the difference. I've decided these reasons are enough for me.

I find 44.1kHz perfectly sufficient but if I am feeling a bit gay and carefree, I might use 48kHz.

But that fact of the matter is that aside from (most) mics and amps and monitors being unable to capture/reproduce much above 20kHz (even on a good day with a prevailing wind), most peoples' ears rarely meet the 20Hz-20kHz theoretical spec either and I'd wager most people here can't hear much (if anything) above 18kHz (and considerably less in many cases). There was a time when age was the determining factor but modern music lovers have loud concerts which can seriously damage hearing in extremis whilst young people (in particular) have the tyranny of earbuds relentlessly pounding away on the cilia strapped, as youngsters are, to their iPods listening to full on music with no dynamic range.

And we're discussing recording at 96kHz to give a (theoretical) bandwidth of 48kHz to play on amps and speakers that can't really reproduce anything over 20kHz to a pair of ears that can't hear much above 18kHz best case? Errmmmm!

Apart from the frequency bandwidth advantages, there must be more to these higher sample rates -
How many times you sample a sound per second must affect the detail of the captured sound - more samples more detail. More detail within the frequencies we CAN hear.
I kept my coat on ..

Stan wrote:Apart from the frequency bandwidth advantages, there must be more to these higher sample rates -
How many times you sample a sound per second must affect the detail of the captured sound - more samples more detail. More detail within the frequencies we CAN hear.
I kept my coat on ..

There isn't anymore "detail" TO hear! 18kHz is 18kHz 19.156kHz IS 19.1di da...So long as they are ALL there (and they are) EVERYTHING is there!

Running tape at 30ips gets you up to 30kHzish (all things else being right!) but that does NOT improve any *^%$g DETAIL at 15kHz!(which BTW we were all pretty happy with when the BBC did it right!).

Because if a sampler has an upper frequency limit of say 20K, it can accurately capture and reproduce ANY transient you can produce using only frequencies within that bandwidth and maximum level limits!
obviously this assumes competent design and no slew rate or power bandwidth limits (neither are a given).
Given a flat frequency response and constant group delay (granted sometimes not as much of a given as it should be), frequency response and transient response are different ways of seeing the same thing!

It is for this reason that all those brain dead "thought" experiments involving square waves being sampled are so bloody stupid, they assume something that does not exist (a square wave).....

And no, adding 'more dots' does not improve the accuracy, more word length lowers the noise floor (and that is all) and more samples increases the upper frequency response limit, but 3 points on a cycle is enough to completely specify a sine wave, and the limit of bandwidth as being LESS then 1/2 the sample rate ensures there are always at least 3 points....

This stuff is counter intuitive, but then much signals theory is.... Negative frequencies (yes really), all sorts of weirdness.

dmills wrote:
And no, adding 'more dots' does not improve the accuracy, more word length lowers the noise floor (and that is all) and more samples increases the upper frequency response limit, but 3 points on a cycle is enough to completely specify a sine wave, and the limit of bandwidth as being LESS then 1/2 the sample rate ensures there are always at least 3 points....

Regards, Dan.

Yes but all the diagrams do is show us people that that's all that's going on and that's why we need to come here to clarify things Dan. 3 points needs the underlying mathematical equations to make it mean anything, otherwise it is just three dots. That's what isn't being told properly and that's why people, like me, get confused and misunderstand things.

Music Manic wrote:
3 points needs the underlying mathematical equations to make it mean anything, otherwise it is just three dots. That's what isn't being told properly and that's why people, like me, get confused and misunderstand things.
Thanks

Unless this forum supports LaTeX (A math typesetting language) in some sane form, I am not getting into trying to write a proof (You really do need to have the correct symbols available!), besides that proof is easy to look up (A level pure maths book, most of them should discuss this), and can probably be done geometrically (but I am NOT going to try it as ascii art!)....

Off duty BBQ lighter AKA Idris wrote:
everyone focuses on upper limit frequency response when talking about higher sample rates....

without considering what it might mean to transient impulse response...

This.

The only transient information one can "hear" is those within our limits anyway. We can feel shock waves below are hearing limit - and I'm sure above. But speakers can't properly reproduce a shock wave - too fast and too loud.

And Dans right - there isn't a such thing as a square wave. Also we only experience impulse responses bound by our own hearing bandwidths. The values we use in digital sampling go along with those natural characteristics.

Dots and samples ?Look up sinc functions and reconstruction filters.

an expert on this stuff i am not, however, my take on it all is this...

if you are doing classical 96k

if you are doing video 48k

everything else 44.1k

any other cogitating upon it is a little OTT and frankly seems to be nothing more then brain/knowledge flexing... not that having an intrinsic and thorough understanding of it all won't subconsciously help in the end product, but, that alone won't do anything to the end product as far as making it better without the creative content part first being there...

As Dan said it's all confusingly counter-intuitive, even to those of us that think we've got a bit of a grasp of it!

Another factor that muddies things further, is that when people say they can hear the difference between 48k & 96k sampling rate, what they are most likely hearing is the performance of crap analogue filters. At the higher sample rate you can have a much more gentle slope which means fewer phase and stability compromises.