[jfaaz]

I have been listening to audio in 192, 176, 88, 48 and 44.1khz.  At first I thought that I could here much more clarity and detail in the higher sample rates.  Then I down sampled some 192 and 96khz audio to 48khz.  I could here no difference in quality.  I use a mac mini and sometimes a Fiio X3 as a dac.  I listen  through ATH-M50 cans and Bose 201 speakers.  When I look at a spectrograph at  the audio files that I have converted to 48khz, the graph looks full and fat.  I am no scientist, but from my ear 48khz sometimes sounds better than 44.1, but no different from anything higher.  Are there more like me that would rather have more free space than extra ones and zeros?

 

[MindsMirror]

A sample rate of 48KHz can perfectly reproduce any sound at frequencies less than 24KHz, that's half the sample rate. The upper limit of human hearing is around 20KHz. It's certainly possible that you can hear up to 22.05KHz or more, so you might be able to hear a difference between 44.1k and 48k sample rates. It's highly unlikely that you can hear anything past 24KHz, so all you are getting with those higher sample rates is inaudible ultrasonic sound.

 

[nick_charles]

There is really no compelling evidence that anything above 44.1 KHz is really required for the delivery end of digital audio as this easily covers the audible range for 99.99% of humans. A substantial number of DBT have failed to show any reliable ability to detect the difference between red book and higher, with the exception of one slightly dubious study. But have you tried DBT'ing the 44.1 and 48.1 versions of files for yourself ? If you find you cannot tell the difference then you can use 44.1 without further worry !

 

[anetode]

Are there more like me that would rather have more free space than extra ones and zeros?

Yes. Sometimes well-mastered versions of recordings are only available through a high-rez download or an SACD. In those cases I convert everything down to 16/48.
 
The most convincing argument I've heard for having sufficient padding above a 40khz sample rate has to do with a choice of gentler digital filters, though it's highly unlikely that there's an audible difference even with filters which are a couple db down in reproducing a 20khz signal. I choose 48khz to gild the lily, going for 88 or higher doesn't really make any sense for playback. However there are a couple of reasons for using a high sample rate in audio recording and production.

 

[SharpEars]

  There is really no compelling evidence that anything above 44.1 KHz is really required for the delivery end of digital audio as this easily covers the audible range for 99.99% of humans. A substantial number of DBT have failed to show any reliable ability to detect the difference between red book and higher, with the exception of one slightly dubious study. But have you tried DBT'ing the 44.1 and 48.1 versions of files for yourself ? If you find you cannot tell the difference then you can use 44.1 without further worry !

 
Actually this covers 100% of humans. A superhuman with the ability to hear beyond 22 kHz has never been found.

 

[castleofargh]

 if you have hirez tracks that are a multiple of 48 then as said it could make sense to convert to 48khz. because there is no calculation needed and no added quantization, you could just remove every other sample of 96khz and be done with it in principle. but then you have to be sure that you will play your tracks at the right rate. if it ends up with windows converting it to 44 on the fly because that's your default output choice, then you gain nothing. and if you set windows to 48 then the 44 files will be converted. but as long as you know what you're doing with your audio software there is nothing wrong with using 48. and even my "worrisome" example surely doesn't have any audible effect.
 
will it sound better than 44khz? I doubt it, the better low pass filter argument to me is a false one. it's not like your DAC is a different one with different filters when you use 48khz, 44 and 48 will be oversampled before reaching the filter so it will work just fine and won't care about the original sample rate anyway.
 
if you "believe" you can ear 20khz and above then surely 48 is better, but I wouldn't bet too much on that.
 
 
 
so my vision about this is that 48 is great, but doesn't actually bring anything useful to the table. just like higher sample rates.

 

[MindsMirror]

   if you have hirez tracks that are a multiple of 48 then as said it could make sense to convert to 48khz. because there is no calculation needed and no added quantization, you could just remove every other sample of 96khz and be done with it in principle.

That's not entirely correct. If you down sample by removing every other sample you will end up with aliasing. The high frequencies you are removing can unintentionally appear as lower frequencies.
 
Sometimes a gentler low pass filter is used rather than a hard sinc filter to ensure that no aliasing will occur. If you look at the spectrum of a few CDs, some have sound all the way up the spectrum and some roll it off more gently at the top. Therefor it is theoretically possible that with certain sources, depending on what sort of filter is used, a sample rate higher than 44.1k could be audibly different, even if you can't hear all the way up to 22.05k.

 

[castleofargh]

 

   if you have hirez tracks that are a multiple of 48 then as said it could make sense to convert to 48khz. because there is no calculation needed and no added quantization, you could just remove every other sample of 96khz and be done with it in principle.

That's not entirely correct. If you down sample by removing every other sample you will end up with aliasing. The high frequencies you are removing can unintentionally appear as lower frequencies.
 
Sometimes a gentler low pass filter is used rather than a hard sinc filter to ensure that no aliasing will occur. If you look at the spectrum of a few CDs, some have sound all the way up the spectrum and some roll it off more gently at the top. Therefor it is theoretically possible that with certain sources, depending on what sort of filter is used, a sample rate higher than 44.1k could be audibly different, even if you can't hear all the way up to 22.05k.

 I hesitated when I wrote it because it's was an obvious oversimplification, that's why I added "in principle". and as it often happens, my attempts in making things simple to grasp end up making me say wrong stuff. you're right and you explained it clearly.

 

[stv014]

Sometimes a gentler low pass filter is used rather than a hard sinc filter to ensure that no aliasing will occur.

 
The advantage of a filter with a gentler slope is actually that it reduces ringing, in addition to being cheaper to calculate. A filter with a steep roll-off still removes aliasing as long as its response falls low enough by the Nyquist frequency.
 
Also, a non-integer resample ratio does not necessarily imply worse quality by any practically significant amount, it just makes the processing computationally more expensive. This can be seen in these tests, where resampling from 44.1 to 96 kHz is more than twice as slow as 44.1 to 88.2, but can still achieve up to 200 dB stopband rejection. A low-pass filter is always needed, and the trade-offs are the same, there is just an additional interpolation step needed, which can be made transparent with good design.

 

[esldude]

   
Actually this covers 100% of humans. A superhuman with the ability to hear beyond 22 kHz has never been found.

Well, no not quite correct.  There have been blind tests where young adults could hear as high as 25 khz.  This with a test tone and with a very high threshold of about 100 db in terms of loudness.  Essentially meaningless with most anything other than test signals.  It is why some researchers wanting the ability to say a system covers 100% of humans will say they favored a 60 khz sampling rate with bandwidth to 25 khz and a 5khz transition band.

 

[bigshot]

That isn't really hearing. That is more like feeling than hearing. It's the detection of sound pressure. They use ultra high frequencies at loud volumes in riot control, and to keep kids from hanging around 7-11s and gas stations. It's uncomfortable.

 

[esldude]

  That isn't really hearing. That is more like feeling than hearing. It's the detection of sound pressure. They use ultra high frequencies at loud volumes in riot control, and to keep kids from hanging around 7-11s and gas stations. It's uncomfortable.

Not according to some testees.  It was in fact heard.  Wish I had the relevant paper handy.  It was about 1% of young adults who had already been screened to have good hearing.   I seem to recall only one reached 25 khz, one about 24 khz and several around 23 khz. Again the threshold was around or just over 100 db so not a lot left before it would become uncomfortably loud.  Also don't know of research showing such was audible even at those levels in the presence of just under 20khz tones.  Again making it likely a non-issue for any normal music rather than test tones.  A by a good margin not relevant to me at my age.

 

[bigshot]

I would bet they were hearing artifacts down lower that the loud sound pressure was causing in their ears. I know when I hear very loud high frequencies, there is a dull ache and thud in my ears that goes along with it. I used to produce the Chipmunks TV show and I had to stuff cotton in my ears at recording sessions because the loud aspirates sped up made my ears feel like a drumhead being pounded on... thud, thud, thud.

 

[ab initio]

 
Not according to some testees.  It was in fact heard.  Wish I had the relevant paper handy.  It was about 1% of young adults who had already been screened to have good hearing.   I seem to recall only one reached 25 khz, one about 24 khz and several around 23 khz. Again the threshold was around or just over 100 db so not a lot left before it would become uncomfortably loud.  Also don't know of research showing such was audible even at those levels in the presence of just under 20khz tones.  Again making it likely a non-issue for any normal music rather than test tones.  A by a good margin not relevant to me at my age.

I have no problem with those very rare six-sigma individuals who can hear to 24--25kHz. I find it comforting to know that any sample rate over 50kHz is theoretically useless in 100% EVERY situation. I consider the (50kHz/44.1kHz) =  1.13 times factor of safety for the remaining 99.99966% of the population an extremely reasonable number to mathematically guarantee every audible sound is encodable to every human on the planet.
 
Cheers

 

[eugenius]

Your dog needs 192kHz.
 
If you put an antenna hat on him, even that's not enough - that's why DSD was invented.