[slwiser]

Not the same thing as the OP requested by interesting none the less. This also has some interesting recordings samples.

A comparison of several high end DACs at 24/96:

Analogue to digital converters comparison

 

[ShanDo]

does it have sense do upsample something by soundcard? (i tried in foobar to upsample to 96Khz but i didnt hear any difference)

 

[b0dhi]

Using a sawtooth, square and other waveforms for comparison between 44.1 and other sampling rates, there was a very obvious difference at 96khz, but beyond that I could not tell any differences. The difference in music should be much less obvious but still "there", even if someone can't pass an ABX test for it (because essentially these tests measure both hearing and sound memory, which is known to be very poor).

 

[hciman77]

Quote:

Originally Posted by HFat /img/forum/go_quote.gif I was thinking about downsampling it (or any other sample) yourself... I wouldn't trust one to be a mere downsampling of the other. In foobar, you need to use replaygain for volume matching in blind tests, right? I'm not sure it's accurate enough though.

When I looked at them using Audacity the waveforms looked identical in shape and intensity until you zoomed it up to absurd levels i.e at the sample level you could make out really small differences in the shapes the intensity was still visually near identical. Audacity has a spectrum analyser but sadly it uses different frequency points for 24/96 and 16/44.1 so I could not use that. Audacity would not let me make a 16/44.1 downsample of the 24/96 track it gave me a 16/96

I will find a way to downsample the 24/96 with suitable dither.

Are there any better (free) ABX programs out there ?

FooBar may be inaccurate as you say but even so it seems to me that the big claims for 24/96 are at least somewhat overstated, I have ordered an Edirol USB sound card that should be superior to my Laptop souncard and has a digital out so I will be able to feed the output to my DAC, perhasp that will be a better test.

 

[HFat]

Quote:

Originally Posted by b0dhi /img/forum/go_quote.gif (because essentially these tests measure both hearing and sound memory, which is known to be very poor).

Only if you can't tell the difference... if you can tell 96k from 44k, you don't need to remember the sound, only which is which. It's about as hard as remembering which of your hands is the right one, except you can't do the L trick you learned in kindergarten.

 

[mr. nice]

http://www.lavryengineering.com/docu...ing_Theory.pdf

above is an interesting read. if he thinks 192k is useless then i need to believe him. personally i think it's all marketing. the bit rate is much more important.

 

[HFat]

Quote:

Originally Posted by hciman77 /img/forum/go_quote.gif I will find a way to downsample the 24/96 with suitable dither.

ssrc (the console utility) is a popular way to do that.

Quote:

Originally Posted by hciman77 /img/forum/go_quote.gif I have ordered an Edirol USB sound card that should be superior to my Laptop souncard and has a digital out so I will be able to feed the output to my DAC, perhasp that will be a better test.

I think so... not so much because the card or even the DAC is better (you might want to try listening to the card itself by the way) but because this should allow you to play 24/96 to being with. I don't know what laptop you have and what Edirol model you're talking about so I'm just guessing but, in my case, I can only listen to 96k without downsampling through my Edirol USB card (and even then, only in "advanced mode" if I'm not mistaken) because my laptop's card only plays 48k (everything is automatically resampled).
In order to ABX with your Edirol, you may need to upsample in software by the way.

 

[HFat]

Quote:

Originally Posted by mr. nice /img/forum/go_quote.gif the bit rate is much more important.

You mean the bit depth?

 

[mr. nice]

bit depth+sample rate= bit rate

bit depth indeed....

 

[HFat]

Just making sure we're on the same page...

I don't hear 16 bits myself (with music which doesn't have crazy dynamics and with my gear anyway) so I wouldn't say it's much more important than anything really.
How many have you been able to hear?

 

[hciman77]

Quote:

Originally Posted by mr. nice /img/forum/go_quote.gif bit depth+sample rate= bit rate bit depth indeed....

I think you mean

bit-depth x sample rate

 

[mr. nice]

i don't hear anything anymore. i got a 16 month old.

seriously, i highly doubt that anything can be heard unless using very high end gear in a treated room and in all seriousness is it all that really big a deal? it's obviously not a concrete definitive "yes it sounds better" answer. some think higher sample rates make the recording sound better and others don't, so in the end we go by math and graphs and theory etc... remember not to forget the music peeps

 

[hciman77]

I downsampled the 24/96 file to 16/44.1. I did an ABX and scored exactly 50%. I ran them both through Audacity and compared the spectra, up to 16K they were within +/- 0.05 db - up to 21360 the files were within +/- 0.14 then they diverged massively to be 17db apart at 21963 (-62db to -79db) , my guess is a problem with the downsampling.

 

[XSAlliN]

If you like Chesky Records then you'd love this: HDtracks high resolution audiophile music downloads

 

[Chippy99]

Quote:

Originally Posted by Joeywhat /img/forum/go_quote.gif What people tend to forget about Nyquist's Theorem just says the highest frequency that can be reproduced...it doesn't mention anything about the lower frequencies. The more times a frequency is sampled, the "better" or more accurate it will be reproduced. So at 44.1 KHz at 20 KHz frequency will only be sampled around two times...but the lower frequencies will be sampled more, and the lowest frequencies will be sampled 100's, if not 1000's of times (I didn't do that math, so I might be off a bit). So how's that relate to higher sample rates? A sample rate of 192 KHz will sample 20 KHz more times then a 44.1 rate, thus producing a more accurate signal.

Unfortunately your argument is flawed. If you take a 1KHz square wave (for example) Fourier shows us that this wave is made up of a 1KHz sine wave plus an infinite series of higher frequency harmonics which when all summed together make up the square wave.

Interestingly, you do need *infinite* bandwidth to produce a *perfect* square wave, since a waveform is perfectly square, then it's amplitude changes infinitely quickly. Therefore you need components of infinite frequency to produce such fast transitions. If you limit the bandwidth to say 20KHz, then the 1KHz cannot be a perfect square - it has little ripples or "ringing" on it.

Now consider my example, but instead of our 1KHz square wave, lets consider a 20KHz square wave. You need harmonics *higher* than 20KHz to try to produce the 20KHz square wave. The trouble is, we (humans) can't hear anything higher than 20KHz, and in fact usually much lower than that.

A 20KHz square wave is in fact a 20KHz sine wave (that with perfect hearing we could just about hear), plus a whole load of harmonics we can't hear. A 20KHz sine wave is just a 20KHz sine wave (obviously). i.e. the two waveforms are *identical* from the point of view of human hearing.

So humans are incapable of telling the different between a 20KHz square wave and a 20KHz sine wave. To do so would need ultrasonic hearing, which we don't have. So it doesn't matter how "lumpy" a 20KHz waveform is. We are unable to discern any of this. Anyone claiming they can *is* mistaken, unless they have the ability to hear frequencies higher than 20KHz, and no-one does!

Incidentally, although all the above is true, the output wave is filtered anyway to smooth out any ripples. So your CD player produces a pretty smooth sinusoidal output at 20KHz anyway.