Sampling Frequency: a scientific study?

[Ptron]

xxkHz vs xxkHz
So when is somebody going to do a true scientific study on this? (or somthing close to it at least) I won't believe either way on this until I see the results of one. Human perception is HIGHLY suggestable. If you played two recordings for someone that were exactly the same but they (and you(double blind))thought one was recorded at 48kHz and the other at 96kHz, they would probably sit there and tell you how they could hear the difference between this aspect and that aspect, ESPECIALLY if they considered themselves to have a good ear. That may sound a little cynical but it's human nature really. I just learned about this kind of stuff in my Intro to Psych course so I'm an expert now

If I were going to stage a test, I would start with 3 or 4 short clips each of a different style of music, recorded in super HiFi analog...maybe 20 seconds long. Then, I would pick two sampling freqs like 48 and 96 and then, all other things being equal, record each clip in both. Then play the recordings for people in pairs of the same music BUT sometimes they would both be 48kHz, sometimes both 96, and sometimes one of each. They would then say what, if any difference they heard between the clips. The test would have to be double blind. You could do that by administering it by computer, which could randomly choose which pairs of clips to play. You might also want to compare results from the average Jack and Jill and the experienced ears.

I hope somebody does someting like this, someday. I'd really like to see the results!

Ptron

 

[alfalfa]

I do not wish to debate the merits or flaws of ABX testing.

Regarding audio theory to back up claims of audible benefits of higher sample rates, I highly recommend you read the paper published by Meridian. This is a summary of the paper I previously wrote:


The Meridian paper (Coding High Quality Digital Audio) is a more complete version of an article published in Audio Magazine which in turn was a subset of the AES paper "Coding Methods For High-Resolution Recording Systems".

I will summarise and paraphase the contents of the Meridian paper with regards to sampling rates. Their analysis calculates the noise spectrum that has the effect of the human hearing threshold. The auditory threshold is then plotted on a dB vs linear frequency 'Shannon plot'. Using Shannons theory and Gerzon-Craven criterion for noise shaping, they determine that the absolute minimum desirable sampling frequency required to replicate the information received by the ear is 52kHz. Hence the minimum rectangular channel to ensure absolute transparency would require a 58kHz sampling rate.

The article concludes that a minimum sample rate of 66.15kHz would be sensible (to provide sufficient margin to move phase, ripple and transition regions further away from the human audibility cutoff)

The full paper can be found here (see page 8-11 for discussion of sampling rates):
http://www.meridian-audio.com/w_paper/Coding2.PDF

 

[Ptron]

I do not wish to debate the merits or flaws of ABX testing.

Neither do I, whatever that is.

Their analysis calculates the noise spectrum that has the effect of the human hearing threshold. The auditory threshold is then plotted on a dB vs linear frequency 'Shannon plot'. Using Shannons theory and Gerzon-Craven criterion for noise shaping, they determine that the absolute minimum desirable sampling frequency required to replicate the information received by the ear is 52kHz. Hence the minimum rectangular channel to ensure absolute transparency would require a 58kHz sampling rate.

The article concludes that a minimum sample rate of 66.15kHz would be sensible (to provide sufficient margin to move phase, ripple and transition regions further away from the human audibility cutoff)

Um...what?

Seriously though, thanks for the link. I will read the parts of the paper that aren't in Martian and see what it has to say.

Ptron

 

[TexRoadkill]

Neither do I, whatever that is.

Um...what?

Seriously though, thanks for the link. I will read the parts of the paper that aren't in Martian and see what it has to say.

Ptron

If you don't understand the answer why ask the question? lol

George Massenburg was also involved in a study and they found that when people listened to recordings at 44.1Khz the brain produced the same type of waves that it does when a person is angry or confused. Apparently your brain does miss the subtle harmonics that are missing from the high frequency spectrum.

 

[drstawl]

the absolute minimum desirable sampling frequency required to replicate the information received by the ear is 52kHz.

"The ear" is quite a vague term given the wide variation in the ability of people to hear frequencies at the upper limits of optimal human hearing.

 

[Doug H]

If you don't understand the answer why ask the question? lol

George Massenburg was also involved in a study and they found that when people listened to recordings at 44.1Khz the brain produced the same type of waves that it does when a person is angry or confused. Apparently your brain does miss the subtle harmonics that are missing from the high frequency spectrum.

Why do I find that very easy to beleive. I love the quality that us home rec'ers get, but the CD was the death of music appreciation. I've f@cking hated them sense the day they came out well, since after I got over the thrill that they don't scratch easily.

It's not digital versus analog, it's the sampling resolution in consumer digital versus analog.

 

[Ed Dixon]

xxkHz vs xxkHz
So when is somebody going to do a true scientific study on this?
Ptron

Most of this was resolved some 50 years ago with Nyquist sampling theory. The original Nyquist paper was published in 1928. There have been scads of studies and works by thousands of others since that time.

Ed

 

[ebeam]

Why do I find that very easy to beleive. I love the quality that us home rec'ers get, but the CD was the death of music appreciation. I've f@cking hated them sense the day they came out well, since after I got over the thrill that they don't scratch easily.

It's not digital versus analog, it's the sampling resolution in consumer digital versus analog.

Well I doubt the average consumer loudspeaker can even reproduce frequencies much higher than 20kHz, especially 15-20 years ago. The death of music appreciation IMO is a direct result of the death of good music - at least in the realm of popular music.