44.1 vs 96khz music - double blind study conducted...

[zircon]

I saw this in Mix magazine this month. Check it out:

http://mixonline.com/recording/mixing/a ... _sampling/

Here's the VERY brief summary. An extremely thorough, properly-conducted double blind test was done wherein several groups of people (audiophiles, professional audio engineers, average college students) were presented with 96khz audio, and that same audio converted to 44.1khz. The goal was to see if people could hear a difference. The findings of the study indicated that not even the audiophiles or the engineers could get above a 50% success rate (or at most, 1 or 2 points) - meaning they would have done just as well simply flipping a coin.

The authors of the study encourage others to try to reproduce it and get the same results, but if you read the methodology, it was pretty damn foolproof (IMO). Another strike against audiophile BS. 44.1khz is just fine for a final delivery format.

[emdot_ambient]

I'm not surprised at all. The only things I'm hesitant about on this study is that it was only 60 subjects and they didn't release details on the systems they used in the trials.

There are just so many variables in the listening experience: the equipment, the placement of speakers, the room acoustics, individuals' ability to hear across the audio spectrum (something that changes with age and life conditions like diseases and conditions such as tinitis that affect your hearing)...the difference in sampling rates is just so likely to be minor compared to most of all the above.

Now, working on music projects is a bit different because aliasing in virtual synths is less at higher sampling rates. But for the final product? I'm in the "CD quality is probably good enough" camp.

[pough]

Nice. My favourite part:

But there's one more reason worth examining, among whose proponents is Ethan Winer ... who is definitely of the "show-me' school of audio theory and is an outspoken critic of "subjectivism" — that school of thought that encourages people to discuss the performance of audio components and systems using vaguely definable and often irrelevant adjectives instead of hard data.

The "vaguely definable and often irrelevant adjective" I like the least is "transparent". If your audio is visible, you're on acid.

[zircon]

Emdot: Well, those factors could influence things, BUT they used multiple listening environments, multiple sets of equipment, and tested individuals with varying ranges of hearing. As you can see in the article, the people with better hearing actually did worse, consistently. If there really is a difference, and no one could accurately detect it under multiple conditions, with high-end gear, there's no way consumers would on their standard at-home listening setups.

[Funkybot's Evil Twin]

They did a great job on that test. I'm not surprised either though to tell the truth. Converter technology has come far enough that 44.1k sounds a hell of a lot better than 44.1k in the 80s or 48k in the time when ADATs ruled. I'm not worried about higher rates anymore and haven't been for a while. Now that my converter oversamples, most of my plug-ins oversample, and my host uses floating point math when summing, I'm good. Now I just need to keep working on making the room sound better.

[pough]

they didn't release details on the systems they used in the trials.

Yes, they did. It says so in the article.

There are just so many variables in the listening experience: the equipment, the placement of speakers, the room acoustics, individuals' ability to hear across the audio spectrum (something that changes with age and life conditions like diseases and conditions such as tinitis that affect your hearing)...the difference in sampling rates is just so likely to be minor compared to most of all the above.

I read the article and came away with the impression that all of those had been addressed. Different (high quality) equipment, testees stay in the same place during testing, testees have hearing tested, variety of ages and genders.

[runagate]

A-doy?

Though, that being said, isn't the point of faster sample rates that you can turn your latency down so long as you have the processor power? I never thought of it as being "better sounding" (you can only hear up to 22kHz no matter what the mix has in it) but that you get more resolution and less jitter, therefore making it important at the mixing stage, not the listening stage. Then again I'm perfectly happy at 16/44 so I'm not exactly defending the "studio wanker" school of thought.

[zeoy]

and they didn't release details on the systems they used in the trials.

The equipment list included amplifiers from high-end manufacturers like Adcom, Carver, Sim Audio and Stage Accompany, and speakers from Snell and Bag End, as well as the oft-worshipped Quad ESL-989 electrostatics, which are supposed to have usable response up to 23 kHz — which is, of course, above the Nyquist frequency of the HHB recorder's converters. The subjects listened to discs that covered a wide range of material and included classical instrumental, choral, jazz, rock and pop, from audiophile labels like Mobile Fidelity, Telarc and Chesky.

pough said the rest

of course I agree with your last point

[zeoy]

A-doy?

Though, that being said, isn't the point of faster sample rates that you can turn your latency down so long as you have the processor power? I never thought of it as being "better sounding" (you can only hear up to 22kHz no matter what the mix has in it) but that you get more resolution and less jitter, therefore making it important at the mixing stage, not the listening stage. Then again I'm perfectly happy at 16/44 so I'm not exactly defending the "studio wanker" school of thought.

I don't thing that latency was ever the main issue. If your CPU can cope with x ms of latency @96K it most probably will do the same with x/2 @44.1

btw cheers Runa

[Jbravo]

well I always figured that if 96khz became the standard for audio and my setup didn't cater to that, I could always just upsample the finished recordings and noone would ever know and the majority wouldn't even care, so that study just confirms that

[emdot_ambient]

they didn't release details on the systems they used in the trials.

Yes, they did. It says so in the article.

Oh, missed that.

[emdot_ambient]

There are just so many variables in the listening experience: the equipment, the placement of speakers, the room acoustics, individuals' ability to hear across the audio spectrum (something that changes with age and life conditions like diseases and conditions such as tinitis that affect your hearing)...the difference in sampling rates is just so likely to be minor compared to most of all the above.

I read the article and came away with the impression that all of those had been addressed. Different (high quality) equipment, testees stay in the same place during testing, testees have hearing tested, variety of ages and genders.

Well they tried to address those, but with only 60 people in the study, how statistically accurate are their findings?

**note that my gut intuition tells me that if the test had been done on 1 million people they'd probably come up with similar reulst**

However, without seeing the actual report it's difficult to determine what to make of the fact that those with better hearing did more poorly, consistently. For example, were those the audiophiles? Or were they evenly spread across all groups, or were they accidentally all young people who were not audiophiles? 60 people is a very small sampling when you're trying to cover that many variables: age, hearing, musical expertise, gender...

I'm in no way trying to nay-say the results, only wishing the sampling had been larger. I'm sure they agree, that's why they are encouraging other studies be done.

[zircon]

However, without seeing the actual report it's difficult to determine what to make of the fact that those with better hearing did more poorly, consistently. For example, were those the audiophiles?

The article says that the people with better hearing were the younger ones, in the 'student' group for example. So these would be your typical listeners.

[Per Lichtman]

http://www.avsforum.com/avs-vb/showthread.php?t=981557

That forum thread continues the discussion with a lot more being added on both sides. It includes comments by several very experienced engineers, some of whom output multi-channel recordings at 24/96. Some of them mention that DVD-A and SACD have an advantage over the compressed surround tracks used with DTS or Dolby Digital even if the higher samplerates are not an issue. As for myself, my post in that thread (#57) also addresses the fact that even if the study is taken as representative, it does not provide evidence that there is no benefit to working at higher samplerates during production but rather only that the final output can be oversampled without the listener losing anything.

Here is an excerpt from my post there:

Old 04-08-08, 03:03 PM #57 | Link
Per Lichtman
New Member

Join Date: Apr 2008
Posts: 1

I am surprised that there hasn't been more discussion about the impact of different samplerates/bitdepths prior to the delivery stage.

Working under the premise that the original study cited is accurate about the end delivery format of Redbook audio being indistinguishable to listeners from 24/96 PCM or other high quality audio formats, that does not mean that it irrelevant at other stages of the production process, whether tracking or mixing. Note, I am not saying that they do make a difference either, just that it is a different question. Fair enough?

Let's imagine for a moment that using the same, high quality A/D converter at 48 kHz and 96 kHz that there is a difference in the digital files created. Let us imagine that the difference is too minute for us to perceive under normal circumstances. What about when that mono or stereo signal is mixed with the other tracks recorded at the same time? Do the these minute differences start to add up as the signals are combined/summed?

As an analog, it is sometimes said the average listener will not consciously perceive less than 3db drop in the volume of sine wave. However, as any mixing engineer can tell you, dropping one or more signals by 2db can have a significant effect on how a mix is perceived or how they relate to each other, and we don't go around arguing that boosting an EQ band by 2db is the same as leaving it at unity.

If the analog applies here (I have not done sufficient empirical testing to confirm or disprove that at this point) the higher sample rates or bitrates could be useful during production even if they were not required in the final delivery. That is before even taking into account digital processing.

A lot of mixing and mastering engineers use some form of digital processing. Not all digital processing applies the same rate (if any) of oversampling. Now that is an area I have done a lot of testing in and I came to the conclusion (as did Massenberg) that there is can be a clear difference when working at higher sample rates for digital signal processing such as EQ and compression. This is comparitively easy to demonstrate because the output files can both be included on the same CD and listened to sequentially.

It seems, as has been discussed here earlier, that one way to test higher samplerate and bit depth differences in the final output that would rule out differences at other stages would be to create the file at 24/96 or 24/192 and then resample that file to Redbook specifications. Of course there are arguments about different resamplers being superior or inferior to each other so maybe it would have to be done multiple times. Thoughts?

[zircon]

Per, I'm not sure I understand what you're saying.

It seems, as has been discussed here earlier, that one way to test higher samplerate and bit depth differences in the final output that would rule out differences at other stages would be to create the file at 24/96 or 24/192 and then resample that file to Redbook specifications.

That is what they did here. The music was at 96khz to begin with and it was resampled to Redbook specifications, eg. 44.1khz.

it does not provide evidence that there is no benefit to working at higher samplerates during production but rather only that the final output can be oversampled without the listener losing anything.

How is any oversampling being done? Again, the source material is 96khz. They are DOWNsampling it to 44.1khz to see if anyone could tell the difference, and no one was reliably able to.

Maybe I'm just not understanding the terminology here.