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

[kmonkey]

Not to mention the fact that most people today listen to most of their music on mp3 players. That 96kHz mix is gonna sound SO awesome when it's compressed to mp3! And when they're not listening to mp3s on crappy earbud headphones, they're listening to either FM radio or CDs in their car. That probably accounts for well over 80% of what people hear today.

But take in account that there is a great chance that 96 khz project created and processed ITB from start, and then converted to 44khz, or mp3 will sound better then same one but started again from scratch in 44khz then converted to mp3. Of course depend on VSTi used, effects used etc.

If you've read the whole thread you'd see that I did say there were reasons in production to work at higher rates. I don't doubt that. But for the vast majority of songs, I still doubt that even you would be able to tell the difference between one recorded at 96 and converted to 44 as opposed to one done at 44 all the way...especially if it was compressed to mp3.

I don't have a problem with people being audiophiles. I have a friend who's totally into the whole monster cable thing and high priced turntables and who eschews any digital audio devices as the spawn of Satan. But I still believe that most of it's all in your head.

Then again, recent studies on how the price of wine is related to people's enjoyment of the wine show that even though the effect is completely mental, the actual fact is that most people really did enjoy expensive wine more. I mean as in the pleasure centers of their brains were more active when they had paid more for the product. So maybe if you believe you can hear the difference between 96 and 44 then the pleasure centers of your brain are actually firing off more even if it's due to the placebo effect.

We ponder.

Nice points. I like that about wine! But there is not any placebo in my example. I am not experiencing any placebo. I am experiencing aliasing. That is most obvious example of working 96 vs 48. Search somewhere on KVR for arkecode VST oversampler thread. If aliasing is placebo and if you can not hear obvious difference with that oversampler i can not convince you even more so just skip my opinion. I am going to sleep now... Great day at KVR

edit: i did it for your pleasure:

http://www.kvraudio.com/forum/viewtopic ... highlight=

[Melkor]

44.1 forever ......

To be honest, I'm not even convinced that 24bit even helps, especially in a 32bit float mixer .... but i still run 24/44.1 because it is now EFFORTLESS to do so ...... and the audio quality is fantastic.

Seriously, sometimes my partner will record vocal takes in the studio and play them back solo'd one at a time and, if im in the other room, i cannot tell if its her being played back through the monitors or her ACTUALLY singing live.

Digital is very convincing when it comes to fooling the ears.

[Sascha Franck]

To be honest, I'm not even convinced that 24bit even helps, especially in a 32bit float mixer ....

You're mixing up things here quite a bit.
One number (24bits) is about recording bit depth while the other is about processing "depth" (or whatever it might be called).
In any case, the advances of using 24bit for recordings are absolutely obvious: It's giving you more dynamic bandwith, hence more headroon, hence it's way less likely to run into digital clippings while still maintaining the best dynamics. With 16bit, offering a total dynamic range of 96dB, you will often stay below the dynamic bandwith proper converters offer, especially once you leave some headroom (which is a must when recording digitally). Using 24bit, offering a dynamic range of 144dB you will almost always have greater dynamics than even the best converters, so leaving quite some headroom is easily possible.
Really, 16 vs. 24 bit is a no-brainer in case you're recording any acoustic signals. Unless you're computer can't handle it, but all remotely modern machines won't have the slightest problems with it.

[yairhol]

May I suggest that you guys run a blind test here on the forum?
One person will make 10 (or more) tracks at 44.1 KHz and 10 (or more) identical tracks but at 96 KHz.
Post these tracks here without labeling which track is at which sample rate and have a few participants guess the sample rate based on quality of sound. The more tracks we have the better our statistics will be.

[whyterabbyt]

May I suggest that you guys run a blind test here on the forum?
One person will make 10 (or more) tracks at 44.1 KHz and 10 (or more) identical tracks but at 96 KHz.
Post these tracks here without labeling which track is at which sample rate and have a few participants guess the sample rate based on quality of sound. The more tracks we have the better our statistics will be.

and how would that prevent people examining the tracks first, thus removing the 'blind' from 'blind test'

[KKeene]

Does anybody know whether such a phenomen happening in the inaudible range does have an effect for the audible range?

Certainly inaudible frequencys above 20kz can be processed and result in frequencys bellow 20khz. Ring modulation creates sum and difference frequencys, so ring modding 5khz, with 21khz, would result in 16khz, and 26khz, the former being audible.

The same can happen with a non linear process like distorition, IE intermodulation distortion.

That said in my experience it doesnt usualy sound good, it usualy sound effing awful.

Here's my argument. I'm familiar with the ring modulation process and it does occur at frequencies higher than audible to produce an audible beat note. It often occurs with high frequency laden sounds such as cymbals and hihats and is regarded as noise. And yes, it can sound awful. But if the ultrasonic information is not recorded or present, the listener doesn't know anything is missing. The argument is that this information and effect can add a sense of 'realism' to the sound if it's captured properly, and I tend to believe that this is the case to a limited extent. For instance, the chiff of a flute or the ambient sound of crickets. The alternate argument is that if it's captured badly, either through the use of a lower sample rate which would poorly represent the information, making it sound crappy and poorly captured, or through the use of higher sample rates with bad or neglectful engineering (in other words, it would be easier to screw up a mix using a higher sampling rate).

I suppose it depends on what the producers are aiming for and how well they can pull it off. But I do see a potential for the less experienced to screw up more with higher sampling rates.

Did any of that make a bit of sense to anyone?

[Kingston]

Great that someone finally tested the delivery medium myth. The one's who originally decided on the rate certainly knew nyquist theorem.

Bad news is that we're actually gone downhill since CD. It's all ~192kbps encoded these days.

The benefits come during the processing at 96000hz

thats why just converting the already processed and mixed song wont make much of a discernable different.

Id like to see a test that double blind tested EQ and comp settings done on a piece of audio, processing at 44 and 96...thats when you would hear the difference because its during the processing the aliasing would become apparent.

thats why a lot of developers over sample there plugins as it moves the aliasing out of the hearable spectrum of sound.

THE above results are most likely true and completely valid...for a consumer looking at an end product purchase....

..but for an engineer it CAN be benefitial to process or synthesise when done digitally at doublerate samplerates.

(i suggest you check this out for yourself on a simple one track with either a synth or an eq curve onto an audio track....export both and use DBX)

bingo!

Of course down-sampling post production will make practically no difference: 44.1Khz reconstructs the audible range completely (which I don't understand the reason for, but despite my tendency not to take authority's word for it, all scholarly sources seem to say so).

However, the results from, say, bit-crushing, distortion, synthesis, et cetera, carry quite a vast difference from 44.1 to 96.

I've done this quite the many times, even posted the results here. The difference can be heard even through +192kbps MP3 conversion, and it was verified here as well.

I always work at 88.2khz because of this.

Any kind of distortion greatly benefits higher rates. EQ's can be a bit so and so. Any decent quality compressor's attack phase will be beneficially affected as well.

Synths.. Let's not even go there.

Id like to see a test that double blind tested EQ and comp settings done on a piece of audio, processing at 44 and 96...thats when you would hear the difference because its during the processing the aliasing would become apparent.

thats why a lot of developers over sample there plugins as it moves the aliasing out of the hearable spectrum of sound.

Yes but that should be done by the plugin designer, it should be part of the design process of the algorythm. We shoudnt need DAW users to run their whole projects at high rates to compensate for shitty dsp coding.

It's extremely CPU intensive to get great quality oversampling into plugins. Might as well skip the interpolation/decimation (which themselves generally only have adverse effects on audio) and work at natively at higher rate. That's what the quad core's are for.

[zircon]

I tried taking one of my massive progressive breaks songs and rendering it at 96khz. The project already maxed out my CPU (e6600 2.4ghz dual core) at 44.1 so at 96 it was unplayable even with a slightly larger buffer. After waiting about 15 minutes for it to render, I listened to it, and A/Bed to the 44.1 render. I'll be honest - outside of greater high end clarity in the 96khz mix, which just sounded like a 3db boost at 16khz or so, I didn't hear any substantial differences. I'd rather work at 44.1 so my projects actually play back in real time.

[DWb]

May I suggest that you guys run a blind test here on the forum?
One person will make 10 (or more) tracks at 44.1 KHz and 10 (or more) identical tracks but at 96 KHz.
Post these tracks here without labeling which track is at which sample rate and have a few participants guess the sample rate based on quality of sound. The more tracks we have the better our statistics will be.

and how would that prevent people examining the tracks first, thus removing the 'blind' from 'blind test'

You could upsample them again?

Are we talking about mixing / rendering or delivery format, here?

[pough]

Are we talking about mixing / rendering or delivery format, here?

My guess is we're muddying the waters and arguing the crossflow. IMO, there's not much point in doing any kind of comparison for ourselves unless we take it to the same level as the initial test. And why bother? They already did it.

The only way to get close in the random, internet world would be to make a custom app that blindly downloads and plays three files (A/B/X), all of which might be the same (just to make things really blind). It could be done, but the files in question would be relatively big so patience is something the testers would need.

Hmm... come to think of it that sounds kinda interesting...

[tony tony chopper]

Post these tracks here without labeling which track is at which sample rate and have a few participants guess the sample rate based on quality of sound. The more tracks we have the better our statistics will be.

I suggest: you take a bit of (effective) 24bit, 96khz audio, then you repeat it in the same, single 24bit, 96khz audio file, in 16bit, 24bit, 44khz & 96khz at random.

AND you'll have to test this properly (and I feel this is gonna go wrong) in a proper playback device. Because if you test something like this in an average player like Winamp at its default 44khz, it will play it back at 44khz with a low-quality resampling, thus the freqs above 22khz will alias & become audible, thus you WILL hear differences, but for bad reasons.

[hseiken]

Audiophiles are idiots. They're more worried about the 'clearity' and 'reproduction' of a sound...why can't they just listen to the music? A good song coming out of crappy speakers is still a good song, I don't care who you are.

[Per Lichtman]

I am seriously considering organizing a follow-up test (or tests) to the one that started this thread to try and address some of the controversy of the initial one. I've already talked with other members of my company about the possibility and the reaction has been positive so far. It is dependent upon whether I can create a large enough hole in my mixing schedule this year but if you are interested, please send me a private message to find out more as it develops.

Unfortunately, online tests would be inconclusive since people listen in different places with different equipment and there are too many variables involved. But, it could be helpful for individuals to determine their own experiences with their own equipment, as long as it is recognized that experiences might vary as any part of that equation did.

Note: I would want people that both believe and those that don't believe that there is a difference between 44.1/48 kHz and 88.2/96kHz for final delivery. The primary qualification for participation wouldl be willing to communicate in a consistently constructive manner so that everyone that puts a lot of time and effort into the project wouldn't get worn out by fighting.

[Melkor]

Yeah Sascha ..... Im aware of the difference ....... and i still state that not only is 44.1 enough to capture all frequency content, but also that 16 bit is enough to capture all amplitude content.

Acoustic instruments sound fantastic on CD's ( I dont care what was used to record it, its still a 16/44.1 wav at the end of the day)

[Sascha Franck]

Acoustic instruments sound fantastic on CD's ( I dont care what was used to record it, its still a 16/44.1 wav at the end of the day)

That's not what things are all about. It's about additional headroom while recording. And I can promise you that ALL people recording acoustic instruments digitally are using 24bit these days, even if they are 100% that their work will only be 16bit in the end.
When you mix that stuff down, you have a lot more time and tools in your hands to carefully adjust things so they fit in a 16bit dynamic range. When you record, you often don't.

I am by no means an audiophile at all. But the benefits of using 24bit over 16bit are massive, even for the mere homestudio owner.