Why would 24 bit / 192 khz flac sound any better than 16 bit / 44.1 khz flac if both are lossless (if at all)?

[Ruben123]

Idk about those imd etc but you sure need a microphone that records those ultrasounds and well many microphones don't actually do that AFAIK

 

[spruce music]

Idk about those imd etc but you sure need a microphone that records those ultrasounds and well many microphones don't actually do that AFAIK

Actually most condenser mics do.  They may have drooping response.  I have several that are only spec'd to 20 khz, but have substantial response to 30 khz or more before dying way off.
 
You can do the old jangling key test.
 
Get two or three metal keys on a keyring.  Jangle them in front of the mic and record at a high sample rate.  They don't sound very loud, but most of the output is ultrasonic.  You get something to at least the high 30 khz or circa 40 khz range that way. 

 

[sonitus mirus]

Idk about those imd etc but you sure need a microphone that records those ultrasounds and well many microphones don't actually do that AFAIK

 
I think there is more to it than simply having a capable microphone.  It would require the instrument to produce ultrasounds, a microphone to record them, a media format capable of playing back these ultrasounds, transducers that could reproduce them with some kind of accuracy, and ears that could benefit from all of this.  If anything in this chain is a limiting factor, we are most likely left with a typical 20-20kHz frequency range.  If the recorded instrument is creating IMD that impacts the audible range, it is already being captured with a typical recording microphone.  The only thing to consider is if the actual ultrasounds are making any kind of difference for the listener, because any audible influence should already be present in the recording.  
 
I would think that system/equipment generated noise would be more likely to occur on a 24/192 audio file than any sounds created from instruments with regards to ultrasounds.  I mean, Led Zeppelin probably wasn't using microphones that recorded ultrasounds, so any HD files probably only include recording equipment noise above 20kHz and not crash cymbal sounds from Bonham.

 

[old tech]

  The post I responded to suggested that no musical instrument produces frequencies even near 20 kHz.
That is simple not true.
 
Frequencies above the upper treshold of our hearing we cannot hear by definition.
But content above this threshold can generate IMD that maps into out audible range.

Yes there are instruments that produce sound frequencies above 20khz, just as there are many consumer electronics and natural phenomena producing such frequencies around us all the time.
 
But does it affect the frequencies within the audible range? Possibly yes but probably, maybe not.
 
I can understand how supersonic or subsonic recorded content can introduce IMD, through interaction with system electronics and speakers.  But, using the dog whistle example, is this relevant to the live acoustic event?  If someone blow a dog whistle next to you while you are chatting with someone else, does it affect the tonality of the conversation or the background noise?

 

[Roseval]

But does it affect the frequencies within the audible range? Possibly yes but probably, maybe not.

Good point.
Indeed IMD does exist but is it loud enough to become audible?
I have my doubts.
Scientific evidence is rare but this survey by Pras indicates that people are able to discriminate between 44 and 88. This is only the case with an orchestral recording.
Maybe because of the IMD?
 
http://www.academia.edu/441305/Sampling_Rate_Discrimination_44.1_KHz_Vs._88.2_KHz
(scroll down a little)

 

[gregorio]

 
[1] Also there are a lot of energy sources but how much are them mechanic? Due to natural selection, our body is very capable of perceiving mechanical energy only if it has enough power. Unlike electromagnetic energy, we really do not need any specific receptors to perceive mechanical energy because almost our entire body is somehow capable of interpreting mechanical stress (if they have enough power).
[2] Beyond this point, scientific works needed.
[3] Considering low levels of ultrasound found in music files due to cropping, bad(?) mastering(?), (I think) It would be impossible to perceive them even indirect form.
[4] If you take hand skin (where energy reflects to epidermal hand cells), yes it needs big magnitude. But ear is very complex organ and there are different type of cells at where stereocilia located. Maybe these cells need less power than hand cells to notice it to brain via neurons?
[5] But I think it like this: When you listen extreme bassy musics, you think like your brain is shaking (maybe not correct term but you got the point). That "shaking feeling" is not audio but it is a intended thing to experience while listening music. Because music makers got that feeling too.

 
1. What do you mean by "mechanic" energy as being different to electromagnetic energy? Mechanic energy IS electromagnetic energy (if we're talking about sound waves).
2. Why? There's no point as far as you are concerned, if you're going to ignore the science (which has already been done) and invent your own theories. And, there's no point as far as science is concerned because science has already done the work!
3. No, it's not due to mastering (good or bad), it's due to the fact that's there's little there in the first place.
4. Science has already done this work, in some cases many decades ago! Science knows the physiological structure of the ear and even what amounts of energy are needed to cause a response in those structures at various frequency ranges.
5. You're free to think whatever you like but this is the science forum and science exists specifically to separate what people may think from the actual facts.
 

Originally Posted by Asuhra /img/forum/go_quote.gif
 
[1] At 48Khz you have much more headroom to put info between the 15.5 and 20Khz Freq range when doing subliminals (it's a mathematical thing).
[2] It has been proven by dog trainers that you can hear at least up to 30Khz because they got affected by the subliminals meant for their dogs @ 30Khz range. [2a] I have proven that on myself time and time again.
[3] As for the resolution, that has also been proven that some people notice differences up till 96Khz.

 
1. A "mathematical thing"? Don't you mean an anti-mathematical thing? The math/science actually demonstrates the exact opposite of what you're stating!
2. Maybe you're confused? This is the science forum, NOT the dog trainers forum.
2a. There's only two options: 1. You've proven that you are not a homo sapien or 2. You've somehow managed to convince yourself but haven't proven anything. If you're going to dispute the accepted science you're going to need exceptional actual proof, not just your personal conviction.
3. Again, science/math proves there is no more resolution, only the same resolution in a wider band of frequencies. Please provide a reference to the proof that people can tell any resolution difference between 96kHz and 44.1kHz with music recordings.
 
Quote:

  [1] Crash cymbals can go far in excess of 20 kHz ...
[2] Actually most condenser mics do.  They may have drooping response.

 
1. Yes but how much ultrasonic energy depends on where we record them, both in terms of the acoustic properties of the room and the distance the mic is from the cymbal. Concert halls, for example, are designed to absorb high frequencies and boost lower frequencies. Additionally, more high frequency energy is lost (absorbed by air molecules) than lower frequency energy, so the greater the distance from the sound source, the less high freq content is present. These two factors are cumulative.
2. Yes but as you say, their response is not flat, they roll off significantly beyond 20kHz (except of course those condensers specifically designed to have a response higher than 20kHz). Additionally of course we have the common use of dynamic (rather than condenser) mics, which roll off earlier than 20kHz.
 

  But does it affect the frequencies within the audible range? Possibly yes but probably, maybe not.
 
If the recorded instrument is creating IMD that impacts the audible range, it is already being captured with a typical recording microphone. The only thing to consider is if the actual ultrasounds are making any kind of difference for the listener, because any audible influence should already be present in the recording.

 
A. For the result of a frequency modulation to be audible, the original frequencies must also fall within the audible band. This is simple enough to test for yourself. With a 96kHz sample rate, use a tone generator to compare say a 1kHz sine wave and a 1kHz square wave, the tonal difference should be obvious. Try the same again but at 12kHz and now you can't hear the 12k square wave, it just sounds like two 12kHz sine waves! This is because an actual square wave cannot exist as an acoustic sound wave. A square wave has to be approximated by frequency modulating sine waves (a fundamental + odd-harmonics) but at 12kHz those odd-harmonics are beyond the audible frequency spectrum and therefore all you hear is the 12kHz fundamental (sine wave).
 
B. This isn't the case with IMD however, which only occurs within electronic systems (rather than with acoustic instruments), when over-driven circuits can generate tones. The energy which caused the circuit to generate the IMD does not have to be within the hearing band but to be audible, the tones generated obviously have to fall within the hearing band (which, as you say, would then be recorded by a mic anyway). If that IMD product is going to modulate with the existing frequency content then to be audible it must fall within the audible band, as per point "A".
 
G

 

[gregorio]

  Scientific evidence is rare but this survey by Pras indicates that people are able to discriminate between 44 and 88. This is only the case with an orchestral recording.
Maybe because of the IMD? http://www.academia.edu/441305/Sampling_Rate_Discrimination_44.1_KHz_Vs._88.2_KHz
(scroll down a little)

 
I haven't had time to go through it in fine detail but there are a few things I noticed:
 
1. Only 3 of the 16 participants demonstrated any ability to differentiate with a significance better than chance/guessing.
2. If I read correctly, although there was significance in the results of those 3, they were actually wrong. They thought the 44.1kHz recording was the higher res version!
3. To be sure one is actually testing what one is aiming to test, all variables other than the variable being tested need to be eliminated. This is considerably more difficult that it usually appears! In this study, the ADC's internal clock was used for the 44.1kHz recording while an external clock was used for the 88.1kHz recording. I'm not sure why they did this, external clocking usually adds jitter to the system. This could explain why the 44.1kHz recording was thought by the 3 to be the higher res version, although with most modern, high quality ADCs the level of additional jitter from external clocking should not normally be audible. This is still a variable which should have been eliminated though, as was using such different recorders.
4. I didn't see where they explained their methodology for down-sampling. Certainly it was fairly common a few years ago that re-sampling software would reduce amplitude of the signal (to eliminate illegal intersample peaks at the high sample rates used for conversion). Generally the reduction was only about 0.2dB or so and most would not detect any difference, a trained listener, with certain types of audio material and certain listening environments might be able to. Volume matching is another variable which is more difficult in practice to eliminate than it may seem to be.
5. Yes, I agree, IMD is another potential variable which I'm not sure was eliminated.
 
G

 

[fuzun]

Who is ignoring science? There is no thing as ignoring science. If you don't believe gravity there is no gravity? The statements I have posted are my opinions (some of them are incorrect) and questions that I posted to wide the discussion.
 
Can you post scientific materials or your opinions instead of arguing?
 
"You're free to think whatever you like but this is the science forum and science exists specifically to separate what people may think from the actual facts." Whats the point of writing this? It was written to give an example in a indirect way.

 

[sonitus mirus]

   
I haven't had time to go through it in fine detail but there are a few things I noticed:
 
1. Only 3 of the 16 participants demonstrated any ability to differentiate with a significance better than chance/guessing.
2. If I read correctly, although there was significance in the results of those 3, they were actually wrong. They thought the 44.1kHz recording was the higher res version!
3. To be sure one is actually testing what one is aiming to test, all variables other than the variable being tested need to be eliminated. This is considerably more difficult that it usually appears! In this study, the ADC's internal clock was used for the 44.1kHz recording while an external clock was used for the 88.1kHz recording. I'm not sure why they did this, external clocking usually adds jitter to the system. This could explain why the 44.1kHz recording was thought by the 3 to be the higher res version, although with most modern, high quality ADCs the level of additional jitter from external clocking should not normally be audible. This is still a variable which should have been eliminated though, as was using such different recorders.
4. I didn't see where they explained their methodology for down-sampling. Certainly it was fairly common a few years ago that re-sampling software would reduce amplitude of the signal (to eliminate illegal intersample peaks at the high sample rates used for conversion). Generally the reduction was only about 0.2dB or so and most would not detect any difference, a trained listener, with certain types of audio material and certain listening environments might be able to. Volume matching is another variable which is more difficult in practice to eliminate than it may seem to be.
5. Yes, I agree, IMD is another potential variable which I'm not sure was eliminated.
 
G

 
What I find most strange, provided I'm reading the data correctly, is that no difference was shown between 88.2kHz downsampled to 44.1kHz.  No difference was found between the downsampled 44.1kHz and the native 44.1kHz files.  But somehow a few individuals consistently managed to select the native 44.1kHz as the better sounding version to the 88.2kHz, or they heard a difference, statistically speaking, but were clearly unsure which file was the 88.2kHz.
 
88.2kHz = file A
88.2kHz downsampled to 44.1kHz = file B
Native 44.1kHz = file C
 
No difference found between files A and B or between files B and C, but there was a statistical difference found between files A and C. (with file C being chosen as the HD version more often)
 
I don't get it.  Seems like a bad test or something wrong with the process somewhere. 
 
Edit:  After typing it out, I suppose one could assume that perhaps file A is too close to the same as file B, and file B is too close to the same as file C, but there is somehow just enough of a difference to be identified between file A and C.  The results still seems a bit odd.

 

[gregorio]

  ... provided I'm reading the data correctly ...

 
I've used this caveat as well and I presume for the same reason, I found the paper quite confusing. For example: "This finding provides support for theories that high-resolution formats better reproduce the details of transients and room acoustics" - Hang on a minute, as the statistically significant results were actually wrong, doesn't that suggest the exact opposite, that 44.1 better reproduces the details? Then there's the apparently illogicality of the results which you highlighted, which could conceivably just be a quirk/fluke of statistics, although by definition, probably not! Then there's the weaknesses in methodology, the failure to eliminate other potential variables, some of which I mentioned. And additionally, some fairly obvious factual errors in the introduction.
 

  Who is ignoring science? There is no thing as ignoring science. If you don't believe gravity there is no gravity?

 
Of course one can ignore the science. If you don't believe in gravity, there is of course still gravity, however one could attribute the effects of gravity to say a magic spell cast by a witch, rather than to the scientific law of gravity. Theorising about the witch and then posing questions about those theories IS ignoring the science!
 

 
Can you post scientific materials or your opinions instead of arguing?

 
There are no scientific materials regarding the witch and my opinion is that she doesn't exist, so no, apart from just not responding, I see no alternative to disputing/arguing. Let's get away from silly metaphors and take one of your actual examples, "And we would like to listen music how it was intended to sound. (Mastering, ... not involved. This is only in theory.)" - That's a contradiction which makes no sense. The whole point of mastering is to affect/change the studio mix so that consumers can "listen to the music how it was intended to sound". Without mastering, the only way of achieving that aim would be to listen to the music in the studio in which it was mixed, which obviously is not practical and why mastering was invented in the first place!
 
G

 

[akg fanboy]

Hey everyone, I don't know if this is the right thread for this, but I do think higher quality (16.flac/24.flac) music do have a purpose for normal people who listen to their music for recreational purposes, even though most people cannot hear the difference through normal means.
I think higher quality music works better with equalization and has less distortion, at least according to the distortion my ears heard. My 256/320 mp3s generally exhibit more noise and distortion than my lossless 16 bit flac music which sound much cleaner. Basically, I normally cannot tell the difference very well between my 320 mp3s and flacs unless I EQ them. Is this just a coincidence with better recording/mastering or can my findings actually be confirmed to be true?

 

[castleofargh]

it could depend on the EQ you're applying. admittedly if you end up with some freqs real close to 0db or even slightly clipped, then mp3 could have some/more clipping than the flac file. and of course that can be more audible. that due to intersample clipping. if the signal is consistently at a good -3db, then you would avoid that problem IMO.
 
some nice peak meter could show that(and they will by oversampling, so that the peaks are closer to the real peaks). if you can get your problem/difference while no peak meter can show potential clipping, then I'd be curious to see a short sample of the 2 versions.
 
otherwise, usually DSPs that really benefit from oversampling will just do so and then downsample back to whatever you originally used on the fly.

 

[Warmuser]

Hi everybody, can you explain me If this album is a true fake? The frequences looks like the .mp3 type. So I want to understand if these album is a fake or I'm wrong. Thanks
 
Here I made the screens of any track: http://imgbox.com/g/0AnaQljpwf

 

[akg fanboy]

it could depend on the EQ you're applying. admittedly if you end up with some freqs real close to 0db or even slightly clipped, then mp3 could have some/more clipping than the flac file. and of course that can be more audible. that due to intersample clipping. if the signal is consistently at a good -3db, then you would avoid that problem IMO.

some nice peak meter could show that(and they will by oversampling, so that the peaks are closer to the real peaks). if you can get your problem/difference while no peak meter can show potential clipping, then I'd be curious to see a short sample of the 2 versions.

otherwise, usually DSPs that really benefit from oversampling will just do so and then downsample back to whatever you originally used on the fly.

So you are saying the frequencies should be at 0dB on the frequency response chart? My music player does show frequencies in real time.

Since DSPs benefit from oversampling, does that mean a 24/192khz file that is already "oversampled" compared to a 16/44 file work better for the DSP? I feel like a true 24 bit music file would have more headroom for equalization with less quality degredation but I would have to find one

 

[castleofargh]

I mean 0db as in digital value, the maximum loudness you can record on a digital PCM file. the music is made of sample points with values all below 0db. a 16bit file goes from 0 to -96db, and a 24bit file goes from 0 to -144db. 0 is always the upper limit you can record on the digital support. and when you reach 0 or try to get over it, a digital system can only register up to 0, so it replaces everything above 0 with 0. that's clipping.
because the samples registered won't always fall exactly at the peak of the signal, you can end up with properly recorded digital values(below 0db), but actual peaks that get above it. look up "intersample clipping" online, it's really not a new problem, but one that can be solved soooo easily.
 
 
 if something works better at a higher sample rate, then as I said above, it will just convert the file on the fly to that sample rate. do you know what sample rate that is for each DSP you own? I don't, so I don't get how I could improve anything by oversampling in advance.
 
whatever you try, check some peak meter(hopefully one that does oversample a good deal to show intersample peaks at a reliable value), and verify first that you're not clipping anything with or without your EQ on the songs that feel wrong to you. if that is really out of the way, then we can look for other reasons why you get what you describe.