Can the human ear tell between 48k and 96k recordings?

  • Thread starter Thread starter T800
  • Start date Start date
T

T800

New member
Since I am still stuck on an old laptop for recording, i was recommended to record at 48k because 96 take up too much memory and that the ear cant tell the difference either way.

But then I was told by other sources that classical music is recorded at 88 and 96. But if the human ear can't ear these frequencies then why is classical music recorded at such high sample rates over other types of music?

Is there any sonic content that can be sensed over 48k?
 
In the Avid forums, somebody said "A good EQ plug will sound very noticeably better when printed at 96 kHz than at 44.1 or 48, even after the file is converted to the lower rate."

Why is this and how is this possible in non-hearable ranges?
 
it would be interesting if you recorded some samples in all the freq's and post them.

or play them yourself for people and see if anyone can tell.
 
My 2 cents? I've never listened to an album and thought to myself "This sounds like shit. I can tell it was recorded at 48k." I think there may be an advantage in the mastering stage if you send the ME 98k files. Something about how it allows him more headroom for processing, effects and such. For the hobby home recording guys I really wouldn't worry too much.

I have experienced plugins that change character a good deal when processing at different frequencies though. My advice would be record at the highest bit rate your gear will allow anyway. My ears may not be as sharp as others and they may hear things I don't.
 
Last edited:
If there is a difference, it would be extremely subtle.

So subtle, in fact, that every other part of the signal chain would have to be first class for this difference to be appreciable. Otherwise, the inherent failings of the recording equipment would overwhelm any quality benefits.

Don't sweat on it. I (and many others) record at 44.1, and as mudersgalore says, no one has chastised me for using that rate.

Personally, I don't think there is a difference. That's only an opinion.
 
Ask yourself, who's your audience, who will be listening to the recording, if it's average Joe, don't worry about it, he'll just crank the bass anyway.
Seriously, unless your music will be used in a surround system in a major application/soundtrack, 48 is plenty good.
My $.02
 
it's not gonna really matter ..... however, there's a pretty good sized body of study that indicates that content above 20k, while not actually heard as content by the human ear, does affect the way we hear stuff that IS audible. Thus, for very super picky guys it can affect things like whether strings sound harsh and stuff like that. So in high-end audio they do hear a difference but it's subtle ..... and it's mostly gonna apply to delicate acoustic stuff.
I'm far too lazy to look for links but there are numerous studies that support the higher rates for ultimate resolution in that type of music ...... so they'll be easy to find if you care to.

But for us ...... 44.1 should be just fine .... it's what I use and I AM an audiophile ....... I just don't obsess over it anymore. Well, I suppose that means I USED to be an audiophile. :D
 
One thing to note: the line where the audio frequency is cut off isn't the sampling rate, it's half the sampling rate (or, to be completely accurate, less than half the sampling rate). In other words: a recording with a 48k samping rate can only accurately reproduce an audio signal that's less than 24k. The reason is that you need to take more than two samples of a waveform to determine its frequency and amplitude.

Of course, nobody can hear a sound with a frequency over 24k anyway (or even over 20k, or for many people, over something more in the 12k-16k range).

So ... possible answers to "does a sampling rate over 44.1 or 48 do any good?:

- Possibly, because (as already mentioned above), though no one can hear a super-20k tone standing alone, when combined with audible frequencies, the super-20k content somehow affects the hearer's preception of the audible range. Further notes: (a) this is, at best, extremely subtle, (b) this is, at worst, non-existent, as it seems inconsistent with how the human ear actually physically works.

- Possibly, because it allows you to move the anti-aliasing filter higher. More explanation: if you want to record at, say, 48k, you ideally should use a brick-wall low-pass filter than cuts off everything above 20k or so. Allowing higher frequency content than you can accurately represent can result in "aliasing," or the appearance of spurious lower-frequency signals in the audible range. Of course, there's no such thing as a brick-wall low-pass filter: any real-world filter will have some imperfections. Depending on the design, it will likely cut frequencies will below the intended cut-off frequency (remember that one octave below 20k is 10k), or "ring" near the cut-off frequency, or otherwise distort or affect the spectral content. Move the filter up to 40k, and you've got an entire octave above the highest possible audio frequency you care about.

- It doesn't make any difference that's audible to a "blind" ear, but if you tell people they're listening to a "better" 96k recording, it'll sound better to them merely because of that. Yes, this is nothing but a placebo effect ... but placebos do, in fact, actually work.
 
I do 44.1. The deal is the audio spectrum that's reproduced is half the sampling frequency so 44.1 comes out to 22.05 Khz. I absolutely believe humans can percieve content above 20k but listening to music damn near anywhere outside of a studio, the listening enviroment is severly compromised, so fuck it.
 
I would imagine it depends on the music you are recording. For critical stuff, more 'information' is captured at higher sample rates, and when the digital audio is converted back to analog, there will be less empty space between samples to smooth over. You should end up with a more accurate reproduction of the source, and it should sound smoother. But is this what you're after?
Those upper range frequencies, or harmonics, might come in handy when recording acoustic instruments, cymbals, or whatever else has that quality. Maybe grand piano or flute? Then again, for the distorted rock guitar most people record, who can hear the difference between 24-bit and 8-bit?
 
The higher sampling rates give a smoother wave form, if you zoom in on a 44.1 wave eventually you get a jagged edge, on a 96 wave you have to zoom further in to get the jagged edge. An analog wave is smooth. The higher sample rates give a sound closer to analog. Classical music tends to show up this more so than rock or dance music due to it's actual nature. A classical orchestra is often recorded with a handful of mics to 96k, the people that listen to it also have very high end play back systems, rock and dance often end up as mp3's.

Can you hear a difference, well yes but only slightly, I would think it's more to do with the feel of it than actually hearing it, 96k is easier on the ear, Analog even easier.

Unless you were in the classical music field 44.1 at 24 bit will sound fine, and the files will be much smaller than at 96k.

Alan.
 
There seems to be general agreement on the practical bottom line, somewhere in the vicinity of "maybe it matters theoretically, but practically: not so much."

Therefore, what follows is nothing more than a digression ... in other words, ignore it if you just are interested in the original question. But I'll go off on the digression anyway.

Actually, I think a higher sampling rate would ordinarily produce a less smooth wave form.

There are quite a few people who worry about "stair-steps" (or even claim to hear them) in digitally-reproduced sound. Do they really exist? In order to get them, you'd need an amplifier that instantaneously changed voltage, then waited for a 1/44,100 seconds, then instantaneously jumped to a new voltage ... not to mention a speaker element that moved a tiny fraction of an inch, then sat still for a fraction of second, then moved another tiny fraction of inch. It would be a challenge - even if, for some reason you wanted to do such a thing - to build such a system.

The difference between what comes out the far side of a 44.1k digital recording and 96k digital recording arises from the fact that the 44.1k system will (at some point, depending on where you set the sample rate) filter out any content over about 20k or so, and the 96k system won't. The 96k system output will have more little tiny short-wavelength wiggles ... well, at least it'll attempt to have more. The extent to which they survive the process of being amplified to a voltage (and with current) sufficient to drive a speaker, then translated into incredibly quick movements of the speaker element, transferred into the air, into your ear, and finally into nerve impulses to your brain is all a bit of an open question.
 
Can the human ear tell between 48k and 96k recordings?

I know that my Bassett Hound can tell.
 
Are your recordings audiophile grade? Probably not. Don't worry about it and record at the lower sampling rate.

Use your ears. The converters in your interface might even sound better at 44.1 than 96. Some converters just don't pull off higher sample rates well.
 
A great 48k converter will often sound better than a cheap 96k converter. It's not a matter of just sound quality, but the more accurate high frequencies represented by a quality 96k recording will contribute to a wider, clearer stereo image, and "openness" in the high end.

That said, if you're running through low to mid end converters/mics/rooms, etc, it's essentially wasted, inaudible bits. Save the space.

If you want to see the difference though, open a DAW session at 44.1, set up a sine wave on an aux track, route it out to an audio track and record 1k, 5k, 10k, and 20k tones onto it. You'll see that 44.1 doesn't just fail to accurately capture a 20k tone, even the 10k tone gets jagged and inaccurate. Even an old man can hear 10k poorly captured in the overtones.

Then repeat the experiment in a 96k session. With 9.6 samples per cycle, 10k will be significantly more "sine wave" like, and 20k will be better. Even if your speakers don't replicate it, what if you end up producing the next Nora Jones breakout hit? I'd hate to have only a low quality recording of that.
 
But:

The difference between a 10k sine wave and some other-shaped 10k wave is composed of harmonics above 10k, most of which aren't audible. As you approach the limit of the highest frequency you can hear, the shape of the wave become inaudible: all your ear can perceive is the amplitude and the frequency. You'd need good hearing to be able to tell the difference between a sine wave and square wave at 10k, much less a sawtooth or whatever (if they're really properly matched in amplitude) - the old man who could hear the differerence would be quite rare.
 
nathaneadam said:
..... Even if your speakers don't replicate it, what if you end up producing the next Nora Jones breakout hit? I'd hate to have only a low quality recording of that.
Click to expand...

That would be an exception. But what if I were recording Tom Waits? He'll sound like a spoon caught in a garburator no matter what. I still say it's up to the sounds you are after, and your level of expertise. If you were a pro recording Norah Jones, and she's sure as Hell wouldn't be laying vocal tracks in my untreated bedroom (I'd have something else for her to lay), there's no argument; I'd have $$$$$ pro equipment. But we're mostly bedroom amateurs, with ¢¢¢ equipment, so arguing pro specs is almost like two school yard kids arguing if Hydrazine or KNO3 would be a 'better' fuel. At their age, and likely lack of science training, does it matter?
 
You must log in or register to reply here.
Back
Top