Sample Rate

[Winterpark]

hey, i'm glad you sorted out your issue.

in all honesty, it's actually good to know that i'm sampling my sounds at 2.26757 millionths of a second! (is that right? i didn't pass maths at year 11)

and btw... i think that this conversation would have been happened with a bit more mutual respect and humour if it had been face to face with a beer in hand

[nuxnamon]

If it's any consolation, Katy perry's firework was recorded in 16/44.1

[Akshara]

For those of you not supporting 192khz sampling rate.. what bit rate and sampling rate are you guys for then?

At this time, based upon the hardware available to me, I have settled upon 24-bit / 44.1khz for basic music tracks and composing; and then 32-bit / 88.1khz for higher quality productions.

Most may already know the following, yet for those newly researching this issue...

The most common recommendations for the best compromise of audio fidelity and resource requirements, and considering current publishing and playback technologies, are to use 24-bit / 44.1khz for material which focuses primarily on the use of conventional sample libraries and loops, with live audio recording as secondary; and then either 24-bit / 48khz or 32-bit / 96khz for material that focuses primarily on recorded live audio, with sample libraries and loops used as secondary, or for material that will be used in professional video and film productions.

When recording at higher sample rates, it is also often recommended to use an external resampling program to convert any samples / loops to the project sample rate, and then to import those into the project or the sample instrument, rather than to use realtime sample rate conversion, which can take up considerable resources. The practicality and time involved with doing this should be considered when deciding upon whether using a higher sample rate is best for your project.

Beyond scientific experimentation, there are two other instances where using higher resolutions are common: one is when using non-sample based virtual instruments, as the generated audio can often sound better when exported at higher sample rates; and then second, when recording dynamic acoustic instruments in a quiet setting, such as a chamber string section or an orchestra, as the increased resolution and noise floor, combined with high end recording equipment, can make a difference.

[mr.ergonomics]

When you sample a recording you create a digital representation of the recording with the values between the digitalized points being approximated by connecting the points. So, when I record something in the 10kHz range with a sampling frequency of 192kHz I'll get approximately 19 digital values(not a lot, but a hell of a lot better than the 2 values you get with a sampling frequency only 2 times the actual frequency) that represent the analog on the same equivalent time scale. The remainder of the values(which is technically infinite because you can reduce the time scale down to whatever you want) are approximated by "connecting the dots" so to speak.

no and that is the point, it's a wrong image of digital audio. a 192khz sampled 10 khz sinus is the same sampled with 44,1 khz. absolutely no increase in derails. the sample points are NOT the signal. it gets reconstructed with this points. your idea about digital audio is definitely not dumb or anything, but it's only 50% of the DA process therefore your assumption is wrong. if you're interested I can explain it. but do what you like, I just want to give you a hint not hinder you to do something.

[3phase]

face meets palm.

anyone else wanna handle this one?

please toni dont.. i envy you living in a world of audio perfection but.. anyway..in this case its correct that 192 k dont brings advantages in theory..its abletons warping and internal fx that benefit from the oversampling fx as higher the rate gets... but this should be handled with a general oversampling option for the program than dealing with so high recording sample rates..

problem with 192 k rates is that you allready start to get the highfrequency radiation shit in your area into the recordings.. ..ok..depends on your area.. lives spectrum analyzer can show you waht is going on there.. and on 192 k recordings you might see some stuff above 50 k you really dont like to feet your mix engine with..

a compromise rate on the 88/9k6 is actually the best to work with when you can effort this on the disk space and performance side of things.. but the benefits inside the recorded files is again little.. the biggest part comes from the better warping on the higher rate..

would be really good if one could switch at least the warping on double rate without effekting the projekt samplerate. so a 44 projekt warps with 88 and an 88k projekt with 196k..

[OneZeroMusic]

no and that is the point, it's a wrong image of digital audio. a 192khz sampled 10 khz sinus is the same sampled with 44,1 khz. absolutely no increase in derails. the sample points are NOT the signal. it gets reconstructed with this points. your idea about digital audio is definitely not dumb or anything, but it's only 50% of the DA process therefore your assumption is wrong. if you're interested I can explain it. but do what you like, I just want to give you a hint not hinder you to do something.

I'm intrigued as to what your explanation will be, mostly because you just explained the exact same thing I said while calling me wrong.

[mr.ergonomics]

no and that is the point, it's a wrong image of digital audio. a 192khz sampled 10 khz sinus is the same sampled with 44,1 khz. absolutely no increase in derails. the sample points are NOT the signal. it gets reconstructed with this points. your idea about digital audio is definitely not dumb or anything, but it's only 50% of the DA process therefore your assumption is wrong. if you're interested I can explain it. but do what you like, I just want to give you a hint not hinder you to do something.

I'm intrigued as to what your explanation will be, mostly because you just explained the exact same thing I said while calling me wrong.

I have only time for an short answer atm so apologies if its a bit fast written..

brief: the sampling theorem says that a bandlimited analog signal that has been sampled can be perfectly reconstructed if the sampling rate is more than 2x higher than the highest frequency in the original analog signal. this is proven by math. every audio signal is just an addition from many sinewaves. a bit simplified.. but the sampling points are "used" to see which sine waves you need to reconstruct the signal. you don't connect just the points from the sampling point you see in the editor.

the so called "more detailed" picture from audio is wrong. when you sample a signal with 192 khz and 44 khz the content below 22 khz (in reality due to not perfect filters a bit lower) is exact the same!

the difference with 192 khz is that you can get higher frequencies (you don't need). more sampling frequency just means higher possible content.


again: a sampled 10 khz sine wave with 44 khz is exactly the same as sampled with 192 khz. when you have a signal with a sine at 10 khz and a sine at 25 khz there is a difference, with 44 khz sampling rate you don't get that 25 khz sine. but we can't hear a 25 khz sine anyway...

it can happen that a da converter sounds better at higher sampling rate, but that is only due to bad and not sharp enough filters in the converter. but even if you consider that you dac has a bad filter... using 96 khz should be absolute enough to prevent this.

[kb420]

192 khz is definitely the way to go. You want your music to sound as crisp and clear as possible. It's absolutely necessary, so that when you blow up, every kid will play your super clean/clear music in their ipod.

[docprosper]

.. and on 192 k recordings you might see some stuff above 50 k you really dont like to feet your mix engine with..

ermmm... wrong. Actually, it's the lower sampling rates that cause high-freq (higher than audible) noise to distort your audio in the audible range, it's called aliasing. In theory, a 192kHz sample rate would protect the audible range from a number of aliasing bins' worth of noise that lower 44kHz sample rates would not protect against. Of course, good hardware filtering right before the ADCs will have a great deal of impact on the recorded quality as well.
-hamish

[xzusa8ky]

I guess the questions you have to ask yourself if you want to insist at recording at 192k are...
how much disk space do you have?
how high can human hearing go?
what are you going to bounce it out at?

i'm sure there could be a bunch of psycho-acoustic stuff that can take place up there in the nether regions where bats and dolphins hear, but i think recording at 192k is a bit overkill...

Oh my..it appears as though I didn't explain myself clear enough. This is a bad analogy but I think it will provide a better understanding. Imagine you have a repeating picture that is 10ft long (a picture of a log for instance). Now imagine that you have a device that only shows you a part of the photo, say once every 10 ft you'll see part of the photo. Based on this you will never see the full photo (sampling frequency equivalent to the frequency that we're sampling). Now imagine you can set the device to view the photo every 5 ft, you get to see more of the photo, but you still don't get to see a good majority of it (sampling frequency being 2x greater than the frequency being sampled). Now imagine you can set the device to see the photo every 1 ft, you get a pretty good idea of a lot of things that are in the photo at this point(sampling frequency being 10x the frequency being sampled). Now, instead of a log that is 10ft long, we have a sound that is 20kHz(approx the max range of the human ear), having a sample frequency of 192kHz is roughly the same as viewing the 10ft repeating picture every 1ft, allowing the software to more accurately reconstruct the sound. That so when it is played back, we can hear a better approximation of the original sound.

I have about 2 TB worth of HDD space, the human hearing range peaks at approximately 20kHZ(as mentioned above) and I'll probably bounce it out at that same sampling frequency and see how big the file is and adjust accordingly.

I hope this is more help in understanding how the sample frequency actually works.

Now does anyone know how to resolve my issue with this?

I understand your explanation, but I've got a better explanation for you... (as i'm actually a music production teacher, who has to teach nyquist theory)

if your sample rate is 44100, then it's actually sampling 44100 times per second.... which you halve to take into consideration that you want to capture the representation of a full cycle. This gives you a magic number of 22050... so according to the nyquist theory, that means if you sample at 44100, theoretically, you are accurately recording sounds that are well above the range of human hearing.

so... 192k is actually sampling 192,000 times per second, which will give you an accurate picture of frequencies up to 88,000Hz, which is 4x greater than any audible frequency heard by humans. I get what you mean about the more snapshots, the more accurate the picture, it's just that i believe that a higher bit rate is more important here, because to use your analogy, every picture of the 'log' you are taking is of a greater quality... like taken with a 3 megapixel vs 10 megapixel camera.

and as i said, i'm not discounting any possible psycho-acoustic stuff that may or may not be occurring, but i'd add a qualification to that... you should check the frequency response of the speakers upon which you are playing back your music, because, i'm pretty sure that they aren't going to extend up to 88,000hz....

so i personally see no point in recording at 192, because my speakers won't be able to play back all that extra detailed top end, that I wouldn't be able to hear anyway with my 30-something year old ears that have played too many punk rock gigs, and if it was going to be released, it's most likely going to end up at 16/44100, or even worse as an mp3.

anyway... i have no solution to your problem, i just felt like i'd like to clarify my position, which is based on some years of experience.

good luck with your recording!

...and the winner is? YOU!

[xzusa8ky]

192 khz is definitely the way to go. You want your music to sound as crisp and clear as possible. It's absolutely necessary, so that when you blow up, every kid will play your super clean/clear music in their ipod.

If you listen to a symphonic orchestra on the most exspensive High Fidelity Top of The Range one million euro top gear you maybe hear some crisp....But normally NO!

[Tone Deft]

Of course, good hardware filtering right before the ADCs will have a great deal of impact on the recorded quality as well.
-hamish

almost got to the point. it's not 'good hardware' it's standard audio hardware which is designed to work in the standard 20-22kHz frequency range. DACs can be applied to more than audio, the automotive, aviation, military and test industries use them to monitor/control equipment but go look for audio DACs (Cirrus Logic, ADI, Ti/Burr Brown are a few), they're made for 20-22kHz.

not to mention your amp/monitors, those are band limited to 20-22kKhz. which one of you guys are using M Audio's 20-100kHz monitors?

there are people out there with custom solutions, home brewed DACs and amps, but there are also people out there that insist on wooden knobs on receivers and Monster cable. those are the same people touting the benefits of ultra-sonic frequencies mixing in the ether to form sub sonic tones we can hear, whoa man... (passes the dutchie 'pon de left hond side.)



what do I use? who cares? think for yourself. I never said the OP shouldn't use 192k, I posted because his reasoning is wrong. it's a natural progression to have his thinking as you learn this stuff. then you dig deeper and learn what things matter and what things don't matter.

I use 48kHz 24 bits. ever CD I've ever heard (which sound amazing) are at 44.1k 16 bit. I like 48 because it works with 96k and 192k when it comes to applying SRC. with my old computer I used 192k to get low latency when recording guitar. my newer computer is a monster and has low latency at 48kHz.

would be really good if one could switch at least the warping on double rate without effekting the projekt samplerate. so a 44 projekt warps with 88 and an 88k projekt with 196k..

the only not half baked part of your post touches on a second thing to consider when choosing a sample rate. consistency. Live will have to do less 'math' on your audio if all your samples are at the same rate. or to put it another way, you can use a simpler work flow with fewer SRC conversions if everything is the same. I stand by this but I'm pretty lazy about it.

[kb420]

192 khz is definitely the way to go. You want your music to sound as crisp and clear as possible. It's absolutely necessary, so that when you blow up, every kid will play your super clean/clear music in their ipod.

If you listen to a symphonic orchestra on the most exspensive High Fidelity Top of The Range one million euro top gear you maybe hear some crisp....But normally NO!

ipod??? get it???

[xzusa8ky]

192 khz is definitely the way to go. You want your music to sound as crisp and clear as possible. It's absolutely necessary, so that when you blow up, every kid will play your super clean/clear music in their ipod.

If you listen to a symphonic orchestra on the most exspensive High Fidelity Top of The Range one million euro top gear you maybe hear some crisp....But normally NO!

ipod??? get it???

Got It! He He..........

[OneZeroMusic]

I have only time for an short answer atm so apologies if its a bit fast written..

brief: the sampling theorem says that a bandlimited analog signal that has been sampled can be perfectly reconstructed if the sampling rate is more than 2x higher than the highest frequency in the original analog signal. this is proven by math. every audio signal is just an addition from many sinewaves. a bit simplified.. but the sampling points are "used" to see which sine waves you need to reconstruct the signal. you don't connect just the points from the sampling point you see in the editor.

the so called "more detailed" picture from audio is wrong. when you sample a signal with 192 khz and 44 khz the content below 22 khz (in reality due to not perfect filters a bit lower) is exact the same!

the difference with 192 khz is that you can get higher frequencies (you don't need). more sampling frequency just means higher possible content.


again: a sampled 10 khz sine wave with 44 khz is exactly the same as sampled with 192 khz. when you have a signal with a sine at 10 khz and a sine at 25 khz there is a difference, with 44 khz sampling rate you don't get that 25 khz sine. but we can't hear a 25 khz sine anyway...

it can happen that a da converter sounds better at higher sampling rate, but that is only due to bad and not sharp enough filters in the converter. but even if you consider that you dac has a bad filter... using 96 khz should be absolute enough to prevent this.

Actually it can't be "perfectly" reconstructed, and it isn't proven by math. I know, I took courses in filter design during university and have done the math by hand, I know my shit. However, you are correct about reconstruction by Sine waves, although I believe there are more ways than this to reconstruct signals. And the more detailed picture is correct, like I said I've done the math by hand and it has gotten to the point where I don't care if your dumb stubborn ass believes it or not.