sample rate ? 44.1 ?88.2?

i know that it really should make no difference but at some primeval level i cant help i just know more is better

what sample rate do u use ?can anyone give me a good reason that 88.2 is worth the extra disk space?

why do i think it sounds better..??

it only samples the wave in more places but its the same wave comes back as at 44.1...

Why not do some truly blind listening tests where you don't know which you're listening to (that you don't know is vital, otherwise pyschological aspects come into play - we hear what we expect or want to hear) and see (hear) if you can reliably tell which is which?

 

One possible advantage of a higher sample rate is your audio interface may operate at a lower latency.

huh?

so i will get 1/2 the latency? sweeeeet

 

a blind listening wont help as my inner cave man wants MORE which is why i guess all the audio interface makers are making such silly sample rates

 

lol

My opinion is if your computer can handle it, and it's not inconvenient to switch, you might as well go up to a higher sampling rate.

 

Some people like knowing that they're getting as many samples as possible. Others like the convenience of staying at 44.1k and not having to worry about messing with buffer settings or anything else that could affect playback and recording.

 

It's up to you.

huh?

so i will get 1/2 the latency? sweeeeet

 

This is not magic: the latency is halved if you use the same size I/O buffer size. Obviously, the size of the I/O buffers you can afford to use depends on the amount of stress you place on the CPU, and higher sample rates put a lot more stress on the CPU (on top of taking twice the hard drive space and throughput), so depending on your machine and session, you might have to raise your buffer size when using higher sampling rate, resulting in same or higher latency.

 

If you have a really powerful machine running at the lowest buffer size without any problem, and it can take even more CPU stress, then you will experience the halved latency.

If you have a really powerful machine running at the lowest buffer size without any problem, and it can take even more CPU stress, then you will experience the halved latency.

 

Also, some cards use a larger buffer at higher sample rates - you still get lower latency, but more like 25% than 50% . . .

Some people like knowing that they're getting as many samples as possible. Others like the convenience of staying at 44.1k and not having to worry about messing with buffer settings or anything else that could affect playback and recording.

 

It's up to you.

 

OK, I take the bait.

 

A higher sample rate than 44.1K does not more accurately allow for the reproduction of all frequencies up to 20KHz.

 

A higher sample rate allows for an extended frequency response, not more accurate response of the audible =>20KHz frequency range.

 

Very high sample rates are more prone to data loss, jitter, and external electronic interference, and so it's definitely not a case of the higher the better, although agreed, that is an easy marketing line to swallow (or an easy line to market. . . )

 

So those people who 'like knowing that they're getting as many samples as possible' are satisfying some psychological need, which is fair enough, as opposed to following sampling theory and science.

 

One definite advantage of higher sample rates is that poor A/D filtering is more easily disguised. There's no excuse for poor filter design, however.

 

Let's just suppose some blind listening test did show a repeatable preference for higher sample rates (I doubt it, personally. No high-profile test has thus far done so - but just suppose one did) then, if the output medium is to be CD, another blind listening test has to be conducted, between 44.1K sampling from start to end vs a 88.1/96K recording which then undergoes a Sample Rate Conversion to 44.1K.

 

You may lose more than you gain in the SRC, and you may not have gained anything anyway. . . .

Ok, I had to google it.

 

http://lavryengineering.com/documents/Sampling_Theory.pdf

 

I didn't read the whole thing from beginning to end yet, but I see what you're saying.

I salute you for following this up.

 

That is an excellent article by Dan Lavry which really does address plenty of 'common sense' misconceptions. . . .

I went through the whole faster sampling rate is better thing and I went back to doing music at 44.1 and TV/movies at 48k. Honestly, I think I wanted it to sound better at 88.2 or 176.4 (for music) but by the time all was said and done and the product is delivered on CD or worse (mp3) I wasn't convinced. SRC being so destructive and then the associated storage and processing limitations tipped the balance for me back to lower sample rates.

Sample rate reduction is a quite smooth thing if the source sample rate is a multiple of the destination sample rate; in other cases artifacts might be introduced. I am using 88.2kHz because some software synthesizers like the ES2 do better anti-aliasing then, but for the final product I go back to 44.1kHz because I agree to the statement more is unnecessary (and a waste of resources).

Sample rate reduction is a quite smooth thing if the source sample rate is a multiple of the destination sample rate;. . . .

 

Is this 'common sense' or do you have some evidence?

Call it "extended common sense".

For sample rate conversion, every destination sample has to be derived from a set of source samples around this point that are "wightedly accumulated", i.e. multiplied by factors depending on their relative position and then added. Obviously, if the source sample rate is not an multiple of the destination sample rate, the positions of the destination samples relative to the source samples' positions change with every destination sample, so the factors have to change with each destination sample, and this is the point where artifacts can be introduced depending on how accurate the algorithm works, how the weighting factors are determined and how many samples from the source are used for the calculation.

If the source sample rate is a multiple of the destination sample rate, there will be no change in the relative sample positions of source and destination, so the weighting factors remain the same all the time, no parasetical frequencies are introduced.

That's it. This is the reason why I think for getting 44.1kHz, you better use 88.2kHz and not 96kHz (if you have a reason at all to use higher sample rates during your creation process).

ho ho its always a good laugh the old sample rate

 

thank you all for your help

 

i am using an apogee rosetta 800

 

seems less harsh at 88.2 but as u say i am probably imaginin it...

 

but hell i have MORE!! so i feel better so i rock more

 

over n out

I have a question: what about processing an audio (like time-stretching or "auto-tuning") ?

 

I think that in this case higher sample rate should be better, right?

I have a question: what about processing an audio (like time-stretching or "auto-tuning") ?

 

I think that in this case higher sample rate should be better, right?

 

Probably you are right, but these things are much more complicated than sample rate conversion since the time stretching and relating algorithms have to do a totally different job: Their nature is to change the input waveforms to fit into new time responses. Some cycles of the original waveform must be skipped or, if the length has to be extended, even be invented based on what is already there, and the result must be smooth and flutter-free, so there is no linear procedure to achieve this. Generally speaking a higher sample rate results in more exact calculations, but we don't really know details here.

Do I dare chime in?

 

Has anybody anywhere done any research on how the supersonic frequencies, which we can't hear and are difficult to record, create sonic artifacts like harmonic beats that alter (for good or for worse) the spectrum that we can hear?

 

Or, is there any discourse on how those supersonic frequencies give our brains the information we need for accurate spatial location and depth perception?

Has anybody anywhere done any research on how the supersonic frequencies, which we can't hear and are difficult to record, create sonic artifacts like harmonic beats that alter (for good or for worse) the spectrum that we can hear?

 

Or, is there any discourse on how those supersonic frequencies give our brains the information we need for accurate spatial location and depth perception?

AFAIK there has been research that shows that you can hear frequencies above 20KHz through bone conduction, so if the sound waves are directly applied to your skull. In that case you're bypassing the middle ear.

 

http://citeseer.ist.psu.edu/images/5f/8a/11/a4/b346680518701486208ca24e80814ce1/1.png

Do I dare chime in?

 

Aren't you're the one that pooh-pooh'd Shannon and Nyquist and the Sampling Theorem

 

Do you take it back yet, eh?

 

Has anybody anywhere done any research on how the supersonic frequencies, which we can't hear and are difficult to record, create sonic artifacts like harmonic beats that alter (for good or for worse) the spectrum that we can hear?

 

If those supersonic frequencies create sonic artefacts in the spectrum we can hear, then we only need to record the spectrum we can hear. Do you need a camera to capture cosmic radiation? No, you need a camera to capture the effects of cosmic radiation (if there weere any) in the visible spectrum

 

Or, is there any discourse on how those supersonic frequencies give our brains the information we need for accurate spatial location and depth perception?

:

 

There was in a massive thread over at PSW - effectively a debunking and deconstructing of existing 'research'. But I'm sure you can find these beliefs very much alive with hi-fi nuts who buy 'oxygen-free' cables from mars, and - uh - scientologists. . . .

now im really confused

 

does 88.2 record freq above 20khz? that 44.1 does not that im not supposed to be able to hear but like the old rubert neve story where he obviously can hear up there (amek 9098 sheen button)

 

coz if thats the case i'll trust rupert and go with 88.2

 

thanks ?

 

blimey

anyway are we all agreed that 192 is a complete waste of space ?

anyway are we all agreed that 192 is a complete waste of space ?

 

No!

 

192 captures more slices of the original analogue sound - so the resulting analogue reconstruction is more accurate. I mean 8 x 2 is 16, right. But it's a more accurate 16 if I know it's also 4 x 4, right?

 

It's like frames/sec or pixels. More is better. Like money, and alcohol and noise. No not like noise. Or alcohol - or maybe, up to a point. . . .

 

Anyway, I'm convinced that there is energy around 50KHz that is somehow vital - I haven't actually got a mic that can capture past 22K - or speakers that can reproduce above 20K but that's not the point. I listened to a 50KHz sine wave for 2 hours last night - I couldn't actually hear anything but it felt good, I swear.

 

I read somewhere about the disproportionate number of errors in higher frequency sampling, an inevitable factor of the physics of speed (not just current designs) but I reckon that's all baloney because Nyquist was sooooo early 20th century, and rock 'n' roll - or at least Logic Pro 7 (when's 8 due, btw?) - hadn't even been invented then, so when Shannon came along along and (allegedly) mathematically proved Sampling Theorem, he was obviously misguided too. Probably out of his head on Mushrooms. . . . which was the first CD by my second band 'Devil's Avacado'. . . . OK, I know that digital technologies such as CD's and digital recording and DTV and satelite telecommunications and Duran Duran are all built on an acceptance of Sampling Theorem, but I'm convinced it's wrong and if I can convince you too, then we might wake up tomorrow and find it's all disappeared. No CD's, no 'Rio' and no awful remix of 'Satelite Of Love' and hopefully all the awful music from that decade referred to as the 80's will disappear too. . . . (who mentioned Duran Duran?)

 

Anyway, it's not just about frequencies, it's about transients. They are NOT one and the same, whatever physics says. I wish I had a name I could pooh pooh for this one, but I don't so I'll settle with 'audio science'. It's all rubbish, all of it, they should use their ears, or rather they should use mine because they obviously haven't got any anyway, and even the mathematical so-called proof which ties transient response to frequency response is just more numbers schnumbers - I know what I hear - or what I feel I hear, or rather what I think I thought I felt I heard. . Rawk.

 

Plus I just spend $9999 on a new interface that samples at 384KHz and I'm sure that it sounds better - less harsh, more open, more sheen, more 'je ne sais quoi. . . '. It just sounds more - uh - expensive. . . . yeah, definitely. More, er, extended, that's it. . . . it sounds like I made the right decision.

 

Phew!

 

I haven't actually done any 'blind' listening tests against my previous interface which only captured frequencies I can actually hear, but 384K just feels better to me - that the main thing - I feel it rawks - so I rawk . . . uh . . . even more. . . and I rawked a lot before, too . . .

 

But I just had a worrying thought. Only an sampling rate of infiinity can capture a true square wave - which would also have a transient time of zero. {they can't be related. . . ?) but if I want to sample infinity, I'm going to need a sample-rate of at least twice - and that means 2.0000000000000000000001 x no that's wrong! That's their theory - I need more slices - the more the better. I need a sample rate of infinity times infinity - infinity squared. That's it. Apogee, MOTU, are you listening?

 

To infinity (squared) and beyond . . . .

...so then the DAC would need to UNSQUARE infinity? Which would make it a circle (which can never be a square wave), so not only wouldn't it be accurate, I would have no corner to sit in when I am bad...which is always...which then means I would be inifitely filled with frustration for not being able to perform my penalty... C'MON PAPA!!! GIVE ME A PENALTY! I WANT MY PENALTY!!!

 

(sorry...have seen Amadeus too many times...)

 

(BTW- Blue- 384 wont accurately capture a butterfly's wings' most minute movements of air properly, which renders the chaos theory incomplete, and then we must all file out of existence, albeit in an orderly fashion. Please, no pushing or shoving, or straying of thought...we are not equiped for capturing that either...but I do know of a rather large goose that could reproduce an amazingly close imitation by breaking wind while using those cute little flipper feet to trot across the surface of a properly sized lake during the northern hemisphere's autmn months. A grand experience, to say the least...and nothing square about it either... Freud and Nietzsche actually agreed upon the goose's psychological drive to learn to do this...wonders never do cease, especially with geese.. this one, btw, named Clarisse, which is interesting since she would ponder quite a bit...and lived to be 451...{connect THOSE dots}...and she was good when sliced thinly, au jue...l'oh mon!)

How does choice of sampling rate impact someone who only uses software instruments to generate audio for songs? Since there's no sampling of an analog signal I'm wondering if the actual stream of 1s and 0s the instrument generates differs depending on song sample rate...

 

Or perhaps a better question is, could someone give a high level explanation of how the audio output of a software instrument is captured by the host during an export or bounce, and how the sampling rate setting for the song relates? Thanks!

blue in the face i am very happy for you but thats exacley the kind of misunderstanding they are using to get us to part with errr $9000

 

its not like pixels or frames..that would be your bit rate..you are reconstucting a wave and mathmatically you only need so many points on the wave to reconstuct it..it really makes no difference past a certain point..as i understand it anyway

 

348!! lol my iner cave man wants MORE!!

 

but it is wrong

 

 

sorry

listened to a 50KHz sine wave for 2 hours last night - I couldn't actually hear anything but it felt good, I swear.

 

 

 

now that is funny

 

reminds me of niel dancing to the tone on the tv....

 

maybe it relaxed your alpha waves and realinded your chakras !

 

hippy

 

having read your post again i have decided to join you on the bus

 

lol

How does choice of sampling rate impact someone who only uses software instruments to generate audio for songs? Since there's no sampling of an analog signal I'm wondering if the actual stream of 1s and 0s the instrument generates differs depending on song sample rate...

 

Nick,

 

it depends.

Imagine a simple sawtooth generator that in fact generates something like a stairway, since it works digitally and produces discrete samples, the steps of the stairway. Now set the frequency of the sawtooth wave to, say, 1000Hz (1000 cycles per second) and the sampling frequency to 44.1kHz (44100 samples per second). How many steps does one cycle of the wave contain? It's not 44 steps since 44 steps times 1000 Cycles is less than 44100 samples; it isn't 45 steps either because 45 times 1000 is more than 44100. What can we do now? - Right: we have to make some sawtooth cycles 45 steps and the rest 44 steps long, but this will introduce a jitter that is audible very clearly, and it doesn't sound good at all.

The maximum error/jitter of the frequency in our example is (45/44)-1 = approx.2.2%, just to state a number. If we decide to double the sampling rate, the error would be reduced to (89-88 )-1 = approx. 1.1% which is obviously better.

The full truth on digital tone generators/synthesizers is more complex, I know. You can keep the height of the steps constant and shift the starting point for a better approach; next is digital filtering, i.e. the synthesizer works with more samples internally and then calculates the best approach for its output. This is called "anti-aliasing" and does the same what the filters in your analog interface prior to the ADCs do when you record an analog signal. If the synthesizer is able to do proper anti-aliasing, you don't need the higher sampling rate for your project, but the simpler the synthesizer's algorithms, the better does a higher sampling rate.

Hi Guys,

 

An important reason to use a high sample rate is that plugins will then work at this rate, this usually gives better result in dsp algorithms.

 

It is fairly simple to do an A/B comparison of a plugin synth or effect at 44.1, 88.2 and 96. The difference is easy to hear for me.

 

Some of the NI plugins allow you to double the sample rate of the host for internal plugin operation, the difference of running at 192 over 44.1 can be quite astounding as well as the hit on CPU

 

Cheers

 

Andy

Hi Guys,

 

An important reason to use a high sample rate is that plugins will then work at this rate, this usually gives better result in dsp algorithms.

 

Yes, this is what I meant... I just tried to give some vivid theory to explain this, but indeed, try it yourself. I believe higher sample rates are more important for sound synthesis than for recording natural sound. Let me recapitulate that in case of good anti-aliasing the synthesizers use higher sample rates internally anyway.