| Author
|
!! Sample Rate !!
|
Koti
Started Topics :
4
Posts :
8
Posted : Jun 19, 2007 02:31
|
How do you regulates Sample rate?
44.100Hz or 96.000Hz ?
whats the best way?
thanxxxx
|
|
|
pilgrim
IsraTrance Junior Member
Started Topics :
19
Posts :
218
Posted : Jun 19, 2007 02:44
|
@koti,
i use 44.100 with 24 bit - only in the very end i reduce bitrate to 16 bit to burn on a cd
I'm not telling it's the best way, if u have a powerful computer and a good interface, why not use higher rates....some people swear to it...
if you're a beginner there's much other stuff to learn and do to get better sound quality then turning up the s-rate or bit-deph
boom & enjoy |
|
|
|
bilbobagginz
Started Topics :
8
Posts :
399
Posted : Jun 19, 2007 02:47
|
you usually want to work with the highest depth & sample rate, and during the final mix you downsample with a high quality resampler, or A/D converter.
but this isn't the law, it's just a guideline.
|
|
|
|
Koti
Started Topics :
4
Posts :
8
Posted : Jun 19, 2007 04:41
|
Hey bros, thanx for the answers!!
But I come perceiving that mine Mix are leaving with a little loss in the regions of acute frequency, I would not be for the fact to be using 24bits/44.1khz?
I work with a Sony vaio intel Core duo, 1gb ram & m-audio ozonic
|
|
|
|
Psynaesthesian
IsraTrance Junior Member
Started Topics :
30
Posts :
557
Posted : Jun 20, 2007 11:12
|
this question just played in my mind this morning after i realised that one of my uploaded tracks went from nice bumpy dance groove to minimal psy after reviews from some very kind souls who took the time to listen to it and give my this valuable feedback.
I always render my tracks from Reason at 44.1kHz / 16 Bit-depth .... but as a wav file it sounds pretty good with the punch and drive still evident. I suppose that the diminishing occurs when you convert them to mp3. Is this correct?
But i've heard mp3 tracks at Samplerate 192 which still sound very good, so i'm somewhat confused.
B'om Shankara!!
 "... b'om ..." |
|
|
|
vipal
IsraTrance Full Member
Started Topics :
123
Posts :
1397
Posted : Jun 20, 2007 12:48
|
maybe the monitors you used listening give more punch then the speakers your friends use.
did you compare the mp3-version with the wav-version on your own system? maybe it will answer your first question.
from what i got from the 'mp3 versus wav'-discussion i never heard of such a dramatic change like you are describing. |
|
|
|
Elad
Tsabeat/Sattel Battle
Started Topics :
158
Posts :
5306
Posted : Jun 20, 2007 12:53
|
|
Upavas
Upavas
Started Topics :
150
Posts :
3315
Posted : Jun 23, 2007 09:46
|
Quote:
|
On 2007-06-20 11:12, Psynaesthesian wrote:
this question just played in my mind this morning after i realised that one of my uploaded tracks went from nice bumpy dance groove to minimal psy after reviews from some very kind souls who took the time to listen to it and give my this valuable feedback.
I always render my tracks from Reason at 44.1kHz / 16 Bit-depth .... but as a wav file it sounds pretty good with the punch and drive still evident. I suppose that the diminishing occurs when you convert them to mp3. Is this correct?
But i've heard mp3 tracks at Samplerate 192 which still sound very good, so i'm somewhat confused.
B'om Shankara!!
|
|
The higher the sample rate and bit depth, the better your resolution. Converting files into lower bitrate and samplerate means your soundquality diminishes as does your dynamic range. On the other hand, how much cpu does your machine have?
An mp3 at 192hz sampling rate will always sound a lot less good then a wave file that is recorded at 192k and 24 bit... It is the same as with a digital photo. The higher the resolution, the better the overall picture quality! In mp3 encoding dcpcm (a kind of pulse code modulation) is being used to compress the filedata. So some frequencies are simply dropped, the frequencies that are being dropped aren't really percieved as missing from the human ear, they simply appear less loud (which they are not), so the ruin outr ears in the long run  . Unlike in the analog domain where every sound is recorded accurately (although usually with a much higher noisefloor), in the digital domain we are always bound to some kind of limitation with the sampling rate, as the "picture" can have a very high resolution and be very near perfect, it is nevertheless always "imperfect", although it has a near nonexistent noisefloor. Like in the movies, where pictures are being displayed of a rate of 24 per second, the eye gets manipulated into seeing a motion picture. The same is the case in the audio domain (44.1khz sampling rate) which is being used to accurately reflect hz ranges of up to 20k (limit of human hearing)on a cd... try rendering a file to 8 bit 22.500 hz, it will sound really weird because the quality is so poor and the high frequencies get aliased (returned improperly) and it will sound somewhat granular, or simply put shitty...
I recommend to render at least 24 bit 44.1k from reason or any other daw.
Upavas - Here And Now (Sangoma Rec.) new EP out Oct.29th, get it here:
http://timecode.bandcamp.com
http://upavas.com
http://soundcloud.com/upavas-1/ |
|
|
|
UnderTow
Started Topics :
9
Posts :
1448
Posted : Jun 23, 2007 18:28
|
Quote:
|
On 2007-06-23 09:46, Upavas wrote:
The higher the sample rate and bit depth, the better your resolution. Converting files into lower bitrate and samplerate means your soundquality diminishes as does your dynamic range. On the other hand, how much cpu does your machine have?
|
|
You have to seperate sample rate and bit depth. Sample rate gives you bandwidth and bit depth gives you accuracy.
Quote:
|
An mp3 at 192hz sampling rate will always sound a lot less good then a wave file that is recorded at 192k and 24 bit...
|
|
With MP3s you talk mainly about kilo-bit rate. A 192 Kbps mp3 isn't the same as 192Khz sample rate.
Quote:
|
It is the same as with a digital photo. The higher the resolution, the better the overall picture quality!
|
|
Comparing audio to video/picture is always a bad idea because with 44.1Khz/24 bit we have the full human frequency and dynamic range covered. That isn't the case with most video/picture resolutions. If you increase the resolution from, lets say regular TV, you will notice a difference because regular TV is way below the resolution of the eyes (and brain).
Quote:
|
In mp3 encoding dcpcm (a kind of pulse code modulation) is being used to compress the filedata. So some frequencies are simply dropped, the frequencies that are being dropped aren't really percieved as missing from the human ear, they simply appear less loud (which they are not),
|
|
It isn't so much that they appear less loud. They are dropped because due to psychoacoustic masking, we don't perceive them. Of course, with lower bit rates, we can perceive the dropped data but for instance with 320 Kbps, we can't really distinguish from the original.
Quote:
|
so the ruin outr ears in the long run .
|
|
Bla. 
Quote:
|
Unlike in the analog domain where every sound is recorded accurately (although usually with a much higher noisefloor),
|
|
Incorrect. If you want accurate reproduction, analogue is not the way to go at all. Digital sampling is much more accurate than analogue recording.
Quote:
|
in the digital domain we are always bound to some kind of limitation with the sampling rate, as the "picture" can have a very high resolution and be very near perfect, it is nevertheless always "imperfect", although it has a near nonexistent noisefloor.
|
|
Incorrect. Sampling rate defines bandwidth, not resolution. With 44.1Khz and good anti-imaging filters, the whole humanly perceived frequency range is covered.
Quote:
|
Like in the movies, where pictures are being displayed of a rate of 24 per second, the eye gets manipulated into seeing a motion picture.
|
|
Bad analogy for the reasons explained above.
Quote:
|
try rendering a file to 8 bit 22.500 hz, it will sound really weird because the quality is so poor and the high frequencies get aliased (returned improperly) and it will sound somewhat granular, or simply put shitty...
|
|
Again, you have to seperate bit depth and sample rate. If you sample rate convert to 22.5 Khz with a good SRC, you don't get aliasing and you still have as a accurate a sound... except with all the frequencies above 11.25Khz removed.
If you render a good mix to 44.1Khz/8bit with a good dither algorithm, you
have all the frequencies but the noise floor is now 8 bits higher at arround -44 dB FS.
And these both don't really sound as bad as you might think!
Quote:
|
I recommend to render at least 24 bit 44.1k from reason or any other daw.
|
|
If any processing needs to happen to the file, yes 24 bits is recommended.
Now having said all of the baove, not all converters and plugins are made equal.
Alot of consumer and prosumer converters will sound better at 96Khz. That isn't because 96Khz has more resolution but because the anti-aliasing filters are not very good. This is only really an issue if you are recording sounds or sending out sounds to be processed externaly with analogue gear. For rendering within a DAW, it doesn't really matter how good your converters are as the signal doesn't go through them when you render. (Although having better converters will help you get a better mix because you can hear more accurately what you are doing).
The same goes for plugins, some plugins sound better at higher sample rates because they alias at lower rates. The solution to this is to choose your plugins carefully. Some will oversample internaly to avoid aliasing issues. Others are designed in such a way that they don't alias reguardless of sample rate.
And 192Khz ia a load of marketing bollocks.
UnderTow |
|
|
|
Upavas
Upavas
Started Topics :
150
Posts :
3315
Posted : Jun 23, 2007 22:08
|
Undertow,
I guess it is self explanatory that 192kbps is not the same as 192k wavefile, as I explained above mp3 is a kind of dcpcm,
the comparison with a digital photograph is an ok method to explain as it will make people understand a little easier, it is mainly intended as a device..., yes we have the full capacity of human hearing, however a 192/24 bit file will always sound better than a 44.1khz 16 bit resolution. I use the word resolution so someone who does not know can understand what I am talking about.
If you cannot distinguish a 320bit mp3 from a 192k wavefile, then your ears must really be shot man, I feel sorry for you...keep saying bla until it is all shot hahaha... actually it is the case that through the dropped frequs the mix appears less loud, which is the reason why so many people crank up their ipods and then wonder where their hearing goes. Analog will always record excactly what is being played. Digital will always be bound to a limitation of bandwidth and therefore never be 100%precisely recorded. True, the human hearing cannot make a difference, however it still is not perfect as analog is as there always is a hz sampling rate that gives a limitation... The question of bandwidth does not arise in the analog domain, hence you are wrong! The noisefloor however does! As to the analogy of hz over fps, I think it is a great analogy... as it helps someone understand the principle of it all as the principal concept is the same. 24 fps is enough to fool the eye or can you see a difference, meaning can you see the single pictures at a rate of 24fps? HDTV runs at a higher frames per second rate, and there are more horizontal picture lines, kind of comparable to bitdepth, making the overall picture so much smoother. The same is the case in the wave domain as is 44.1khz and the bit depth, simply describing a sample with more 1s and 0s and therefore more precise... And 192khz is used in the professional film industry nowadays, so be a bit more careful before you call it marketing bolloks mate.
Upavas - Here And Now (Sangoma Rec.) new EP out Oct.29th, get it here:
http://timecode.bandcamp.com
http://upavas.com
http://soundcloud.com/upavas-1/ |
|
|
|
UnderTow
Started Topics :
9
Posts :
1448
Posted : Jun 23, 2007 23:23
|
Quote:
|
On 2007-06-23 22:08, Upavas wrote:
Undertow,
I guess it is self explanatory that 192kbps is not the same as 192k wavefile, as I explained above mp3 is a kind of dcpcm,
|
|
I've never heard of dcpcm. Not much of an explanation then I guess.
Quote:
|
the comparison with a digital photograph is an ok method to explain as it will make people understand a little easier, it is mainly intended as a device...,
|
|
Unfortunately it just confuses the issue.
Quote:
|
yes we have the full capacity of human hearing, however a 192/24 bit file will always sound better than a 44.1khz 16 bit resolution.
|
|
This is where you are wrong. The quality of the converters is much more important than the sample rate. On cheap converters, you might hear a difference. On high quality converters, you probably won't hear any difference.
Quote:
|
I use the word resolution so someone who does not know can understand what I am talking about.
|
|
But the term resolution is absolutely wrong. You do not increase resolution by increasing the sampling rate. That isn't how sampling works.
Increasing sampling rate only increases the bandwidth. Once you have the full bandwidth that humans can hear covered, there is no need to go higher except to allow less steep anti aliasing filters in the converter decimators. That is why cheaper converters with less good filter designs might sound better at higher sampling rates.
Still, 192Khz is just way overboard and will not even improve on the filters compared to 96Khz or 88.2Khz.
[quote]
If you cannot distinguish a 320bit mp3 from a 192k wavefile, then your ears must really be shot man, I feel sorry for you...
[quote]
Aaah the age old argument... have you ever participated in a double blind test between 96Khz and 192Khz?
If you think you can distinguish a difference between these sampling rates, your brain is shot man, I feel sorry for you...
Quote:
|
keep saying bla until it is all shot hahaha...
|
|
hahaha indeed.
Quote:
|
actually it is the case that through the dropped frequs the mix appears less loud, which is the reason why so many people crank up their ipods and then wonder where their hearing goes.
|
|
As numerous double blind test show that people can't distinguish between 320Kbps MP3 (properly converted) and 44.1Khz/16bit waves, your suggestion that it sounds less loud is far fetched.
The reason people turn their iPods up too loud is because they are not listening to full-range speakers so they don't feel the bass. It is also often to drown out background noise.
Anyway, this whole iPod comment is a red herring. Who is talking about iPods?
Quote:
|
Analog will always record excactly what is being played.
|
|
You have no idea. What analogue format are you suggesting? No analogue format is as accurate as current day digital audio sampling quality.
Quote:
|
Digital will always be bound to a limitation of bandwidth and therefore never be 100%precisely recorded.
|
|
So what? We are not recodring for bats. We don't need that extra bandwidth. The only thing that counts is what humans can hear.
Quote:
|
True, the human hearing cannot make a difference, however it still is not perfect as analog is as there always is a hz sampling rate that gives a limitation...
|
|
Yes a bandwidth limitation but you said it yourself: "the human hearing cannot make a difference". That is all that counts.
Quote:
|
The question of bandwidth does not arise in the analog domain, hence you are wrong! The noisefloor however does!
|
|
Again, what analogue format are you suggestion that has unlimited bandwidth? I'd like to hear about this new format that defies the laws of physics.
Quote:
|
As to the analogy of hz over fps, I think it is a great analogy... as it helps someone understand the principle of it all as the principal concept is the same.
|
|
But it doesn't help anyone understand anything. It just supports the myths. It is not the same thing and it is not how sampling works.
Quote:
|
24 fps is enough to fool the eye or can you see a difference, meaning can you see the single pictures at a rate of 24fps? HDTV runs at a higher frames per second rate, and there are more horizontal picture lines,
|
|
There you go again, mixing two different parameters and confusing the issue. The frame rate in video is indeed analoguous to the sample rate but not in the way that you describe. With higher frame rates you can also avoid aliasing in video.
An example of aliasing in video is when you see a car accelerate from a standing start. The wheel will start moving forward and at some point, when the spinning frequency of the wheels reach half the frame rate, they will start slowing down again and then move backwards... Frequency aliasing. But increasing the frame rate will not make the image sharper!
And again, with a 2r or 25 frame per second rate, we are below what the eye can perceive so it makes sense to increase the frame rate as there is a net gain. Not so with audio sampling rates.
Quote:
|
The same is the case in the wave domain as is 44.1khz and the bit depth, simply describing a sample with more 1s and 0s and therefore more precise...
|
|
More 1's and 0's, yes. That is bit depth. Increasing bit depth increases accuracy. That is not the case with sampling rate.
Of course there is not point in going beyond 24 bits for two reasons: 1) 24 bits covers the whole human dynamic range. 2) Even the best converters can't achieve true 24 bit dynamic range due to the self noise of the components. 21 bits is more or less as good as it gets. The lower bits are just filled with noise.
Quote:
|
And 192khz is used in the professional film industry nowadays,
|
|
Film audio is usually data compressed before the audience gets to hear it... Anyway, most films are 48Khz. The film industry is a good example of NOT being fooled by the marketing. Of course, there are always exceptions...
Quote:
|
so be a bit more careful before you call it marketing bolloks mate.
|
|
It is always good to be carefull but in this case I am right. Most of the top engineers in the audio world agree that a sampling rate of arround 60Khz would be ideal. (I mean electrical engineers. The people that design the converters. Not audio engineers that often don't understand the technology). There is a very simple reason for that: With a 60Khz sampling rate, you can have an anti-aliasing filter that has a similar slope to the natural high frequency roll-off of air itself. No one is complaining about the way air sounds as far as I know.
It is very tempting to believe that higher numbers will automaticly give you better quality. It is a natural human tendency. That is exactly why this is being marketed this way. But that doesn't mean that the actual reality of how things work follows this base human way of thinking. Not everything is as intuitive as it might seem at first glance.
UnderTow |
|
|
|
UnderTow
Started Topics :
9
Posts :
1448
Posted : Jun 24, 2007 03:20
|
Here is a much more usefull analogy: You only need two points to define a circle. With two points, there is only one possible circle that can pass through those two points. Adding any more points will not define that circle any better. It is still the same circle.
If anyone is wondering how this relates to sound, it is good to know that a sine wave is just a circle drawn out over time. This is nicely portrayed in this nice animation: http://www.rkm.com.au/ANIMATIONS/animation-sine-wave.html
Coincidentaly (or not ), Nyquist found that you only need two sampling points to perfectly sample (and reproduce) a sinewave.
The Nyquist-Shanon sampling theorem that is the basis of digital audio states: "Exact reconstruction of a continuous-time baseband signal from its samples is possible if the signal is bandlimited and the sampling frequency is greater than twice the signal bandwidth.
The bandlimiting is what we have been talking about and is half the sampling rate. So with 44.1Khz sampling rate you can theoreticaly perfectly sample frequencies up to 22.05Khz.
The problem here is the bandwidth limiting. That means you need to let through all the frequencies (sound) up to the band limit (half the sample rate) but absolutely no frequencies (sound) above that limit. In audio terms that means a very steep filter.
In practical terms, there is no infinitely steep filter without any artefacts. So what happens is that the filter starts working at arround 20Khz and by the time it reaches 22.05Khz (or half of whatever the sample rate is) it doesn't let any sounds through. This filter is one of the weak points of cheaper converters.
If you have any doubts about the frequencies above 20Khz, here is a simple experiment you can try if you have 96Khz (or more) capable converters: Record an instrument that has frequencies above 20Khz. (Most will do but good examples are bells, cymbals, tambourine etc) or take a well produces 96Khz mix.
Take this mix or recording and sample rate convert it to 48 Khs with a good sample rate converter. (You can get the free version of R8Brain from http://www.voxengo.com). Now load the original 96Khz file with the newly converted 48Khz file into your sequencer. Make sure they both start at the beginning of the timeline (they need to be perfectly aligned!) and reverse the polarity of the 48Khz file. (Also called phase inversion). Send both track at full level to a bus of the master output. You will notice that if the original file had sound above 20Khz, the output/bus meter will show there is signal. But can you hear it? If the experiment was performed correctly, you won't hear anything!
The other issue with sampling theory is that to reproduce the original sampled wave EXACTLY, you need infinite bit depth. Anything less than infinite bit depth will cause distortion of the wave form. This is called quantization distortion.
Normaly the quantization distortion is dependant on signal and the sample rate. This distortion, when audible, is unpleasant. Luckily. there is a way to decorrelate the distortion from the signal and sample rate. This process is called dithering.
Dithering is a kind of randomization of the quantization grid used for the sampling. In practise what happens is that random noise (Triangular Probability Distrubution Function noise to be exact) is added to the signal so that instead of randomizing the quantization grid, we slightly randomize the signal.
Random signals are perceived as white noise. So adding dither increases the noise floor slightly but this is better than signal dependant quantization distortion which sounds ugly. (We are very used to ignoring noise as it is everywhere in nature). Also the peak level of this random noise is lower than the peaks of any potential quantization distortion. We have a win win situation!
So then the next question is, how many bits do we need? Well as allready mentioned in the previous post, we only need to cater for the range that we as humans can hear (without going deaf).
A jet fighter at close range produces about 140 dB while silence (basicly the sound of air molecules bouncing arround in our ears) is 0dB. So that tells us that we need about 140 dB of dynamic range.
24 bits gives us 144 dB of dynamic range! Hurray hurray! (See my note in the previous post about converters never actually achieving true 24 bit).
So now our dither noise (and ambient noise and electronics self noise etc) is down some 130 dB below the maximum level we can produce.
So there we have it. 44.1Khz sampling rate and 24 bit bit depth.
Ideally, for filter slope reasons, a bit more than 44.1Khz would be nice but the best converters these days are there allready. How is this possible?
In the real world of modern converters, the signal isn't sampled at 44.1Khz (or 48, 88.2, 96...) It is sampled at 64 or 128 times that rate! So we are actually sampling the analogue waveform at 2.8Mhz or more!!! As you can imagine, this doesn't need a very steep anti aliasing filter any more. (That filter used to limit the bandwidth, remember?) We only need to remove frequencies below 1.4Mhz! Easy peasy.
Once the signal is in the digital domain, it is decimated down to the actual rate we are going to use, 44.1Khz. These digital decimation filters when well designed will do their job without any audible artifacts in the audible range.
So back to the point of this post, with good converters, 96Khz is allready overkill. 192Khz is marketing bullshit and that is understood by anyone that understand sampling theory and modern audio converters. Yet the marketeers will pray on the ignorance that is rife within the audio market...
Further reading: http://www.lavryengineering.com/documents/Sampling_Theory.pdf
If you really want to learn more about all the details of digital audio I recommend "Digital Audio Explained (for the audio engineer)" by Nika Aldrich. It doesn't go too deeply into the math yet gives clear and concise explanations of how this all works.
UnderTow
|
|
|
|
shamantrixx
Started Topics :
7
Posts :
549
Posted : Jun 24, 2007 07:00
|
Oh here you are! Good old Undertow with a classical piece on sample rates and how stupid are all others who dare to believe any different than what he says.
Do you have any pleasure in life beside calling someone a no brain? I guess not!!!
"It occurred to me by intuition, and music was the driving force behind that intuition. My discovery was the result of musical perception"
Albert Einstein, speaking about his theory of relativity |
|
|
|
Elad
Tsabeat/Sattel Battle
Started Topics :
158
Posts :
5306
Posted : Jun 24, 2007 07:52
|
hehe i bet he have some pleasure deal with audio!!
tnx undertow , great explain
nice to know the resolution term is incorect when it comes to this..
but one Q , if i notice the better quality 48khz to 44.1khz sampling rate , is it simply bad convertor or placebo ??
www.sattelbattle.com
http://yoavweinberg.weebly.com/ |
|
|
|
Upavas
Upavas
Started Topics :
150
Posts :
3315
Posted : Jun 24, 2007 10:06
|
Now this is an explanation like out of the book. With this I can live...
And yes the quality of the converters is important. However most converters in the mid price range do just nicely nowadays, and as to the sample rate as being less important, I don't think so and both arguments, yours and mine have their right. As to the reason why ipod users crank their volumes, yes and no. When they are listening with headphones yes, when they are listening with a speaker system no. Just recently I played a set at Organix and remember the dj after me saying he had trouble keeping up with the overall volume and was wondering how I had the system that loud without distorting everything. Came out he was playing mp3's and I was playing wavs... particularly the bass was what made him trouble. Go figure...
DCPCM or Delta Pulse Code Modulation is generally being used in mp3 compression, ADPCM is what you get in video games, 48k sampling rate in the general movie industry with a tendency to go to 192, especially now with newer technology and new new surround sound modes and now that computers have finally become a bit faster again. And yes, with the right lowpass filter any aliasing gets safely removed. I still feel that 48k like dat sounds nicer on a big system than cd 44.1k... oh and who told you that film data is data compressed before it reaches the audience?
No way man. My boss would kill anyone doing such a thing... as a matter of fact he would probably be really pissed off...hahaha...
Also remember that when you change the sampling rate digitally (not using DA/AD converters the sound will be out of sync with the original... meaning faster or slower depending which way you go...which always is a pain in the @$$ when you have to convert sound with film to tv format... especially in North America where we have 29.97 fps as opposed to the nice even 25fps they have pretty much everywhere else in the world... Most audio for tv film actually is recorded with a sampling rate of 48048 khz, so when it is reduced by .1% which has to be done in North America because the film itself is reduced in Speed by.1% to match the damned 29.97 frames per second...it will be playable on most systems (48k)...It is so much easier to convert when you have even numbers... as to analog recording, the recording is always more accurate than digital recording, the reason why we use digital is because of the static noise that comes with analog recording, the hiss on a tape if you will. Since we don't have an option for infinite sampling rates, which you explained well above the recording can never be as accurate as analog, where we get a straight 100% recording. What are you referring to when you speak of analog format btw. ? You mean the device (Tape, Nagra etc.) ?
Upavas - Here And Now (Sangoma Rec.) new EP out Oct.29th, get it here:
http://timecode.bandcamp.com
http://upavas.com
http://soundcloud.com/upavas-1/ |
|
|
|
|
