| Author
|
Sample Rate question
|
PoM
IsraTrance Full Member
Started Topics :
162
Posts :
8087
Posted : Feb 7, 2013 20:16
|
|
edited cause useless post |
|
|
astro_traveller
IsraTrance Junior Member
Started Topics :
13
Posts :
65
Posted : Feb 7, 2013 23:41
|
it is digital representation of the wave...
|
|
|
|
knocz
Moderator
Started Topics :
40
Posts :
1151
Posted : Feb 8, 2013 03:47
|
Quote:
|
On 2013-02-07 14:37, Colin OOOD wrote:
When you change the speed of a recording in your DAW, whether it's with a modulated delay, or a sampler, or any other way of replaying those recorded samples out at a different rate, remember that the samplerate of your DAW does not change, so you're still hearing a stream of samples at 44.1kHz or whatever. Whatever technique you're using to change the speed of your recording has absolutely no effect on the number of samples per second you're actually listening to.
|
|
Yes I realize that man, but still the technique must need to do some sort of dithering when slowing a sound down: it's feed an input (in mono) with a certain sample rate, and the DAW's expecting to receive the output with the same sample rate, but the effect must recalculate the input sample positions in order to achieve the slow down effect.
For instance, in one sample comes the value 0.1, and the next sample has a 0.2 -> slowed down by half, the first sample should be 0.1, the third sample 0.2 and the second sample needs some value in between. And it's this something that I'm referring to. In this example, no matter what sample rate chosen, the processor must do dithering, but with a really high sample rate this dither should be less noticeable. The slower the stretch, the more dithering that's induced, thus with smaller sample rates we risk hearing the dithering without going so far.
Other than that, it's like what Upavas says and I agree, when the sound is played, everything over 20 kHz is gone we can't hear it, the equipment doesn't play it, and the tune shouldn't have it -> therefor 44.100 Hz is enough for listening. I'm just defending that it can make a difference on the processing stage, in particular circumstances.
 Super Banana Sauce http://www.soundcloud.com/knocz |
|
|
|
Upavas
Upavas
Started Topics :
150
Posts :
3315
Posted : Feb 9, 2013 02:49
|
[quote]
On 2013-02-08 03:47, knocz wrote:
Quote:
|
Yes I realize that man, but still the technique must need to do some sort of dithering when slowing a sound down: it's feed an input (in mono) with a certain sample rate, and the DAW's expecting to receive the output with the same sample rate, but the effect must recalculate the input sample positions in order to achieve the slow down effect.
For instance, in one sample comes the value 0.1, and the next sample has a 0.2 -> slowed down by half, the first sample should be 0.1, the third sample 0.2 and the second sample needs some value in between. And it's this something that I'm referring to. In this example, no matter what sample rate chosen, the processor must do dithering, but with a really high sample rate this dither should be less noticeable. The slower the stretch, the more dithering that's induced, thus with smaller sample rates we risk hearing the dithering without going so far.
|
|
No!
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/ |
|
|
|
knocz
Moderator
Started Topics :
40
Posts :
1151
Posted : Feb 9, 2013 03:24
|
|
knocz
Moderator
Started Topics :
40
Posts :
1151
Posted : Feb 9, 2013 03:40
|
Lets see if you can visualize it this way: if you use a normal camera to make a film (movie), and play it into slow motion, the image will jump frame by frame. Shoot it with a high speed camera, and you can slow it down but still see a fluid image. With a super high speed camera you can slow it down as much as you can fluidly see light travelling!
In either situations the video is shown by the screen at about 60 Hz, and at 30 frames per second we perceive a fluid image.
Same thing goes for audio. If you have a higher sampling rate, you have a higher resolution, and with slowing down the samples the higher (faster) frequencies become slower and in the audible range.
Super Banana Sauce http://www.soundcloud.com/knocz |
|
|
|
Upavas
Upavas
Started Topics :
150
Posts :
3315
Posted : Feb 9, 2013 13:10
|
Quote:
|
On 2013-02-09 03:24, knocz wrote:
Why not?
|
|
Because the turning wheel of the car in the movie is not shown, at least not in newer productions.
Higher sample rates translate into higher audio frequencies being played back without aliasing, you still cannot perceive them because your ear is human (provided you have set an lpf at 20.000 hz frequency of at least -6dB per octave, a 44.100 hz sampling rate will be the same to your ear as 192.000 hz sampling rate)... .
If you lower the sample rate of a wav file the sound will still have the same tonal height, but it may be aliased... depending on the height of it's frequency...
Your analogy is flawed btw. because if the wheel of a moving vehicle was shown turning at 24rps in a cinematic movie format it would stand still, thus your eye can perceive it. I love lp filters ;-)
You can twist and turn your sound, but as long as the sampling rate in your daw is set at 44.1khz, that's what it is...
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/ |
|
|
|
knocz
Moderator
Started Topics :
40
Posts :
1151
Posted : Feb 10, 2013 22:11
|
Ok, then I'll show you.
I took a synth (Operator, with 4 oscillators FMing each other, so 1 feeds into 2 that feeds into 3 that feeds into 4, which is the only oscillator to go to the output), using only sine waves for oscillators. the first FM oscillator was 42 octaves high, the second 24 octaves high, the third 12 octaves high. I played a simple C3 note, and recorded it.
I took the recording and let it play for half a bar, than for each half bar I multiplied it by 2, making it twice as slow. I notice that I set the resample method to Repitch, so you'll have the same effect as slowing down a vinyl. I've done this for 2 bars, so you'll hear the original, than multiplied by 2, than by 4, than by 8. At the end and after the 2 bars, I let the sound go for half a beat to feed a tape delay, and automated the delay time from 1 ms to 1000 ms for another 2 bars.
I've done the recordings 6 times (was just going to do it twice, but I though what the heck I might as well do it a couple more times), each in different Sample Rates. This sample rate was consistent among each run, and at the end I rendered the track in 44.1 kHz, cause that's what we're going to hear at the end. In the master I put an EQ to LP and HP 40 Hz and 20 kHz, just so the super sonic frequencies aren't present in the final rendering mixdown.
You can listen to the results here:
22.05 kHz https://dl.dropbox.com/u/6226974/Audio/Sample%20Rate%20Test%2022.05kHz.wav
44.1 kHZ https://dl.dropbox.com/u/6226974/Audio/Sample%20Rate%20Test%2044.1kHz.wav
48 kHz https://dl.dropbox.com/u/6226974/Audio/Sample%20Rate%20Test%2048kHz.wav
88.2 kHz https://dl.dropbox.com/u/6226974/Audio/Sample%20Rate%20Test%2088.2kHz.wav
96 kHz https://dl.dropbox.com/u/6226974/Audio/Sample%20Rate%20Test%2096kHz.wav
192 kHz https://dl.dropbox.com/u/6226974/Audio/Sample%20Rate%20Test%20192kHz.wav
As we all know, you can hear dramatic changes comparing the 22.05 kHz file with any other file, as with 22 kHz sampling rate can only accurately store frequencies up to 11 kHz (Nyquist theorem -> sampling rate / 2).
If you listen to the rest of the files sequentially, you might not hear much differences, but they are there and are quite noticeable comparing the 44.1 kHz file to the 192 kHz file.
In the first 2 bars, where the synth output is samples, then multiplied by 2, 4 and than 8, you can hear some difference. And not only on the multiplied parts, in the original synth resampling in the beginning there is a big difference in the FM tone.
You can hear major differences in the end, with the tape delay effect. In the end, the 192 kHz file has much more of a low end compared to the 44.1 kHz, where at the end you can only hear a single harmonic coming down, going toc toc toc. In the 192 kHz file, the end of the file has a full low end frequencies.
As I stated before, with lower sampling rates we cant store higher (super sonic) frequencies. And although we cant actually hear them, it doesn't mean that the sound isn't affected by them, and that we cant make use of them. With the first two bars we are hearing the effect of higher harmonics modulating the last oscillator, affecting the tone. However, this tone is mostly affected in the higher frequencies, and with a small increase in the sampling rate it becomes unnoticeable. By multiplying the source, we are bringing down all of the frequencies in the file effectively in half, so everything that was in the 20 kHz to the 40 kHz now gets crammed into the 10 kHz to 20 kHz and comes into out audible range, and by multiplying more the same procedure goes on. In the last 2 bars with the tape delay effect, the upper limit of each sampling rate is exposed through time: since with a 44.1 kHz sampling rate we can only store frequencies up to 22.05 kHz, that top harmonic (22 kHz) is the highest being pulled down. At the end, after so much stretching, that top harmonic is the only one you can hear, resulting in the toc toc toc. However, with the 192 kHz sampling rate file, there are much more harmonics above the 22.05 kHz imposed by 44.1 kHz sampling rate, and in the end there are still many harmonics to fill out our audible frequency range.
Now, if I had used a HP and LP filter before recording (or used a microphone that naturally has a frequency limit), this wouldn't happen as the recorded source would be input limited to a specific range.
Now, take this information, or leave it. In most situations there isn't a difference between sample rates, I was just exploring one where it does, by the definition of the sampling theory, and in the exposed results we can perceive quite a big difference.
Super Banana Sauce http://www.soundcloud.com/knocz |
|
|
|
PoM
IsraTrance Full Member
Started Topics :
162
Posts :
8087
Posted : Feb 11, 2013 00:35
|
that don't mean you will notice the same thing with all synths,often that why they have a quality switch, some oversample and run at way higher sample rate .
i say buslhit as it depends many factors,how it s implemented ect...but for some process to compare to analog you may need 100 or 200 times the sampling rate we use or even more
|
|
|
|
PoM
IsraTrance Full Member
Started Topics :
162
Posts :
8087
Posted : Feb 11, 2013 01:07
|
edit ,read it wrong..but your test is a bit flawded cause more factors come into play than just the sampling rate.
about using higher sampling rate in many case ,best /easiest is to use tools that sound good at 44,1 to me |
|
|
|
Colin OOOD
Moderator
Started Topics :
95
Posts :
5380
Posted : Feb 11, 2013 01:52
|
|
Alien Bug
IsraTrance Junior Member
Started Topics :
27
Posts :
682
Posted : Feb 11, 2013 02:39
|
|
PoM
IsraTrance Full Member
Started Topics :
162
Posts :
8087
Posted : Feb 11, 2013 13:53
|
nice article sum it up well |
|
|
|
PoM
IsraTrance Full Member
Started Topics :
162
Posts :
8087
Posted : Feb 11, 2013 13:55
|
Quote:
|
On 2013-02-11 02:39, Alien Bug wrote:
a lot of old school hardware have 12bit and lower than 44,1kHz sr and most of people will say that it sounds better than newest 24bit, 96khz stuff. so higher not always mean better
sometimes is good to leave numbers away and start to trust our ears 
|
|
ahah yes i was thinking some old nords can sound far from analog lushness in the highs but we like them for this for example.
it remind me vintage/modern analog with some stuff sounding cleaner, people don t like it usually but technically it s better, lower thd, more bandwitch ect..
i m sure lot of old digital stuff will be classic, i think it is already.. when i listen a old nords, it s like listening to a sh101 to me..i feel a bit the same thing, it sound kind of vintage/classic . |
|
|
|
Upavas
Upavas
Started Topics :
150
Posts :
3315
Posted : Feb 12, 2013 10:38
|
The clocking gets more accurate with higher sampling rates, ok, that is if you are clocking via sample rate... that said, are you??? Or are you using midi clock timecode? Which most of us are... And even if you are clocking via sample rate, I would suggest that 44.100 times clock adjustment per second is enough, do you really need it more precise? No, the reason, you can't tell a difference!
Fixing the filters
is arbitrary at best, most of us don't hear more than about 18khz frequency, and many psy heads had their ears ruined with too much spl, so some won't even go over 15khz. But for the sake of what a new born baby would hear, it is hairsplitting at best... he suggests that himself...
Oversampling for dsp
name me 2 processors that neglected to include oversampling in their code and which you use today... I don't know of any in this day and age...
When More is Better: Converter Design
Is it audible? I doubt it will make any difference at all...
On another note, the higher the sampling rate, the more computing power needed. If you really want it to sound better, get a better converter, provided you are actually converting something... as suggested in his post...
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/ |
|
|
|
|
