Fox2232:

Are you by chance having 192kHz as default sample rate set for soundcard? Try with less, I experienced few games in past having troubles with it. One would not even start.

Irenicus:

WTF. Why would you ever try and use 192khz as default? Even the highest of high end A/V receivers can't handle that. It's very niche. 48khz is what you need for games, anything more is not supported by 99% of games.

Fox2232:

WTF. Why would you ever try and use 44khz as default? Even the highest of high end A/V receivers can't handle that. It's very niche. 11khz is what you need for games, anything more is not supported by 99% of games. - - - - But on serious note. At what sampling rate do you want to mix few dozen of sounds when each can start with microsecond offset from each other. Since microsecond offset implies time unit of 1 milionth of second, you want 2MHz sampling rate to prevent loss of information. There is difference between output quality and mixing quality. You could as well say that older person does not hear sounds above 16kHz, so they should use 32kHz sampling rate, right? Wrong!

Noisiv:

You are confusing frequency domain behavior of a system with its time domain response. Please get yourself acquainted with the sampling theorem. A sampled waveforms contains ALL the information without any distortions, when the sampling rate exceeds twice the highest frequency contained by the sampled waveform. Do note that the theorem is not dealing with approximations. It states a necessary condition for the faithful representation, ie. a perfect reconstruction of the signal.

Fox2232:

FYI Having two 1Hz signals mixed with 0.5s delay makes them 2Hz signal requiring double sampling rate. to prevent loss of information.

But on serious note. At what sampling rate do you want to mix few dozen of sounds when each can start with microsecond offset from each other. Since microsecond offset implies time unit of 1 milionth of second, you want 2MHz sampling rate to prevent loss of information

Fox2232:

So, when you have half dozen of signals each at same sampling rate, mere time delay is going to increase requirements on sampling rate to prevent information loss.

Fox2232:

For output device, nobody really cares, but mixing... That's why there is still analog mixing in studios.

Noisiv:

So you are willingly choosing to simply disregard a well established theoretical work and follow your intuition instead. OK lets follow the intuition: No sh1t? That's exactly what Nyquist theorem says: 2 * 1Hz = 2Hz sampling rate needed. But that's not all you had said in your previous post, and it's not what you're saying all along. You're saying a tiny time shift somehow requires a sampling rate increase. The smaller the time shift, the bigger sampling rate increase needed: Forget everything and just think from a informational point of view how stewpid would be that the one time introduction of a constant (a timeshift,a one dimensional scalar) should require a stupendous amount of the sampling rate increase. ---> It doesn't. It's just a time shift ie: Omega*t -> Omega (t+ TimeShift)

Noisiv:

Again: Nyquist sampling theorem is NOT an approximation. It states that in order to perfectly reconstruct the signal you need twice the bandwidth of the highest frequency.

Fox2232:

Yes, I am saying that if you have freaking 2 signals shifted by 0.00000000001s, there is no way to mix them without loss of information of F*ing 44kHz sampling rate. Simple as that is, that's reality.

Fox2232:

Fact is that you know words, but do not understand model. that one beat mixed with another delayed by 0.001s really makes it 1000Hz signal. Have beat mixed with another in 0.000001s, and it is 1MHz. Moment you do not have control over delays nor source sampling rates of each signal, you need to go up. A lot as you could have realized from 10 & 11Hz mixup.

Noisiv:

y1+y2 = A*sin(W*t)) + A*sin (W*t - phase) I have just done it. Principle of superposition Q.E.D You are confusing FFT original domain frequency with its representation in the frequency domain. But even if you are not acquainted with FFT or sampling theorem it should be intuitively obvious that adding phase shifted identical waves should not be expensive. It should not be expensive from the informational or the computational or the storage needed point of view; you're merely introducing a simple constant and then adding the waves. In other words: it's cheap as f*k and it does not require stupendous increase of the needed bandwidth

Fox2232:

Well, do yourself a favor. Take one of those lovely math modeling applications. Way you write hints that you have some at hand or accessible. Take simple small part of song. Do SRC from 44.1kHz to 44.101kHz (by 1Hz up). There should be no loss, no distortion right? Then repeat by 1Hz again and again till you are at 44.2kHz. Finish by SRC to 48kHz for comfortable playback on any player. Compare the two. Or compare them in mathematical way. Take that original 44.1kHz and do SCR directly to 48kHz. If there is no loss of information then 100step conversion will be exactly same as 1 step conversion.

Noisiv:

To what purpose? Besides, I have no idea what you've just said. Your idea that adding time shifted signal on top of the identical signal leads to explosion of the required bandwidth, eventually going to infinity as the time shift approaches zero, ie the idea that adding two almost identical signals would lead to requirement of infinite bandwidth, is in disagreement with every logic and intuition. Not to mention with the fundamental theorem of information theory, with superposition principle, with sampling theorem etc etc. Adding two identical harmonics which are infinitesimally time shifted results in the harmonic of exactly the same frequency, and almost double the amplitude! A sin(kx - ωt) + A sin(kx - ωt + φ) = 2A cos (φ /2) sin (kx - ωt + φ/2) ~ 2A sin(kx - ωt) More generally adding two identical harmonics which are time shifted, results in the harmonic of the same frequency, and the amplitude modified by a factor of 2*cos (φ /2). Not to explosion of the required bandwidth.

Fox2232:

Having 2 pulses/signals that close together as in #31 leads to very high frequency. In that case well above regular sampling rates.

Noisiv:

The dude forgot to do his filtering. Page 19: Filtering to avoid aliasing Page 21: Over sampling http://www.atomhard.byethost24.com/pub/Sampling_Theory.pdf

Fox2232:

In less than minute this video explains clearly that to capture and reconstruct 10Hz signal you need sampling rate to be 20Hz (2x of sampled content), but at same time it is not sufficient for proper representation of all other frequencies.