Modartt user forum » Pianoteq user forum » How it can affect me recording 48 kHz instead of 192 kHz?
I mean the real facts. Not just "use it on mixing stage not to lose information after processing".
I would admire some examples you experiensed from working with less sample rate and what it costed if you did.
Thanks.
Here you can find some answers:
https://www.youtube.com/watch?v=-jCwIsT0X8M
Here you can find some answers:
Thanks, appreciate
That is one of the best and clearest descriptions that I have seen of analog curve fitting, Nyquist frequencies, mathematically produced noise, etc.
Thanks!
I took a 5500Hz sine wave and made a smooth slide down to 5512,5Hz. I did this using sampling rates of 11025Hz, 22050Hz and 44100Hz. Here's what happens:
It can be seen that approaching half the sampling frequency, you accurately transmit information about the signal frequency, but you acquire distortions in the phase and amplitude of the signal. It can be seen that using the common sampling rate of 44100Hz, there are slight noticeable amplitude distortions even around 5512Hz. There is information about this in this video:
Therefore, I believe that it takes at least 5 samples per wavelength for reliable signal transmission. I believe that phase and amplitude distortions can ruin your spatial experience in stereo signals that have been digitized at an insufficient sampling rate. I can still hear 16000Hz and I find that 24 / 96kHz is adequate for music playback(necessary minimum). Also, for physical modeling in the time domain using the finite element method, it is recommended to create a mesh of the model, in which there will be at least 5-6 spatial elements per wavelength for an adequate solution. Also, ears suggest that 96kHz recording is more fluffy and spatial (if all stages of delivering sound to the end user were followed with due attention). .. a couple of cents ..
Hello @scherbakov.al, I find the video you posted very interesting. I don't understand a single word of russian, but I've seen the waveforms on the oscilloscope. Anyway, I believe it all depends on the DAC quality, since Nyquist theorem only asks the sample rate to be twice the maximum frequency you want to record. I did the same experiment some time ago, and waveforms of 20 kHz were perfectly reconstructed at 44.1 kHz. It is worth mentioning that the waveform is not reconstructed only sending the voltage matching the recorded value, the signal is the sample value convoluted with the sync function, ideally from -infinite to +infinite, but as sync function trends to zero for positive and negative values, in practice DACs use smaller time windows and approximated functions.
If I have time I will repeat the experiment and post it here.
Edit: I've just found a video wich demostrates what I said: https://www.youtube.com/watch?v=pWjdWCePgvA