I'm sure this is a silly question but
Is there any point of converting cd audio files 16bit 44.1kHz (wav-->FLAC) to 48kHz? (Other than you have software that can't play 44.1kHz which seems pretty unlikely)
There is no increase in music clarity, fidelity or anything like that, correct? Once down-sampled to 44.1kHz for CD purposes from whatever it was mastered at, the data is lost, there is no point in just adding zeros back in?
Or is the sampling rate based on a calculation done to other stored audio data in the file and thus resampling actually will have a real positive effect?
Thanks
If you print a book on larger pages, do you get more words?
> there is no point in just adding zeros back in?
Fun fact: upsampling is not adding zeroes in, which would positively ruin the audio. The smooth waveform that comes from the 44.1KHz samples is recalculated with more samples, which doesn't add any new information.
Theoretically: no.
Practically: it depends. Some DACs perform a bit better at multiples of 48 kHz. You can also take control over the anti-imaging filter.
I wouldn't convert the files. I'd resample during playback if you need to.
Thanks for the info...I was pretty sure that is how it works but like anything else in life had a debate with someone at work.
That someone is surely a person to argue the hot water will boil faster than cold water debate, lol!!
thanks again
Ben
xnor: Thanks for the response.
xnor or anyone else: I have some 24/96kHz files that I'm gonna downsample to 16bit 48kHz and am going to use foobar cause its free and I have it. I've installed the SoX component, should I just use the normal version, use 95% passband (what is a passband by the way and when would use use lower), 50% (linear) phase response and enable aliasing-imaging.
Or is one of the other two mods better?
You mention take control over the anti-imaging, what does that mean?
I'll leave the 44.1kHz CD files alone...seem to play fine everywhere.
Thanks
Ben
Theoretically: no.
Practically: it depends. Some DACs perform a bit better at multiples of 48 kHz. You can also take control over the anti-imaging filter.
I wouldn't convert the files. I'd resample during playback if you need to.
D/A converters usually employ a low pass filter near half the sampling rate. If you resample 44.1 to a higher sampling rate, then the converter's filter will operate at higher frequencies instead of ~20 kHz. You've effectively bypassed it and replaced it with your configurable resampler's filter.
D/A converters usually employ a low pass filter near half the sampling rate. If you resample 44.1 to a higher sampling rate, then the converter's filter will operate at higher frequencies instead of ~20 kHz. You've effectively bypassed it and replaced it with your configurable resampler's filter.
I suppose that's true only for non oversampling DACs. or am I missing something?
95% passband (what is a passband by the way and when would use use lower)
An ideal lowpass filter to prevent aliasing would perfectly preserve 100% of the frequencies below Nyquist (half the sample rate) and would discard 100% of the frequencies at/above that rate, without introducing any noise. Real-world filters, even digital ones, make a tradeoff where they perfectly preserve 95% (for example) and then begin rolling off to zero at Nyquist. In other words, the uppermost 5% of frequencies are attenuated on a slope that reaches silence at the very end. This is like using a graphic equalizer with the lower bands all at zero (no change) and with the bands from Nyquist and up set to negative infinity (silence).
When the slope is too narrow, there's a risk of audible "ringing", although I'm not sure what that actually sounds like. I believe 5% is considered a reasonable margin. When downsampling to 44100 with 95% passband, frequencies up to a little bit under 21 kHz would be perfectly preserved before the rolloff begins.
There's no harm in using a softer slope, although you then risk attenuating audible bands.
I personally don't understand it in any more depth than this. Monty's videos at xiph.org are a big help.
D/A converters usually employ a low pass filter near half the sampling rate. If you resample 44.1 to a higher sampling rate, then the converter's filter will operate at higher frequencies instead of ~20 kHz. You've effectively bypassed it and replaced it with your configurable resampler's filter.
I suppose that's true only for non oversampling DACs. or am I missing something?
Its true either way, in that you'll basically be substituting the DAC's normal oversampling filter for your own oversampling filter combined with whatever one it uses at the higher sampling rate (which will probably do very little given that your oversampled file will have no higher frequencies anyway).
I guess you could do this if you thought there was some problem with the DAC's digital filter, although I think its pretty unlikely to be very helpful in practice.
Well, there will be image frequencies even if the signal has been sufficiently low-passed, though the DAC will likely be over sampling, rendering the issue moot.
This is a great help, thanks.
So in theory if downsampling to 48kHz, a 90% passband should be fine...(48000/2*.95) = 21600 which is well above my and probably anybody's hearing?
Thanks
Ben
95% passband (what is a passband by the way and when would use use lower)
An ideal lowpass filter to prevent aliasing would perfectly preserve 100% of the frequencies below Nyquist (half the sample rate) and would discard 100% of the frequencies at/above that rate, without introducing any noise. Real-world filters, even digital ones, make a tradeoff where they perfectly preserve 95% (for example) and then begin rolling off to zero at Nyquist. In other words, the uppermost 5% of frequencies are attenuated on a slope that reaches silence at the very end. This is like using a graphic equalizer with the lower bands all at zero (no change) and with the bands from Nyquist and up set to negative infinity (silence).
When the slope is too narrow, there's a risk of audible "ringing", although I'm not sure what that actually sounds like. I believe 5% is considered a reasonable margin. When downsampling to 44100 with 95% passband, frequencies up to a little bit under 21 kHz would be perfectly preserved before the rolloff begins.
There's no harm in using a softer slope, although you then risk attenuating audible bands.
I personally don't understand it in any more depth than this. Monty's videos at xiph.org are a big help.
Pointless if the original file is 44.1 in the same way showing a 720dp film on a 1020dp screen won't give you a better image. The only vaguelly possible scenario might be if you have a dodgy sou card that gets along better with 48 than 44.1
D/A converters usually employ a low pass filter near half the sampling rate. If you resample 44.1 to a higher sampling rate, then the converter's filter will operate at higher frequencies instead of ~20 kHz. You've effectively bypassed it and replaced it with your configurable resampler's filter.
I suppose that's true only for non oversampling DACs. or am I missing something?
You are missing something. A proper DAC has to apply a low pass filter whose corner frequency is just below half the sample rate. Before oversampling this was often a multistage analog filter.
This is a great help, thanks.
So in theory if downsampling to 48kHz, a 90% passband should be fine...(48000/2*.95) = 21600 which is well above my and probably anybody's hearing?
Yes. The frequency above which cutting off all response with a very steep filter still can't be heard can be as low as 16 KHz and often is.
The frequency band between the Nyquist frequency (half the sample rate) and the high frequency cutoff is called the Transition Band. This band is typically about 2 KHz or more wide in reaql world DACs.
Really steep filters can reduce the transition band to a few hundred Hz. However there is some controversy over whether or not really narrow transition bands can cause other audible artifacts.