Digital Facts and realities?

[Clive]

Trying to find a simple explanation of what each 16 bit sample means is surprisingly difficult. Is each sample for the whole frequency spectrum across both channels? Is there one sample per each frequency? This is really basic stuff. There must be a dummies guide somewhere.

[mikmas]

I expect that is true is reality but it seems 44,100hz isn't overkill, it's just enough but here I'm conflating frequency extension with the gaps between samples.

I think it has been mentioned a number of times, but worth repeating. The sampling process is continuous - so where would these 'gaps' that are continually being alluded to come from?

It's not like the computer gets really tired between samples and needs a little sit down to recover ..

[NRG]

There still seems to be some misunderstanding, there is nothing between the samples. The signal is bandwidth limited. The xiph.org video explains this. Now if you believe that there are signals above 20khz that you can hear or affect the signal below 20khz and you want to capture them, then you need to raise the Nyquest frequency to encompass those signals, let’s say up to 24Khz...then you need a Nyquest sampling frequency of 48khz and the analogue signal to be bandwidth limited to 24Khz.

[Clive]

I think it has been mentioned a number of times, but worth repeating. The sampling process is continuous - so where would these 'gaps' that are continually being alluded to come from?

It's not like the computer gets really tired between samples and needs a little sit down to recover ..

But does continuous mean every single millisecond is accounted for? If so then even the lollipop diagrams are not correct. Indeed it's closer to analogue in terms of sampling but with restricted digital levels....though less restricted than analogue!

[struth]

one of the issues with 24/96 or higher is the room it takes up. It may or may not improve things, depending, but its subjective. I usually upsample but i fear there is little poit and have reduced this. I do quite like the dsd effect of upsampling that way tho. Not sure why, nor do i especially care tbh. hearing results is results to me

[Firebottle]

Trying to find a simple explanation of what each 16 bit sample means is surprisingly difficult. Is each sample for the whole frequency spectrum across both channels? Is there one sample per each frequency? This is really basic stuff. There must be a dummies guide somewhere.

A 16 bit sample is taken of each channel, 44100 times per second.

That's two 16 bit samples taken every 22.6 microseconds.

Imagine any music waveform you like (band limited) and the instantaneous level is taken every 22.6 microsecond, on a zero to 65536 scale. This gives the 16 bit sample.

[NRG]

Trying to find a simple explanation of what each 16 bit sample means is surprisingly difficult. Is each sample for the whole frequency spectrum across both channels? Is there one sample per each frequency? This is really basic stuff. There must be a dummies guide somewhere.

You need to sample each channel L+R individually for stereo. At the sample point the signal is converted into a number, for 16 bit that’s anywhere from +/- 32,767 Now this throws up an issue with ADC and that’s quantisation error where the sampled signal doesn’t fit exactly into the granularity of 16bits. This is where dither comes into play to add a randomisation to the sample point via the use of shaped noise to reduce this quantisation error. The video explains this as well.

[NRG]

Maybe also worth pointing out that this all refers to the ADC not the playback. As enthusiasts and music lovers we can do sod all about what happened in the ADC process, it’s out of our control. All we can obsess over or worry about is the playback side.

[Clive]

Thanks guys, that clears up where I was hearing conflicting statements.

[Jimbo]

At a rate of 44100 samples per second, how big would you expect any 'gap' to be?

As an experiment you might try cutting a hundredth of a second slice out of a guitar note wave form (using an audio editor)

Listen back to the result - do you hear the jump?

But the sampling rate may not be enough so when sampled at a higher rate you may be able to hear the difference

Have a look at hearing sensitivity and response to timing with lower 16 bit sampling rate. Higher sampling rates may be closer to the timing issues our hearing may percieve?