st31276a wrote on 2026-03-20, 14:59:

High sampling rate is beneficial for recording, as it relaxes the requirements for the analogue anti aliasing filter on the inpu […]
Show full quote

High sampling rate is beneficial for recording, as it relaxes the requirements for the analogue anti aliasing filter on the input side significantly.

No human can hear the difference between a straight wire and a good 44.1kHz 16-bit AD/DA section in ABX testing, unless you set the gain insanely high and listen to the noise floor characteristics in the absence of signal.

Once sampled at 192kHz, a digital brickwall filter can be applied cutting everything above 20kHz, whereafter 3/4 samples can be discarded, resulting in a 48kHz stream. No difference will be measurable within the audible range of the spectrum.

16-bit sample precision is sufficient for storage and playback of normal audio signals too, as modern multibit delta sigma DA converters shape the quantization noise in a way that it can be filtered out, getting noise floors as good as -118dB, where running it straight would otherwise set the noise floor at the quantization level of -96dB.

Generally speaking, yes, but not entirely.
The DAC’s low-pass filter must have a linear phase and constant group delay (a Bessel filter) to minimise its impact on the signal shape.
By it has a slow roll-off.
Therefore, the sampling rate is set as high as possible to ensure the necessary suppression of the DAC’s carrier frequency using an analogue post-filter.
To achieve this, a digital upsampler (filter) is used within the DAC itself, which performs interpolation.
Increasing the original sampling rate allows the interpolated (fictional by the digital filter) values to be replaced with real ones.
A 192 kHz sample rate is more than enough for modern DACs to function properly (they typically operate at a 384 kHz conversion rate), i.e. 2x interpolation.
384 kHz is the upper limit for a reasonable sample rate.
I have no idea what the marketing people will come up with next.
Increasing analogue bit depth (ENOB) is far more difficult than increasing the sampling rate.
Only a handful of end products have actually reached 20 bits.
The Asus Essence STX measures in at something like 12-bit.
Although the maths is cheap enough, there’s nothing stopping anyone from making a DAC that accepts 64/128-bit inputs data.

https://www.audiosciencereview.com/forum/inde … ound-card.4915/
When designing analogue circuits, the fact that a designer has copied the technical specifications from the components datasheet onto the his devices page does not in any way imply that they have actually managed to achieve those specifications in their end-user device.

Interesting observation about the Bessel filter on the output.

With the upsampling/oversampling happening in the modern DA converter (as opposed to clocking out some samples through a resistor network at sampling rate) the sampling rate of the digitally stored signal does not need to be as high, as the signal should not contain components above 20kHz anyway.

I wonder if the pass band phase characteristics of a garden variety Butterworth is audible at all. Moving phase and moving your head physically a couple of inches should have about the same effect.

I totally agree that the advertised bit depths and qualities of equipment is mostly meaningless. That’s why I saly, if you can actually manage 16 bits from medium to air, you would have an exceptional experience.

st31276a wrote on Yesterday, 06:50:

With the upsampling/oversampling happening in the modern DA converter (as opposed to clocking out some samples through a resistor network at sampling rate) the sampling rate of the digitally stored signal does not need to be as high, as the signal should not contain components above 20kHz anyway.

The audio data itself does not, of course, contain any data above 20 kHz.
However, the delta-sigma modulators in DACs operate at a frequency that is a multiple of the sampling rate.
And this frequency will, in any case, be present at the DAC output.

The main problem for designers, including Creative, is that they simply copy the datasheet word for word.
Even though the approaches are constantly changing.
Analog Devices, for example, has changed its approach in its datasheets three times regarding the connection of digital and analogue ground planes. 😀