Sample rate/Bit depth blind test

[roman.i]

Hello!

I'm trying to understand if there is a real difference in sound quality between different sample rates and bit depths.
Since I'm already biased (I know the position of the best quality), please give you opinion on this.
Here is the mix, every two beeps are bounced in a different quality.

https://srv-file10.gofile.io/download/rfvpeQ/mix.wav
(mirror https://file.io/2fi1ZA)

I also added a poll, so you can vote what sounds best.

Thanks!

[Burillo]

what are you testing here, exactly? project sample rate, or audio file sample rate? mixing/rendering the entire project in 96k and then downsampling to 44k is not the same as mixing/rendering entire project in 44k. there will be no audible difference between original and downsampled audio file, but there may be a different between processing in 44k and 96k because a lot of plugins will sound different depending on sampling rate (less need for anti-aliasing filters, no cramping for badly coded plugins, etc.).

so, exactly what are you comparing?

[roman.i]

what are you testing here, exactly? project sample rate, or audio file sample rate? mixing/rendering the entire project in 96k and then downsampling to 44k is not the same as mixing/rendering entire project in 44k. there will be no audible difference between original and downsampled audio file, but there may be a different between processing in 44k and 96k because a lot of plugins will sound different depending on sampling rate (less need for anti-aliasing filters, no cramping for badly coded plugins, etc.).

so, exactly what are you comparing?

I'm comparing the same exact "project" once in 44.1Khz exported 16bit and 24bit WAVs, then changed it to 88.2Khz, exported it in 16bit and 24bit. Sources are plugins, so I guess they're changing accordingly when the project sample rate being changed.
So there are 4 samples, 44.1/16, 44.1/24, 88.2/16 and 88.2/24 exported to WAV.
Then, another project in 88.2/24 combining these four WAVs exported to 88.2/24 WAV file.
So actually I'm comparing if there any noticeable difference in these four samples.

I hear some slight difference in the stereo image being more "fluid" in 88.2Khz in comparison with 44.1Khz, and I don't hear any difference between 16bit and 24bit.

[Tj Shredder]

The result you hear is reasonable. No need to verify...
That you can’t hear a difference between the bit depth is also clear. A finished mix simply does not need headroom and the noise floor of your listening environment would mask any difference if there would be any anyway... (you could measure it though, but only in silent passages...)

[Burillo]

what you hear is exactly what you'd expect to happen - processing at different sampling rate yields slightly different results depending on the plugins that you're using (and the number of them), while mixdown at different bit depths is not going to affect anything (in fact, you can pretty much get away with doing a mixdown into 12-13bit file and still not hear any differences).

however, if you render your 88K project and then downsample the mixdown to 44.1K, you won't hear any differences. so, it's not the format that's creating the difference, it's the processing.

[roman.i]

So any reason why recording done today in 24bit and not in 16bit if we can't hear the difference?

[WasteLand]

So any reason why recording done today in 24bit and not in 16bit if we can't hear the difference?

headroom...

[Gamma-UT]

So any reason why recording done today in 24bit and not in 16bit if we can't hear the difference?

It's about having a margin of safety.

24bit recording means, in general, the noise floor of the equipment is above the noise floor of the digitised audio - that noise may also be less objectionable than quantisation-error noise. You can get away with less amplification during recording to get the signal into the sweet spot and so less risk of even more objectionable hard clipping.

Also, noise introduced by quantisation is generally pushed down to inaudible levels by working at 24bit or higher (yer average DAW is doing 32bit or 64bit floating point for anything that isn't printed) up to final conversion. Processing repeatedly at 16bit will tend to introduce more obvious quantisation noise.

[Burillo]

So any reason why recording done today in 24bit and not in 16bit if we can't hear the difference?

don't confuse recording with playback. it makes no difference for playback but it does make a difference for processing, especially if you're piling on gobs of distortion and/or compression.

[codec_spurt]

wrong.

[burnt circuit]

Hello!

I'm trying to understand if there is a real difference in sound quality between different sample rates and bit depths.
Since I'm already biased (I know the position of the best quality), please give you opinion on this.
Here is the mix, every two beeps are bounced in a different quality.

https://srv-file10.gofile.io/download/rfvpeQ/mix.wav
(mirror https://file.io/2fi1ZA)

I also added a poll, so you can vote what sounds best.

Thanks!

If it really mattered, people wouldn't choose to listen with ear buds, laptop speakers, phones, desktop speakers and televisions over a flat response room with flat response monitors or headphones.

Most of the world population listens to streaming .MP3s, while others debate about which sample/bit rate matters (to them)... usually the engineer types.

44.1k is more then enough... (they're listening to the production.)

[Burillo]

basically, given a 24-bit noise floor, the dynamic range available is the difference between a whisper and a jet engine. even the most dynamicest music ever (the poster children of "muh dynamics" and "mun subtleties" are usually things like jazz and classical) does not approach these kinds of dynamics. it doesn't even approach 16-bit dynamics. you can test it yourself by taking the most dynamic recording you have, with all of its subtleties and everything, and run it through a bit crusher plugin. you'll only start hearing audible differences at 12-13 bits (and with a lot of music made today, even 10 bits is enough).

[jonljacobi]

What you observe is largely correct. Some processing will be more accurate at higher sample rates if the plugin is not already oversampling (using a higher sample rate) internally. This is because of reduced latency in the stream.

You won’t notice the difference in bit depth upon recording or maybe at all, but the larger numbers and therefore greater accuracy reduce quantization errors which can add up to digital noise.

[roman.i]

Thanks, this explains a lot.
I had another question about bit depth and audio interface, I'll squeeze it here.
My audio interface - Presonus Studio 26c, which is considered as a home recording level sound card.
It got 114dB DAC/ADC, 110dB range in the instrument input(which I use for recording), and 108dB in the line output(for the monitors).
108dB in bit depth is around 18bit depth, which covers the range for playback,
but for recording it doesn't cover the 24bit range and there are audio interfaces that have 120dB range for both recording and playback. My monitors have 118dB dynamic range, if this matters.
So, if I upgrade the audio interface to a high end one then it will only give more room for the recording, and not affect the playback at all.
Is this correct?

[Burillo]

it may affect playback as there are other things to consider (THD, jitter, etc.).

that said, home recording interfaces of today are top grade $10000 systems of yesteryear. if you have a problem, it isn't with the quality of the interface.