Audio Ramblings

The only way to truly verify it is to perform a double blind test (ABX test).

It's really hard to ABX test two different platforms. The primary problem is volume. Your brain will always pick the louder option.

Your brain can feel differences as low as 0.2 dB. Streaming services can differ in volume by as much as 40 times that number.

To properly perform an ABX test across platforms, you have to rip it properly and gain-match them to an accuracy of <0.1 dB.

This is really hard to do accurately. If you did do all that and setup a proper ABX test with good gear, the results will disappoint you. There'll be no difference.

No, you can't.

Go to abx.digitalfeed.net and do the test. Don't be disappointed when they sound exactly the same. Almost none of our population can tell the difference. But why?

It's an even smaller number that can tell the difference between 320kbps and lossless.

Lossless audio means every bit is preserved as it is. This is a must for producing and archiving music. But not for listening.

We've read that humans can hear between 20Hz to 20kHz. Can we?

As we age, our hearing capabilities get worse. The average adult can barely hear above 16 kHz.

Lossy codecs optimize for human hearing. It throws away all the data that >99.9% of the population can't hear. Even for the ones who can, it's incredibly difficult to distinguish.

Those who can need to know exactly what to look out for to tell them apart. They also need to have very good audio gear and be in a quiet, proper environment.

TLDR: Stick to Spotify. Don't switch solely because they offer lossless. It's a marketing gimmick.

You're shooting yourself in the foot. Hi-res lossless introduces noise at the upper frequency range. Let me start with high sample rates.

The kHz in audio files refers to the sample rate. If the file contains data all the way till 20 kHz frequency, the sample rate will at least be 40 kHz. Refer to the Nyquist–Shannon sampling theorem.

We've all read that our hearing range is 20 Hz to 20 kHz. It gets worse as you age. The average adult can barely hear past 16 kHz, and it dies out well before 20 kHz.

A transducer (device that converts energy from one form to another, electric to sound in this case) will always have some amount distortion at the lowest and highest frequency range.

When the audio has a very high sampling rate, the transducer might attempt to reproduce those high-frequency sounds. These high-frequency sounds, which are not audible to us, can interact with frequencies that are audible to us and introduce distortion. This was verified with a proper experiment, here's the paper.

This problem is amplified with very high sample rates, i.e. >44.1 kHz.

"Ok, but what about >16 bit audio? They have higher dynamic range!"

Yes, it does. But to actually hear the increased dynamic range, you have to play the audio at volumes loud enough that it will cause permanent hearing loss in a few seconds. In other words, you'll hear it for sure, but you'll never hear anything else for the rest of your life.

TLDR: Stick to Spotify. Lossless is a scam for listening.

"Why don't you upload delimited music in lossless formats?"

I've already established that you objectively cannot tell the difference between transparent lossy and lossless.

Lossy audio is still the better experience. Particularly, Apple's AAC codec. Let me tell you why.

Apple knows what they're doing when it comes to audio, we can all agree on that. They have their own codecs. ALAC is their lossless codec, and they have their own implementation of AAC. This, in my opinion, is the best lossy codec (sorry, Opus). This is what I refer to as Apple's AAC. They use a unique process to encode their AACs, they call it Apple Digital Master.

Audio should always be/are recorded with >44.1 kHz and >16-bit. It's best to have the additional headroom when recording/mixing/mastering. It's bad only for listening.

The conversion process to 16-bit, 44.1 kHz audio involves two processes: dithering (to reduce bit-depth) and downsampling (to reduce sample rate).

Dithering adds a little bit of noise to compensate for the distortion that's unavoidable in the process. But is it really unavoidable?

Apple converts higher bit depth audio to 16-bit without dithering. They mention it in Apple Digital Masters documentation. It's unclear how they do this, but I trust Apple. So should you. They have the best autists working on it. :P

Does this make a noticeable difference? Not for most people. But this entire thing is about "getting the best experience". Dithering adds noise. This doesn't. This is the best experience.

"But what about Bluetooth transcoding? I don't want a lossy to lossy transcode!"

Good point. AAC is just as good as lossless.

Apple's AAC is extremely resilient to generational loss. Even back in 2013, Apple's AAC had zero perceivable loss after transcoding an AAC file to AAC 100 times. 100. The quality remained almost exactly the same with no audible difference after 100 passes. You're just transcoding it once. And it's 2024. They've done incredible things in the audio space since 2013.

Bandwidth is an important factor, too. Lossless files are huge. If you're just listening to them, you're better off saving space/data.

TLDR: Don't use lossless for listening. Use Apple's AAC instead.