Hacker News new | past | comments | ask | show | jobs | submit login

In the box processing uses 32-bit or 64-bit float. Fixed-point DSP processing was a thing maybe ten years ago, and even then the standard was 56-bit. 24-bit is nowhere close to good enough for ITB DSP.

That aside - the bit depth part of this article is silly and wrong. With an unprocessed acoustic recording, the difference between 16-bit and 24-bit sources is fairly easy to hear on professional equipment.

By the time rock/pop/IDM/etc has been mixed and mastered, the dynamic range can be so limited you might as well distribute it at 8-bits. (Barely an exaggeration, BTW.)

This is not even close to being true of jazz, orchestral, and folk recordings. Typically recording engineers allow somewhere between 10dB and 20dB for peaks, which means the actual recorded resolution of sustained non-peaky instruments and quiet sections is somewhere around 12-bits - comfortably low enough to hear quantisation errors, even with dither.

So for some genres, 16-bits is plenty. For others it's nowhere near good enough.

In 2019, there's really no practical reason not to distribute music as 24-bit FLAC for high-end use. If you're listening on mobile you may as well use one of the better compressed formats. But for home playback, 24-bit is master-tape quality with no significant downside.

Sampling rate is a more complex issue. 48k is significantly better than 44.1k for the reasons mentioned.

Vinyl can go up to 100k or so, although not very accurately, and some people - including some very highly respected professional audio equipment designers, like Rupert Neve - believe that makes a difference.

But it's very hard to record ultrasonics "just in case" because the microphone->preamp->ADC chain has to handle them accurately, and that rarely happens. So there's very little of value up there in most recordings anyway - although maybe more on vintage tape masters than on modern digital recordings.

Personally I'm equally happy with 48k or 96k. The 192k recordings I've heard have been disappointing, possibly because of the intermodulation effects, but also because jitter becomes more of a problem at high rates.

Very inadequately. There was a quadrophonic vinyl system that failed commercially, which played back surround speakers using modulation of a 30K carrier tone. You had to use a special (aka 'good') stylus, and it sort of worked. The resulting carrier tones would go from 18 kHz to 45 kHz and the fact that this worked at all is evidence that vinyl goes up that far if you let it: wear will tend to scrub off that information unless it's a high energy transient, in which case there's a big chunk of plastic refusing to be worn off (but you'll dull it).

> That aside - the bit depth part of this article is silly and wrong. With an unprocessed acoustic recording

It's neither silly nor wrong—the article's title literally excludes it from consideration. This is about music downloads, not music production.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact