Hacker News Reader

Comment by splitdisk | Hacker News Reader

From my experience, what matters more than sample rate is 24 bit vs. 16 bit sampling in the recording/production process. Using heavy compression and EQ can mean that very quiet sounds can become louder, in this case 24 bit recording is ideal. Sample rate wise, anything above 40khz is fine for most ears (I've probably lost a few khz in the upper range anyways) Another note is that most converters operate at a multiple of 48K, so it makes sense to use 48/96khz if you are recording. It all comes down to how much disk space you have, and want to use up.

View on HN

I can still hear around the 20k range, so let's not exclude a few listeners just because we wear hearing protection when it matters. In practice, 20k audio content means 44k or higher sample rates, due to the fact that actual filters have finite transition bands. There's an unfortunate history of engineers with poor hearing who inflict pain on others, such as the horizontal retrace on NTSC TVs which still annoys me when I encounter one.

24bit also means we don't have to record at 0dBFS, which saves a lot of time.

I'm very jealous of your healthy hearing range... I now wear hearing protection, but the damage has been done. When it comes to file storage, I will take a 44.1/48K 24 bit FLAC happily, since it usually comes out to the same size as a CD-Q WAV file anyways. I see no reason why everything shouldn't be in this format, but CD's have made a serious dent in formatting standards/habits.

24bit has been shown to be excessive in actual tests, and it doesn't really pass the "bullshit test". 24 bit allows you to switch between a whisper in a library and a jet engine or shotgun blast in a single recording—yes, it would instantly damage your hearing.

By comparison, 16 bit audio can "only" record a whisper in a library and a motorcycle or jackhammer.

Double blind tests show that 8 bits are not enough, but 14 bits are.

Citation? The article is talking about dithering, which you really don't want to do, not least because the end result will probably compress worse than the higher-bit-depth version. The fact that they suggest it at all implies, to my mind at least, that 16 bits isn't enough.

You absolutely do want to do dithering. Dithering converts distortion (error correlated with the signal, which is bad) to noise (error uncorrelated with the signal, which is less bad). This means that even though the noise floor is higher, you can recover more of the original signal. There is virtually no case where that is not desirable.

You're right of course that it will compress less well, but that's to be expected because you've lost less information!

Or because you've added randomness?

Store the 24-bit signal, and you could do a dithered downsample to 16-bit on playback if you think that's a good idea. Wouldn't that be better all round?

It's my understanding that 24 bit audio can capture the quieter sounds in greater detail than 16 bit. So say if you EQ out a frequency range in order to "zoom in" on a much quieter range, like the motorcycle vs. whisper, you can hear the whisper in as much detail as a full CD quality recording.

For playback though, I agree that 14 bits are probably enough. Even high quality mastering tape has the equivalent of about 12 bits of dynamic range, which is fine. Many fabled analog pieces of equipment have terrible signal-noise characteristics, but are still valued for other reasons (coloration, distortion etc...)-which is all fine by me.

Telephone audio is 8khz, so recording at a multiple of 8k helps with downsampling for IVR prompts or hold music. With dithering, it isn't terrible to downsampling from 44.1k to 8k, but it's nice to avoid it.

Have you tried listening to SACD? The high sample rate might not give you more reproduction of audible frequencies, but the difference in arrival times it can encode makes well recorded stereo stuff more interesting to listen to, in my limited experience.

I know this seems counterintuitive, but there is literally no difference in arrival times (over audible frequencies) that can be encoded at higher sampling rates. Digital sampling does not quantize over the time domain for any frequency below the Nyquist frequency.

If you have the time, watch the two videos that xiph.org did a few years ago[0]. There's a great in-depth explanation, as well as a hands on demonstration to demonstrate this reality.

[0] https://xiph.org/video/

This is directly addressed by the article under the "Sampling fallacies and misconceptions". You don't lose "arrival time" (AKA phase) when you use a lower bitrate. They have a video that explains it very well: http://xiph.org/video/vid2.shtml

I would be very curious to listen to SACD on some good headphones in a quiet room. Not sure if I've ever even seen a SACD player aside from maybe in the Sony store 10 years ago. The trick would be to find something that would be mastered for the format.

Are you sure? Many so-called "universal" bluray players can play SACDs.

I got a Denon one. I haven't played any SACD on it yet (I got it for bluray), though I guess I could easily find some at that video rental store (in Tokyo).

Reply on news.ycombinator.com