Hacker News Reader

Comment by andrewzah | Hacker News Reader

That doesn't make any sense. RIAA EQ is a serialization format, that is then decoded by your audio equipment. That's what the phono jack is for on a receiver, pre-amp, etc.

What people "hear" regarding LPs is generally their biases + their particular analog setup. My setup is going to sound totally different than yours, regardless if it's from a CD or an LP.

There are some cases where an LP had a better -master- than a CD for the same release, but that is due to the publisher/owner of the master tracks and not the actual medium itself.

View on HN

I found out when digitizing vinyl that LP's before 1977, when digital mixing and mastering was not the norm that there is more data in the frequencies between 20 and 50kHz. I do my recording standard at 192kHz and analogue LP's often reach 48kHz of data while digital mastered ones do only 22kHz. After that date often the frequency is cutoff at 20-22kHz, probably to save memory room. Only after 2000-2010 there is enough memory to record at insane sample rates and bits per sample. Most awful are cds having only 16 bits and 41kHz sample rate, by modern standards really poor.

"Most awful are cds having only 16 bits and 41kHz sample rate, by modern standards really poor."

44.1kHz is all that one needs for -perfect fidelity-, given that humans cannot hear higher than ~20kHz. See: Nyquist-Shannon Sampling Theorem. [0] 16 bits of quantization gives a Signal to Noise Ratio (SNR) of 96dB, and noise dithering increases that [1]. More bits are not necessary unless one is doing music production with that audio and warrants the extra headroom, before mixing back to 44.1kHz/16bit for distribution. [1]

The only thing that higher bits/sampling rates do for a listener is waste CD and drive space, so it is odd to lament 44.1kHz/16bit as being "really poor". Even if your speaker setup can actually output extremely high frequencies, only your dog will be able to hear it...

[0]: https://en.wikipedia.org/wiki/Nyquist%E2%80%93Shannon_sampli...

[1] https://web.archive.org/web/20181109203430/https://people.xi...

Some humans can hear much higher than 20khz. Very little of our recording and playback equipment does well above that point, however, so efforts to reproduce that part of the spectrum most often fail miserably.

CD's are certainly good enough for now.

The only source I could find about humans hearing above 20 kHz is this paper reporting that some people could detect tones up to 28 kHz: https://asa.scitation.org/doi/10.1121/1.2761883

But that is for a pure tone under perfect laboratory conditions. Also, they report that the minimum hearing threshold at those frequencies was above 90 dB. So I don't think any human in this planet can actually hear anything above 20 kHz under normal circumstances, much less for music.

Couldn't beat[1] in the audible range occur from signals above 20kHz?

[1]: https://en.wikipedia.org/wiki/Beat_(acoustics)

Yes, but that'll be captured in the CD range with the rest of that signal. My feeling is that few instruments, no music and little transducer equipment exists for that range (>20khz), and so few are aware of it without the intermediation of technical assistance, that theres no point in attempting to record or reproduce any of that. all it can do is add noise.

cf musopen. https://musopen.org/music/

They've got some lovely high bitrate and bandwidth recordings there, and what they've caught is room and audience noise. They'd have benefited from more mics and mixing and less "throw bits at it"

I am doing production with these recordings, tools are more precise at higher bit and sample rates. Also quality of listening equipment is a factor as hearing is not a flat frequency response, things as ambiance and resonance play a role, music is not a single frequency tone but a combination of complicated frequency additions, subtractions, multiplications and divisions. So more frequency room can produce nicer highs and clearer sound, especially when music is uncompressed. It is simplistic to state that 44.1/16bit is enough, if that is the case then any combination of recording, amplifier and speaker would suffice. This just isn't the case, also that we would need awesome priced equipment isn't really worth it. But there really is a difference between a phone with a music app playing a 16bit/44.1kHz song on a medium quality headphones and a dedicated player with a good DAC chip playing a 96kHz song on a good quality headphones. People that cannot hear the difference should not invest in such equipment and will be happy with the need to invest less money for their entertainment.

>It is simplistic to state that 44.1/16bit is enough, if that is the case then any combination of recording, amplifier and speaker would suffice.

No, that doesn't follow. It just means that recording at those bit rates and sampling frequencies isn't the "bottleneck" for the sound in the system.

>But there really is a difference between a phone with a music app playing a 16bit/44.1kHz song on a medium quality headphones and a dedicated player with a good DAC chip playing a 96kHz song on a good quality headphones.

When everything you compare to is better quality then of course it sounds better. But it's not due to the 96kHz vs 44.1kHz. Nyquist-Shannon is mathematical reality.

You have to compare the same high quality headphones, the amp with the same DAC, with the same recording, being played through a high quality filter bringing everything down to 44.1kHz/16bit.

"You have to compare the same high quality headphones, the amp with the same DAC, with the same recording, being played through a high quality filter bringing everything down to 44.1kHz/16bit."

I agree. If there is a methodologically sound study showing that people can actually distinguish (>50%) between 44.1kHz/16bit and 192kHz/24bit on the same hardware [at exactly the same dB SPL], then I will be really surprised. It's not physically possible to hear those differences, as far as I understand it.

I'd also be surprised. Even more so with grown-ups, most of them don't even hear >20kHz.

44.1/16 is enough for distribution. This is not simplistic. It simply can reproduce waveforms up to the nyquist frequency perfectly within a defined noise range. That's a mathematically proven fact. Of course you can botch that with bad listening equipment, but there's nothing a ridiculously oversampled signal would improve given the same listening equipment. You do speak about the studio processing right?

There is an extremely nice "Digital show and tell" video by Monty at

https://xiph.org/video/vid2.shtml

(part 1 at https://xiph.org/video/vid1.shtml is also very well worth watching.)

Reply on news.ycombinator.com