Ripping HDCD in iTunes?

[Dusty Chalk]

See, I don't think of it that way. Maybe I don't want to use integer math. Fractions would represent an analog waveform that much better.

So, to turn your analogy back around on you, it's like going from integer math to floating point math to represent an analog waveform.

[grawk]

Except it doesn't matter how you want to think of it, when the way it's stored is the way I described it. The formats are designed to be backwards compatible, so If you lop off the last 8 bits of each word of a 24 bit recording (say by sending a 48/24 spdif signal to a dac that only supports 16 bits, it decodes properly. 24 bits provides the same resolution, with more headroom. In video, more bits = more resolution. In audio, more bits is just more bits. A 24 oz bottle holds 8 oz more water than a 16 oz bottle, but the water itself isn't any different.

[Dusty Chalk]

Alright, I have to admit, I don't understand your argument starting at post 25 -- could you go back and explain it again, or at least point me to a wiki page?

I mean, if I understand you correctly, you seem to think that something is making you use only 16 bits of resolution in a 24 bit word -- why would you think that? It's amplifying data, the additional 8 bits would have to be below the LSB of the 16 bit word in order for it to be compatible with the 16 bit DAC, not above the MSB.

If you're saying something else, then please explain it to me.

[grawk]

Audio bit depth - Wikipedia, the free encyclopedia

By increasing the sampling bit depth, smaller fluctuations of the audio signal can be resolved (also referred to as an increase in dynamic range). The 'rule-of-thumb' relationship between bit depth and dynamic range is, for each 1-bit increase in bit depth, the dynamic range will increase by 6 dB (see Signal-to-noise ratio#Fixed point). 24-bit digital audio has a theoretical maximum dynamic range of 144 dB, compared to 96 dB for 16-bit; however, current digital audio converter technology is limited to dynamic ranges of ~120 dB because of 'real world' limitations in integrated circuit design.

[Dusty Chalk]

Except that's based on dynamic range, and I'm not talking about dynamic range, I'm talking about resolution. I won't argue that you can't hear a whisper next to a jet engine, but with complex music, I don't see how a high noise floor would prevent you from hearing the difference. I mean, to extrapolate on what you're saying, if I raise the noise floor high enough (what is it, -48db?), you're saying you wouldn't be able to tell the difference between 8 bit and 16 bit audio, and I think you would. I agree you wouldn't be able to hear a 1-bit (out of 16) signal with a -48db noise floor ("dynamic range"), but I bet you could still tell the difference between an 8-bit signal and a 16-bit signal ("resolution").

[grawk]

It's not about telling the difference, it's about having enough space for the data. If your loudest note is less than 96db above the quietest spot, then more bits aren't needed. Perception doesn't enter into it.

[The Monkey]

All I want is for the pretty pretty lightses to turn on on my DACSes.

[grawk]

your lights will turn on, if you don't screw something up dinny

[blessingx]

Not surprising in the least, but as it just accidently occurred to me while this conversation was occurring, if you're iTunes, Play, Airfoil, etc. volume is slightly off 100%, the HDCD light goes dark.

[grawk]

Right, change the volume and it's no longer bit perfect.

[Icarium]

Digital volume control for the lose

[morphsci]

Right, change the volume and it's no longer bit perfect.

Digital volume control for the lose

Unless the volume control works with a greater bit depth than the source material. You can still end up not bit perfect but you do have wiggle room. To me, that's the only real good reason for the Sabre32 being 32 bit. Actually makes the digital volume control useful.

[Icarium]

Nod.. honestly I use it anyways for convenience.. I'm not sure that I can tell the difference from time to time.. but generally I have to be pretty lazy to use it since volume knobs are within reaching distance. Filburt says he can hear the diff when using foobar2k vol control and other times its maybe too subtle to call.

Meh ;p

[grawk]

Eh, who cares, bit perfect matters for transfers but for listening, bits you can't here don't matter. Analog or digital, once it's too quiet to hear, it's gone.

[morphsci]

Eh, who cares, bit perfect matters for transfers but for listening, bits you can't here don't matter. Analog or digital, once it's too quiet to hear, it's gone.

Agreed and the threshold rises every year, at least according to my wife.

[Filburt]

HDCD works by sticking data in the LSB as a control signal to operate a set of digital filters. When you use volume control, that data drops out and hence you lose the HDCD decoding. Probably more than anything, though, the major advantage with HDCD was the hardware. The Pac Micro Model 2 mastering unit had excellent filters and ADCs in it and the PMD-series digital filters were some of the best available. So, even without HDCD itself, the end-user was probably getting a substantial boost in fidelity versus other systems.

As far as my intermittent issues with foobar volume control goes, I've had some trouble with the EQ making very audible distortion/artifacts and it's bad enough that I don't use it even when, otherwise, it would ostensibly help on some recordings I have. This happens even with CDA or lossless, so it's not just me defeating the psy model on a lossy codec (which is a problem you can run into with even a perfectly good EQ). So, maybe something related to this problem is going on with the volume control *shrug*. I don't have the latest version of foobar, though, so I don't know if this is still an issue...nor am I sure that this isn't something specific to my setup

As far as 16 vs 24 bits, there are more discrete levels between bit 24 and 23 than there are between 16 and 15. However, in both cases, the steps between MSB and MSB-1 are so numerous I doubt it really makes much of a difference. At lower bits, though, I guess its relevance is conceivable, though I'm not sure I'd really weigh it all that heavily. That is, for example, 8 bits down (48dB), there is a significant disparity in granularity between the two. Mainly where this would even mean anything is if you have a recording with a lot of dynamic range, where using 24 bits helps keep anything audible well away from the region near the LSB, and so you may benefit from better low-level linearity. Do I worry about it much? No, not really. I've honestly found using a good digital filter and, even more importantly, making sure the output stage is well designed makes more of a difference. It's something that probably matters more in digital imaging. Ears aren't nearly as sensitive or discriminating as eyes, and so I'm just not sure how much it would matter even if there is some sort of minor transient error present. Really, if anything, my biggest criticism with the Redbook standard is that 44.1KHz Fs made the task of filtering unnecessarily burdensome for then-existing hardware. However, there are some good filters that were made since then, and modern hardware should be powerful enough to make an adequate filter, provided one has the requisite knowledge.

[Dusty Chalk]

It's not about telling the difference, it's about having enough space for the data. If your loudest note is less than 96db above the quietest spot, then more bits aren't needed. Perception doesn't enter into it.

It is about telling the difference, if you can. And I'll leave it at that.

[grawk]

You're perceiving differences, because you're comparing unequal things. Take a 96/16 and 96/24 or 44/16 and 44/24 or 48/16 and 48/24, with proper conversion from 24 to 16 bits, and there is no difference. Comparing something mastered for 96/24 to something mastered for 44/16, and there are a LOT more variables.

[Dusty Chalk]

Fair point (and agreed that they are different), I will see if I can figure out a way to make that comparison.

[CarlSeibert]

Hmm. Coming from the visual world, I never thought this through. Let me see if I've got this straight..

In imaging, you've got n-number of steps between two absolute values - black and white, or some arbitrary ink coverage values that you're printer has chosen to represent them. If you have 8 bits, you have 256 usable steps; 16 bits and you have 64K-odd usable steps. (8 bits is generally fine for output, BTW, but doesn't have enough headroom to allow for much editing - editing used here in the sense of tonal manipulation).

If I'm hearing grawk right, in audio, only one end of the scale is fixed. You go down from 0db in steps that are fixed in size, rather than being proportional to some absolute value for "really very quiet". With a longer word length, you get a different value (in terms of energy, say) for your "really very quiet" value, and thus, if the "really very quiet" value is lower than the 80 or so db down that real gear can produce (or the 6 db down that pop recordings might have), then your longer word is wasted. The waveform drawn in the usable part of the dynamic range would come out identical either way.

Filbert? grawk? Is this what you're saying?

[grawk]

That's what I'm saying.

[Dreadhead]

That's what I'm saying.

QFT.

Just to muddy the waters a bit: I was talking with one of the Smyth brothers about why they used 24bit and it appears that at least in the movies where things have a bit more dynamic range (ie the explosions are 120db while the conversation can be at 40-60db) that 16bit could mean that the noise floor was audible when the volume was set for the conversation and that's why it's useful there. Not saying that this applies to music where the real dynamic range is generally a lot smaller but it was an interesting point.

As far as resampling from 96-44.1 if it's done right there is lots of evidence that it's inaudible too but compared to dithering from 24-16bit the there is even more crap to mess up and A LOT of programs do it horribly, even ones that you would expect not to like Adobe. Apparently both Apple and Microsoft took this rather seriously the last go around and the built in re-samplers are pretty good if not very good (at least that's my understanding). At least in the windows case for XP the resampler and was a hot mess.

[blessingx]

As it's at least slightly related, I've always been curious why B&W's Society of Sound, which really seems to strive hard for quality recordings, releases 24/48 files instead of 24/88.2 or 24/96? Can't think of anyone else doing this. Any guesses or is it likely simply a file size issue?

[Pars]

I'm still trying to verify if an HDCD compatible player I have is working or not. I haven't been able to get the HDCD light to come on with anything I have put in it. Supposedly Roxy Music-Avalon is an HDCD, but my copy was licensed by BMG, so I don't know if it really is an HDCD or not.

[blessingx]

Do you have any other HDCDs?

HDCD list - Head-Fi: Covering Headphones, Earphones and Portable Audio