HDCD

[Andreas]

I really do not understand HDCD. Please do not provide me with www.hdcd.com links. They did not solve my confusion.

Okay, what I understand is the following:

Within the regular redbook audio stream (16 bits/44.1 KHz), some additional information is contained, namely in the least significant bit (but probably not on every sample).

If you play it back on a CD player without HDCD capabaility, this additional information is lost. Also, the HDCD info in the least significant bit is incorrectly interpreted as audio, but won't affect the outcome audibly.

If you play it back on a HDCD player, the additional information is used to increase the bitrate and/or the dynamic range (here the info gets a bit cloudy). I have gathered that the top 3 db are expanded to 9 db, so that 6 db of dynamic range are gained.

Now, when I extract a HDCD to a .wav file and analyse it with a wave editor, the HDCD information will not be recognized.

Could it be that the waveform seems to be clipped and truncated severely, but that these clippings are recovered during HDCD playback?

Could it be that the harshness that I hear on the Friends 2001 remaster (and other HDCD discs) simply comes from the fact that these take full advantage of the dynamic range of HDCD, and are consequently clipped/truncated if no HDCD player is used?

(I don't have a HDCD player).

[krabapple]

Great question. I've wondered that too. I think I'll go ask on the prosoudn web board.

In the meantime, [url=http://groups-beta.google.com/group/rec.audio.high-end/browse_frm/thread/4864ac8e1d314e2b?q=%27how+hdcd+works%27]
here's[/url] a thread about how HDCD works

[Andreas]

...it has a mode which allows for signal peaks to exceed 0 dB, something that other converters don't usually allow because that is a hard clipping point. The HDCD encoder (unless the dynamic range extension mode is not used) thus allows a higher average level to be mastered, with the peaks compressed to prevent digital clipping, and a small amount of compression of low-level signals as well. Producers will do just about anything to get a "loud"-sounding CD, and the HDCD encoder seemingly makes that possible, going beyond what analog compressors alone can do.

That is what I have read elsewhere, so I guess it is true. However, it does not really answer my question. I still think that digital clipping cannot be recovered by a HDCD player

[krabapple]

I haven't asked yet on PSW, because I want to look at waveforms of a few more HDCDs first. One question is, do they *always* show digital clipping in .wavform view?

Tell me if my reasonign is good here: if the HDCD decoder reallly recovers samples that only *appear* clipped in a digital rip, re-digitizing the analog output (making sure to mointor levels so you aren't introducing clipping during digitzation) and viewing the .wavform should yield normal-looking peaks, right?

However, if the peaks are not recovered, then the analog output will still appear flat-topped, even when digitzed at levels such that peak is < 0 dB FS. Right?

Right now i don't have any standalone HDCD decoders -- the only one available to me is what's included in Windows Media Player. And I haven't figured out how to capture the decoded stream of that digitally .
I suppose I *could* route the analog out of my soundcard back into the analog ins, and digitize that....

[dudelsack]

Right now i don't have any standalone HDCD decoders -- the only one available to me is what's included in Windows Media Player. And I haven't figured out how to capture the decoded stream of that digitally .
I suppose I *could* route the analog out of my soundcard back into the analog ins, and digitize that....

FWIW, I've heard Windows Media Player doesn't actually decode HDCD, despite the HDCD light going on. I have no experience with it, myself...

[lukpac]

However, if the peaks are not recovered, then the analog output will still appear flat-topped, even when digitzed at levels such that peak is < 0 dB FS. Right?

I'm not sure about that, at least if the "clipping" is mild.

[lukpac]

FWIW, I've heard Windows Media Player doesn't actually decode HDCD, despite the HDCD light going on. I have no experience with it, myself...

No idea, but it can certainly detect the difference between HDCD and non-HDCD, even on the same CD.

[dudelsack]

FWIW, I've heard Windows Media Player doesn't actually decode HDCD, despite the HDCD light going on. I have no experience with it, myself...

No idea, but it can certainly detect the difference between HDCD and non-HDCD, even on the same CD.

Microsoft owns HDCD. So I'll admit what I said above doesn't make much sense, UNLESS there is no software decoding solution for HDCD. HDCD would be an easy thing to detect, I think, given that the information is in the LSB every time...

[krabapple]

However, if the peaks are not recovered, then the analog output will still appear flat-topped, even when digitzed at levels such that peak is < 0 dB FS. Right?

I'm not sure about that, at least if the "clipping" is mild.

Well, I'm pretty sure I've seen flat-topped .wavs that top out at less than 0 dB FS. I can generate these myself by taking a clipped .wav and bringing the level down.

[lukpac]

Well, I'm pretty sure I've seen flat-topped .wavs that top out at less than 0 dB FS. I can generate these myself by taking a clipped .wav and bringing the level down.

Sure, if you simply change the gain digitally. But I'm not sure if you'll see the same flat tops by doing a D/A -> A/D conversion. I actually did this a bit recently. I should see if I still have the files and check it out.

[krabapple]

Cool, post them if you can.

I looked at rips from some Joni Mitchell HDCD and King Crimson HDCD wavs tonight. Even the loudest JM tracks I could find didn't show evidence of flattops -- suggesting that rips from HDCDs needn't *necessarily* look clipped when viewed as .wavs. The often-loud KC tracks varied with how recently they were remastered; the most recent one showed some flattops (but not much -- something like 11 possibly clipped samples if the track is normalized to 0 db). Then I loaded one of the Japanese Yes HDCD tracks from 'Big Generator'. Jesus. It's an almost solid block of green with a few cracks in it. And the bastards dropped the peak level down digitally to -0.5 dB in case the hundreds of samples at full scale might choke some players.

Btw, some 'flat top' areas resolve into rounded peaks you blow them up enough to view individual peaks...which accords with the KC material only showing 11 clipped samples; it 'looks' worse than it is. Others, like the Yes, show gross flat-topping even at high magnification.

Given that mastering engineers seem to have a few options they can enable or disable when using HDCD, it's hard to say what causes what I see. The digitized analog capture of a decoded vs nondecoded HDCD would be interesting.

[Patrick M]

The often-loud KC tracks varied with how recently they were remastered; the most recent one showed some flattops (but not much -- something like 11 possibly clipped samples if the track is normalized to 0 db).

Could you elaborate on this? I was thinking they all came out within the same year.

[krabapple]

'USA" and 'Earthbound" came out significantly after the others-- 2002 vs 2000.

I compared a track on 'USA' (Larks II) to a few on 'Red' (Red, One More Red Nightmare) and there was clipping on the former, none on the latter.

[Andreas]

Steve's Audio Fidelity Gold HDCDs did not show any clippings, as far as I remember.

[lukpac]

Steve's Audio Fidelity Gold HDCDs did not show any clippings, as far as I remember.

Wouldn't that simply mean he didn't use dynamic range extension?