HDCD-encoded versions - how are they?

[Bean Bag]

If you like the sound of a HDCD, it is most certainly because you like the mastering.

Nothing is more important than the mastering, that's for sure...

The reason why Friends in particular sounds better on the HDCD than on the 1990s CD is because it was mastered from the absolute master tape which had better stereo separation and simply more information than the dub used before.

A good source is at least as important or more than the mastering...

A HDCD still uses 16 bits for each of its samples.

Yes, but it encodes the "equivalent of 20 bits worth of data in a 16-bit digital audio signal"

A HDCD decoder will use the least-significant-bit information to recreate the 20 bit source, but based on only 16 bits of information.

Yes...the least significant bit is set aside for this extra data.  However, it would not be based on 16bits of information -- it would be based on the quality of the source.  And I would assume they transfered the master tapes to hi-res (at least 24bit, 196k) for mastering. OR -- just mastered directly from the master tape.  So...it's source would not be 16 bit...it would be the master tape, or the lossless hi-res copy of it.



But I do agree, above all else - the source and the quality of the mastering are where you have to start.  The last car in the train (other than your stereo equipment) is the format; LP, CD, SACD, DVD-Audio or HDCD....the last three of which best redbook CDs to varying degrees of betterment. 

[Andreas]

Yes, but it encodes the "equivalent of 20 bits worth of data in a 16-bit digital audio signal"

That is mathematically not possible.

I know I am in the minority on this one, but I think HDCD offers no advantage over regular redbook. If the peak extension feature is used, it's actually a degradation of the signal if you play it on a non-HDCD player.

[Bean Bag]

Yes, but it encodes the "equivalent of 20 bits worth of data in a 16-bit digital audio signal"

That is mathematically not possible.

Well, it might be possible if one of those 16 bits has a compressed algorithm of sorts -- that is then decoded by the HDCD capable player.  The decoded information is then added as the remaining bits and the whole thing is read as 20 bit.

I know I am in the minority on this one, but I think HDCD offers no advantage over regular redbook. If the peak extension feature is used, it's actually a degradation of the signal if you play it on a non-HDCD player.

But it is only to a minute degree, so they say.  The bulk of the mastering data, as processed by the HDCD chip during mastering, is comparable to any other high-quality mastering system.  So, with any degradation--it still probably on par with the best systems.

[Andreas]

Well, it might be possible if one of those 16 bits has a compressed algorithm of sorts -- that is then decoded by the HDCD capable player.  The decoded information is then added as the remaining bits and the whole thing is read as 20 bit.

As I wrote above, mathematically not possible. It will be an approximation of the 20 bit source with the accuracy of 16 bits.

[Wirestone]

This isn't the way I read it.

The Wikipedia entry:

HDCD encodes the equivalent of 20 bits worth of data in a 16-bit digital audio signal by utilizing custom dithering, audio filters, and some reversible amplitude and gain encoding; Peak Extend, which is a reversible soft limiter and Low Level Range Extend, which is a reversible gain on low-level signals.[1][2][3][4]

HDCD encoding places a control signal in the least significant bit of a small subset of the 16-bit Red Book audio samples (a technique known as in-band signaling). The HDCD decoder in the consumer's CD or DVD player, if present, responds to the signal. If no decoder is present, the disc will be played as a regular CD.

In itself, the use of the first bit in the dithered least significant bit stream does not degrade the sound quality on a non-HDCD player; it only decreases the signal-to-noise ratio by a minuscule amount. HDCD Peak Extension, if chosen in HDCD mastering, will apply compression to the peaks which will be audible in playback on a non-HDCD system which does not apply the approriate expansion curve.

HDCD provides several digital features, which the audio mastering engineer controls at his/her own discretion. They include:

    * Dynamic range compression and expansion, with which virtually 4 more bits of dynamic range can be added to the musical signal.[citation needed]
    * Precision digital interpolation filtering with multiple modes of operation, which can reduce alias distortion and temporal smearing, resulting in a more natural, open, and accurate sound reproduction.[citation needed]

[Bicyclerider]

This isn't the way I read it.

The Wikipedia entry:

 HDCD Peak Extension, if chosen in HDCD mastering, will apply compression to the peaks which will be audible in playback on a non-HDCD system which does not apply the approriate expansion curve.

HDCD provides several digital features, which the audio mastering engineer controls at his/her own discretion. They include:

    * Dynamic range compression and expansion, with which virtually 4 more bits of dynamic range can be added to the musical signal.[citation needed]
    * Precision digital interpolation filtering with multiple modes of operation, which can reduce alias distortion and temporal smearing, resulting in a more natural, open, and accurate sound reproduction.[citation needed]

So if you already have a heavily compressed (loud) mastering, and it's HDCD peak extension encoded, it will sound even MORE compressed on a non HDCD player.  That could explain why the 2001 remasters sound so much more compressed and less dynamic than the 1990 remasters.  Or it could be that without the HDCD peak extension, the mastering wouldn't sound so compressed.  The problem with excessive compression of course is that dynamic range is limited.

[wgolly]

Can anyone give a list of the "Easter Egg" hidden tracks on the 2000 remasters?

there's a california girls sessions snippet at the end of the party!  cd.  Don't remember any others though.

[Dr. Tim]

I agree with Bean Bag that if you're going to listen to HDCD discs on your computer using the decoder in Windows Media Player, use the analog outputs ot you won't hear the benefit.  The Y-connectors one uses to hook an MP3 player into a stereo via the headphone mini-jack will do fine.  The other correspondents are of course right that a good set of D/A circuits can play a big role in how good the analog playback sounds, HDCD or not.  Some computer sound cards are better than others, but the more recent they are, the better they are.  After all, they expect people to WATCH BIG EXPLODING MOVIES on these laptops, etc.  so they can't sound too bad anymore. 

That said, I also agree the best option for good sound is to pick up an HDCD  D/A converter and hook it up to the S/PDIF digital output of  your best CD player.  Audio Alchemy made some good ones pre-Microsoft and they can be found on eBay fairly readily.

[Bean Bag]

So if you already have a heavily compressed (loud) mastering, and it's HDCD peak extension encoded, it will sound even MORE compressed on a non HDCD player.  That could explain why the 2001 remasters sound so much more compressed and less dynamic than the 1990 remasters.  Or it could be that without the HDCD peak extension, the mastering wouldn't sound so compressed.  The problem with excessive compression of course is that dynamic range is limited.

That's not right.  Here's the quote...

HDCD Peak Extension, if chosen in HDCD mastering, will apply compression to the peaks which will be audible in playback on a non-HDCD system which does not apply the approriate expansion curve.

This refers to "peaks" not audible in 16bit.  Since that format does not apply the appropriate curve to this extended data--it's lost in space.  If this "peak extension" feature is chosen during HDCD mastering--this plug-in will compress this data and it will become audible even on non-HDCD systems.  Otherwise -- you don't get them, and/or they're not in their proper context.  So, this is good.  It's extending the peaks for you by essentially cramming this otherwise lost data into the limited space of the 1980s-technological wonder of 16bits.

On a HDCD-capable player, however, you get them in their proper context.  This is even better.

[Bean Bag]

Well, it might be possible if one of those 16 bits has a compressed algorithm of sorts -- that is then decoded by the HDCD capable player.  The decoded information is then added as the remaining bits and the whole thing is read as 20 bit.

As I wrote above, mathematically not possible. It will be an approximation of the 20 bit source with the accuracy of 16 bits.

Well it is possible.  Think of a 100mb file -- zipped or compressed, if you will -- down to 80mb.  It's smaller and it's only 80mb.  But I'll swear to you that there is still, somehow, 100mb of data crammed in there.  No only that, but it is also not compromised.

What they're doing is finding similar information across the spectrum and throwing it out.  Think of this posting; and let's throw out every letter "e."  If there's 20 letter "e's" we'll call it e20.  Rather than use all the space required to create 20 separate e's -- we'll make one "e" and just tell the "decoder" to stick this one identical "e" wherever we tell it.  This negates the need to store 20 freekin' e's.

So all that is mathematically possible...if you believe....in tomorrow.  Sorry.  It's Friday.   


Now you might say -- "yes, Mr Bag...but am I listening to the compressed 80mb file, sans-"e" with a non-HDCD capable player?"  Well, yes you are.  But what it's throwing out are not all important "e's."  It's tossing things you would not be able to hear anyway and beyond 16bit definition.  But if chosen, don't forget -- the peak extension feature will allow this to be heard by compressing it into the 16bit source.  So it's not compressing the bulk of the signal (a la "dynamic range compression") but rather just those things with which we're not audible to you in the first place.  And on an HDCD player...they're decompressed and returned to their original position in the signal and therefore  E X T E N D I N G  back to 20 bit glory.

[Bean Bag]

double post...

[Andreas]

Well it is possible.  Think of a 100mb file -- zipped or compressed, if you will -- down to 80mb.  It's smaller and it's only 80mb.  But I'll swear to you that there is still, somehow, 100mb of data crammed in there.  No only that, but it is also not compromised.

Ahem...I know about compression algorithms (Ph.D. in mathematics). You are of course right that you can fit certain information in less space with a clever lossless data compression algorithm. FLAC can reach a 1:2 lossless compression ratio, but it does so by encoding the entire information. In the case of HDCD, the claim is that 4 additional bits of information can be encoded in 1 bit of the redbook scheme, and that 1 bit (least significan bit) is, per the wikipedia article you referenced, not even used all the time. That is why I say it is mathematically not possible.

I have not found any real description of the HDCD algorithm since it is not open source; otherwise I could explain at which step they are losing information.

[Bean Bag]

Ahem...I know about compression algorithms (Ph.D. in mathematics).

 

You're right,...that's why I say...If you believe.  We don't know exactly what they're doing.  But if they're doing what I understand they're telling us ...it is more or less possible.

So with that...I then move to my ears.  HDCDs tend to sound more spacious and full.  There's a resemblance to DVD-Audio in the taught-ness and accuracy of the natural timbre.

The Beach Boy 2-fers, as I said, are not a good example of A-B'ing the technology.  What was done in 1990 is not what was done in 2000, even excluding the HDCD process.  There's different sources, different levels, noise reduction, etc etc...and who knows what else.

[The Heartical Don]

If I may (temporarily) interrupt a most interesting discussion -

thanks guys! This is more than I'd hoped would be posted... and be sure that this is one of the few threads that I will print out.

Now go on!

[Bicyclerider]

Quote
HDCD Peak Extension, if chosen in HDCD mastering, will apply compression to the peaks which will be audible in playback on a non-HDCD system which does not apply the approriate expansion curve.


This refers to "peaks" not audible in 16bit.  Since that format does not apply the appropriate curve to this extended data--it's lost in space.

This doesn't make sense to me.  The quote says that if HDCD peak extension is chosen in mastering (and was it chosen for the Beach Boys 2001 2fers) the compression "will be audible in palyback on a non HDCD system."  My non HDCD does not have an "appropriate expansion curve" whatever that is so yes, the added compression IS audible on non HDCD players.  If the peaks are not audible in 16 bit why does the quote say they will be audible?

[Bean Bag]

This doesn't make sense to me.  The quote says that if HDCD peak extension is chosen in mastering (and was it chosen for the Beach Boys 2001 2fers) the compression "will be audible in palyback on a non HDCD system."  My non HDCD does not have an "appropriate expansion curve" whatever that is so yes, the added compression IS audible on non HDCD players.  If the peaks are not audible in 16 bit why does the quote say they will be audible?

The problem with that quote is that it is a tad too brief, it doesn't properly explain its own terms and is prone to be taken in other ways.  If you add 'then' -- it all makes easier sense.

"HDCD Peak Extension will apply compression to the peaks which will then be made audible in playback on a non-HDCD system, which does not apply the appropriate expansion curve."


So the peaks will be made audible not more compression as we think of it.

[mjd180]

This doesn't make sense to me.  The quote says that if HDCD peak extension is chosen in mastering (and was it chosen for the Beach Boys 2001 2fers) the compression "will be audible in palyback on a non HDCD system."  My non HDCD does not have an "appropriate expansion curve" whatever that is so yes, the added compression IS audible on non HDCD players.  If the peaks are not audible in 16 bit why does the quote say they will be audible?

The problem with that quote is that it is a tad too brief, it doesn't properly explain its own terms and is prone to be taken in other ways.  If you add 'then' -- it all makes easier sense.

"HDCD Peak Extension will apply compression to the peaks which will then be made audible in playback on a non-HDCD system, which does not apply the appropriate expansion curve."


So the peaks will be made audible not more compression as we think of it.

Wait. So are you saying "peak extension" gets used specifically for the benefit of playback in a non-HDCD environment, so as not to interfere with the original 16-bit dynamic range? If this is the case, then why is it even an optional feature in the HDCD encoding process? Why would it not be "built-in" to the original design? To save money? 

[Bean Bag]

Wait. So are you saying "peak extension" gets used specifically for the benefit of playback in a non-HDCD environment, so as not to interfere with the original 16-bit dynamic range? If this is the case, then why is it even an optional feature in the HDCD encoding process? Why would it not be "built-in" to the original design? To save money? 

Frankly, I don't know why Peak Extension is an "option" either.  As Andreas said...few people know "exactly" what's going on here anyway.  I'm just taking them for their word.  And their word on this (as is Microsoft's, who bought the technology) is really vague.  I would guess, the more options left to the mastering engineer, the better. ?

But Peak Extension, as they explain it, is for the benefit of both HDCD and non-HDCD players.  It sounds like what they're doing with this feature is  a) for non-HDCD players, it allows for expanded ranges or peaks to be encoded or approximated or compressed or dithered down (something any good later-day mastering units would do) and, b) for HDCD players it allows those peaks to be restored to full, non-dithered 20-bit glory via the "Zipped-up" info stored in the "unused" bit.

 

[Bean Bag]

Also...for what it's worth, everything depends on how it gets down-sampled.  If it's a CD -- it's all going to be 16-bit.  It's just that with HDCD, they're claiming that they're able to "restore" to 20-bit (or 24), by un-zipping that remaining data stored on that least used bit, with a HDCD-player.

I think they're doing some of that.  I have quite a few HDCDs most of which don't have the logo on the covers.  In almost all circumstances I can tell.  It will sound a little fuller and richer, and sure enough, when I look down...the little HDCD light will be on.

[Andreas]

But Peak Extension, as they explain it, is for the benefit of both HDCD and non-HDCD players. 

That's not correct. Peak extension is only for the benefit of HDCD players, because additional dynamic range in the top level area will be uncovered (in the form of "extended peaks"). If played back on a non-HDCD player, the effect will be audbile as very harsh digital compression, which is not a benefit.

Most HDCDs do not use peak extension because it really degrades the signal if not decoded.