A little exploration of the benefits of HDCD

[DrCWO]

Hi *,
I recently listened to the album "Romantically Helpless" by "Holly Cole" and liked it.
Next I saw that I could buy an HDCD (>$50) in Japan, which might sound better than the 44/16 version offered by Qobuz.

The question arises: Will that be worth the price?

I have some HDCD albums decoded by dBpoweramp to 24Bit so I evaluated quickly what might be the improvement.
The album I picked for evaluation was "Local Hero" by Marc Knopfler.

This is what I did:
- Decoded the original WAV (44/16) with dBpoweramp to 44/24.
- Opened both, the original WAV and the decoded in a wave editor.

Starting with track six "The Rocks and the Thunder" the 16 and 24 bit version were exactly the same but with different volume.
I recognized that this is quite a silent track with all peaks below 0.45 (1.0 equals 0dBFS).

Next I analyzed track tree "Freeway Flyer" and there I found some few differences between the two versions.

* The red signal is the original CD rip (44/16).
* The brown signal is the 24 Bit output of dBpoweramp after a volume increase of 2.5dB to fit both signals as good as possible.
* The light green signal is the difference between both.

We see that the HDCD carries additional information that was suppressed by mastering the 44/16 version.
Results:
* The 44/24 signal created by dBpoweramp HDCD decoding reverts compression done for the 44/16 signal above -0.5 dBFS.
* The 44/24 signal is 2.5dB quieter so you can get a disappointing 0.5 bit higher resolution (which means 16.5 bit) which is far away from advertised 20 Bit for HDCD!

As there is a lot of negative talk in the net regarding the quality of the dbPoweramp's HDCD decoder I also tried with HDCD.exe that can be downloaded here.

This is the result with the same Track:

* The red signal is the original CD rip (44/16) as above.
* The green signal is the 24 Bit output of HDCD.exe after a volume increase of 6dB to fit both signals as good as possible.
* The blue signal is the difference between both.


As you can see, the result is absolutely identical. The only difference is that HDCD.exe lowers the signal by 6dB and you could theoretically gain 1 bit instead of 0.5 bit.

So the answer to my original question is clear to me:
I don't think it's worth paying $50 for the original CD for that half, probably inaudible, bit.

 

[Blumlein 88]

@amirm can give you some good information about HDCD I think.

 

[Gregss]

Hi *,
I recently listened to the album "Romantically Helpless" by "Holly Cole" and liked it.
Next I saw that I could buy an HDCD (>$50) in Japan, which might sound better than the 44/16 version offered by Qobuz.

The question arises: Will that be worth the price?

I have some HDCD albums decoded by dBpoweramp to 24Bit so I evaluated quickly what might be the improvement.
The album I picked for evaluation was "Local Hero" by Marc Knopfler.

This is what I did:
- Decoded the original WAV (44/16) with dBpoweramp to 44/24.
- Opened both, the original WAV and the decoded in a wave editor.

Starting with track six "The Rocks and the Thunder" the 16 and 24 bit version were exactly the same but with different volume.
I recognized that this is quite a silent track with all peaks below 0.45 (1.0 equals 0dBFS).

Next I analyzed track tree "Freeway Flyer" and there I found some few differences between the two versions.

View attachment 256712
* The red signal is the original CD rip (44/16).
* The brown signal is the 24 Bit output of dBpoweramp after a volume increase of 2.5dB to fit both signals as good as possible.
* The light green signal is the difference between both.

We see that the HDCD carries additional information that was suppressed by mastering the 44/16 version.
Results:
* The 44/24 signal created by dBpoweramp HDCD decoding reverts compression done for the 44/16 signal above -0.5 dBFS.
* The 44/24 signal is 2.5dB quieter so you can get a disappointing 0.5 bit higher resolution (which means 16.5 bit) which is far away from advertised 20 Bit for HDCD!

As there is a lot of negative talk in the net regarding the quality of the dbPoweramp's HDCD decoder I also tried with HDCD.exe that can be downloaded here.

This is the result with the same Track:
View attachment 256714
* The red signal is the original CD rip (44/16) as above.
* The green signal is the 24 Bit output of HDCD.exe after a volume increase of 6dB to fit both signals as good as possible.
* The blue signal is the difference between both.


As you can see, the result is absolutely identical. The only difference is that HDCD.exe lowers the signal by 6dB and you could theoretically gain 1 bit instead of 0.5 bit.

So the answer to my original question is clear to me:
I don't think it's worth paying $50 for the original CD for that half, probably inaudible, bit.

Hello,
If the source files were older, they likely were not HDCD quality, but lower quality upsampled to a higher bit rate. Unlikely to make any real difference as you are limited to what info the source files had.

 

[MrSoul4470]

All the HDCDs I have (mostly Neil Young) sound better than other pressings of the same album. I'm not saying that is because they are HDCDs, but maybe just a different master. No matter why, they sound better to me.

 

[voodooless]

Oh, I know: “let’s compress these tracks into clipping, then encode the information to fix this a bit by stealing one of those precious 16 bits.”

“… but can’t we just not have the track clipping?”

“… no!”

 

[MrSoul4470]

The album I picked for evaluation was "Local Hero" by Marc Knopfler.

Are you sure that HDCD uses the peak extend feature? Did you check it? If not then it's basically a fake HDCD (many of them out there) with no audible improvement.

 

[DrCWO]

Oh, I know: “let’s compress these tracks into clipping, then encode the information to fix this a bit by stealing one of those precious 16 bits.”

“… but can’t we just not have the track clipping?”

“… no!”

That‘s what it seams to be here as I did not find any low volume expansion.
Might have been better to produce a normal CD with 3dB less volume…

 

[DrCWO]

Are you sure that HDCD uses the peak extend feature? Did you check it? If not then it's basically a fake HDCD (many of them out there) with no audible improvement.

It does, look at the diagrams I put in.

 

[kemmler3D]

The 44/24 signal created by dBpoweramp HDCD decoding reverts compression done for the 44/16 signal above -0.5 dBFS.

Wait, are we sure they weren't just two different masters? What we are seeing here is the HDCD peak extension feature? I had never heard of that before, it seems unnecessarily clever...

 

[DrCWO]

Wait, are we sure they weren't just two different masters?

I used ONE disk as I wrote and compared the 16bit output with the HDCD decoded 24bit result.
So MUST be the same master.

 

[DrCWO]

What we are seeing here is the HDCD peak extension feature?

Yes it is

 

[MRC01]

The value of this is a bit more subtle. By raising the level of the quiet parts, it increases their amplitude resolution in 16 bit encoding. Decoding is done in greater than 16 bit, so when it pushes them back down they don't lose resolution.

Put differently, when you have a quiet part of the music, the most significant bits are all zero and being "wasted" and the signal is encoded with fewer bits. You can shift it up/left into those bits to gain more resolution signal to noise ratio, then shift it back down/right on playback to avoid dynamic range compression. Or don't shift it down, if you consider that compression desirable. That's what happens when you play it back without HDCD encoding.

Of course this is much ado about 1 measly bit. But it does work.

HDCD also has commands to apply different digital filters dynamically as the music plays, as I recall it was linear vs. minimum phase depending on the shape of the waveform. For example use linear phase most of the time but shift to minimum phase for highly transient/dynamic sections like castanets, string plucks or cymbal crashes. I don't know how often that is actually used.

 

[levimax]

Isn't the elephant in the room the fact that NO popular music track past or present has ever even come close to using all the dynamic range a regular 16 bit CD can offer?

 

[DrCWO]

Of course this is much ado about 1 measly bit. But it does work.

I think they only can use one bit. Otherwise shifting the signal higher up will result in too high compression distortions at play back on non- HDCD devices.

 

[MRC01]

Isn't the elephant in the room the fact that NO popular music track past or present has ever even come close to using all the dynamic range a regular 16 bit CD can offer?

True, however some rare tracks use 50+ dB of dynamic range, and shifting these quiet parts up one bit increases the resolution and lowers the noise a little. So while the full range could be captured in standard 16 bit, HDCD slightly improves the SNR of the quietest parts. So the effect may conceivably be audible in some tracks.
However, I agree that even though HDCD is based on legit engineering, it was mostly a marketing gimmick. Most CDs that use the dynamic range extension could simply have been shifted down 1 bit or 6 dB quieter and the full range captured on CD. The compression and HDCD encoding was not necessary.

 

[jcarys]

HDCD was correctly reviewed at the time of release as having no improvement over regular CD. The only reason I bought them was because Reference Recordings used it as their release format. I bought the discs because I liked the quality of the recordings, artist and repertoire. If there are higher resolution original recordings available, my default assumption would be that Qobuz or one of the other hi-rez streamers would offer it.

 

[MRC01]

HDCD was correctly reviewed at the time of release as having no improvement over regular CD. ...

Technically speaking, HDCD did have an improvement over CD. Measurable and based on sound engineering. And pretty clever too, using the LSB to encode extra data, folding dynamic range in a way that can still be played and sound reasonable without a decoder. It's just that the benefit was of marginal value and had little or no practical use.

 

[Chrispy]

All are using specific hdcd playback chain?

 

[MRC01]

If you play an HDCD on a player that doesn't decode, then the extra HDCD information is simply part of the dither noise down around -90 dB (since it's embedded in the LSB). Many HDCD encoded discs did not actually use any HDCD features, so they play back exactly the same with or without a decoding player. It happens that Pacific Microsonics made equipment that was used in the recording chain at some studios, and this equipment was HDCD enabled, so it embedded the HDCD codes in the LSB even when none of the HDCD features were used. When played on an HDCD player the HDCD icon lights up even though it's not doing anything.

Other discs did use HDCD features like dynamic range expansion and filter selection. Several from Reference Recordings, as mentioned earlier. Some of these, for example Eiji Oue directing the Minnesote Orchestra playing Pictures at an Exhibition (RR-79CD), actually do use HDCD features like dynamic range extension. Others do not. Seems kinda random whether or not they did. However, I'm glad my Oppo BDP-83 decodes HDCD, because I prefer the full uncompressed dynamic range. The difference is easily audible (quiet parts 6 dB louder - yuck).

 

[tmtomh]

Hi *,
I recently listened to the album "Romantically Helpless" by "Holly Cole" and liked it.
Next I saw that I could buy an HDCD (>$50) in Japan, which might sound better than the 44/16 version offered by Qobuz.

The question arises: Will that be worth the price?

I have some HDCD albums decoded by dBpoweramp to 24Bit so I evaluated quickly what might be the improvement.
The album I picked for evaluation was "Local Hero" by Marc Knopfler.

This is what I did:
- Decoded the original WAV (44/16) with dBpoweramp to 44/24.
- Opened both, the original WAV and the decoded in a wave editor.

Starting with track six "The Rocks and the Thunder" the 16 and 24 bit version were exactly the same but with different volume.
I recognized that this is quite a silent track with all peaks below 0.45 (1.0 equals 0dBFS).

Next I analyzed track tree "Freeway Flyer" and there I found some few differences between the two versions.

View attachment 256712
* The red signal is the original CD rip (44/16).
* The brown signal is the 24 Bit output of dBpoweramp after a volume increase of 2.5dB to fit both signals as good as possible.
* The light green signal is the difference between both.

We see that the HDCD carries additional information that was suppressed by mastering the 44/16 version.
Results:
* The 44/24 signal created by dBpoweramp HDCD decoding reverts compression done for the 44/16 signal above -0.5 dBFS.
* The 44/24 signal is 2.5dB quieter so you can get a disappointing 0.5 bit higher resolution (which means 16.5 bit) which is far away from advertised 20 Bit for HDCD!

As there is a lot of negative talk in the net regarding the quality of the dbPoweramp's HDCD decoder I also tried with HDCD.exe that can be downloaded here.

This is the result with the same Track:
View attachment 256714
* The red signal is the original CD rip (44/16) as above.
* The green signal is the 24 Bit output of HDCD.exe after a volume increase of 6dB to fit both signals as good as possible.
* The blue signal is the difference between both.


As you can see, the result is absolutely identical. The only difference is that HDCD.exe lowers the signal by 6dB and you could theoretically gain 1 bit instead of 0.5 bit.

So the answer to my original question is clear to me:
I don't think it's worth paying $50 for the original CD for that half, probably inaudible, bit.

Thanks for doing the legwork to produce all this information - much appreciated!

As you and others have noted, the HDCD you analyzed here uses both the Peak Extend feature and the Low Level Gain feature. The former is responsible for the higher, un-clipped/un-limited peak in your example, while the latter is responsible for the 3-4 lower-volume quiet parts in your example.

I agree with you that these small differences are likely not worth paying good money for an HDCD version, but even more than that, not a whole lot of HDCDs out there use both, or even one, of these features. So there are plenty of HDCDs where your graph would show no difference at all.

Even worse, there are CDs out there - some marked as HDCD on the package or disc, some not - where the CD was produced from an HDCD master source (or I suppose from a rip of a an HDCD disc), but it was remastered or otherwise messed with during production, which scrambled some of the HDCD flags or encoding. The result with such discs is that some tracks on them will trigger HDCD decoding in an HDCD-capable player, while others will not. So in some cases you'll have an album on CD and the overall volume level will suddenly change from track to track as the CD player detects, and then doesn't detect, the HDCD flag during playback.

HDCD's goal of fitting 20 bits into a 16-bit format seems to me to have been a pretty good concept, but the reality was unfortunately what @levimax notes above: even when peak extend and low level gain were enabled, it was not to fit hugely dynamic content into the 16-bit CD format - it was to provide an option to undo needless compression and limiting applied during the mastering stage.