HDCD in Practice

[MRC01]

I found an example among my CD collection so I'll share it here. It is Reference Recordings RR-79, Pictures at an Exhibition, starting on track 9. I played on my Oppo BDP-83, whose settings permit enabling or disabling HDCD encoding, and captured the digital output on a Tascam recorder at recording level "max" so it wouldn't change the bits. The recorded non-decoded bitstream shows as "HDCD enabled", so the Tascam made bit-perfect copies.

I'll cut to the chase with some before/after pictures, hopefully worth 1,000 words:
First, the non-decoded waveform - what you get without an HDCD player. It still rates a decent DR15.

Next, the decoded waveform, which rates DR19.

In both cases (with and without HDCD decoding), peak levels are a smidge below 0 dB. With HDCD disabled, the quiet/average levels are 6 dB louder and the dynamic peaks are softly compressed. With HDCD enabled, it shifts the quiet parts / average levels down 6 dB quieter which makes room for the expanded dynamic peaks. Another way to put this: when you enable HDCD encoding in settings, it plays much quieter.

With quiet parts shifted 6 dB (1 bit) lower, HDCD seems to trade resolution in quiet parts for bigger dynamic range. That is, shifting the quiet parts down by 6 dB means they have 1 bit less resolution, in order to make the dynamic peaks 1 bit louder. Perhaps the low level extension feature was intended to fix that and restore resolution in the quiet parts. Or, perhaps the DAC was supposed to do all of this in 20 or 24-bit so it doesn't lose resolution when it shifts the quiet parts down.

Note: The BDP-83 decodes HDCD on both analog & digital outputs. But my DAC says the digital outputs are still 16-bit even when HDCD is decoded. Looks like Oppo realized (or assumed) that none of the HDCD discs actually use more than 16 bits of dynamic range, so they didn't use 24-bit output. This player does emit 24-bit data for DVD-A, so the 16-bit output for decoded HDCD seems like a software choice rather than a hardware limitation.

 

[JoachimStrobel]

Tidal has quite a few Reference Recordings titles that first came out as HDCDs. I have read that these need no HDCD decoding for playback--presumably, the label supplied Tidal with unencoded 20- or 24-bit "masters" (whatever that may mean in this context). Does anyone here know more details? Thanks.

I see HDCD CDs on Qobuz that are not labeled as such and are only provided as 16bit. Tidal seems to be a step further there.

 

[MRC01]

The big question is: have they been decoded? That is: are these services streaming the "raw" bits from the CD? If so then they have not been decoded, and the device you're using to listen to them may be able to decode them (so long as the bits haven't been changed, for example with digital volume controls). Or are these services decoding the HDCD info and streaming the resulting bits?

If you're curious, if your streaming service has the album I showed above, you could capture it while streaming and compare to answer the question.

On a related note, my Oppo BDP-83 also decodes pre-emphasis in the digital outputs. That is, it applies the treble attenuation digitally, which seems to belie the purpose of pre-emphasis, which was to increase the bit depth of the treble (it was supposed to attenuate in analog after D-A conversion). However, I believe the original reason for pre-emphasis was that early ADC and DAC weren't fully 16-bit, but more like 14-bit. With full 16-bit resolution, applying the attenuation curve digitally seems harmless. Unlike HDCD, which you can turn on and off in settings, detection and decoding of pre-emphasis is automatic with no way to disable it.

 

[JoachimStrobel]

So true, my old DAC has a light once DeEmphasize is on. I saw it only on some mid-80s CDs.

 

[MRC01]

I found an HDCD in my collection that seems just "wrong". It's RR-68CD, Mozart's piano concertos played by Istomin.

Why "wrong"? When HDCD encoding is enabled the overall level is reduced 6 dB to make room for extended peaks. However, those peaks never actually happen. When decoded, the highest album peaks are -9 dB (-3 dB when not decoded). If I shift the HDCD encoded version up 6 dB it is exactly the same as the non-decoded version. So essentially, they gave up 1 bit of resolution across the entire album for no good reason, since the MSB is always zero (each sample essentially shifted down 1 bit). One could say they gave up 2 bits because the LSB is used for HDCD encoding. With this album, HDCD is counterproductive, actually lowering the resolution.

So, some HDCDs are best left not decoded! Whether to decode depends on the album. For this one, I ripped the files, boosted them by 3 dB so the peaks are just under 0 dB and this will also stomp the HDCD encoding. Then re-burned a new CD.

 

[MRC01]

PS: as I analyze more of the HDCD in my collection I can sum it up in one pithy phrase: HDCD is not beneficial often enough to always decode it, and it's not detrimental often enough to always ignore it.

This puts it into the worst-case category: you need to decode each HDCD disc and compare it with the non-decoded version to decide what is best. Repeat for each individual disc, because they all use it differently.

 

[Guermantes]

If you are decoding to a 24-bit target format or data stream, I don't see the problem with decoding the HDCD via software, even if the peaks are down. It is frustrating that the mastering engineer decided not to use the peak extension or gain adjustment features, thus wasting 1 bit of dynamic range for the sake of an HDCD label.

I just checked one HDCD I have: Laurie Anderson's "Life on a String" in Foobar2000 and it seems that no extension features or filters were enabled by Bob Ludwig when he mastered it. So I decoded a couple of tracks to 24 bit WAV with Foobar and the peak levels are exactly the same in the undecoded and decoded versions. In this case nothing was lost by decoding to 24 bit LPCM -- but nothing was gained either.

 

[MRC01]

Yes that's what I'm finding - most of my HDCD discs don't actually use any HDCD features so it's just a marketing label.

I found another that uses HDCD to shift the level down 1 bit, but don't actually use peak extension: Reference Recordings RR-2101. It's a 2-disc set. Disc 1 shifts does the above (loss of 1 bit for no good reason), the second does nothing - the encoded data is exactly the same as decoded.

Back to the topic of this thread: I'm finding that in practice, most of the time HDCD is just a harmless marketing label that does nothing (no features used). Sometimes it actually lowers the resolution of the music. Sometimes it uses peak extension. But even in both of these cases, the music itself doesn't have anywhere near 16 bits of dynamic range, so it doesn't matter. There's no loss or gain of information or resolution either way. When you shift it down you don't lose anything because the bottom few LSBs were already essentially random noise or dither. When you use peak extension you don't gain anything because the original uncompressed wave would have fit into 16-bit without compressing the peak and using HDCD to re-expand it.

 

[JoachimStrobel]

HDCDs provided early audiophiles the illusion that there was progress in reproduction technology. It eventually lead to DVDAUDIO, SACD, Blu-ray and 96/24 streaming. I would not be surprised if HDCD did not accomplish anything sound wise, but the same can be said for 3Star Michelin Restaurants (foodwise).

 

[Angsty]

I still use my 12-yr old CD player and I own some HDCD discs. I’ve switched to an outboard Modi 3 DAC for that player.

How would I know if a contemporary external DAC supported HDCD decoding? I never see it mentioned anymore. I’m not pining for it, but just curious.

 

[mhardy6647]

I still use my 12-yr old CD player and I own some HDCD discs. I’ve switched to an outboard Modi 3 DAC for that player.

How would I know if a contemporary external DAC supported HDCD decoding? I never see it mentioned anymore. I’m not pining for it, but just curious.

You're not the only one who is curious about such things -- if that makes you feel any better

 

[tmtomh]

I still use my 12-yr old CD player and I own some HDCD discs. I’ve switched to an outboard Modi 3 DAC for that player.

How would I know if a contemporary external DAC supported HDCD decoding? I never see it mentioned anymore. I’m not pining for it, but just curious.

I'm not aware of any currently in-production outboard DACs that support HDCD decoding.

Since HDCD decoding happens purely in the digital domain, it is possible that a CD player like yours could still be decoding the HDCD source in the digital domain before outputting the digital signal that feeds your DAC. Does your CD player have an indicator light or something on its on-screen display to show when it's playing decoded HDCD content? (Even if it does, it could be possible that it doesn't light up or show that indicator when it's outputting through its digital outputs, even if it is doing the decoding.)

I don't know exactly how these things work, but for example the Oppo 203 and 205 universal players don't support HDCD while the older 103 and 105 units did, and that was not because of the DACs in the units - it was because the new generation of the central processing circuitry (the MediaTek board Oppo and others used in their players) stopped supporting it. That suggests to me that HDCD compatibility comes before the DAC in the signal chain.

I also just Googled and found a thread in another forum in which one person hypothesizes that if you use an HDCD player's digital outputs you lose HDCD capability because "the decoding happens in the DAC." However, another person responds to that saying they have an HDCD capable Sony player and when they use its optical digital outputs, it does output an HDCD-decoded digital signal to an outboard DAC. (They say it doesn't do so over HDMI, which implies they are talking about a Sony DVD, Blu-Ray, or universal player.)

But again, I'm not an expert and I could be completely upside-down on this. I am guessing someone will come along soon to give us a more definitive answer.

 

[MRC01]

... Since HDCD decoding happens purely in the digital domain, it is possible that a CD player like yours could still be decoding the HDCD source in the digital domain before outputting the digital signal that feeds your DAC.
...
I am guessing someone will come along soon to give us a more definitive answer.

I've A/B tested to confirm that my Oppo BDP-83 decodes HDCD in the digital domain. That is, if you run its digital outputs (toslink or coax) to a DAC, you'll get the HDCD decoded version of the music.

PS: I tested by capturing the digital output of the BDP-83 while playing an HDCD disc twice: one with HDCD enabled in the BDP-83's settings, the other with it disabled. Then comparing the waveforms.

This seems "wrong" because when the BDP-83 does this, it maintains 16-bit digital outputs. The point of HDCD was to increase resolution, so decoding while keeping it as 16-bit undermines that goal. That said, no HDCD music that I've ever seen actually uses 16-bits of resolution even when decoded, so it shouldn't hurt anything. Yet since standard 44-16 can store the decoded version, it shows that HDCD really serves no purpose. Other than marketing: licensing it and having an icon pop up to make audiophiles feel better knowing it's "high definition".

I find it disappointing because that means HDCD was counterproductive. They apply dynamic range compression or soft limiting, so that the full dynamic range can only be heard on players that paid for the HDCD license. Yet that full resolution easily fits in standard 44-16 encoding, so it's entirely unecessary. The HDCD label might as well read, "WARNING: the full resolution of this CD was intentionally obscured so it can only be heard on HDCD players."

So the good news is that most CDs having the HDCD label don't actually use any HDCD features. The HDCD label only means a Pacific Microsonics AD converter was used; it doesn't mean they actually applied any HDCD encoding features, and they usually don't.

 

[MRC01]

PS: incidentally, the BDP-83 also detects and decodes CD preemphasis in the digital domain, also in 16-bit resolution.

 

[DSJR]

Going back twenty odd years, I found a few of the HDCD discs we had on dem all had a kind of 'reverberant acoustic' reproduction, as if reverb was added in an inviting 'nice analogue' way (as in a vintage Koetsu cartridge played through a top ARC preamp, minus the background noise) and not as natural to me as a nice 'dry' master recording. Maybe our UK sourced discs were sh*t and maybe so were the players advertising this (Arcam mostly I recall), but we all thought it was a bit of a gimmick and not worth pursuing frankly. Maybe more recent playback devices do it better or even properly? I can't remember the title, but one trendy disc of the time that advertised HDCD wasn't! - or at least it didn't switch the front panel indicator lamp on...

 

[DrCWO]

Just came back from digging trenches in our garden for sprinkles so definitely not that willing at the moment.

One line summary of what HDCD aimed was to encode 20 bit content in 16 bit CD format. They took the advantage of the fact that the dynamic range of music is not always 0 to max. So they used a sliding scale to shift the available 16 bit of data up and down (volume expansion). This signal which is a few Kilohertz/second is encoded in the rightmost bits. The decoder extracts these bits and applies the range shifting automatically. Non HDCD decoders just lose the low order bits to noise.

I disagree, please look at that. This cannot be done by a low frequency volume shift

A little exploration of the benefits of HDCD

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...

www.audiosciencereview.com