HDCD list
Jan 14, 2011 at 11:09 PM Post #514 of 595
I don't know if anyone will read this. This is very important information for this list of HDCDs.
 
The problem is that there are a lot of "fake" HDCDs out there. I have around 2000 albums on my hard drive. About 15 or 20 were HDCDs. But of those discs, over half were "fake" HDCDs. Please allow me to explain.
 
Pacific Microsonics (PM) invented the HDCD process in the early '90s. After that they started selling a special A/D converter that would include the HDCD codes. This converter was called the PM Model One and was limited to 44.1/16 output. Then in the early 2000s they introduced the Model Two that would go all the way up to 192/24. That was the only substantive difference.
 
These A/D converters were the best sounding converters available at the time and are still considered among the best today. They were designed by Keith Johnson, who designs most of the Spectral super high-end audio gear. In addition to just sounding incredible, they applied the HDCD encoding process, which can be broken down into the following items:
 
a) On-the-fly selection of the low-pass anti-aliasing filter. The machine would analyze the content of the music being converted and choose one of two filters to achieve the best sound. Contrary to the internet rumor machine, this does NOT require any decoding. There never has been and never will be multiple playback filters for the HDCD process.
 
b) Selectable dither levels and shapes. This was quite advanced for its day as most people were not aware of the importance of proper dither in capturing low-level signals. Again, there is NO decoding required or even possible for this step.
 
c) Perhaps the heart of HDCD was the use of two separate "compansion" (compression/expansion) methods. The idea is that the dynamic range of the music is compressed during the recording phase and then expanded back upon playback. One of these is Low-Level Extension (LLE). When the peak music level reaches -45 dBFS, the system boosts the gain by 0.5 dB. As the level drops the boost increases, to a maximum of 4.0 dB at -63 dBFS.
 
In theory this should be decoded for proper playback. In practice it almost never is used. Even quiet passages of classical music are louder than -45 dBFS. I have seen a handful of discs where the LLE becomes active for a few seconds as the song fades out. I have seen one classical disc where the LLE was active during the music. There were solo piano chords that faded away to nothing before the next chord was struck. Even with these I doubt that one would hear much difference between decoding them or not.
 
d) The other half of the equation is Peak Extend and it is active at very high signal levels. It really should be called "Peak Compression", because that is what it does. It takes the top 9 dB of the recording and compresses it down to fit into 3 dB on the CD. Then an HDCD decoder will expand it back to the full 9 dB of range. This is probably the only feature that might require decoding to achieve the best sound on an HDCD disc. I say "might" because unless the decoded music reaches a level of -9 dBFS, the decoder won't do anything. About half of the albums I examined that use Peak Extend (PE) don't ever reach levels that high. Examples include most of the early Joni Mitchell catalog. The other half of the albums that use PE have peaks that reach as high as -1 dBFS and will definitely benefit from decoding.
 
The reason that I know all of this is because there is a plug-in for Foobar (the Windows-based music player) that will identify which features were used on any album. Configuring the display is a little tricky. I posted some instructions at:
 
http://diyhifi.org/forums/viewtopic.php?f=2&t=2124
 
Now here is the kicker. When the mastering engineer is making an album, he can choose to use PE, LLE, both, or neither. Many, many albums are made using NEITHER. But the PM A/D converters still put the secret "subcode" into the bitstream that will light up the "HDCD" light even if there is NOTHING to decode!
 
That is why I said that about half of the "HDCD" albums I have (that light up the "HDCD" light) are "fakes". They will sound exactly the same on any CD player because there is nothing to decode. (Well, actually every CD player sounds different, but that is another topic for another day.)
 
So I hate to break the news to you. The fact is that this list of HDCD discs is just plain wrong. There are many titles that, while made with the PM A/D converter, were not made using any of the HDCD features and do not require decoding. For example on the first post in this thread, "Patsy Cline's Greatest Hits" was listed. This is not an HDCD disc and requires no decoding whatsoever.
 
Simply putting a disc in and seeing the light is not enough. The only way to know for sure is to use Foobar with the latest HDCD plug-in. PE is very simple to interpret -- it's either on or off. LLE is a bit trickier. Watch the gain setting in between tracks for a quick indication of whether it was used.
 
Sorry to break the bad news. It is just part of PM's slimy marketing practices. Rather unfortunate, as they had the best engineers in the world along with some of the sleaziest marketing guys in the world. But finally there is a way to know for sure how a disc was recorded and if it actually needs decoding. (This is why there are so many discs with no marking on the jacket that will light up the "HDCD" light. The mastering engineer used the PM A/D converter because he liked the sound, but didn't use any of the HDCD features because so few players have the decoders.)
 
Jan 15, 2011 at 12:53 PM Post #515 of 595
Exactly right. Thank you for the incredibly informative post, CHansen! Especially interesting was the part about most of the advantages of the PM A/D converter (low-pass anti-aliasing filter, dither levels and shapes) only taking place during the encoding stage rather than the decoding stage.
 
Have you tried hdcd.exe to decode HDCD? With what you know about how the HDCD process was actually used in practice, is hdcd.exe good enough?
 
Jan 15, 2011 at 5:21 PM Post #516 of 595
Thank you for the kind words. I'm glad that they were helpful.
 
When I looked at HDCD.exe, I plugged the values for Peak Extend into a spreadsheet so that I could look at the graph. My sense was that they were created by trying to read the values from the graph in the AES paper or something. They seemed close but not exactly perfect.
 
The Foobar plug-in did not use HDCD.exe, but apparently some libraries that were found when reverse-engineering Windows Media Player. Presumably that should be perfectly accurate, but again looking at the values for Peak Extend in a spreadsheet didn't give me confidence they were correct.
 
Either one may be, but they are certainly not the values I would select for encoding the Peak Extend curve. But who knows what the original designers did? I hope at some point to analyze the output of the PMD-100 digital filter to see for sure what the correct decoding values are, but that is not a trivial project.
 
The other thing that is critical for proper decoding is to know the time delay between reading a "subcode" packet and subsequently applying the gain change. This is only important for Low-Level Extension, as Peak Extend is either "on" or "off" for the entire song, and normally for the entire album.
 
But the take home lesson is that just because the "HDCD" light turns on, doesn't mean that the disc is HDCD encoded. So there are going to be many false positives in the current list. There are a couple of indirect and cumbersome ways to tell if PE was applied, but the latest Foobar plug-in will give a direct read-out instantly on PE, and a direct readout on LLE with only a little bit of work (looking at the silence between tracks).
 
Jan 16, 2011 at 3:18 PM Post #517 of 595
 Thanks a million for all the info! This helps me understand why so many hi-end/high priced CD players and DACs don't bother with HDCD decoding (including my Cambridge 840C which is as high end as I can afford). I was concerned that I might be missing something important. I know there are now higher rez sources, but so much music has been released only on CD.
 
On the other hand, what you explained about the pre-playback processing of HDCD strikes me as very worthwhile. The early literature on HDCD did claim that the majority of the benefits were built into the recordings and compatible with any CD player. It is also a big clue that the recording engineers and producers were aware of and concerned with sound quality. This is very frequently not the case unfortunately. HDCDs are, in my experience, associated with good sound. The Reference Recordings CDs I have are contenders for the best sounding CDs I have ever heard, though that has something to do with Keith Johnson and not just HDCD I'm sure.
 
Hence, I hope the list keeps getting updated and I see a few entries that I want to purchase.
 
P.S.  When you say "On-the-fly selection of the low-pass anti-aliasing filter", does that mean that it can optimize this in real time? (as opposed to setting the filter at the beginning of a track?)
 
Thanks again.
 
Jan 16, 2011 at 11:12 PM Post #518 of 595
Yes, the PM A/D converter has two anti-aliasing filters. These change according to some unknown algorithm based on the music being converted. One filter is optimized for flat frequency response and the other for improved transient response (less ringing). The only differences show up above 15 kHz or so. I would assume that it uses the filter with better transient response unless there is content above 15 kHz and then it would switch to the other filter.
 
If you load the HDCD plug-in into Foobar and write the proper commands to the display section, it will show  which filter is being used as the music plays back. Some instructions are available here:
 
http://diyhifi.org/forums/viewtopic.php?f=2&t=2124
 
You will still need a basic understanding of Foobar, which is not quite a plug-and-play installation.
 
Feb 3, 2011 at 3:46 PM Post #519 of 595
Thanks CHansen for your post(s); it helped me tweak my Foobar - now I can see the HDCD features implementation.
Here's an update to a few discs mentioned earlier in the thread, after a play-through with HDCD enabled Foobar2000...
 
Classical/Jazz
Jacques Loussier Trio -
Satie Gymnopedies - Gnossiennes * (Telarc CD-83431)  PE, LLE not noticed
Jacques Loussier Trio Plays Debussy * (Telarc CD-83511)  PE, LLE several tracks
Beethoven Allegretto from Symphony No.7 * (Telarc CD-83580)  PE, LLE not noticed
Mozart Piano Concertos 20/23 * (Telarc CD-83628)  PE, LLE not noticed 
 
( * = HDCD not indicated on disc, annotations, PE = Peak Extend, LLE = Low Level Extension)
 
Feb 6, 2011 at 5:11 PM Post #520 of 595
Yes, Foobar can be somewhat inscrutable. I was only able to figure out the correct "commands" by reading through a thread on Hydrogen Audio, which is the official support outlet for Foobar. I immediately received a warning from an administrator because I said that the Pacific Microsonics encoder was one of the best sounding ones ever made.
 
It turns out that it is forbidden to say that something is good or that you like something. You can only say that it measures better or worse than other products... I had to be very contrite and promise not to do it again to avoid being banned!
 
Anyway, it looks like the Telarc crew (which has now changed hands) was definitely using the PM A/D encoder. It seems a little odd that they would use the LLE on some discs and not others. But it may be that there were two different mastering engineers. It's also tricky looking for LLE because it rarely comes on, even when the mastering engineer turned it on. I usually just look at the last five seconds of a track, the silent time between tracks, and the first five seconds of the next track. If it was used, this is when it will show up.
 
You can turn on some really nice peak level indicators in Foobar by going to File - Preferences - Display - Default User Interface. Then press the "Quick Setup" button. Choose the bottom item on the list, "Visualisation + Cover Art + Tabs". The cover art is small, which is annoying.
 
Below the main playlist window is an area with a "Spectogram" or some such nonsense. Go to View - Layout, and select "Enable Layout Editing Mode". Then you can Right-Click in the box with the Spectrogram and select "Replace UI Element". There will be a huge list of items. Scroll down to "Playback Visualisations" and select "Peak Meter" ("VU Meter" is slowed down and is more of an averaging meter).
 
Now you can see the peak recording levels on the disc. (I guess it's actually a copy of the disc!) There are even some "peak-hold" bars that make it easier to see the highest peak levels. PE doesn't kick in until the level reaches -9 dBFS. The higher the level goes, the more effect decoding has on PE. A lot of classical pieces will often be cut at very low levels, with only a few passages that reach -9 dBFS (or above), so that HDCD decoding ends up not doing all that much.
 
But there is no question that it is better to have it when it is needed than not. It is also interesting to see how rarely it is needed and how many times there will be "fake" HDCD discs that require no decoding whatsoever. This Foobar plug-in is truly a great tool!
 
 
Feb 17, 2011 at 11:51 PM Post #521 of 595
Continuing my review of discs on my shelf for HDCD implementation...
 
Soundtrack
Music by/featuring Mark Knopfler -
Local Hero  (Mercury/Vertigo 811 038-2)
Cal  (Mercury/Vertigo 822 769-2)
The Princess Bride  (Mercury/Vertigo 832 864-2)
Last Exit To Brooklyn  (Mercury/Vertigo 838 725-2)
Wag The Dog  (Mercury 314 536 864-2)
Metroland  (Mercury/Warner 9 47006-2) 
 
The six above shows up with Peak Extend on. Wag The Dog and Metroland has a number of tracks triggering Low Level Extension (LLE); the other four didn't seem to have LLE.
The first four are non-USA pressings.
 
Feb 24, 2011 at 11:20 AM Post #522 of 595
Spinning this one while working...

Jazz
Duke Ellington: Piano In The Foreground * (Columbia/Legacy CK 87042)

While remastering this disc, it looks like the engineer used the Pacific Microsonics AD Converter (evidenced by the HDCD indicator showing up) on a number of tracks, but not all of them. No PE or LLE applied on the HDCD tracks, but the Transient Filter indicator flickers YES very often. Still, it's a great disc!
 
Feb 26, 2011 at 2:21 PM Post #523 of 595


Quote:
Jazz
Duke Ellington: Piano In The Foreground * (Columbia/Legacy CK 87042)

While remastering this disc, it looks like the engineer used the Pacific Microsonics AD Converter (evidenced by the HDCD indicator showing up) on a number of tracks, but not all of them. No PE or LLE applied on the HDCD tracks, but the Transient Filter indicator flickers YES very often. Still, it's a great disc!


The PM A/D converter is still used by many mastering engineers. I would assume that there were at least two different engineers used on this disc, as normally the same converter would be used on all of the tracks. If the PM converter is used, the Transient Filter will always flicker because it automatically selects one of two anti-aliasing filters depending on the spectral content of the music from moment-to-moment. But no decoding is required for this, nor is it even possible. No HDCD decoder has more than one playback filter.
 
We can only speculate as to why they included which anti-alias filter was used in the "sub-code", but nobody has ever figured out any use for that information whatsoever.
 
Nov 1, 2011 at 4:36 PM Post #525 of 595
Where it states the best of chris isaak.,is this the 2006 release.??
I would like to add two more to the list which do have the hdcd logo clearly marked on both the 'outer' packaging & the disc itself:
Chris Isaak;Baja Sessions
Bed of Roses;Soundtrack,from the Motion Picture
 

Users who are viewing this thread

Back
Top