1) The ffmpeg HDCD filter is very much based on the Kode54 foobar work, but it has advantages over foobar in being multiplatform and part of a large and otherwise useful audio processing framework.
For an example you could:
ffmpeg -i "Silver_&_Gold.flac" -af hdcd /run/shm/OUT.flac
and get (along with a 24-bit version) these interesting stats:
HDCD detected: yes, peak_extend: enabled, max_gain_adj: -4.0 dB, transient_filter: detected, detectable errors: 0
The foobar scanner is nice, but it is not script-able, and Windows only, and there's no way to export the data. So for me, ffmpeg is where the data comes from, and there is even more interesting data if using -v verbose. And this is the kind of thing that interests me.
Thanks for the info. I was unaware of the capabilities of ffmpeg. It sounds like some nice advantages are there:
a) Multi-platform
b) Can scan the entire track
However if you see a max gain adjustment of 4 dB in a track, you can be certain it was just a fade-out in a song. I have looked at many,
many tracks with Foobar, and the
only time I've see the LLE come on with non-classical music was a fade-out.. Foobar has some really nice bar-graph style peak level meters and you can see the levels. LLE doesn't engage until the level reaches -45 dBFS, which essentially never happens except with fade-outs in non-classical music. From the PM operator's manual:
There are two modes of Low Level Extension, “Normal” and “Special”. Normal mode begins
to affect the input signal 45 dB below peak level, gradually raising the gain 4 dB as the
level drops over an 18 dB range. Special mode begins to affect the input signal 39 dB
below peak level, and gradually raises the gain 7.5 dB over a 26 dB range. Normal mode is
optimized to provide the best combination of decoded dynamic range and resolution and
undecoded compatibility. Special mode is designed to provide the best possible decoded
dynamic range and resolution at some potential expense of undecoded compatibility. To
access Special mode, from the Operating Menu select (SETUP/OUTPUT/HDCD_16/LOWLVL/
SPECIAL). Typically, Special mode is used only for HDCD 16-bit master tracking with the
assumption that the recording will be decoded by the Model Two to a 24-bit or 20-bit word
length for digital post production before being re-encoded to HDCD 16-bit using Normal
mode to produce a release master.
If you see a level of greater than 4.0 of gain adjustment, there was an error made in the disc where the mastering engineer
mistakenly used "Special Mode". This is not only wrong, but there is no consumer equipment that can even decode this level of gain adjustment! Only by playing this back through the PM Model One (44 and 48 kHz only) or the PM Model Two (added dual- and quad-rate sampling rates) could this file be properly decoded...
The LLE does not begin until the level reaches -45 dBFS and then reaches a maximum adjustment of 4 dB as the peak signal level reaches -63 dBFS. This
only happens with fade-outs and is not worth decoding. On classical music I have seen very low-level passages with one or two instruments playing quietly where the gain adjustment changes between 0.5 dB, 1.0 dB, and 1.5 dB. I have never seen a case where the gain adjust reaches even 2.0 dB. It takes -18 dB to trigger a 4 dB gain adjustment, which means that I have never seen even a super high-fidelity classical recording reach a level of -9 dB below -45 dBFS = -54 dBFS. It is certainly possible that such a recording exists, as although I've examined scores of HDCD recordings, I've certainly not examined every single one ever made.
But remember that a CD only has a maximum S/N ratio of roughly 96 dB. By the time the recorded level gets down to -54 dBFS, the signal-to-noise ratio would only be 42 dB - hardly "high fidelity". That is precisely why modern formats use 24 bits instead of 16. The entire premise of HDCD was largely market-driven and misleading. Even in the very best case, HDCD could only add about 1.4 bits of resolution, and rarely added even 1 bit of resolution.
The real reason HDCD recordings sound so good is that the PM A/D converter was the first really good sounding A/D converter on the market. When it came out, the only competition it had were the Sony units, which were in large part responsible for much of CD's reputation for "harsh" sound. By the late '90s there were a few more good sounding A/D converters available, and today there are many,
many more. But the PM is still one of the better choices available, even today. However in 2009 when The Beatles CDs were remastered, the folks at Abbey Road chose the Prism as the best sounding one in their DAC shootout.
There are two reasons to care about HDCD at all in 2016. First is that if it was recorded with a PM A/D, it will likely sound very good - especially for recordings made between 1998 and the early 2000s. (By the mid-2000s there were many other good sounding choices available.) It's the same thing with all audio equipment. Some amplifiers sound better than others. Some CD players sound better than others. Some portable players sound better than others. Some software players sound better than others. We don't always understand the reasons why, but audible differences exist in pretty much everything.
4) Thanks for confirming that nothing is to be done with TF, but...
Regarding only-TF being "fake" HDCD: As I see it, the HDCD encoding is still there, even if no features are used. Even if every HDCD control code signals PE off and gain adjust 0.0, that is different from a CD that has no HDCD codes at all. The control code has a built in error check to confirm a valid code. The no-op is still a code, and it still must have passed through HDCD encoding equipment to have tens of thousands of such error-free no-ops embedded. In this case the HDCD encoding may add no value, but it is still there. For that reason I included them on the list, but made their HDCD detected "Yes" yellow instead of green, as described in the key at the top of the page.
In support of my assertion:
Neil Young - Silver and Gold: 30040 error-free HDCD packets in each channel, all signaling at least PE.
Silverchair - Diorama: 34279 error-free HDCD packets in each channel, all no-ops.
Beck - Midnight Vultures: (chan0:49738, chan1:50051) error-free HDCD packets, either no-op or TF only.
521 other non-HDCDs combined: 12 apparently error-free HDCD packets, and 8 broken ones, all by random chance.
Thanks again for your information. I'm guessing you are the founder of Ayre Acoustics, which I only know because I wrote a stub article about that company at Wikipedia last year sometime while actually writing about the Pono Player technical specs. I hope you don't take my minor quibble about HDCD decoding still being possible (even if useless) on TF-only discs as an insult. I have no doubt that you are an expert.
-bp
No, the flashing TF are still "fake" HDCDs simply because there is no player on the planet (including the PM Model One and Model Two professional units, which were combination A/D and D/A converters) that had two different playback (reconstruction) filters for HDCD. While the PM A/D used two different anti-aliasing filters while performing A/D conversion (depending on the level of high-frequency content), they were precluded from using two different playback filters as Ed Meitner had already patented that idea for Museatex (now out of business). Since there is
nothing to decode, one does not need any special type of player or software to achieve proper playback of the HDCD files when the TF flag triggers.
As noted in the thread, it is mildly interesting to know that a CD was made with the PM A/D converter. However
any disc made with the PM A/D converter will light up the "HDCD" light, as they always injected the "subcode"
even when there was nothing to decode. In my opinion this was likely a misleading tactic deliberately taken by PM to scare people into buying an HDCD-equipped CD player.
Further, the actual use of LLE only occurs in non-classical music during fade-outs, and only extremely rarely during classical music - to the point where it arguable whether it is even worth decoding. The
only mode that arguably benefits from decoding is PE. Decoding PE will restore the full dynamic range of the recording, as PE "squishes" the top 9 dB of signal into 3 dB of codes on the disc. If not decoded, it is like adding "compression" in the studio - a tool that many people
like the sound of. Again, from the PM operator's manual:
Using Peak Extension allows very high average recording levels without “clipping” or gen-
erating “overs”. This approach can be used to get the “hottest” possible sound (almost no
dynamics) during undecoded playback for air play, with decoding restoring normal dynam-
ics for home listening.
However, because Peak Extension limiting has an “easy over” curve that begins to affect
the signal at - 3 dBfs, it usually shouldn’t be used with highly compressed source material
that will almost always be in the limiting curve, unless a highly limited or distorted sound
is desired during undecoded playback.
Typically, Peak Extension recordings do not have the “crunch” or “edge” produced by hard
clipping that is sometimes desired for certain types of rock material.
Even Pacfic Microsonics recommended against using PE for many types of pop music! The advantage of decoding it is that you will hear it exactly as intended. The disadvantage is that after decoding, the signal level can be reduced by anywhere between 0 dB and 6 dB - depending on how the mastering engineer set the levels during the final mastering process. If you decode an HDCD track that had PE engaged, you may want to increase the volume level slightly. Based on the many tracks I have analyzed, I would suggest doing it by ear, but usually +1 dB or +2 dB is sufficient. I've not seen any tracks where I felt that the PE compressed the music so much that it would require a full +6 dB of gain adjustment.
But I always find that each album requires a different playback level anyway. I always play Led Zeppelin at higher volumes than Pete Seeger, for example. For that reason I never use "volume leveling" features of any kind. I suppose for "background listening" it might be useful, but I never listen to music as "background noise". In fact my experience is that a good stereo makes "background listening" impossible, as Tyll Hertsen noted in his review of the Pono Player:
I had started with a few very familiar test tracks and quickly determined they sounded
pretty good and I should just move on to some casual listening as I wrote. I pulled up
an old favorite, Lyle Mays' (keyboard player on many of Pat Metheny's albums) "Street
Dreams." Three or four tracks into it I realized that my eyes were closed and my fingers
had been hovering over the keyboard unmoved for the last ten minutes...I had been
sucked into an old favorite and was
way inside this particular album again for the first time
in a long time.
"What the...."
I shook it off, and tried another old favorite; this time Carmen McRae's "Carmen Sings Monk".
First cut on the album is a live recording of "Get it Straight." There's an instrumental break
in the middle of this short 3:55 tune, and a little bass solo that starts at 2:12 that has a
spine tingling run of notes starting at 2:28 in which you can hear audience members gasp
at the crescendo of notes and intensity. I was swept right along until, as the song wound down,
I opened my eyes to see, yet again, my fingers hovering motionless above the keyboard.
"Damit, I'm not going to get anything done with this Pono gadget playing in my ears."
YMMV
Hope this helps!
And I can confirm what CHansen says that the file size difference is really not that big: a 16-bit HDCD-encoded FLAC of 30.2MB becomes 45.6MB when decoded to 24-bits with the method I described. Thank you for the new list, bpzero! We really needed a list that separates the "true" from the "fake" ones! Also, foo_hdcd has been updated and has fixed some audio-channel errors with some HDCDs (e.g. some that were falsely decoded with a louder left channel), so I'd suggest everyone to download the latest update!
