[CHansen]

Quote:

 
 
When using the foobar script [%__bitspersample% bits] and playing HDCD encoded 16 bit files or the HDCD CD itself, 24 bits is displayed in the status bar (for me anyway). Do you think this is because of that 1 bits worth of extra range and foobar shows 24bit because it is not capable of displaying 17 or 20, etc?

Dear 51-50,
 
Yes, that is exactly the case. There was a period when playback DAC chips could only reach 16 bits. Then I believe it was Denon that released the first 18-bit chip. This started a numbers race. Within only a few years, EVERY DAC was claimed to be at least 24 bit, even the very least expensive and lowest actual performance chips. There was a "numbers" race that lasted only for a few years. For the last ten years or so, one cannot find an audio chip that is not rated at least at 24 bits of resolution.
 
Today a new "numbers" race is starting, and many DAC chips are claiming that they will produce "32-bit resolution". This is fairly silly in one aspect. 32 bits of resolution would reach a dynamic range of over 192 dB!!! This is absolutely not possible. But on the other hand it also means that it can accept up 32 bits of data from the oversampling digital filter. It is possible that this would sound very slightly better than the way that the same data is presented to a 24-bit DAC. The length of the data exceeds 24 bits, so it must be reduced to the correct length either by truncating, rounding, or dithering. All of these have some audible disadvantages, so it may sound better to reduce it to 32 bits, whereas the audible disadvantages of reducing the sample length from the digital filter all the way to 24 bits may have an audible disadvantage.
 
In addition, the processing inside the software player is done with bytes. One byte equals 8 bits. So it requires a lot of extra work to display the number of bit is increments less than 8 bit increments.
 
Hope this helps,
Charles Hansen
Ayre Acoustics, Inc.
www.ayre.com

 

[fiftyonefifty]

Quote:

Dear 51-50,
 
Yes, that is exactly the case. There was a period when playback DAC chips could only reach 16 bits. Then I believe it was Denon that released the first 18-bit chip. This started a numbers race. Within only a few years, EVERY DAC was claimed to be at least 24 bit, even the very least expensive and lowest actual performance chips. There was a "numbers" race that lasted only for a few years. For the last ten years or so, one cannot find an audio chip that is not rated at least at 24 bits of resolution.
 
Today a new "numbers" race is starting, and many DAC chips are claiming that they will produce "32-bit resolution". This is fairly silly in one aspect. 32 bits of resolution would reach a dynamic range of over 192 dB!!! This is absolutely not possible. But on the other hand it also means that it can accept up 32 bits of data from the oversampling digital filter. It is possible that this would sound very slightly better than the way that the same data is presented to a 24-bit DAC. The length of the data exceeds 24 bits, so it must be reduced to the correct length either by truncating, rounding, or dithering. All of these have some audible disadvantages, so it may sound better to reduce it to 32 bits, whereas the audible disadvantages of reducing the sample length from the digital filter all the way to 24 bits may have an audible disadvantage.
 
In addition, the processing inside the software player is done with bytes. One byte equals 8 bits. So it requires a lot of extra work to display the number of bit is increments less than 8 bit increments.
 
Hope this helps,
Charles Hansen
Ayre Acoustics, Inc.
www.ayre.com

 
Makes perfect sense CHansen, thanks for chiming in. I'm curious as to what you think about a flagged 16 bit HDCD file where there is no PE or LLE, just the TF ticks on and off. The foobar script [%__bitspersample% bits] displays this as 24 bit as well.

 

[CHansen]

Quote:

 
Makes perfect sense CHansen, thanks for chiming in. I'm curious as to what you think about a flagged 16 bit HDCD file where there is no PE or LLE, just the TF ticks on and off. The foobar script [%__bitspersample% bits] displays this as 24 bit as well.

 
Hello 5150,
 
There have been many attempts by several different people to reverse engineer the HDCD decoding algorithm. Every single one of them were successful in finding the three different flags in the "subcode" one for PE, one for LLE, and one for the TF. And of course, they all have "activity", so I think everyone just assumed that they needed to be decoded.
 
I have known about the sleazy marketing tactics of HDCD since before 2000, so when every one of the reverse engineering people were saying there was activity in the TF flag but that they couldn't figure out how to decode it, I started to look into it. I read every single piece of literature they put out (including the patents), and without exception, they implied that there were two conjugate filters, but they never actually flat-out said so.
 
This set my alarm bells ringing. Why not brag a bout such a feature if it really had it? My suspicions grew and I made several posts in public forums where I expressed by skepticism. Nobody every came out and correct or contradicted me, which made me even more suspicious. So I finally contacted one of the developers and asked him and he admitted that there was only one playback filter. (Although the Spectral 4000 CD player claims that an improved version of this single filter was developed by Keith Johnson for this machine. Please refer to http://spectralaudio.com/bulletins/0806.pdf )
 
I contacted him again to ask why they had included the flag? Were the planning to include two filters but run out of time and money before it could be implemented? This didn't make sense, as when the PMD-200 was introduced they could have easily have added it.
 
My own suspicion is that they wanted the HDCD system to be a single patented package. And Ed Meitner, then at Museatex, now of EMM Labs, had beaten them to the punch! Ed had already patented a two-filter playback system. So they could only patent the two-filter scheme on the record side! My guess is that they had planned to implement a two filter playback scheme, had already set up the encoding bits, and then either found out about the patent or possibly knew about the patent and were hoping to license but were unable to.
 
The bottom line is that just as the TF bits n the subcode are a left over legacy of some unknown yet unimplemented plan, the fact that Kode 54's HDCD plug-in decoder for Foobar expands to 24 bits are part of a left over legacy code. If Kode 54 gets bored some day, he could easily make some changes so that ONLY HDCD titles with PE and/or LLE are expanded. And 18 bits would be MORE than enough. It's just that there is no standard for an 18 bit (or even a 20 bit) format.  The next jump that most software players will handle is 24 bits.
 
Hope this helps,
Charles Hansen

 

[fiftyonefifty]

Charles - That helps alot. Great info there. Thanks so much!

 

[evolance7]

If you are still updating the list. Here is an album I owned with HDCD, Artist: September (Petra Marklund) - Album: September. 

 

[wedgeworld]

Don't know if your still adding but here are some notes. I have the king crimson 2004 COURT OF THE CRIMSON KING remaster and it is not HDCD, at least it doesn't say so. All the rest of those remasters are.
My Chris Isaak best of is not HDCD, again, at least it does not say so. This is the CD with DVD of music videos on Reprise UPC ending in 8129.

Here are some to add
Camel - Rajaz
Camel - A nod and a wink
Chris Isaak - Speak of the Devil
Chris Isaak - Baja Sessions
Bryan Ferry - I see you have his name under Roxy and just say entire catalog, but there are a few more recent ones that are not. I know Boys and Girls, Bette Noir are for sure. I can get details if you would like.

 

[Chris J]

Bruce Springsteen:
"Tracks" 4 CD boxed set
"18 Tracks" basically a single CD sampler of "Tracks".

 

[MickeyMouseDrum]

Jacky Cheung - Vol. 2 "A Whole New World" (3-disc set). It is labeled as HDCD and confirmed encoded on my H/U as well as outboard DAC. 

 

[MickeyMouseDrum]

BT - Movement in Still Life (Nettwerk Production) 
 
*USA and Canadian version, original front page list only mentions American version*

 

[CHansen]

Hello All,
 
Just a reminder -- there are lots of discs that were mastered using the Pacific Microsonics (PM) A/D and D/D converter, as it is a very good one. (In fact for many years, it was the only good one available.) Every single one of these discs will light up the HDCD light on a stand-alone CD player, but many of them require NO decoding, and therefore should not be listed as HDCD discs.
 
There are three features of the PM converter, PE (Peak Extend), LLE (Low Level Expansion), and TF (Transient Filter).
 
PE and LLE are optional, and can be engaged or disengaged by the mastering engineer at his discretion. TF will always be on, but there is nothing to decode (as explained a few posts above), so it can be completely ignored.
 
This means that many "HDCD" discs require NO decoding and can be treated like any other disc. In essence, they just use a very good A/D converter and that is all.
 
The only way that I know of to tell exactly what is going on is to use Foobar. (Make sure to have the HDCD plug-in installed.)
 
Go to "File - Preferences - Default User Interface" and copy and paste the following into the middle long box titled "Status Bar":
 
%codec% | %samplerate% Hz | $info(bitspersample) bits | %channels% | %playback_time%[ / %length%] | %bitrate% kbps | HDCD = $if(%__hdcd%,'yes | ','no')$if(%__hdcd%,PE: %__hdcd_peak_extend% LLE: %__hdcd_gain% TF: %__hdcd_transient_filter%)
 
Then when playing a file, the HDCD information will be at the end of the line. Non-Pacific Microsonics files will say "HDCD = No".
 
Files made with the Pacific Microsonics converter will have three values: PE, LLE, and TF . The transient filter will always be changing, especially when cymbals are being used. You can watch this while listening and easily see it change when there are high levels of energy in the top octave (10 kHz to 20 kHz). You can ignore this as there is nothing to decode.
 
If LLE is used on popular music, you will only see it briefly during a fade-in or fade out. It is only active when the level drops to -45 dBFS, which is VERY quiet. Occasionally you will see it come on during classical music, but only during a very, very quiet passage when only one or two instruments are playing. It engages so rarely that if it is used, it is really not worth the bother to decode it, IMO (if PE is engaged you will want to decode the disc for that and LLE will also be decoded at the same time).
 
If PE is used, it will be on for the entire track. This is the ONLY feature that is worth decoding, IMO.
 
The list of "HDCD" discs would be 1000x more useful if people would verify that a disc uses PE. If it does, one should decode it for best sound quality. If not, it really isn't worth the bother, IMO.
 
Thanks!

 

[interpolate]

It appears any Doors CD remastered and pressed in August 1999 and any subsequent SACD/DSD reissue is 24-bit capable in the right system. I just played the The Doors first album remaster through 24-bit and certainly it does sound a bit more dynamic. Using Windows Media Player 12 with Windows 7 Professional.

 

[Chris J]

  It appears any Doors CD remastered and pressed in August 1999 and any subsequent SACD/DSD reissue is 24-bit capable in the right system. I just played the The Doors first album remaster through 24-bit and certainly it does sound a bit more dynamic. Using Windows Media Player 12 with Windows 7 Professional.

 
The CD cannot be 24 bits.
It may have come from a 24 bit mastering process, but even HDCD is limited to, what? 17 bits?

 

[CHansen]

   
The CD cannot be 24 bits.
It may have come from a 24 bit mastering process, but even HDCD is limited to, what? 17 bits?

 
Excellent point. Pacific Microsonics (notorious prevaricators) always claimed "20 bits of resolution" for HDCD. That is nonsense.
 
The Peak Extend was 6 dB, or 1 bit, giving 17 bits.
 
The Low-Level Extension was a maximum of 4 dB, but this was never realized in pop music. The process doesn't even kick in until the signal level reaches -45 dBFS. Show me one single pop album with a level this low (besides a fade-out), and I will give you a dollar. Even on classical music the LLE only starts to kick in on VERY quiet passages with one or two instruments.
 
While this extends the total theoretical of HDCD to around 17.5 bits, in practice it virtually never happens. So where do they get off claiming an extra 2.5 bits of resolution that don't exist?
 
Dither.
 
They claimed that adding dither allows for signals below the noise floor to be retrieved. This is certainly true for steady-state signals. (It is less clear regarding transients.) But any dither will do the same thing. Even the earliest converters that did not have built in dither were dithered by the noise of the analog tape that was used as a source, if not the analog electronics themselves. Every modern converter has dither also, but none of them claim "28 bits of resolution" or any nonsense like that.
 
So 17 bits is about right for some HDCD. But remember that in the Operator's Manual, they recommend NOT to use PE for already dynamically compressed pop music (undecoded playback would be super compressed)....
 
There are probably MORE "HDCD" discs that had neither PE nor LLE turned on (therefore requiring NO decoding) than actual discs made that did require decoding. Most discs that light up the "HDCD" light require no decoding. The only way to know is to use Foobar. Put the following text in the status bar to see which features are ACTUALLY in use:
 
%codec% | %samplerate% Hz | $info(bitspersample) bits | %channels% | %playback_time%[ / %length%] | %bitrate% kbps | HDCD = $if(%__hdcd%,'yes | ','no')$if(%__hdcd%,PE: %__hdcd_peak_extend% LLE: %__hdcd_gain% TF: %__hdcd_transient_filter%)
 
The Transient Filter will change states with the amount of high-frequency energy, but there is nothing to decode. Ed Meitner beat Pacific Microsonics to that patent.
PE is either on or off for the entire song.
LLE will come on at varying levels (0 to -4 dB), but ONLY during extremely quiet passages.
 
PE is really the only thing that is worth decoding, but as long as you are at it you may as well go ahead and decode the very rarely used LLE.
 
Hope this helps.

 

[interpolate]

Of course, 24-bit is just a nice marketing number used by manufacturers. The reality is anything below -100dBFS will be barely audible at normal listening listening levels. I'll look at the foobar thing when I get a chance.

 

[CHansen]

  Of course, 24-bit is just a nice marketing number used by manufacturers. The reality is anything below -100dBFS will be barely audible at normal listening listening levels. I'll look at the foobar thing when I get a chance.

 
What you are saying make sense. Unfortunately I've had many experiences contrary to that which are inexplicable to our current understanding. For example I once fed a 24-bit DAC chip an audio signal with the LSB either rounded, with ±0.5 bits TPDF dither added, or with ±1.0 bits TPDF dither added. The latter is considered to be the "gold standard" of dithers, with the least amount of added noise that achieves no noise-modulation of the signal.
 
I was expecting to hear zero difference between any of these. At the 24th bit (-144 dB), the DAC chip had no resolution, the ear is claimed to have no sensitivity, the ambient noise in the room was far higher, and the noise in the analog circuitry was far higher. Yet during blind listening tests, I was able to correctly identify all three signals with ease, even when using 16-bit source material! Go figure...
 
There is far more to our hearing than is currently understood. They have just found in the last decade that the nose works by using quantum-based spectrometers to "listen" to the vibrational frequencies of the side chains of molecules. There is evidence that a similar mechanism is involved in hearing.