HDCD feedback and information thread

[vortecjr]

14 hours ago, CatManDo said:

I recently ripped my entire CD collection. There were some HDCDs among them, some with peak extension, some without.

 

I found this ripping procedure (using EAC and CUE Tools) on the Steve Hoffman forum and it worked well.

 

 

https://forums.stevehoffman.tv/threads/ripping-hdcd-sacd.301966/#post-8265734

 

Maybe there are quicker procedures, but for my collection (about a dozen HDCDs among 3000 CDs) it was fast enough. I haven't yet done a comparison between the non-HDCD rip and the 24/44 rip.

I used Foobar2000 to scan the library and list the albums with the Peak Extend feature enabled. I then used DBPowerAMP to decode the files. 

[vortecjr]

10 hours ago, manisandher said:

 

Will have a look Jesus, but not sure if I'll be able to find any - it was a while ago now, and I gave up on 16/44.1 pretty quickly from what I can recall. Leave it with me...

 

Mani.

Thanks.

[tmtomh]

2 hours ago, vortecjr said:

I actually have a lot of HDCDs that have the Peak Extend feature enabled. This thread is not really about if it's worth it or not. The fact is that the content exists and needs to be dealt with properly in a collection.   

 

Well, how worthwhile it is on balance, given the overall number of HDCDs with Peak Extend on the market, is on-topic for a thread whose title includes the phrase "HDCD feedback."

 

But putting that aside for a moment, my comment clarified the bit-depth of the decoded HDCD stream, and that is substantive information. 

[vortecjr]

15 minutes ago, tmtomh said:

 

Well, how worthwhile it is on balance, given the overall number of HDCDs with Peak Extend on the market, is on-topic for a thread whose title includes the phrase "HDCD feedback."

 

But putting that aside for a moment, my comment clarified the bit-depth of the decoded HDCD stream, and that is substantive information. 

Again, that is not the purpose of this discussion. However many there are on the market they exist and I'm looking for comments...do you like them or not and under what circumstances. 

 

I didn't take exception to your comments on the technical aspects. 

[tmtomh]

41 minutes ago, vortecjr said:

Again, that is not the purpose of this discussion. However many there are on the market they exist and I'm looking for comments...do you like them or not and under what circumstances. 

 

I didn't take exception to your comments on the technical aspects. 

 

Your first post asked, "what is your experience with it?" My comment was responsive to that. If you really want to keep going back and forth about what's on-topic, we can do that, but it seems beside the point. I certainly agree an HDCD thread is useful, and I'm glad you started one.

[vortecjr]

8 hours ago, tmtomh said:

 

Your first post asked, "what is your experience with it?" My comment was responsive to that. If you really want to keep going back and forth about what's on-topic, we can do that, but it seems beside the point. I certainly agree an HDCD thread is useful, and I'm glad you started one.

I'm trying to keep this thread from becoming a MQA type tread. People are free to post their feedback on it and experience without making an argument for or against it. The intent here is to discuss if people like them or not and discuss possible reasons why or not. Also, some people don't even know they have them and the goals is to show people how to identify them and use them properly so we get feedback that matters.    

[vortecjr]

Moving on...I noticed something interesting from the Foobar2000 scan of my library. The scan did not identify any hi-resolution content with the HDCD markers. I would have thought that the Reference Recording HRx tracks would be identified. According to the Reference Recording website, "Currently we record at 176.4 kHz/ 24 bits, with HDCD, and that is what we put on our HRx discs." 

 

https://referencerecordings.com/format/hrx

 

So this brings up some concern. The research I have done suggests that the HDCD recordings on CD are distorted unless you play them back via a hardware or software decoder. So naturally, I have to wonder if the HRx HDCD recordings on DVD are also distorted unless you play them back via a hardware or software decoder. More research is needed to know if the software decoder even works on the Reference Recording HRx tracks. 

[tmtomh]

3 hours ago, vortecjr said:

Moving on...I noticed something interesting from the Foobar2000 scan of my library. The scan did not identify any hi-resolution content with the HDCD markers. I would have thought that the Reference Recording HRx tracks would be identified. According to the Reference Recording website, "Currently we record at 176.4 kHz/ 24 bits, with HDCD, and that is what we put on our HRx discs." 

 

https://referencerecordings.com/format/hrx

 

So this brings up some concern. The research I have done suggests that the HDCD recordings on CD are distorted unless you play them back via a hardware or software decoder. So naturally, I have to wonder if the HRx HDCD recordings on DVD are also distorted unless you play them back via a hardware or software decoder. More research is needed to know if the software decoder even works on the Reference Recording HRx tracks. 

 

This sounds very strange (not criticizing you; rather, referring to the Reference Recordings web site blurb). HDCD was designed for 16-bit recordings and I'm not aware of any existing decoder chips that look for HDCD data except in the 16th bit. In other words, even if Pacific Microsonics makes/made an ADC capable of applying HDCD encoding to a 24-bit data stream and burying the HDCD data in the 24th-bit, I don't think any consumer playback device could decode that.

 

The way the Reference Recordings blurb is written, it is not clear what they mean when they say they still currently use HDCD in the production of their high-res recordings. If they do, I would guess that probably means they use Pacific Microsonics ADC units - which happen to be HDCD capable - to digitize their recordings. But I would be very surprised if they enable the substantive features of it - Peak Extend and Low-level Range Extend. It would hardly be necessary with high-res content, since high-res provides more bits and more dynamic range than HDCD (24 bits vs 20 bits), and again, AFAIK no consumer playback software or hardware could decode those features if they were embedded in the least significant bit of a 24-bit stream.

 

In other words, one can use a Pacific Microsonics ADC for recording, and enable HDCD but not actually use any HDCD features - which means the resulting disc/file will trip an HDCD flag but there's no actual HDCD data or features in the digital content. In fact, I believe one can even use a PM ADC and simply turn off its HDCD features altogether, meaning one can record non-HDCD content using a PM unit. In that case, one might still say "we use HDCD" when they really just mean "we use PM equipment to digitize our recordings." I suspect one of these two scenarios is what happens with Reference Recordings' high-res material.

 

So that leaves the digital filters - both the filters used in the ADC during recording, and the digital filters built in to any HDCD-certified/capable hardware playback device. Part of HDCD's claim was that the digital filters built in to HDCD hardware players were special/high-precision/very good. And presumably they were considered to match up well, or be complementary, to the filtering used by the Pacific Microsonics ADC.

 

Software HDCD decoders/players do not, AFAIK, necessarily have such filters - or if they have multiple filters choices, I don't know that they automatically switch to HDCD filters when playing back files that were ripped from HDCD discs.

 

Finally, as to your concern about HDCD recordings on CDs being "distorted" unless played back on HDCD hardware or software, there are two sources of distortion:

1. Increased noise, because the HDCD data is buried in the 16th bit, meaning it's in the dither, and because HDCD data is by definition not random, it therefore makes the dither not totally randomized - and since the entire function of dither is to reduce the perception of quantization noise by randomizing it, HDCD encoding makes dither less effective. Again, this is not an issue with high-res content (because it hardly matters if it has a 23-bit effective noise floor vs 24), but theoretically an undecoded HDCD will have only a 15-bit noise floor, while a decoded HDCD will have a 20-bit noise floor. This is not distortion per se, but it does reduce S/N ratio.
2. Distortion of peaks if Peak Extend was used. Peak Extend is a reversible soft limiter - basically just like the widespread digital limiters used (and overused) in mastering today, except HDCD Peak Extend saves the original peak info and it can be restored during playback by an HDCD-capable device or software. So playing an HDCD with Peak Extend on a non-HDCD playback device will result in the peaks being limited. That's unfortunate, but it's not going to result in audible distortion in most cases, because the peaks are soft limited rather than hard-clipped: the peaks will be altered and their dynamics reduced, but the entire point of soft limiting is that it avoids audible distortion (unless the mastering engineer pushes the levels so hard and abuses the limiter so much that it can't round off the peaks properly).

[R1200CL]

@Miska Does HQPlayer (Embedded) support HDCD, or any plan to implement support ?

[manisandher]

5 hours ago, vortecjr said:

So naturally, I have to wonder if the HRx HDCD recordings on DVD are also distorted unless you play them back via a hardware or software decoder.

 

If you use a PM1/2 to play back the HRx 24/176.4 files from RR, the HDCD LED will illuminate. However, don't be fooled by this - there are no HDCD processes beyond 16/44.1. So you can play back the HRx files on any DAC with no more added distortion than there would be playing them back on a PM1/2.

 

(I used to have a DAC with a PMD200 filter, which also illuminated the HDCD LED when playing back HRx files. But DACs with this filter (as opposed to the regular PMD100) are few and far between.)

 

Mani.

[Miska]

1 hour ago, R1200CL said:

@Miska Does HQPlayer (Embedded) support HDCD, or any plan to implement support ?

 

No, as far as I know there's no free implementation available as source code, nor public specifications. It goes into exactly same category as MQA...

 

[Miska]

2 hours ago, tmtomh said:

Distortion of peaks if Peak Extend was used. Peak Extend is a reversible soft limiter - basically just like the widespread digital limiters used (and overused) in mastering today, except HDCD Peak Extend saves the original peak info and it can be restored during playback by an HDCD-capable device or software. So playing an HDCD with Peak Extend on a non-HDCD playback device will result in the peaks being limited. That's unfortunate, but it's not going to result in audible distortion in most cases, because the peaks are soft limited rather than hard-clipped: the peaks will be altered and their dynamics reduced, but the entire point of soft limiting is that it avoids audible distortion (unless the mastering engineer pushes the levels so hard and abuses the limiter so much that it can't round off the peaks properly).

 

I have successfully ripped all my few (3 Mark Knopfler albums in total) HDCD encoded CDs into 24-bit AIFF using dbPoweramp and the HDCD decoder plugin. The plugin's gain setting just works exactly opposite way than what at least I understand from the description. So check the results afterwards in Audacity for correctness, that you don't get extra clipping! I needed to do the ripping twice because of this.

 

[tmtomh]

1 hour ago, Miska said:

 

I have successfully ripped all my few (3 Mark Knopfler albums in total) HDCD encoded CDs into 24-bit AIFF using dbPoweramp and the HDCD decoder plugin. The plugin's gain setting just works exactly opposite way than what at least I understand from the description. So check the results afterwards in Audacity for correctness, that you don't get extra clipping! I needed to do the ripping twice because of this.

 

 

Interesting - I've never used the dbPoweramp/HDCD plugin combo. I've only used hdcd.exe in command line mode (in a Windows emulator on my Mac).

 

If I understand your comment correctly, it sounds like the HDCD plugin restored the limited peaks, but did not perform the accompanying - and necessary - reduction in overall gain (aka makeup gain) of the waveform. resulting in all those peaks going into clipping. Is that right? If so, I find that a little strange - to my knowledge, hdcd.exe automatically applies makeup gain (or I should say makeup attenuation) of 6dB to compensate for the louder peaks that occur when the full peaks are restored during HDCD decoding. I'd always assumed that this was just a hard-wired (so to speak) part of the HDCD spec and that any hardware or software HDCD decoder would properly adjust the gain as a matter of course.

 

So if you did it twice, it sounds like the 2nd time you changed a setting in the decoder plugin, to make it perform the gain reduction so the restored peaks would all be below digital max aka 0.0 and avoid clipping - yes?

[Miska]

37 minutes ago, tmtomh said:

If I understand your comment correctly, it sounds like the HDCD plugin restored the limited peaks, but did not perform the accompanying - and necessary - reduction in overall gain (aka makeup gain) of the waveform. resulting in all those peaks going into clipping. Is that right?

 

It has option for that, but it works inverse to the description (or maybe it's just my English skills).

 

37 minutes ago, tmtomh said:

If so, I find that a little strange - to my knowledge, hdcd.exe automatically applies makeup gain (or I should say makeup attenuation) of 6dB to compensate for the louder peaks that occur when the full peaks are restored during HDCD decoding.

 

It should be -6 gain, not 6 dB gain...

 

So the data needs to be extended downwards, not upwards because there's no space to extend it up. IOW, content needs to be bit-shifted down by one step before peak extension. This is where the setting becomes confusing.

 

When used correctly, peaks don't run into clipping, if the option is used incorrectly, it makes results worse than without decoding.

 

Quote

So if you did it twice, it sounds like the 2nd time you changed a setting in the decoder plugin, to make it perform the gain reduction so the restored peaks would all be below digital max aka 0.0 and avoid clipping - yes?

 

Yes, exactly. IIRC, the option needed to be disabled for correct operation. But this is easy to check with audio editor.

 

[Miska]

In my opinion, HDCD is best dealt with at ripping time, decoding as necessary and thus get rid of the problem at once for good. It may require bit of extra hassle at that point, but then it is done...

 

[vortecjr]

6 hours ago, tmtomh said:

 

This sounds very strange (not criticizing you; rather, referring to the Reference Recordings web site blurb). HDCD was designed for 16-bit recordings and I'm not aware of any existing decoder chips that look for HDCD data except in the 16th bit. In other words, even if Pacific Microsonics makes/made an ADC capable of applying HDCD encoding to a 24-bit data stream and burying the HDCD data in the 24th-bit, I don't think any consumer playback device could decode that.

 

The way the Reference Recordings blurb is written, it is not clear what they mean when they say they still currently use HDCD in the production of their high-res recordings. If they do, I would guess that probably means they use Pacific Microsonics ADC units - which happen to be HDCD capable - to digitize their recordings. But I would be very surprised if they enable the substantive features of it - Peak Extend and Low-level Range Extend. It would hardly be necessary with high-res content, since high-res provides more bits and more dynamic range than HDCD (24 bits vs 20 bits), and again, AFAIK no consumer playback software or hardware could decode those features if they were embedded in the least significant bit of a 24-bit stream.

 

In other words, one can use a Pacific Microsonics ADC for recording, and enable HDCD but not actually use any HDCD features - which means the resulting disc/file will trip an HDCD flag but there's no actual HDCD data or features in the digital content. In fact, I believe one can even use a PM ADC and simply turn off its HDCD features altogether, meaning one can record non-HDCD content using a PM unit. In that case, one might still say "we use HDCD" when they really just mean "we use PM equipment to digitize our recordings." I suspect one of these two scenarios is what happens with Reference Recordings' high-res material.

 

So that leaves the digital filters - both the filters used in the ADC during recording, and the digital filters built in to any HDCD-certified/capable hardware playback device. Part of HDCD's claim was that the digital filters built in to HDCD hardware players were special/high-precision/very good. And presumably they were considered to match up well, or be complementary, to the filtering used by the Pacific Microsonics ADC.

 

Software HDCD decoders/players do not, AFAIK, necessarily have such filters - or if they have multiple filters choices, I don't know that they automatically switch to HDCD filters when playing back files that were ripped from HDCD discs.

 

Finally, as to your concern about HDCD recordings on CDs being "distorted" unless played back on HDCD hardware or software, there are two sources of distortion:

1. Increased noise, because the HDCD data is buried in the 16th bit, meaning it's in the dither, and because HDCD data is by definition not random, it therefore makes the dither not totally randomized - and since the entire function of dither is to reduce the perception of quantization noise by randomizing it, HDCD encoding makes dither less effective. Again, this is not an issue with high-res content (because it hardly matters if it has a 23-bit effective noise floor vs 24), but theoretically an undecoded HDCD will have only a 15-bit noise floor, while a decoded HDCD will have a 20-bit noise floor. This is not distortion per se, but it does reduce S/N ratio.
2. Distortion of peaks if Peak Extend was used. Peak Extend is a reversible soft limiter - basically just like the widespread digital limiters used (and overused) in mastering today, except HDCD Peak Extend saves the original peak info and it can be restored during playback by an HDCD-capable device or software. So playing an HDCD with Peak Extend on a non-HDCD playback device will result in the peaks being limited. That's unfortunate, but it's not going to result in audible distortion in most cases, because the peaks are soft limited rather than hard-clipped: the peaks will be altered and their dynamics reduced, but the entire point of soft limiting is that it avoids audible distortion (unless the mastering engineer pushes the levels so hard and abuses the limiter so much that it can't round off the peaks properly).

Someone knows a lot about HDCD:) 

[vortecjr]

5 hours ago, manisandher said:

 

If you use a PM1/2 to play back the HRx 24/176.4 files from RR, the HDCD LED will illuminate. However, don't be fooled by this - there are no HDCD processes beyond 16/44.1. So you can play back the HRx files on any DAC with no more added distortion than there would be playing them back on a PM1/2.

 

(I used to have a DAC with a PMD200 filter, which also illuminated the HDCD LED when playing back HRx files. But DACs with this filter (as opposed to the regular PMD100) are few and far between.)

 

Mani.

My Oppo is suppose to decode HDCD, but it will not via network playback. I can't even try it here with a cd because the disks are stored elsewhere.

 

I contacted Reference Recording and they emailed me saying, "Built into the program for HDCD is a filter morphing ability, which seamlessly looks ahead in the music to handle changes in timbre or transients. That is always on and in use if the HDCD flag is on.  

 

I also contacted Jim Web who wrote the articles in the technical section I added to the top post. Jim was kind enough to look at a sample HRx file for me. On quick look he didn't find the usual flagging, but is looking into it more.

[vortecjr]

4 hours ago, Miska said:

 

No, as far as I know there's no free implementation available as source code, nor public specifications. It goes into exactly same category as MQA...

 

What about source code from ffmpeg? 

[vortecjr]

5 hours ago, Miska said:

 

I have successfully ripped all my few (3 Mark Knopfler albums in total) HDCD encoded CDs into 24-bit AIFF using dbPoweramp and the HDCD decoder plugin. The plugin's gain setting just works exactly opposite way than what at least I understand from the description. So check the results afterwards in Audacity for correctness, that you don't get extra clipping! I needed to do the ripping twice because of this.

 

I get the same results..I think. The bottom one is the ripped track. I did not select the checkbox to avoid any clipping.  

 

[Axiom05]

5 minutes ago, vortecjr said:

What about source code from ffmpeg? 

Ayre developed their own decoding protocol as there is nothing out there.

[vortecjr]

2 hours ago, Miska said:

In my opinion, HDCD is best dealt with at ripping time, decoding as necessary and thus get rid of the problem at once for good. It may require bit of extra hassle at that point, but then it is done...

 

I agree based on your circumstance because you have two albums. My collection was ripped many years ago and I recently found almost 1000 tracks with the embedded HDCD flags. I located them, moved them into a work folder, converted them with the DBPowerAMP converter, and then moved them back into my library, but what a pain that was to do. Also, I only converted the titles that had peak extension enabled and this takes a little more time to sort out.  

[vortecjr]

Here is an interesting update for HDCD decoding with DLNA streams.

 

I emailed Simon N with MinimServer and explained the current state of affairs and included some links for him to read. Simon was very receptive to the idea as adding HDCD decoding and now has it working in his lab. In an upcoming release he will add HDCD decoding via ffmpeg to MinimServer/MinimStreamer.   

[Miska]

22 minutes ago, vortecjr said:

What about source code from ffmpeg? 

 

Do they have something for that too? It doesn't mean it is usable in a commercial product though. It is massive amount of work to clean up ffmpeg such way that it can be used in a commercial product... It could be that would have to go too...

 

[Miska]

27 minutes ago, vortecjr said:

I get the same results..I think. The bottom one os the ripped track. I did not select the checkbox to avoid any clipping.  

 

 

Yes, the lower one looks correct result...

[vortecjr]

5 minutes ago, Miska said:

 

Do they have something for that too? It doesn't mean it is usable in a commercial product though. It is massive amount of work to clean up ffmpeg such way that it can be used in a commercial product... It could be that would have to go too...

They sure do. HDCD is both nowhere and everywhere if you look for it:) I didn't look at the code...

https://ffmpeg.org/ffmpeg-filters.html#hdcd