Topic: HDCD volume jump when next track is not HDCD (Read 3181 times)

sean6011 · « **on:** October 06, 2021, 03:35:57 am »

Hi all,
I've been though plenty of thread heading back to the start of the HDCD implementation and haven't come across a solution but see this seems to have been noticed in a few previous builds so forgive me if the answer is already out there.

The issue I'm having is where a HDCD track is played as part of a playlist where the next track does not have HDCD information the volume jumps at the end of the track. At about 10-20 seconds remaining Process HDCD disappears from the audio path and volume leveling target shows that of the next track however it doesn't appear to apply anything as the output volume shifts upwards. Disabling Volume leveling and peak normalise doesn't change this behaviour.

Obviously this isn't an issue if you're just playing albums but is awful when you have playlists that bounce up and down in output between track and also during the track.

Server is Focal Fossa with MC28 and client is windows 10 with MC28.

Any help much appreciated as I'd rather keep the HDCD functionality.

Dawgincontrol · « **Reply #1 on:** October 06, 2021, 08:07:52 am »

Ran into this problem a while back. Had to quit using HDCD.

Thread below. Guess it's still not fixed.

https://yabb.jriver.com/interact/index.php/topic,126223.msg877558.html#msg877558

sean6011 · « **Reply #2 on:** October 06, 2021, 08:26:24 am »

Thanks for the reply, I'm glad its not just me being daft.
I'll follow the other thread in case there is a resolution. Looks like hdcd is staying off for now.

Matt · « **Reply #3 on:** October 06, 2021, 12:36:05 pm »

Would you be willing to mail me two files that show the jump? Presumably one HDCD and one not.

I was trying in my library and it seemed fine.

I'm matt at jriver dot com.

Thanks.

Matt · « **Reply #4 on:** October 06, 2021, 01:44:13 pm »

I maybe reproduced it. Investigation underway.

Matt · « **Reply #5 on:** October 06, 2021, 01:49:15 pm »

With volume levelling engaged, we would add 6 dB to HDCD files. But I think that was from a past where there was a 6 dB shift in HDCD files.

So I'll try removing the shift next build and let me know how it works.

Thanks.

Dawgincontrol · « **Reply #6 on:** October 06, 2021, 02:06:30 pm »

I'll check it out on the next build for you, Matt. Thanks.

sean6011 · « **Reply #7 on:** October 07, 2021, 01:43:55 am »

Great stuff. Thanks Matt and Dawgincontrol.

Dawgincontrol · « **Reply #8 on:** October 10, 2021, 02:45:28 pm »

Matt,

Just tried playing a mix of HDCD and non-HDCD files in the latest version with HDCD checked in options. I tried with loading the file into memory and not loading the file into memory.

Result is the same. The non-HDCD file will have a slight volume reduction about 5 seconds or so before the HDCD file and the HDCD file will have a volume increase at about the same time on my system in both scenarios. You don't notice it if the file has faded out before the time mentioned above.

It may have something to do with when it processes the next file.

Attaching a log file. I skipped to the end of the affected files in playback.

Hendrik · « **Reply #9 on:** October 10, 2021, 04:22:55 pm »

You may need to re-analyze the HDCD file to properly catch its volume level if you use volume leveling. I'm not sure if that offset was reflected in the analyzed data.

Dawgincontrol · « **Reply #10 on:** October 10, 2021, 10:14:26 pm »

Re-analyzed all my HDCD files and same problem. Thought it might be tied into volume leveling, but it was the same whether I had that checked or not.

Back to unchecking the HDCD play where it doesn't happen.

Matt · « **Reply #11 on:** October 11, 2021, 11:56:49 am »

Quote from: Dawgincontrol on October 10, 2021, 10:14:26 pm

Re-analyzed all my HDCD files and same problem. Thought it might be tied into volume leveling, but it was the same whether I had that checked or not.

Back to unchecking the HDCD play where it doesn't happen.

Want to mail me two files that show the volume jump? I can't explain it since the shifts are out and our volume level matches the reference HDCD decoder.

I'm matt at jriver dot com.

Thanks

Dawgincontrol · « **Reply #12 on:** October 11, 2021, 04:35:09 pm »

Matt,

I went through a few alternating tracks and between 5 and 6 seconds from track end is when each track does the volume change. Then a few locked up the system. I have had a similar problem when in a playlist and the bitdepth of the succeeding files goes from 16 to 24 and vice versa as far as a temporary freeze. Don't know if it's related, but it happens at the same time. I had HDCD processing for the test log.

Attaching the log of the latest including the system locking ups (there were a few at the aforementioned time.

In DSP studio i have the following checkmarked:

Output Format, Volume Leveling, Effects, and Equalizer. Not running any outside plug-ins.

Under Output format Output encoding is "none". Channels is "Source Number of Channels. The only sample rates different are 88,200 to 96,000 as it doesn't work correctly otherwise.

Audio configuration is CXR-200 (NVIDIA High Definition Audio) WASAPI (My receiver with built-in DAC is Cambridge Audio CXR-200) via HDMI from a DELL computer with plenty of memory.

Bitdepth - Automatic, Buffering - 100 milliseconds (both recommended) I have played around with the bufferring in regard to the freezes without success.

Open Device for exclusive access checked.

Memory playback - Load decoded file into memory (any changes to this have made no difference

Using SOX for resampling (changing this made no difference.

Switch tracks: Gapless, Do not play silence (leading and trailing) checked (have tried both on and off with no change)

Can't think of anything else. The other guy stated his happens with 10 seconds to go, mine is always between the 6-5 seconds left transition, both volume change and freeze when it happens.

If you need something else let me know

JimH · « **Reply #13 on:** October 11, 2021, 05:51:01 pm »

When you get a "lock", try power cycling the DAC.

Try playing all lower res files and try playing no multi-channel files.

Matt · « **Reply #14 on:** October 11, 2021, 06:27:27 pm »

Disable volume levelling. Still a problem? I think sending me some files would be good. Thanks 👍

Dawgincontrol · « **Reply #15 on:** October 12, 2021, 10:14:49 am »

Files should be in 2nd attachment above.

Volume leveling makes no difference.

Only checking the "Play as HDCD" box matters in the volume increase and only when alternating between HDCD and HDCD tracks in a playlist.

If two or more HDCD tracks are played in a row there is no problem until it reaches a non-HDCD track.

Much like the problem with different bitdepths. It only happens when transitioning to different bitdepths (16 to 24 and vice versa). Except that happens whether HDCD is checked or not. Something processes there at the 5-6 second left transition.

dtc · « **Reply #16 on:** October 12, 2021, 12:28:21 pm »

This gets a little tricky with software decoding.

HDCDs are recorded at a normal level. When the signal is send to a hardware HDCD chip, the chip drops the signal level by 6 dB so the decoded signal, which has a wider dynamic range, does not clip. Many people increase the volume control on their amplifier to compensate for the lower level, but that does run the risk of clipping.

When a software decoder is used, when it expands to 24 bits (actually 20 bits padded to 24)) it should drop the level by 6 dB, just like the hardware chip does. Some decoders have an option to increase the level by 6 dB, which brings it back to the original volume level, at the risk of clipping. That is basically the equivalent of turning up the volume on the hardware chip system.

Replay gain is tricky, depending on whether it is calculated on the original file or on the decoded data. If the replay gain is calculated on the original file without decoding then applied to the decoded data the volume will still be 6 dB too low. Adding back 6dB will get it back to the original level, but again at the risk of clipping.

Replay gain should really be calculated for the decoded data if the software decoder is going to be used. However, that means doing the decoding before doing the replay gain calculation.

It seems to me that replay gain can only be correct for one of the two possibilities - for the original file with no decoding or for the data after decoding. Applying replay gain on the fly when there is an option to decode or not, will only work if replay gain is also calculated on the fly, which is not possible without a significant delay while the calculation is done.

It seems that the best way to handle this is to decode the file to the 24 bit form and store that. Then, apply replay gain to that file. That will also avoid clipping problems, although it might not be load enough given the expanded dynamic range. That is what I do. Then there is no ambiguity. Unfortunately, I do not think MC can create a decoded file. I use dBpoweramp to do that and have since before MC had HDCD capability.

As far as I can tell, the Cambridge Audio CXR-200 does not decode HDCD, so the non-decoded file should not be sent to the DAC if you want HDCD decoding to be done. Which means replay gain will not work for you if it is calculated on the original non-decoded file.

Hope I got all that right.

Dawgincontrol · « **Reply #17 on:** October 12, 2021, 09:17:31 pm »

Thanks dtc. Interesting stuff.

Going by your info, shouldn't the decoded file loaded into memory work (which is what I use and have tried all the other memory/nonmemory options). Or is it because at the 6 seconds to go interval MC is trying to load the next file and puts out the compensation for the next to play at that time instead of, say adding a second or two to complete the algorithm at the tracks/files end for affected transitions?

This seems why the sequencing of 16 and 24 bit tracks has trouble at the exact same time whether HDCD or not as the program is trying to load the next and having trouble with the decoding.

I have tried numerous combinations to no avail. I get what you're saying and very much appreciate the info.

Would like to hear from sean6011 and see if the update changed/fixed his problem.

dtc · « **Reply #18 on:** October 12, 2021, 09:49:56 pm »

What 24 bit file are you loading into memory? If it is an actual 24 bit file that has been decoded from HDCD by dBpoweramp or another program and if that 24 bit file has been run through audio analyze, then volume leveling should work on it. However, even then if the HDCD decode has increased the dynamic range, then the main part of the file may still be at a lower level than you would like.

If you are just loading a 16 bit file with HDCD encoding into memory, then when decoding it I am not sure whether MC is doing volume leveling before or after the decoding. And, I do not know if audio analyze is determining its value before or after decoding.

MC starts loading the next file a few seconds before the next file begins. That may be what you are seeing on that issue.

Dawgincontrol · « **Reply #19 on:** October 12, 2021, 10:59:13 pm »

Don't want to conflate and confuse you with two different things.

The change between differing bit depths is independent of the HDCD change. It just causes problems at the same time in the transition. These are not HDCD files. There is no volume change, just a temporary freeze in a playlist when going from one bitdepth to another at the 6 second mark from the end. Has really nothing to do with HDCD processing as it happens whether HDCD is turned on or off.

HDCD file problems only happen when HDCD coding is selected. When turned off (which is what I have done) there is no HDCD processing and no transitional volume problem. When HDCD is turned on, then there is a volume change in the transition between HDCD and non-HDCD tracks at the 6 second mark from end.

Hope this isn't confusing. Just, two different problems that happen at the same time period which leads me to believe there's processing going on in the program.

dtc · « **Reply #20 on:** October 13, 2021, 08:23:41 am »

OK, lets leave HDCD out of this.

When I play a 16 bit file, then a 24 bit file, then a 16 bit file in either Playing Now or in a Playlist, I do not see any extra pauses between tracks. I do see that MC starts loading the next track about 6 seconds before the end of the currently playing track, but it does not start the next track until the currently playing track finishes. You can see that using Audio Path and watching the timer on the track. I am using memory play with Play Full File (not decoded).

Do you see a longer than expected pause between tracks of different bit depth compared to tracks of the same bit depth, or are you just seeing and hearing the normal transition between tracks? Are you hearing that any of the currently playing track is cut off on transition?

Matt · « **Reply #21 on:** October 13, 2021, 10:15:21 am »

I think I found the problem.

When you would enable HDCD processing, it wouldn't take effect during audio analysis.

So next build:
Changed: Audio analysis will engage HDCD processing if it's enabled (you might need to reanalyze your HDCD files).

Dawgincontrol · « **Reply #22 on:** October 13, 2021, 10:30:49 am »

Thanks Matt.

In relation to the other problem, I changed to "Playing full file (not decoded) from memory" as "dtc" had in their set-up and the bit transition seems to be working no problem. I thought I had tried that previously, but it appears to have made a difference from trying to load the decoded file or not using memory. Will see if it remains that way.

Was about to send you a file or two, but will wait.

Thanks to dtc for his time.

dtc · « **Reply #23 on:** October 13, 2021, 06:07:19 pm »

Quote from: Matt on October 13, 2021, 10:15:21 am

I think I found the problem.

When you would enable HDCD processing, it wouldn't take effect during audio analysis.

So next build:
Changed: Audio analysis will engage HDCD processing if it's enabled (you might need to reanalyze your HDCD files).

How will that work with audio analyze on auto import? That might be confusing.

dtc · « **Reply #24 on:** October 14, 2021, 07:17:25 am »

I also tried no memory playback and memory playback of decoded track and did not see any additional pause between 16 bit, 24 bit and 16 bit tracks in my test playback. Playback was the same as with memory playback of not decoded track.

Matt · « **Reply #25 on:** October 14, 2021, 08:22:17 am »

Quote from: dtc on October 13, 2021, 06:07:19 pm

How will that work with audio analyze on auto import? That might be confusing.

Hendrik pointed out a similar concern that we wouldn't know a file was HDCD until after analysis. We'd save the volume information at the same time.

So I made setting HDCD to 1 empty all the volume information.

You'll need to analyze again, but not a big deal.

If you change your HDCD option from off to on you would also need to re-analyze. We're not flagging that so you just have to do it manually.

dtc · « **Reply #26 on:** October 15, 2021, 08:01:12 am »

Yes, the only way to insure the volume data is in sync with the HDCD field seems to be to redo the analysis automatically or at least warn the user when a change to the HDCD field is made. I am not sure how easy it is to trigger a dialog when a field is changed, but it would be nice to warn the user about the volume data, if it is set, when the HDCD field is changed. If you are already changing the volume data, then a dialog box seems appropriate.

Thanks for paying attention to this. HDCD is not used that much, but the whole 6dB thing is confusing to many people.

Dawgincontrol · « **Reply #27 on:** October 15, 2021, 09:21:11 am »

Windows update and problems with the bit change transfer returned after it installed itself.

I know it's probably unique to certain motherboards or memory usage, but Windows updates always cause some new problem. Unfortunately on Windows 10, you can only delay them before they eventually make their mark.

Anyway back to square.

Matt · « **Reply #28 on:** October 16, 2021, 06:08:06 am »

Please test build 76 and let us know. Be sure to analyze audio again. Thanks 👍

Cinelder · « **Reply #29 on:** October 16, 2021, 11:58:00 am »

As I'm a little confused, could you please advise if after the updates it is now advised that everyone who has HDCD tracks manually redo audio analysis for all HDCD files? I only use JRiver to decode them (don't have any hardware that does it) and usually play HDCD tracks as albums, so don't often mix HDCD and non-HDCD files in sequence.

Thanks

Matt · « **Reply #30 on:** October 16, 2021, 12:04:21 pm »

Yes, analyze again.

Frobozz · « **Reply #31 on:** October 16, 2021, 12:07:23 pm »

I tested build 76 with an HDCD album that has peak extend

First time re-analyzing the album the HDCD field got set to 0
Second time re-analyzing the album the HDCD field got set to 1 but the fields for R128, replay gain, DR, peak level all got cleared.

Matt · « **Reply #32 on:** October 16, 2021, 12:08:57 pm »

When we detect HDCD the volume information will be reset. You'll have to analyze again.

Frobozz · « **Reply #33 on:** October 16, 2021, 12:12:03 pm »

When I re-analyze a third time the HDCD field gets set back 0
But the analysis fields for R128, DR, peak level and such all get new values that are correct for the HDCD decoded version

But the HDCD field is now 0 so won't get played with HDCD decoding

Matt · « **Reply #34 on:** October 16, 2021, 12:32:12 pm »

I can't explain that. Could you send the file to matt at jriver? Thanks.

Frobozz · « **Reply #35 on:** October 16, 2021, 01:14:07 pm »

I think I see what's happening in my setup

I have both "Analyze for HDCD" and "Play as HDCD if possible" set
When both are set I get the behavior described above

When I clear the "Analyze for HDCD" option and have "Play as HDCD if possible" set, then I get the expected behavior and the HDCD field gets reset and set and reanalyzed as expected.

What is the "Analyze for HDCD" option supposed to do in build 76? It seems it is redundant now and "Play as HDCD if possible" is the option that controls HDCD analysis.

Cinelder · « **Reply #36 on:** October 16, 2021, 02:15:40 pm »

Quote from: Matt on October 16, 2021, 12:04:21 pm

Yes, analyze again.

Many thanks for the quick response.

Cinelder · « **Reply #37 on:** October 16, 2021, 04:36:19 pm »

Quote from: Frobozz on October 16, 2021, 01:14:07 pm

I think I see what's happening in my setup

I have both "Analyze for HDCD" and "Play as HDCD if possible" set
When both are set I get the behavior described above

When I clear the "Analyze for HDCD" option and have "Play as HDCD if possible" set, then I get the expected behavior and the HDCD field gets reset and set and reanalyzed as expected.

What is the "Analyze for HDCD" option supposed to do in build 76? It seems it is redundant now and "Play as HDCD if possible" is the option that controls HDCD analysis.

This exact same behavior also happened to me after I created a playlist from files in my library that were previously analyzed and tagged HDCD and reanalyzed them. They all show HDCD=0 after re-analysis with those settings (Audio Analysis for HDCD = 1, Play as HDCD if possible set). Please advise on what I should do now and what specific settings or combination I should use. Re-analyzing known HDCD files with the current build of JRiver (76) will now no longer recognize them as HDCD. This has affected nearly 100GB of files in my library.

Matt · « **Reply #38 on:** October 16, 2021, 05:14:54 pm »

Could you mail one of the files having the problem to matt at jriver? I'll try to figure it out Monday. Thanks.

Frobozz · « **Reply #39 on:** October 16, 2021, 05:35:12 pm »

Quote from: Cinelder on October 16, 2021, 04:36:19 pm

This exact same behavior also happened to me after I created a playlist from files in my library that were previously analyzed and tagged HDCD and reanalyzed them. They all show HDCD=0 after re-analysis with those settings (Audio Analysis for HDCD = 1, Play as HDCD if possible set). Please advise on what I should do now and what specific settings or combination I should use. Re-analyzing known HDCD files with the current build of JRiver (76) will now no longer recognize them as HDCD. This has affected nearly 100GB of files in my library.

Don't panic. It can all be recovered.

One way would be to install an older version (before 76) and rescan and reanalyze all the files.
Another would be to wait for a fixed build or a proper understanding of what's going on. And then rescan and reanalyze with that build.

Then there is also the option of restoring from a library backup from before this happened. Then select all files and do a Library Tools >> Update Tags (from library).

Frobozz · « **Reply #40 on:** October 16, 2021, 05:38:47 pm »

Quote from: Matt on October 16, 2021, 05:14:54 pm

Could you mail one of the files having the problem to matt at jriver? I'll try to figure it out Monday. Thanks.

Will do.

Cinelder · « **Reply #41 on:** October 16, 2021, 06:29:16 pm »

Quote from: Frobozz on October 16, 2021, 05:35:12 pm

Don't panic. It can all be recovered.

One way would be to install an older version (before 76) and rescan and reanalyze all the files.
Another would be to wait for a fixed build or a proper understanding of what's going on. And then rescan and reanalyze with that build.

Then there is also the option of restoring from a library backup from before this happened. Then select all files and do a Library Tools >> Update Tags (from library).

Thank you! Very much appreciated.

Postscript: I rolled the MC install all the way back to build 66 (just to be on the safe side). Using the first option, I was dismayed to discover after conducting audio reanalysis, MC would *not* recognize as HDCD some of the known HDCD files. In my library, the HDCD files that got "missed" (and which show as HDCD=0 in tags after audio reanalysis) were about 1/3 of the known HDCD files - over 1000 tracks in my library.

Happily, what did work very well was the third option you suggested. I used a library restore from before the HDCD incident today using build 66. I then used this restored library to find all of the files previously tagged as HDCD. I then used Library Tools/Update Tags (from library). Now the tags do show as HDCD=1 all previously known HDCD files that MC had analyzed and identified as HDCD. So I'm now back to where I started, and while that sometimes defines not getting anywhere, in this case I consider it as major progress.

So thanks again!

Frobozz · « **Reply #42 on:** October 17, 2021, 07:59:03 am »

Here are some albums that would be interesting test subjects for unusual HDCD behavior.

I have a handful of albums that have HDCD but only a few of the tracks on those albums need HDCD decoding. Because only some of the tracks on the album have peak extend and/or gain. The rest of the tracks have no peak extend and no gain.

Depending on how the HDCD detection analysis is done you can end up with MC flagging only one or two of the tracks as HDCD. Which could result in volume differences for those tracks due to HDCD decoding.

Pete Seeger "Playlist: The Very Best of Pete Seeger"
Track 12 has peak extend. The rest of the tracks have no peak extend and no low level gain

Pete Seeger "The Essential Pete Seeger"
Same as "Playlist: The Very Best of Pete Seeger", just a different pressing and album cover

Beck "Mutations"
Last track has peak extend, but no audible difference I can note if decoded. Maybe peak extend got accidentally enabled during mastering.

Keb' Mo' "Slow Down"
Tracks 1 and 6 have gain. All other tracks have no peak extend and no gain.

Tool "Lateralus"
First track has gain. Every other track has no peak extend and no gain.

Neil Young "Sugar Mountain: Live At Canterbury House 1968"
All tracks have gain except track 4 which has no peak extend and no gain.

Matt · « **Reply #43 on:** October 17, 2021, 09:24:43 am »

Thanks for the file. I could reproduce the problem. I'll hopefully get it fixed tomorrow. Thanks again.

Matt · « **Reply #44 on:** October 25, 2021, 09:11:41 am »

Dawgincontrol sent me a couple HDCD files and it helped me find that when a track transition would start we could jump volume going between HDCD and not HDCD files.

I ended up reworking HDCD decoding to more tightly hook to a given file for a coming build.

So hopefully everything HDCD is wonderful from now on

Dawgincontrol · « **Reply #45 on:** October 25, 2021, 09:20:33 am »

Great work Matt. Thanks for your time.

INTERACT FORUM

Author Topic: HDCD volume jump when next track is not HDCD (Read 3181 times)