jimfisheye
2K Club - QQ Super Nova
- Joined
- Jan 8, 2010
- Messages
- 3,044
Jump to the end if you are uninterested in the nerdy testing discussion!
Reaper DAW used for testing.
Test track: Jethro Tull - Aqualung (as it was available in both LPCM and DTS2496)
Sources:
LPCM 4.1 master of quad mix from bluray edition
DTS 2496 4.1 master of quad mix from DVDV edition
decoder codecs tested:
ArcSoft
ffmpeg
Kodi media player
core-only decoder in DVD Audio Extractor
First test:
ffmpeg codec = kodi player codec : 100% null
Second round of tests:
Arcsoft vs ffmpeg tests:
arcsoft codec
Lfe channel is 1.45db hotter but not a perfect null even when normalized for that
It’s at the closest to nulled with the 1.45db offset normalized.
Arcsoft cancels with ffmpeg down to -38db subtracted from ffmpeg decoded for Lfe channel
all main channels are the same level but not perfect null
Arcsoft cancels down to -60db when subtracted from ffmpeg decoded for all main channels
Listening:
You still hear the -38db difference signal in the Lfe channel (normalized for the 1.45db) by ear with system volume at full scale (which would incinerate speakers with full program). You hear nothing at same full system volume in all the main channels from the -60db difference signal.
Listening to music, A/B sounds identical with the 1.45 Lfe difference normalized. Naturally, you hear the 1.45 hotter Lfe channel out of the box if listening critically.
Third round of tests:
Comparing Arcsoft and ffmpeg decoded dts to LPCM from bluray and core-only dts decode:
(Core-only decoded dts upsampled to 96k with SOX for the following tests)
Lfe channel is 0.35db lower in arcsoft decoded dts than LPCM from bluray
Lfe channel is 1.8db lower in ffmpeg decoded dts than LPCM from bluray
Null tests have best results normalizing gain for that as noted above.
Main channels are 0.02db lower in both arcsoft and ffmpeg decoded dts than LPCM from bluray
The smallest adjustment I have available in Reaper is 0.04db. Adjusting either direction gives a larger signal after subtraction. It’s at the closest to nulled with the 0.02 offset.
-20db difference signal subtracting core-only from LPCM from bluray (all channels up with Lfe normalized)
music program clearly heard with system at normal listening levels
-40db difference signal subtracting ffmpeg decoded dts from LPCM from bluray (all channels up with Lfe normalized)
barely audible whisper of cymbal top frequencies heard with system at normal listening levels
-42db difference signal subtracting arcsoft decoded dts from LPCM from bluray (all channels up with Lfe normalized)
same barely audible whisper of cymbal top frequencies heard with system at normal listening levels
Listening to music, A/B sounds identical on my system with the Lfe difference normalized as noted between the LPCM and the Arcsoft or ffmpeg decoded dts.
Conclusions:
The most reasonable conclusion is that the same 4.1 master presented on the LPCM bluray was used for the dts2496 DVDV edition. The very close level match and the A/B match when listening to the LPCM vs the Arcsoft or ffmpeg decoded dts would not be possible if they were different masterings (eg. differently eq’d). Meters and ears are in agreement with physics.
The Kodi media player app uses the same codec as the ffmpeg utility. A 100% null of the digital audio streams can only be possible if they are exactly identical. This is the one 100% conclusion here.
The meter results alone might suggest that the Arcsoft decoder codec is very slightly more accurate than the ffmpeg decoder. At least regarding the Lfe channel level. Listening tests don’t call out one or the other as “wrong”. By themselves, neither presentation with the Lfe level as decoded sounds incorrect. Comparing them to the LPCM does suggest the Arcsoft decoder produces a more true Lfe channel level. However, the difference is not enough to recommend using it over the ffmpeg decoder IMO.
The original media creator suite the Arcsoft decoder came from appears to no longer be sold. I think this was around 2012? The full suite was expensive (I believe in the $1200 range). The free media player version that used the codec was a Windows only demo. It wasn’t intended to be freely available and said demo was never offered for OSX or Linux. Only the paid for full suite. Installing and using it is fully obnoxious. Requiring Windows versions of the freeware apps eac3to and Audiomuxer running with Wine. Hacking the codec into eac3to was required and it was the full blown ‘Windows experience’ with critical order of events installing, uninstalling, adding and deleting (Wine simulated) Windows registry entries, reinstalling, and so forth.
ffmpeg is a freeware audio/video utility and Kodi media player is a freeware media player which both use the same codec. Both are available for OSX and Linux as well as Windows. Simple software installs that simply just work (in OSX anyway).
Tip:
Convert straight to FLAC from dts with ffmpeg. (Not to wav first.) This utility defaults to 44.1k and 16 bit output when converting to wav unless you give the proper commands with all the extra cryptic flags. For FLAC output, the default is to match the source.
The command is:
ffmpeg -i source_file.dts output_file_name.flac
I believe you can use ffmpeg to extract the dts streams from the VIDEO_TS folder of a DVDV disc if you want to stay with freeware. I still have a copy of DVD Audio Extractor that I used with. Selecting the direct stream demux option to output dts files.
In conclusion, you could use the Kodi media player with confidence that you are getting the fullest near-lossless playback of the dts2496 program. Or convert to FLAC as noted. Stand-alone hardware disc player users will have to rip and convert with the computer and prepare FLAC files or burn DVDA or bluray discs as needed for their system to hear the fullest fidelity from these dts2496 DVDV discs. Note that the free Kodi media player is not very elegant! This might be the clumsiest and most garish GUI I've seen in any software! Better to convert to FLAC and use VLC or your favorite media player.
I notice that the more recent Jethro Tull discs are starting to put a disclaimer in the booklet that some hardware bluray players do not have the ability to fully decode the dts2496 audio stream. I believe this extends to all DVD (non-bluray) hardware players.
Reaper DAW used for testing.
Test track: Jethro Tull - Aqualung (as it was available in both LPCM and DTS2496)
Sources:
LPCM 4.1 master of quad mix from bluray edition
DTS 2496 4.1 master of quad mix from DVDV edition
decoder codecs tested:
ArcSoft
ffmpeg
Kodi media player
core-only decoder in DVD Audio Extractor
First test:
ffmpeg codec = kodi player codec : 100% null
Second round of tests:
Arcsoft vs ffmpeg tests:
arcsoft codec
Lfe channel is 1.45db hotter but not a perfect null even when normalized for that
It’s at the closest to nulled with the 1.45db offset normalized.
Arcsoft cancels with ffmpeg down to -38db subtracted from ffmpeg decoded for Lfe channel
all main channels are the same level but not perfect null
Arcsoft cancels down to -60db when subtracted from ffmpeg decoded for all main channels
Listening:
You still hear the -38db difference signal in the Lfe channel (normalized for the 1.45db) by ear with system volume at full scale (which would incinerate speakers with full program). You hear nothing at same full system volume in all the main channels from the -60db difference signal.
Listening to music, A/B sounds identical with the 1.45 Lfe difference normalized. Naturally, you hear the 1.45 hotter Lfe channel out of the box if listening critically.
Third round of tests:
Comparing Arcsoft and ffmpeg decoded dts to LPCM from bluray and core-only dts decode:
(Core-only decoded dts upsampled to 96k with SOX for the following tests)
Lfe channel is 0.35db lower in arcsoft decoded dts than LPCM from bluray
Lfe channel is 1.8db lower in ffmpeg decoded dts than LPCM from bluray
Null tests have best results normalizing gain for that as noted above.
Main channels are 0.02db lower in both arcsoft and ffmpeg decoded dts than LPCM from bluray
The smallest adjustment I have available in Reaper is 0.04db. Adjusting either direction gives a larger signal after subtraction. It’s at the closest to nulled with the 0.02 offset.
-20db difference signal subtracting core-only from LPCM from bluray (all channels up with Lfe normalized)
music program clearly heard with system at normal listening levels
-40db difference signal subtracting ffmpeg decoded dts from LPCM from bluray (all channels up with Lfe normalized)
barely audible whisper of cymbal top frequencies heard with system at normal listening levels
-42db difference signal subtracting arcsoft decoded dts from LPCM from bluray (all channels up with Lfe normalized)
same barely audible whisper of cymbal top frequencies heard with system at normal listening levels
Listening to music, A/B sounds identical on my system with the Lfe difference normalized as noted between the LPCM and the Arcsoft or ffmpeg decoded dts.
Conclusions:
The most reasonable conclusion is that the same 4.1 master presented on the LPCM bluray was used for the dts2496 DVDV edition. The very close level match and the A/B match when listening to the LPCM vs the Arcsoft or ffmpeg decoded dts would not be possible if they were different masterings (eg. differently eq’d). Meters and ears are in agreement with physics.
The Kodi media player app uses the same codec as the ffmpeg utility. A 100% null of the digital audio streams can only be possible if they are exactly identical. This is the one 100% conclusion here.
The meter results alone might suggest that the Arcsoft decoder codec is very slightly more accurate than the ffmpeg decoder. At least regarding the Lfe channel level. Listening tests don’t call out one or the other as “wrong”. By themselves, neither presentation with the Lfe level as decoded sounds incorrect. Comparing them to the LPCM does suggest the Arcsoft decoder produces a more true Lfe channel level. However, the difference is not enough to recommend using it over the ffmpeg decoder IMO.
The original media creator suite the Arcsoft decoder came from appears to no longer be sold. I think this was around 2012? The full suite was expensive (I believe in the $1200 range). The free media player version that used the codec was a Windows only demo. It wasn’t intended to be freely available and said demo was never offered for OSX or Linux. Only the paid for full suite. Installing and using it is fully obnoxious. Requiring Windows versions of the freeware apps eac3to and Audiomuxer running with Wine. Hacking the codec into eac3to was required and it was the full blown ‘Windows experience’ with critical order of events installing, uninstalling, adding and deleting (Wine simulated) Windows registry entries, reinstalling, and so forth.
ffmpeg is a freeware audio/video utility and Kodi media player is a freeware media player which both use the same codec. Both are available for OSX and Linux as well as Windows. Simple software installs that simply just work (in OSX anyway).
Tip:
Convert straight to FLAC from dts with ffmpeg. (Not to wav first.) This utility defaults to 44.1k and 16 bit output when converting to wav unless you give the proper commands with all the extra cryptic flags. For FLAC output, the default is to match the source.
The command is:
ffmpeg -i source_file.dts output_file_name.flac
I believe you can use ffmpeg to extract the dts streams from the VIDEO_TS folder of a DVDV disc if you want to stay with freeware. I still have a copy of DVD Audio Extractor that I used with. Selecting the direct stream demux option to output dts files.
In conclusion, you could use the Kodi media player with confidence that you are getting the fullest near-lossless playback of the dts2496 program. Or convert to FLAC as noted. Stand-alone hardware disc player users will have to rip and convert with the computer and prepare FLAC files or burn DVDA or bluray discs as needed for their system to hear the fullest fidelity from these dts2496 DVDV discs. Note that the free Kodi media player is not very elegant! This might be the clumsiest and most garish GUI I've seen in any software! Better to convert to FLAC and use VLC or your favorite media player.
I notice that the more recent Jethro Tull discs are starting to put a disclaimer in the booklet that some hardware bluray players do not have the ability to fully decode the dts2496 audio stream. I believe this extends to all DVD (non-bluray) hardware players.