Post Reply 
 
Thread Rating:
  • 0 Votes - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Volume loss converting DSF to Wav with Convolution
23-05-2019, 22:39
Post: #11
RE: Volume loss converting DSF to Wav with Convolution
(23-05-2019 20:54)simoncn Wrote:  How did you do on-the-fly transcoding for the DSD version of this test track? The DSD downloads are DFF, not DSF. Only DSF can be transcoded by MinimStreamer and FFmpeg.
Sorry, no accurate enough with my posts. All files are DSFs ripped from SACDs.
Stream.options: convOut= -i /media/music/pa/convolve/FiltersM*.wav -lavfi afir=gtype=gn
Steam.transcode: flac:wav,dsf:wav24;176
Find all posts by this user
Quote this message in a reply
24-07-2019, 14:05
Post: #12
RE: Volume loss converting DSF to Wav with Convolution
And what are your IR files? The convolve/FiltersM* files?
Usually IRs are provided with their typical gain to be applied to them prior to process further, so no need to use gtype=gn (which might be just pure guessing).
If you know max typical gain of your impulse file, use something like: afir=gtype=none:irgain=X
where X is actual gain of impulse file used in convolution.
Find all posts by this user
Quote this message in a reply
24-07-2019, 23:53
Post: #13
RE: Volume loss converting DSF to Wav with Convolution
(24-07-2019 14:05)richardpl Wrote:  And what are your IR files? The convolve/FiltersM* files?
Usually IRs are provided with their typical gain to be applied to them prior to process further, so no need to use gtype=gn (which might be just pure guessing).
If you know max typical gain of your impulse file, use something like: afir=gtype=none:irgain=X
where X is actual gain of impulse file used in convolution.
The IR files are wav files and yes, I can adjust the gain either in Minimserver or when I generate the impulse files.

However, the audio quality of the conversion from DSF to wav could be better. I no longer convert DSF files on the fly; have converted all DSF files to PCM using another tool.
Find all posts by this user
Quote this message in a reply
25-07-2019, 09:23
Post: #14
RE: Volume loss converting DSF to Wav with Convolution
What you found wrong with DSF conversion, and how you did it?
Find all posts by this user
Quote this message in a reply
26-07-2019, 08:46
Post: #15
RE: Volume loss converting DSF to Wav with Convolution
(25-07-2019 09:23)richardpl Wrote:  What you found wrong with DSF conversion, and how you did it?
In critical listening, the converted 24/176.4 tracks had less body and considerably less bass. Even CD 16/44.1 was better than ffmpeg on-the-fly conversion!

I ended converting all the DSF files using Tascam Hi Res Editor. It was a manual conversion and tags had to added with MP3Tag using its tag copy capability.
Find all posts by this user
Quote this message in a reply
26-07-2019, 17:00
Post: #16
RE: Volume loss converting DSF to Wav with Convolution
Which ffmpeg command you used to do conversion to PCM? Have you used soxr resampler instead? (Need ffmpeg build with swresample and soxr support)
Find all posts by this user
Quote this message in a reply
26-07-2019, 20:30
Post: #17
RE: Volume loss converting DSF to Wav with Convolution
The ffmpeg command is generated internally by MinimStreamer. For 24/176 output, it is:

ffmpeg -v 0 -analyzeduration 0 -f dsf -request_sample_fmt flt -i filename.dsf -vn -ar 176400 -f s24le -acodec pcm_s24le -sample_fmt s32 pipe:

or with soxr enabled:

ffmpeg -v 0 -analyzeduration 0 -f dsf -request_sample_fmt flt -i filename.dsf -af aresample=resampler=soxr -vn -ar 176400 -f s24le -acodec pcm_s24le -sample_fmt s32 pipe:

I have tried a few combinations just now. In my system, increasing the output resolution to 32/352 improves the sound. Enabling soxr also improves the sound but not by as much as changing from 24/176 to 32/352.

Any suggestions for changing the ffmpeg command to get better results would be much appreciated.
Find all posts by this user
Quote this message in a reply
28-07-2019, 08:29
Post: #18
RE: Volume loss converting DSF to Wav with Convolution
(26-07-2019 20:30)simoncn Wrote:  I have tried a few combinations just now. In my system, increasing the output resolution to 32/352 improves the sound. Enabling soxr also improves the sound but not by as much as changing from 24/176 to 32/352.

Any suggestions for changing the ffmpeg command to get better results would be much appreciated.

Does using 32/176 helps somehow?
Check with astats filter max peak value for overall and for 24/176 and 32/352.
Do they differ?

Can you share DSF file where it is obvious that bass and body is somehow hurt?
Find all posts by this user
Quote this message in a reply
28-07-2019, 10:31
Post: #19
RE: Volume loss converting DSF to Wav with Convolution
(28-07-2019 08:29)richardpl Wrote:  Can you share DSF file where it is obvious that bass and body is somehow hurt?
You can use these files to test, which all share the same master; there are 16/44.1, DSD64 and other versions.
http://www.2l.no/hires/
Find all posts by this user
Quote this message in a reply
28-07-2019, 21:17
Post: #20
RE: Volume loss converting DSF to Wav with Convolution
(28-07-2019 08:29)richardpl Wrote:  Can you share DSF file where it is obvious that bass and body is somehow hurt?

From this page, download "Vision of Her" in DSD 64 format. If you play it on a renderer that can handle native DSD, it has good bass and body. Now transcode it to PCM 32/176 using FFmpeg. The bass and body are noticeably impaired.

I will do an astats test soon and post the results.
Find all posts by this user
Quote this message in a reply
Post Reply 


Forum Jump:


User(s) browsing this thread: 1 Guest(s)