Post Reply 
 
Thread Rating:
  • 0 Votes - 0 Average
  • 1
  • 2
  • 3
  • 4
  • 5
Volume loss converting DSF to Wav with Convolution
29-07-2019, 19:24 (This post was last modified: 29-07-2019 19:25 by richardpl.)
Post: #21
RE: Volume loss converting DSF to Wav with Convolution
The various converters may apply additional normalization and even compression (like in that 44.1k variant) after conversion. FFmpeg command you use does not do this. Normalization cant be done in single pass, you need to find value with astats/volumedetect and apply it with volume filter. For best output you may use double pcm variant of codec in intermediate steps for best final result.
So when you do conversion of DSD64 to pcm, I'm afraid you will need to first convert to wav and double sample format and than after normalization convert to final output which is s24 FLAC. (or even better keep it in float pcm)

When resampling you may use this additional parameter: -af aresample=44100:isf=flt:osf=dbl:precision=33

This is with my limited listening test.
Find all posts by this user
Quote this message in a reply
29-07-2019, 19:33
Post: #22
RE: Volume loss converting DSF to Wav with Convolution
The astats filter max peak value is the same for 24/176 and 32/352.
Find all posts by this user
Quote this message in a reply
29-07-2019, 19:42
Post: #23
RE: Volume loss converting DSF to Wav with Convolution
(29-07-2019 19:24)richardpl Wrote:  The various converters may apply additional normalization and even compression (like in that 44.1k variant) after conversion. FFmpeg command you use does not do this. Normalization cant be done in single pass, you need to find value with astats/volumedetect and apply it with volume filter. For best output you may use double pcm variant of codec in intermediate steps for best final result.
So when you do conversion of DSD64 to pcm, I'm afraid you will need to first convert to wav and double sample format and than after normalization convert to final output which is s24 FLAC. (or even better keep it in float pcm)

When resampling you may use this additional parameter: -af aresample=44100:isf=flt:osf=dbl:precision=33

This is with my limited listening test.

Thanks very much for this. How much of this (if anything) can be done by a one-shot real-time on-the-fly conversion? This is a constraint of how MinimStreamer works. Would adding isf and osf to the aresample filter help?
Find all posts by this user
Quote this message in a reply
01-08-2019, 08:27
Post: #24
RE: Volume loss converting DSF to Wav with Convolution
(29-07-2019 19:42)simoncn Wrote:  Thanks very much for this. How much of this (if anything) can be done by a one-shot real-time on-the-fly conversion? This is a constraint of how MinimStreamer works. Would adding isf and osf to the aresample filter help?

It would not help much regarding less overall volume level you get when resampling to much lower sampling rate.
Find all posts by this user
Quote this message in a reply
01-08-2019, 12:00
Post: #25
RE: Volume loss converting DSF to Wav with Convolution
What about applying a volume filter to increase the gain as part of the single-pass conversion? The gain would be chosen by the user as an average value. Not perfect, but probably better than applying no gain at all.

Another option (harder to implement) might be to first convert a short section of the audio file to come up with an approximate normalization value, then convert the whole file using this value. Does this make sense?

You also mentioned compression for 44.1 files. Is there an FFmpeg filter that can do this?
Find all posts by this user
Quote this message in a reply
Post Reply 


Forum Jump:


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