Hello,
From observation and previous forum research, dBpoweramp's FLAC encoder feeds libFLAC a blocksize of 4608. This was libFLAC's own default until FLAC 1.2.0 (2007) changed it to 4096. 4608 remains fully subset-compliant at CD rates so this is a configurability question, not a defect report.
The reason it matters in practice: I have isolated a decode failure in Aurender's MQA Core Decoder that is triggered specifically by blocksize-4608 FLAC files (untagged MQA-CD rips are detected in-stream, the decode attempt fails, and the track is ejected from the queue). Re-encoding the identical audio at blocksize 4096 — decoded MD5 verified unchanged — decodes correctly.
The defect is Aurender's (their FLAC-frame-to-MQA-decoder handoff appears to assume 4096-sample buffers, and I have reported to them), but their decoder will have been validated overwhelmingly against 4096 files, and other integrated decoders may share the assumption. The official MQATagRestorer utility bundles in libFLAC and re-encodes to 4096. I discovered this problem in a project where I rewrote that utility in python for forward compatibility with new versions of macOS and also to eliminate some other changes that utility would do that had nothing do to with MQA.
Questions:
Thanks as always for the support
From observation and previous forum research, dBpoweramp's FLAC encoder feeds libFLAC a blocksize of 4608. This was libFLAC's own default until FLAC 1.2.0 (2007) changed it to 4096. 4608 remains fully subset-compliant at CD rates so this is a configurability question, not a defect report.
The reason it matters in practice: I have isolated a decode failure in Aurender's MQA Core Decoder that is triggered specifically by blocksize-4608 FLAC files (untagged MQA-CD rips are detected in-stream, the decode attempt fails, and the track is ejected from the queue). Re-encoding the identical audio at blocksize 4096 — decoded MD5 verified unchanged — decodes correctly.
The defect is Aurender's (their FLAC-frame-to-MQA-decoder handoff appears to assume 4096-sample buffers, and I have reported to them), but their decoder will have been validated overwhelmingly against 4096 files, and other integrated decoders may share the assumption. The official MQATagRestorer utility bundles in libFLAC and re-encodes to 4096. I discovered this problem in a project where I rewrote that utility in python for forward compatibility with new versions of macOS and also to eliminate some other changes that utility would do that had nothing do to with MQA.
Questions:
- Is the blocksize currently exposed anywhere in the FLAC encoder settings (command-line box or otherwise)?
- If not, would you consider either exposing it or aligning the default with libFLAC's own 4096?
Thanks as always for the support
Comment