#81 Proper bitdepth reporting with DSD

[Anonymous]

Hi Maxim, could you please include $info(decoded_bitspersample) in your foo_input_sacd component?
I think that $info(bitspersample) should always show 1bit no matter if we set output to DSD or PCM and the $info(decoded_bitspersample) should show values depending on the chosen output format (1 for DSD and 24(?) for PCM). Now the $info(bitspersample) changes with the output, which is weird as the file is still 1bit.

[Maxim V.Anisiutkin]

Added in foo_input_udsd v0.0.11. Convert gets 24 bits.

[Anonymous]

Thank you, I have never used the foo_input_udsd component. Could you also add those "decoded_bitspersample", "dsd_bitspersample" and "dsd_samplerate" parameters to foo_input_sacd component as most people use that one?

[Maxim V.Anisiutkin]

It's a replacement for foo_input_sacd component, which is retired now.

[Anonymous]

I am sorry to hear that because foo_input_sacd works great for me both in stereo via USB-ASIO (as Native) and 5.1 via HDMI (as PCM) and consumes little resources.
I tried foo_input_udsd v0.0.11 and it stutters, produces some weird noises, stops etc. even with stereo. It also consumes much of procesors power even in stereo. As to 5.1 I can't even listen to it as it really almost hangs my laptop. Plus it cuts to early on the last track of playlist. I tired 32 bit, 64 bit and full and portable - it does not work properly. I just can't use it so it would be great if you decided to include those 3 commands to the foo_input_sacd.

[Anonymous]

Just to confirm - other user also said the newer component behaves strange. He said: "Your stutter comment made me run a quick performance test between foo_input_sacd vs foo_input_udsd and in my test the old component managed to convert DSF file to PCM five (5) times faster than the new one. All decoder options are identical (defaults). DSD is slow and inefficient to begin with but this new component seems to require much stronger hardware to perform well. "

[Anonymous]

I am sorry to bother you but since there is new FB version 2.24 out and it supports new commands - could you also add %channel_mask% and "stream_count" (though I am not sure if the second one is applicable to DSD)?

[Maxim V.Anisiutkin]

"channel_mask" is added in v0.0.12. "stream_count" is not relevant because for all playable DSD streams is equal to one. This version should have approximately the same performance as the latest foo_input_sacd.

[Anonymous]

It still stutters (even crashed foobar once) and consumes about 96% of processor (foo input sacd consumes 8-20%). It still cuts the ending of each song, the track changes and the ending of the previous song is heard as the beginning of next. Because there is nothing after the last song in the playlist, the ending of last song is just cut off (I think it's about 1 second). The cutting only happens with DSD ( I tried one mp3 and one flac and it seemed ok). The "channel mask" worked for stereo SACD but not for 5.1 (produced empty space).

[Maxim V.Anisiutkin]

Could you share that crash dump file please?

[Anonymous]

Unfortunately there is no crash dump - I checked the crash report folder and it is not there. It happened on 32 portable 2.24 (my test FB). FB just closed suddenly and when restarted it asked if I want to open it in safe mode or normally but no crash dump. I played sacd again but it didn't crash this time. The processor load is so high that even my mouse cursor jumps/stutters across the display (again 96%).

[Anonymous]

reinstalled the portable version of 2.24, this time I made it 64bit and managed to crash FB with foo_input_udsd.
I also discovered that while "normal" dsd eats more resources than with foo_input_sacd but is responsive. But when it is DST and 5.1 it reaches 96% while with input_sacd it was no more than 40-45-50.
Crash log attached.

[Maxim V.Anisiutkin]

Thank you. I added parameters to foo_input_sacd v 1.5.12. Interesting, what difference in performance it has compared to foo_input_udsd.

[Anonymous]

Thank you, foo_input_sacd v 1.5.12 works well.
Two minor "issues":
1. "channel mask" does not produce anything for 5.1 SACD's but it works well with stereo.
2. "decoded_bitspersample" always shows 24 no matter if I change the output mode to DSD, DSD+PCM or PCM. What does it show then? If DSD is chosen the signal shouldn't be decoded and the "meters" do not work so there is no "fake" PCM stream for visualisation. What do those 24 bits mean then?

[Maxim V.Anisiutkin]

1. Even after Properties->Tools->Reload info?
2. The "natural" format for both PCM and DoP is 24-bit.

[Anonymous]

1. Reloading does not help - the field is still blank for 5.1 SACDs (I checked different albums).
2. Does it mean that "decoded_bitspersample" takes data after being DoPed and before deDoPed? I use ASIO+DSD for Native Output.

[Anonymous]

I really tried everything to make "channel mask" work with 5.1 sacd (including reloading info) but it still stays empty. Could you please check it?