How Long to Convert SACD ISO Files?

[artwwweb]

View attachment 46766

I have installed the SACD plugin to Foobar so can convert ISO to FLAC but I have some questions about the settings. I'm at the very edge of my knowledge here so I imagine that all the options ought to work, but I thought it was best to check before converting all my ISOs. (I will prob then delete the ISOs as they are quite large, so I want to get it right.)

I'm trying to use the same settings as @skherbeck quoted above (partially from research and partially from sheep mentality.)

PCM Sample rate: The default is 44100. If I change it to 88k or 196k the playback speed from Foobar doubles or quadruples! That makes some sort of sense but can't be right. (I was expecting it to be 196k as a default which I would prob change to 88.)

DSD2PCM mode. I've chosen one of those with a low pass filter because I believe that's what actual SACD players have to do (as the DSD creates noise in the ultra high frequencies?) But again, which is really the best option?

PCM Volume. I've chosen +6db. I read that is best because, again, it's what a player does. (I think the reason is this: The head room given by DSD's one bit approach means that it converts to a low level PCM signal.) But on a few ISOs this results in severe distortion. I'm happy to use +6db if that is the norm, and then re-do the few that don't work. Is there a way to tell which I should use?

LFE I've left "as is" rather than a boost or drop.

Should I be using a DSD processor? It seems to work without choosing that option even though I also have the DSD Processor installed in Foobar.

So... what do others do? Is there a way to definitively tell what to do by looking at the properties of the ISO?

Do I need to even bother with these options? I don't really care about ISO playback on foobar. Perhaps the conversion works on its own parameters and ignores these?

When I choose convert, I get a warning that foobar does not support DSD passthru, so the conversion is not lossless. Of course, that looks worrying, but is it just the basic problem that any conversion from DSD to PCM is not strictly lossless? So although it's technically correct, I shouldn't worry too much? (I was also wondering if there are some conversion options that are less lossy than others.)

I'm not expecting a complete, definitive answer to all this, but if anybody has a comment on even part of this it might help me feel more confident in choosing my options. (Tbh, I probably wouldn't be able to tell the difference, but I like to go with what's best to give my ears a chance to notice.)

 

[HomerJAU]

There’s at least two posts here on QQ with screenshots of recommended settings. I’ve just spent 20 minutes looking for them and failed... 88.2kHz is a good compromise for file size/quality. You will need a high pass filter for sure.

Edit: Here is one:
https://www.quadraphonicquad.com/fo...ith-oppo-pioneer-bd-players.22388/post-381149

 

[artwwweb]

There’s at least two posts here on QQ with screenshots of recommended settings. I’ve just spent 20 minutes looking for them and failed... 88.2kHz is a good compromise for file size/quality. You will need a high pass filter for sure.

Edit: Here is one:
https://www.quadraphonicquad.com/fo...ith-oppo-pioneer-bd-players.22388/post-381149

Thanks for taking the time to answer @HomerJAU. I did quote @skherbeck from this very thread, but the image of his settings is not shown directly, so it's not that obvious. Interesting to see that in the post you point to, also by @skherbeck, the settings are the same except for the PCM volume boost.

I think I am going to use the +6db along with the other settings, but I'd still love to feel more certain that they are the best. It feels a bit like I am "pressing random buttons until it works" rather than truly knowing and then doing what is best. Maybe I need to ask on a Foobar forum.

Like I said before, I'm not expecting anyone to be able to give a full, definitive answer, but if anybody has learned anything about the differences between the settings it might help improve our joint understanding.

 

[sukothai]

...I think I am going to use the +6db along with the other settings, but I'd still love to feel more certain that they are the best...

Every SACD is different; so blindly adding 6dB isn't recommended. See this post:

https://www.quadraphonicquad.com/forums/threads/not-streaming-dedicated-pc.24967/post-373858
and this:

https://www.quadraphonicquad.com/forums/threads/not-streaming-dedicated-pc.24967/post-373869

 

[ar surround]

Every SACD is different; so blindly adding 6dB isn't recommended. See this post:

https://www.quadraphonicquad.com/forums/threads/not-streaming-dedicated-pc.24967/post-373858
and this:

https://www.quadraphonicquad.com/forums/threads/not-streaming-dedicated-pc.24967/post-373869

Yes, tell me about it. I had to redo a few conversions because I blindly added the 6dB. Consequently, I never add more than 3dB and simply live with the result.

 

[jimfisheye]

Using the correct formats and transcoding approach is actually important!
The 32:1 decimation setting that results in 88.2k sample rate PCM files is a virtually lossless transcode. Anything else can be altering to different degrees. Clipping and really mangling it is very possible to do.

If you aren't tech savvy (DAW savvy), convert it to 88.2kHz, 24 bit with the -6db option.

DAW savvy?
32:1 decimation which results in 88.2k files but go to 32 bit floating point instead of 24 bit fixed.
The transcode from DSD to PCM here can result in a peak level anywhere from -6dbfs to +6dbfs. Anything above 0 clips in PCM fixed point.

DSD -> 88.2k 32 bit floating point -> normalize to -0.01 -> render to 24 bit fixed

Use your favorite DAW app to normalize the program to just under 0db. Render new 24 bit files.

* That blanket -6db approach I suggested above would give you 22 bits of data instead of 24 in the worst case scenario. Use your volume control.

 

[artwwweb]

After some experimentation I have discovered how to check for clipping during the foobar conversion process. Just in case others do not know, I thought I'd share:

In short: Open a console window in Audacity, then convert, then search the console output for the word 'overload'.

More detail:
In Audacity, choose View/Console. This opens a console window. Tick the Write Log checkbox and it will prompt you to choose/create a log file. (You could just read the console, but it's easier to search if saved as a file.)

Run your conversion(s).

Open the log file. Search for the word 'overload'. If you find nothing, all is OK.

If there are overloads it will tell you the exact time at which the peaks occur. You then know to re-convert at a lower level.

I have converted 75 SACDs at +3db and had 3 which had a few overloads. (Blood On The Tracks, Paul Weller - Studio 150 and The Stones' Sympathy EP. Two of those are 'modern' recordings - the Weller and The Stones remixes, which I kind of expected to be louder. About half my SACDs are Dutton Vocalions - of course! - and they were all fine - of course!) Reconverting at +0db produced no clipping.

I've been converting in batches of 10, many of which are DV twofers, and it takes about 70 mins per batch on my not very fast PC (2.9GHz, 8Gb RAM) so about 5 mins per album. Also, doing 10 at a time means that although it took a while, being an unattended process meant it wasn't onerous.

There is an option in the SACD options dialogue to create an error log if the conversion causes clipping. I'm not sure what this does - I've not found a separate conversion log. This should be the obvious thing to use rather than using the console but as I couldn't find a log I had no idea if there were errors or not. However, I have kept the option ticked in case it is that which writes the clipping info to the console. I could turn it off and test but there was no incentive to do so.

 

[artwwweb]

After some experimentation I have discovered how to check for clipping during the foobar conversion process. Just in case others do not know, I thought I'd share:

In short: Open a console window in Audacity, then convert, then search the console output for the word 'overload'.

More detail:
In Audacity, choose View/Console. This opens a console window. Tick the Write Log checkbox and it will prompt you to choose/create a log file. (You could just read the console, but it's easier to search if saved as a file.)

Run your conversion(s).

Open the log file. Search for the word 'overload'. If you find nothing, all is OK.

If there are overloads it will tell you the exact time at which the peaks occur. You then know to re-convert at a lower level.

I have converted 75 SACDs at +3db and had 3 which had a few overloads. (Blood On The Tracks, Paul Weller - Studio 150 and The Stones' Sympathy EP. Two of those are 'modern' recordings - the Weller and The Stones remixes, which I kind of expected to be louder. About half my SACDs are Dutton Vocalions - of course! - and they were all fine - of course!) Reconverting at +0db produced no clipping.

I've been converting in batches of 10, many of which are DV twofers, and it takes about 70 mins per batch on my not very fast PC (2.9GHz, 8Gb RAM) so about 5 mins per album. Also, doing 10 at a time means that although it took a while, being an unattended process meant it wasn't onerous.

There is an option in the SACD options dialogue to create an error log if the conversion causes clipping. I'm not sure what this does - I've not found a separate conversion log. This should be the obvious thing to use rather than using the console but as I couldn't find a log I had no idea if there were errors or not. However, I have kept the option ticked in case it is that which writes the clipping info to the console. I could turn it off and test but there was no incentive to do so.

Aaagh! I've just realised I made a 'typo' in the above post. When I mention Audacity I meant to say Foobar, of course. (I can't edit the post.)

After some experimentation I have discovered how to check for clipping during the Foobar conversion process. Just in case others do not know, I thought I'd share:

In short: Open a console window in Foobar, then convert, then search the console output for the word 'overload'.

More detail:
In Foobar, choose View/Console. This opens a console window. Tick the Write Log checkbox and it will prompt you to choose/create a log file. (You could just read the console, but it's easier to search if saved as a file.)

Run your conversion(s).

Open the log file. Search for the word 'overload'. If you find nothing, all is OK.

If there are overloads it will tell you the exact time at which the peaks occur. You then know to re-convert at a lower level.

I have converted 75 SACDs at +3db and had 3 which had a few overloads. (Blood On The Tracks, Paul Weller - Studio 150 and The Stones' Sympathy EP. Two of those are 'modern' recordings - the Weller and The Stones remixes, which I kind of expected to be louder. About half my SACDs are Dutton Vocalions - of course! - and they were all fine - of course!) Reconverting at +0db produced no clipping.

I've been converting in batches of 10, many of which are DV twofers, and it takes about 70 mins per batch on my not very fast PC (2.9GHz, 8Gb RAM) so about 5 mins per album. Also, doing 10 at a time means that although it took a while, being an unattended process meant it wasn't onerous.

There is an option in the SACD options dialogue to create an error log if the conversion causes clipping. I'm not sure what this does - I've not found a separate conversion log. This should be the obvious thing to use rather than using the console but as I couldn't find a log I had no idea if there were errors or not. However, I have kept the option ticked in case it is that which writes the clipping info to the console. I could turn it off and test but there was no incentive to do so.

 

[par4ken]

I have found that you can get much better DSD to PCM results by using one of the filters available here.

http://s-audio.systems/dsd-filter/?lang=en
S.Audio.Systems DSD Filter and S.Audio.Systems DSD Short Filter MP Filter. The short filter sounds the best to me but you have to use 176.4 KHz or higher (overkill I know). The other filter will work with 88.2 KKz.
They can be installed in Foobar and many others.

 

[lukpac]

Foobar does multi threaded conversions. An 8 core processor does 8 files at at same time, but HDD write speed is the major bottleneck with even an 8 core processor.

Note for DSD to PCM conversion, you *don't* want to do conversions multi-threaded, as track breaks will be discontinuous. If one track segues to the next, you're likely to get a pop or click. Doing them single-threaded (with the correct settings) will avoid that. Details here:

https://sourceforge.net/p/sacddecoder/bugs/134/

 

[par4ken]

Note for DSD to PCM conversion, you *don't* want to do conversions multi-threaded, as track breaks will be discontinuous. If one track segues to the next, you're likely to get a pop or click. Doing them single-threaded (with the correct settings) will avoid that. Details here:

https://sourceforge.net/p/sacddecoder/bugs/134/

I don't understand about the multi thread part, I don't see why that would make a difference. I do understand that it might be preferable to convert a single DSD file to a single PCM file (with a cue sheet) though, as track breaks in the wrong place can cause clicks and pops.

Years ago I had that same problem when burning CD's from my vinyl rips, they had clicks between each track. Then I started splitting tracks with CD Wave Editor, which is now a free program. I still use it!
MiLo Software

 

[lukpac]

I don't understand about the multi thread part, I don't see why that would make a difference. I do understand that it might be preferable to convert a single DSD file to a single PCM file (with a cue sheet) though, as track breaks in the wrong place can cause clicks and pops.

Years ago I had that same problem when burning CD's from my vinyl rips, they had clicks between each track. Then I started splitting tracks with CD Wave Editor, which is now a free program. I still use it!
MiLo Software

It's because (as far as I'm aware) of the nature of DSD being a delta-sigma format. That is, it doesn't encode absolute values, just change from the previous value. At the beginning of a track, there is no previous value to reference, so the PCM conversion will start at digital silence even if it shouldn't. When converting single-threaded (with the proper settings and foo_input_sacd >= 1.2.5), the previous track is referenced, so the samples at the start of each track are correct.

The issue with CDs is different: track breaks have to occur at frame boundaries, and if they don't (and your burning software isn't smart enough to automatically adjust), there will be silence from the end of the track to the frame boundary.