Converting The Thriller SACD to Music Files

QuadraphonicQuad

Help Support QuadraphonicQuad:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

ssully

2K Club - QQ Super Nova
QQ Supporter
Since 2002/2003
Joined
Jul 2, 2003
Messages
3,841
Location
in your face
I've found a few interesting things while ripping all my SACDs to 88/24 PCM, mainly that DSD digital levels can be all over the place. Some have to be ripped with 0dB in order not to overload as 88/24 PCM, some can take +6dB of boost, most are in between.


Among the strangest is the 1999 single-layer, nonhybrid Sony SACD of Thriller (Michael Jackson) -- for that one, even with +0 dB boost, there are still some digital 'overs' ,on three tracks (The Girl is Mine, Beat It, Billie Jean), when converted to 88/24. This is the only SACD of mine so far (out of a few dozen) that this is true for.

chain is Pioneer BDP-80 ethernet out --> iso2dsd (Raw iso output) --> Foobar2k w/foo_input_SACD (88khz PCM output, Multistage 32fp DSD2PCM mode) --> wav-->flac (output is always 24 bits)

Checking the 'Log Overloads' option and using View--> Console when using foobar_input_sacd is a convenient way to document overloads.
Replaygain scanning of the original DSD64 track of 'Billie Jean' embedded in the iso says it peaks at +1.062; converting to PCM at the above settings produces 35 'overs', all within the first 18 seconds.
Changing PCM output to 44 kHz reduces the number of overs to 4;, changing it to 176 kHz produces way more than 35, so I assume it's an effect of ultrasonic noise.
Using the 88kHz output setting with Direct 32fp 30Khz lowpass DSD2PCM mode reduces overs from 35 to 33, suggesting that ultrasonic noise between 22 and 30khz is the main culprit.

I've tried fiddling with drastically reducing preamp settings in foobar, and with changing f2k output device settings (DS vs DSD vs WSAPI), changing output format to 32 bit where possible, but they seem to have no effect on the level of the signal going into the converter, so nothing changes.

It's curious but not disastrous. The overs are brief and not extremely numerous, unlikely to be audible, on the three tracks.
 
I applaud your patience on that ssully. Honestly, I just don't have so much patience with this stuff. I just want to rip it, and listen. I find it all annoying that they all don't behave at least very similarly. I wish I didn't know this... :mad:@:
 
Thriller has some glitches in the DSD data stream. Convert to PSM and clean up those errant peaks with software like you would use to remove pops from an LP rip.

I've found a few interesting things while ripping all my SACDs to 88/24 PCM, mainly that DSD digital levels can be all over the place. Some have to be ripped with 0dB in order not to overload as 88/24 PCM, some can take +6dB of boost, most are in between.


Among the strangest is the 1999 single-layer, nonhybrid Sony SACD of Thriller (Michael Jackson) -- for that one, even with +0 dB boost, there are still some digital 'overs' ,on three tracks (The Girl is Mine, Beat It, Billie Jean), when converted to 88/24. This is the only SACD of mine so far (out of a few dozen) that this is true for.

chain is Pioneer BDP-80 ethernet out --> iso2dsd (Raw iso output) --> Foobar2k w/foo_input_SACD (88khz PCM output, Multistage 32fp DSD2PCM mode) --> wav-->flac (output is always 24 bits)

Checking the 'Log Overloads' option and using View--> Console when using foobar_input_sacd is a convenient way to document overloads.
Replaygain scanning of the original DSD64 track of 'Billie Jean' embedded in the iso says it peaks at +1.062; converting to PCM at the above settings produces 35 'overs', all within the first 18 seconds.
Changing PCM output to 44 kHz reduces the number of overs to 4;, changing it to 176 kHz produces way more than 35, so I assume it's an effect of ultrasonic noise.
Using the 88kHz output setting with Direct 32fp 30Khz lowpass DSD2PCM mode reduces overs from 35 to 33, suggesting that ultrasonic noise between 22 and 30khz is the main culprit.

I've tried fiddling with drastically reducing preamp settings in foobar, and with changing f2k output device settings (DS vs DSD vs WSAPI), changing output format to 32 bit where possible, but they seem to have no effect on the level of the signal going into the converter, so nothing changes.

It's curious but not disastrous. The overs are brief and not extremely numerous, unlikely to be audible, on the three tracks.
 
I've found a few interesting things while ripping all my SACDs to 88/24 PCM, mainly that DSD digital levels can be all over the place. Some have to be ripped with 0dB in order not to overload as 88/24 PCM, some can take +6dB of boost, most are in between.


Among the strangest is the 1999 single-layer, nonhybrid Sony SACD of Thriller (Michael Jackson) -- for that one, even with +0 dB boost, there are still some digital 'overs' ,on three tracks (The Girl is Mine, Beat It, Billie Jean), when converted to 88/24. This is the only SACD of mine so far (out of a few dozen) that this is true for.

chain is Pioneer BDP-80 ethernet out --> iso2dsd (Raw iso output) --> Foobar2k w/foo_input_SACD (88khz PCM output, Multistage 32fp DSD2PCM mode) --> wav-->flac (output is always 24 bits)

Checking the 'Log Overloads' option and using View--> Console when using foobar_input_sacd is a convenient way to document overloads.
Replaygain scanning of the original DSD64 track of 'Billie Jean' embedded in the iso says it peaks at +1.062; converting to PCM at the above settings produces 35 'overs', all within the first 18 seconds.
Changing PCM output to 44 kHz reduces the number of overs to 4;, changing it to 176 kHz produces way more than 35, so I assume it's an effect of ultrasonic noise.
Using the 88kHz output setting with Direct 32fp 30Khz lowpass DSD2PCM mode reduces overs from 35 to 33, suggesting that ultrasonic noise between 22 and 30khz is the main culprit.

I've tried fiddling with drastically reducing preamp settings in foobar, and with changing f2k output device settings (DS vs DSD vs WSAPI), changing output format to 32 bit where possible, but they seem to have no effect on the level of the signal going into the converter, so nothing changes.

It's curious but not disastrous. The overs are brief and not extremely numerous, unlikely to be audible, on the three tracks.

I haven't heard that disc but I would be very curious to check Byrds - Greatest Hits SACD, it's my poster child for SACD "distortion" but I have no way to measure it. So I wonder if the level was simply too hot.
 
Thriller has some glitches in the DSD data stream. Convert to PSM and clean up those errant peaks with software like you would use to remove pops from an LP rip.


You mean PCM, not PSM?

These glitches aren't audible when I play the output tracks, so they don't behave like LP pops, thankfully.

Yes, I can go in and de-clip with an editor but I'm curious as to what happened here. This must be one of the earliest SACDs ever offered, I guess there was a learning curve?

foo_input_sacd has only positive PCM volume settings from +0 to +6dB. I wonder if negative settings would allow a 'fix' for this? I've put in a question to the developer on sourceforge.
 
I've found a few interesting things while ripping all my SACDs to 88/24 PCM, mainly that DSD digital levels can be all over the place. Some have to be ripped with 0dB in order not to overload as 88/24 PCM, some can take +6dB of boost, most are in between.


Among the strangest is the 1999 single-layer, nonhybrid Sony SACD of Thriller (Michael Jackson) -- for that one, even with +0 dB boost, there are still some digital 'overs' ,on three tracks (The Girl is Mine, Beat It, Billie Jean), when converted to 88/24. This is the only SACD of mine so far (out of a few dozen) that this is true for.

chain is Pioneer BDP-80 ethernet out --> iso2dsd (Raw iso output) --> Foobar2k w/foo_input_SACD (88khz PCM output, Multistage 32fp DSD2PCM mode) --> wav-->flac (output is always 24 bits)

Checking the 'Log Overloads' option and using View--> Console when using foobar_input_sacd is a convenient way to document overloads.
Replaygain scanning of the original DSD64 track of 'Billie Jean' embedded in the iso says it peaks at +1.062; converting to PCM at the above settings produces 35 'overs', all within the first 18 seconds.
Changing PCM output to 44 kHz reduces the number of overs to 4;, changing it to 176 kHz produces way more than 35, so I assume it's an effect of ultrasonic noise.
Using the 88kHz output setting with Direct 32fp 30Khz lowpass DSD2PCM mode reduces overs from 35 to 33, suggesting that ultrasonic noise between 22 and 30khz is the main culprit.

I've tried fiddling with drastically reducing preamp settings in foobar, and with changing f2k output device settings (DS vs DSD vs WSAPI), changing output format to 32 bit where possible, but they seem to have no effect on the level of the signal going into the converter, so nothing changes.

It's curious but not disastrous. The overs are brief and not extremely numerous, unlikely to be audible, on the three tracks.

Yes, the Thriller SACD like some others was mastered out of ScarletBook specifications which only very occasionally allow the maximum peak to reach +3.1 dB with respect to the SACD 0dB audio reference level.
 
Yes, the Thriller SACD like some others was mastered out of ScarletBook specifications which only very occasionally allow the maximum peak to reach +3.1 dB with respect to the SACD 0dB audio reference level.

The Thriller SACD was discussed before on QQ. And it turns out that the disc is indeed within the Scarlet Book spec.
 
Yes, the Thriller SACD like some others was mastered out of ScarletBook specifications which only very occasionally allow the maximum peak to reach +3.1 dB with respect to the SACD 0dB audio reference level.

Based on my tests the SACD DSD2PCM conversion of the Thriller may generate ≥ +6dB (SACD) artificial peaks (glitches) at several positions. Having removed those glitches the maximum peak of the album is of ca. +5dB (SACD).
 
Based on my tests the SACD DSD2PCM conversion of the Thriller may generate ≥ +6dB (SACD) artificial peaks (glitches) at several positions. Having removed those glitches the maximum peak of the album is of ca. +5dB (SACD).

That thread is confusing and inconclusive

1) are you/they talking about the *stereo* DSD tracks? (I didn't even know there was a multchannel SACD version of "Thriller" -- mine is stereo-only

2) for me , it's not track 7 that is the problem for the foo_input_sacd/foo_converter combo; it's tracks 3, 5 and 6 that report 'overs' even when PCM Volume boost is set to +0dB . Track 7 is fine, no overs.

2) when you examined the disc were there peaks *greater than* maximum SACD spec versus just *equal to* maximum?
 
Last edited:
That thread is confusing and inconclusive

1) are you/they talking about the *stereo* DSD tracks? (I didn't even know there was a multchannel SACD version of "Thriller" -- mine is stereo-only

2) for me , it's not track 7 that is the problem for the foo_input_sacd/foo_converter combo; it's tracks 3, 5 and 6 that report 'overs' even when PCM Volume boost is set to +0dB . Track 7 is fine, no overs.

2) when you examined the disc were there peaks *greater than* maximum SACD spec versus just *equal to* maximum?

Thriller SACD is a stereo-only disc, of course. Try to do a DSD2PCM conversion of Thriller applying -2dB gain. In the wave form you may see the glitches I mentioned.
 
Thriller SACD is a stereo-only disc, of course. Try to do a DSD2PCM conversion of Thriller applying -2dB gain. In the wave form you may see the glitches I mentioned.

Here you are (Saracon D2P 24/88.2 conversion, TPDF, -2dB gain):

Thriller.jpg
 
Yes, those are the three tracks that foo_input_sacd flags as having overs in my setup.

Are people still claiming Thriller is within Scarlet Book spec?

(Unfortunately foo_input_sacd does not offer negative gain options for its DSD to PCM conversions. I have submitted a feature request ticket to its developer.)
 
Yes, those are the three tracks that foo_input_sacd flags as having overs in my setup.

Are people still claiming Thriller is within Scarlet Book spec?

(Unfortunately foo_input_sacd does not offer negative gain options for its DSD to PCM conversions. I have submitted a feature request ticket to its developer.)

In another forum forum member Black Elk wrote:

Thriller was mastered only once for SACD by Bernie Grundman, however, as I wrote in this thread:

http://www.stevehoffman.tv/forums/showthread.php?t=87013

it was mastered louder than the Scarlet Book spec. permitted due to a lack of good SACD metering at the time. This led to some issues between Sony and other Japanese SACD player manufacturers (especially the ones doing DSD to PCM conversion) with regard to output stage and/or converter clipping. It is possible that Sony in Japan re-modulated the DSD master, padding its level down a few dB, for later production runs. Maybe someone who bought both the US and Japanese editions at initial release can confirm that they have the same level?

Btw, I wish I could analyze another rare single-layer stereo-only SACD - Aerosmith / Rocks :yikes :scream .
 
Yes, those are the three tracks that foo_input_sacd flags as having overs in my setup.

Are people still claiming Thriller is within Scarlet Book spec?

(Unfortunately foo_input_sacd does not offer negative gain options for its DSD to PCM conversions. I have submitted a feature request ticket to its developer.)

It is obviously out of spec.

But, it wasn't "mastered" to be out of spec. There are glitches that cause it to be out of spec.

If you decode it at the gain level appropriate for an in spec SACD, then clean up the glitches, everything is fine.

So, it is kind of semantics. It isn't like it was intentionally mastered too loud, it has errors.
 
Back
Top