Quad SACDs: incorrect dynamic range readings and how to interpret them

QuadraphonicQuad

Help Support QuadraphonicQuad:

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

steelydave

Super Moderator
Staff member
Moderator
Moderator
Since 2002/2003
Joined
Apr 21, 2002
Messages
3,078
Location
Toronto, ON
(Warning, technical stuff ahead.)

Ever since Audio Fidelity started releasing old quad masters on SACD a few years ago, it always vexed me that the dynamic range of the quad mix on the multichannel layer registered lower on the dynamic range meter than its stereo counterpart. It didn't seem to logically follow that an audiophile label would take the time to compress the dynamic range of the quad mix when they didn't do it to the stereo mix. When I noticed the DR readings of the quad mixes on the Dutton Vocalion quad SACDs were even worse, I knew something was amiss. I don't have Steve Hoffman's golden ears, but some of the quad mixes on the D-V discs were registering as low as DR7 (which is earbleeder territory for me) and I was sure I would have noticed something like that while listening to these discs. The other thing that was giving me cognitive dissonance was that the Sony Japan quad SACDs of the Jeff Beck albums had much higher DR ratings on the quad layers (similar to the DR of the stereo mix) despite sounding just as good to my ears as the AF and D-V discs.

Just to backtrack for a moment: I've ripped my entire music collection to FLAC, both CD's and high-res PCM material (both stereo and multichannel), along with having a pretty big collection of SACD ISOs. So I have a relatively large collection of digital files - I'm not sure how many at the moment but it's in the tens of thousands. One of the things I do as part of my ripping process is do a dynamic range reading using foobar's DR meter - I know DR readings aren't a be all and end all for sound quality, but it gives me a sort of baseline way of checking if something is going to sound decent or not, and also if new remasters come out I can see if they're worth investigating. ie if I have a CD rip of something that's DR14 and a new remaster comes out that's DR9, I know it's probably not worth my time. So I do DR readings for everything, and on discs that have stereo and multichannel layers I do seperate readings, and append '2.0' or '5.1' (or whatever the channel configuration is) to the log filename, and then dump the log files in the folder with the album artwork.

As I've been ripping all my AF and D-V SACD's over the last couple of years, as I said, I was troubled that the dynamic range of the quad mixes was always lower than the stereo mixes. AF discs seemed to be 4 or 5dB lower, and the D-V discs seemed to be worse, sometimes up to 6dB lower than their stereo counterparts, which didn't correlate at all to what I was hearing with my own ears.

The DR logfiles didn't seem to offer any insight either, and everything seemed like it was functioning properly, no errors or anything. When you do a DR analysis on a whole album, it shows the DR for each track, and then gets an overall rating by adding the rating for each track together and then dividing it by the number of tracks, rounding to the nearest whole number.

Then I had a small flash of genius (I have these about once a year, so I guess I've peaked early this year) and remembered that if you do a DR reading on a single track, it'll show you the dynamic range readings for each individual channel. I thought maybe in quad mixes that the front speakers were registering really low DR numbers due to only having drums and bass or something and skewing the results. What I actually discovered was that unlike PCM playback formats (BluRay, DVD-A etc.) I guess SACD doesn't have 4.0 channel configuration in its specs, so discs are authored as either 5.0 (AF discs) or 5.1 (D-V discs) with silent center or center and LFE channels.

The impact of this highlights a shortcoming in the way the DR meter calculates dynamic range. For a 5.0 track, it takes the dynamic ranges of the 5 channels (FL, FR, C, SL, SR) adds them together, and then divides by 5 to get the DR rating, rounding to the nearest whole number. So on these quad SACDs, the silent channels register as DR0, which lowers the overall rating! In the case of AF discs the center channel registers as DR0, and on D-V discs both the center and LFE channels both register as DR0. Let me demonstrate from logfiles:

This is the DR for the track 'Lahaina' from the Loggins & Messina 'Full Sail' disc that AF put out last year.
FL: 10.23 dB
FR: 12.98 dB
C: 0.00 dB
SL: 11.54 dB
SR: 11.30 dB

Official DR Value: DR9

Samplerate: 2822400 Hz / PCM Samplerate: 88200 Hz
Channels: 5
Bits per sample: 1
Bitrate: 14112 kbps
Codec: DST64

As you can see, the individual channels register between 10.23 and 12.98 yet the calculated DR is DR9, with the DR0 of the center channel skewing things.


Things get even more skewed on the D-V discs because of the empty LFE track. Here's a DR rating for 'Caravan' from the Hugo Montenegro 'Others By Brothers' SACD from a couple of years ago:

FL: 12.48 dB
FR: 12.93 dB
C: 0.00 dB
LFE: 0.00 dB
SL: 14.88 dB
SR: 11.71 dB

Official DR Value: DR9

Samplerate: 2822400 Hz / PCM Samplerate: 88200 Hz
Channels: 6
Bits per sample: 1
Bitrate: 16934 kbps
Codec: DST64

Again, you can see that the individual channel values are really high - one nearly as high as DR15 - but it clocks in overall at a measly DR9.


Compare that with 'Black Cat Moan' from the Beck Bogert & Appice quad SACD that Sony Japan put out - these have content in both the center and LFE generated from the other four channels, and thus you get a higher DR rating even though the individual channels are lower than the two previous examples:

FL: 12.24 dB
FR: 12.61 dB
C: 13.48 dB
LFE: 11.16 dB
SL: 9.29 dB
SR: 9.70 dB

Official DR Value: DR11

Samplerate: 2822400 Hz / PCM Samplerate: 88200 Hz
Channels: 6
Bits per sample: 1
Bitrate: 16934 kbps
Codec: DST64


So if you want to get a better idea of the dynamic range of AF and D-V quad SACD's there's a simple equation you can do.

For AF discs, take the DR reading you get, multiply by 5 (the number of channels on the disc), and divide by 4 (the number of channels with actual sound). So for the Loggins & Messina track above for example, DR9 x 5 / 4 = DR11.25 which is about right.

For D-V discs, you do DR reading x 6 / 4. So for the Montenegro track above, DR9 x 6 / 4 = DR13.5


I'm not sure if this is of any interest to anyone but me, but it felt reassuring somehow to be able to somewhat scientifically verify that the quad layers on all these AF and D-V SACDs were just as dynamic as their stereo counterparts.
 
Genius!! I have wondered the exact same thing Dave!! But I wasn't in any position to understand it. Thanks for figuring that out!!!


Sent from my iPhone using Tapatalk
 
@Dave,

I convert an SACD album to PCM applying the gain resulting the maximum peak level of the album to be at -0.01 dBFS. During the DSD2PCM conversion the conversion script that I use removes silent channel(s) (if they exist). I always run foobar DR meter on flacs containing no silent/muted channels.
 
Maybe HDMI to receiver compatibility?
I doubt it? Most classical SACDs are 5.0 (SACD allows layers to be 2.0, 5.0, or 5.1). Having no LFE means you're not telling the AVR to supply no sub, but instead allowing it to use bass management, as I understand it. Not a huge deal and shouldn't add much space to have a blank channel, but...
 
I doubt it? Most classical SACDs are 5.0 (SACD allows layers to be 2.0, 5.0, or 5.1). Having no LFE means you're not telling the AVR to supply no sub, but instead allowing it to use bass management, as I understand it. Not a huge deal and shouldn't add much space to have a blank channel, but...
I only use 5.1 analog from Oppo 205, my receiver has no HDMI.
 
Last edited:
I doubt it? Most classical SACDs are 5.0 (SACD allows layers to be 2.0, 5.0, or 5.1). Having no LFE means you're not telling the AVR to supply no sub, but instead allowing it to use bass management, as I understand it. Not a huge deal and shouldn't add much space to have a blank channel, but...
I cant think of any other reason beyond compatibility. Isn't bass management engaged regardless of if there is an LFE or not? maybe we are saying the same thing.
 
I convert an SACD album to PCM applying the gain resulting the maximum peak level of the album to be at -0.01 dBFS.
You'll almost certainly get true peaks above 0dBFS that way due to intersample peaks. If you have a true peak-aware normalizer, you could normalize to -0.01dBTP, but I wouldn't even recommend that. The new peak target for files uploaded to streaming services is -2dBTP, which is my new go-to. Absent true peak metering, I certainly wouldn't normalize above -1dBFS, because even that is fairly likely to result in intersample clipping.
 
Some time ago I also noticed this anomaly, when ripping my SACDs to my HDDs. So since then I take a simpler approach to it. I generally don't bother with FooDR on any Mch, and simply rely on FooDR for the stereo layer to tell me what DR to expect, knowing that the Mch layer will have a higher DR than the stereo. IMHO the DR meter was designed for stereo only and no thought was given to Mch.. However it remains a great tool, once one accounts for it's anomalies.
 
If you convert SACD to PCM the Music Media Helper’s (MMH) Channel Volume tool corrects the DR readings as per Dave’s findings.

It also resolves incorrect issues for all MCH files that have extremely low channel peaks and very high channel DR caused by spikes that give abnormally high DR readings (including some SACDs and streaming conversions).
 
This whole thread is so interesting, right up my alley.

Garry, I rip my SACD's with the Sonore software to a DSF file. Can MMH do what you say above with a DSF file?
I also assume that if I rip to a DSF file, MMH can convert to a FLAC file?
All my files as you know go to JRiver and as they import to JRiver I have the Audio Analyzer tool working and I get the DR's for each track.
But this whole thread makes me think just by nature of what Dave says Post #1, that all the SACD files are off?
Interesting for sure.
 
Back
Top