Sound of DSD vs. PCM? DeltaWave Comparison of La La Land Soundtrack

[GXAlan]

La La Land (2016) is a really fun movie that won a bunch of Academy Awards including Best Original Score and Best Original Song.

The jazz songs were recorded at Conway Studios which features nice Genelec monitors

and the final mixes were done with Neumann KH monitors

Inside Track: La La Land

Achieving a naturalistic sound on a Hollywood film set isn’t easy — but was central to director Damien Chazelle’s vision for La La Land.

www.soundonsound.com

Justin Hurwitz,the composer of La La Land also famously uses JBL 3 series speakers.

If you're interested in the soundtrack, you'll see that the "high-resolution" version on streaming is 24-bit/44.1 kHz. If you want physical media, the only high resolution option is a SACD released in 2023!

Spoiler: DeltaWaveLog

DeltaWave v2.0.13, 2024-09-14T15:27:02.0361363-07:00
Reference: 05 - Herman's Habit.dsf[L] 9848832 samples 88200Hz 24bits, mono, MD5=00
Comparison: 05. Justin Hurwitz - Hermans Habit.flac[L] 4923912 samples 44100Hz 24bits, stereo, MD5=00
Settings:
Gain:True, Remove DC:True
Non-linear Gain EQ:False Non-linear Phase EQ: False
EQ FFT Size:65536, EQ Frequency Cut: 0Hz - 0Hz, EQ Threshold: -500dB
Correct Non-linearity: False
Correct Drift:True, Precision:30, Subsample Align:True
Non-Linear drift Correction:False
Upsample:False, Window:Kaiser
Spectrum Window:Kaiser, Spectrum Size:32768
Spectrogram Window:Hann, Spectrogram Size:4096, Spectrogram Steps:2048
Filter Type:FIR, window:Kaiser, taps:262144, minimum phase=False
Dither:False bits=0
Trim Silence:True
Enable Simple Waveform Measurement: False

Resampled Reference to 44100Hz
Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -0.013dB Comparison: 0dB
Initial RMS values Reference: -14.929dB Comparison: -14.928dB

Null Depth=13.472dB
Trimming 273 samples at start and 1 samples at the end that are below -90.31dB level

X-Correlation offset: 10337 samples
Trimming 1 samples at start and 0 samples at the end that are below -90.31dB level

Drift computation quality, #1: Excellent (0μs)


Trimmed 423 samples ( 9.591837ms) front, 0 samples ( 0.00ms end)


Final peak values Reference: -0.013dB Comparison: 0.154dB
Final RMS values Reference: -14.919dB Comparison: -14.918dB

Gain= -0.0006dB (0.9999x) DC=0 Phase offset=234.393601ms (10336.758 samples)
Difference (rms) = -47.03dB [-45.85dBA]
Correlated Null Depth=66.72dB [63.92dBA]
Clock drift: 0 ppm


Files are NOT a bit-perfect match (match=1.75%) at 16 bits
Files are NOT a bit-perfect match (match=0.01%) at 24 bits
Files match @ 49.5625% when reduced to 8.75 bits


---- Phase difference (full bandwidth): 3.33510941694337°
0-10kHz: 4.92°
0-20kHz: 3.50°
0-24kHz: 3.34°
Timing error (rms jitter): 2.8μs
PK Metric (step=400ms, overlap=50%):
RMS=-45.8dBr
Median=-46.6
Max=-40.2

99%: -40.98
75%: -44.58
50%: -46.61
25%: -49.0
1%: -60.19

gn=1.00007013560789, dc=-1.46849237377185E-06, dr=0, of=10336.7578204906

DONE!

Signature: 1a7eb58e9fc8c2e9fa6768c9c4608e7c

RMS of the difference of spectra: -118.030933852178dB
DF Metric (step=400ms, overlap=0%):
Median=-37.5dB
Max=-24.1dB Min=-71.2dB

1% > -52.94dB
10% > -45.06dB
25% > -41.86dB
50% > -37.5dB
75% > -34.15dB
90% > -30.88dB
99% > -25.29dB

Linearity 17.7bits @ 0.5dB error

Let's compare the DSD version of track 5 against the 24/44.1 kHz PCM version. @pkane 's DeltaWave converts DSD to PCM using high precision mathematics, but it's up to the user to choose the resampling frequency and cut off frequency. I chose 88 kHz, 50 kHz cut off, and 5kHz bandwidth.

As expected, we get a very good fit

The integrated loudness matches overall

The aligned spectrum looks good but there is a difference well above the audible threshold but note that the SACD (blue) rolls off faster (?)

What I don't get is that the PK Metric is -45.8 dBr, which means that the two recordings should be different.

 

[Old_School_Brad]

So there is some differences in the content between 2 - 4 kHz?

And, by the looks of it, quite large phase variations? -Or what does that "Delta Phase" plot show?

 

[GXAlan]

So there is some differences in the content between 2 - 4 kHz?

And, by the looks of it, quite large phase variations? -Or what does that "Delta Phase" plot show?

That’s what I am seeing. I defer to @pkane if this is real or not.

If this is real, this is interesting. I bought the SACD just expecting this to be a physical media version of the 24-44.1 streaming.

But it’s not. And DeltaWave says the difference might be audible.

It’s not advertised as being remastered.

Scenarios
1) Bug in processing during production where different source files are used

2) Intentional change that’s a result of some sort of fancy seasoning to modify the sound, like DSEE does, but this is done during the SACD encode to make the disc sound different.

3) Result of DSD to PCM conversion algorithms and parameters. But if this is true and the PKMetric is accurately describing that the sound is different, it might suggest how different SACD players can affect the sound.

I personally haven’t had time to compare the two critically yet, but will do so in an unblinded fashion.

 

[bennetng]

Universal's Audible Watermark

This is my personal blog. The views expressed on this page are mine alone. Update, April 8, 2013: Can you hear it? Take the watermark listening test. Update, January 14, 2015: Check out the results of the listening test. A while ago, I wrote about my confusion regarding Weird Spotify Compressio

www.mattmontag.com

Oldfield watermarked at qobuz?

Oldfield watermarked at qobuz?

hydrogenaud.io

 

[somebodyelse]

Back when watermarking was proposed for DVD-Audio someone posted the fundamental problem as they saw it: to be effective at surviving conversion to lossy data formats the watermark must be audible, because lossy formats work by removing things that aren't audible. Putting an audible watermark on a format whose whole point was offering purportedly superior sound quality was going to make it a hard sell. Bob Stuart confirmed that this was indeed the core of the problem. It doesn't look like the situation has changed much.

 

[pkane]

That’s what I am seeing. I defer to @pkane if this is real or not.

If this is real, this is interesting. I bought the SACD just expecting this to be a physical media version of the 24-44.1 streaming.

But it’s not. And DeltaWave says the difference might be audible.

It’s not advertised as being remastered.

Scenarios
1) Bug in processing during production where different source files are used

2) Intentional change that’s a result of some sort of fancy seasoning to modify the sound, like DSEE does, but this is done during the SACD encode to make the disc sound different.

3) Result of DSD to PCM conversion algorithms and parameters. But if this is true and the PKMetric is accurately describing that the sound is different, it might suggest how different SACD players can affect the sound.

I personally haven’t had time to compare the two critically yet, but will do so in an unblinded fashion.

Seems real to me. If there were other errors across the spectrum, I’d investigate some mastering differences or some conversion/setting issues. Considering how well the spectra match except for a very small frequency range, this is intentional. As has been pointed out, possibly used for watermarking purposes.

 

[GXAlan]

@bennetng @pkane
The watermark is definitely going to be it! La La Land *is* a UMG release.

Hmm. I have always said “La La Land” soundtrack is an example of why I like tubes amps sometimes. I like the music composition but said the recordings were flawed. I even blamed it to the singers being actors/actresses rather than singers and on very transparent systems it didn’t sound as good as

This might be a reason why I like physical media more than streaming!

What I probably need to do is to rip the vinyl LP and see how the LP compares to the two digital sources and see what the CD layer looks like.

It also puts into context what -30 to -40 dBR PKMetric means for audibility.

 

[pkane]

@bennetng @pkane
The watermark is definitely going to be it! La La Land *is* a UMG release.

Hmm. I have always said “La La Land” soundtrack is an example of why I like tubes amps sometimes. I like the music composition but said the recordings were flawed. I even blamed it to the singers being actors/actresses rather than singers and on very transparent systems it didn’t sound as good as

This might be a reason why I like physical media more than streaming!

What I probably need to do is to rip the vinyl LP and see how the LP compares to the two digital sources and see what the CD layer looks like.

It also puts into context what -30 to -40 dBR PKMetric means for audibility.

PKMetric is weighted by equal loudness curves, so frequency errors/differences around 3kHz will be weighted a lot more heavily than similar errors at 100Hz or at 15kHz, as an example.