Analysis of MQA/non-MQA recordings through QRONO d2a

[GXAlan]

Content: Chesky Ultimate Demonstration Disc, original CD and MQA-CD versions; HDTracks version of The Raven
Source: Bluesound Node Icon
Recorder: E1DA Cosmos ADC, default digital filter, 24/88.1 kHz recording

1. Comparison of Source Files (CD vs. MQA-CD)

Spoiler: DeltaWave Log

DeltaWave v2.0.13, 2025-06-08T09:03:03.5702987-07:00
Reference: 03 - Rebecca Pidgeon - Spanish Harlem.flac[?] 9596160 samples 44100Hz 16bits, stereo, MD5=00
Comparison: 03 Spanish Harlem MQA.flac[?] 9507960 samples 44100Hz 16bits, 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

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -1.551dB Comparison: -1.562dB
Initial RMS values Reference: -21.648dB Comparison: -21.632dB

Null Depth=10.015dB
Trimming 18 samples at start and 88182 samples at the end that are below -90.31dB level

X-Correlation offset: -3749 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Drift computation quality, #1: Excellent (1.91μs)


Trimmed 1811 samples ( 41.06576ms) front, 220 samples ( 4.988662ms end)


Final peak values Reference: -1.551dB Comparison: -1.542dB
Final RMS values Reference: -21.604dB Comparison: -21.61dB

Gain= -0.0186dB (0.9979x) DC=0 Phase offset=-85.017517ms (-3749.273 samples)
Difference (rms) = -50.32dB [-53.87dBA]
Correlated Null Depth=59.91dB [53.44dBA]
Clock drift: 0.02 ppm


Files are NOT a bit-perfect match (match=0.75%) at 16 bits
Files match @ 50.0012% when reduced to 9.37 bits


---- Phase difference (full bandwidth): 62.9998727416267°
0-10kHz: 6.74°
0-20kHz: 58.43°
0-24kHz: 63.00°
Timing error (rms jitter): 14.3μs
PK Metric (step=400ms, overlap=50%):
RMS=-78.8dBr
Median=-79.2
Max=-73.5

99%: -74.29
75%: -77.13
50%: -79.2
25%: -82.13
1%: -103.15

gn=1.00214944387847, dc=-9.3989119362255E-07, dr=2.41113738775602E-08, of=-3749.27250302069

DONE!

Signature: ddc23dc6a809895a2841e2bdd5a7fbe3

RMS of the difference of spectra: -95.5622352690719dB
DF Metric (step=400ms, overlap=0%):
Median=-35.4dB
Max=-14.1dB Min=-51.4dB

1% > -49.71dB
10% > -42.64dB
25% > -39.23dB
50% > -35.39dB
75% > -31.85dB
90% > -27.63dB
99% > -18.55dB

Linearity 18.9bits @ 0.5dB error

2. Comparisons of Analog Recordings (24/88) from Bluesound Node Icon (CD vs MQA-CD)

Spoiler: DeltaWave Log

DeltaWave v2.0.13, 2025-06-08T09:42:59.9203404-07:00
Reference: Standard.flac[?] 19274766 samples 88200Hz 24bits, stereo, MD5=00
Comparison: MQA.flac[?] 19076096 samples 88200Hz 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

Discarding Reference: Start=1.7s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.685dB Comparison: -4.688dB
Initial RMS values Reference: -24.859dB Comparison: -24.81dB

Null Depth=12.912dB
Trimming 16786 samples at start and 1710 samples at the end that are below -90.31dB level

X-Correlation offset: 20219 samples
Trimming 175 samples at start and 0 samples at the end that are below -90.31dB level

Residual error too large 1.8180377. Trying alternative drift correction method

Drift computation quality #2: Poor (16.18μs)

Drift found with alternative method: 0.2275ppm


Trimmed 26 samples ( 0.294785ms) front, 4 samples ( 0.045351ms end)


Final peak values Reference: -4.685dB Comparison: -4.692dB
Final RMS values Reference: -24.806dB Comparison: -24.807dB

Gain= 0.0044dB (1.0005x) DC=0 Phase offset=229.240363ms (20219 samples)
Difference (rms) = -32.31dB [-32.79dBA]
Correlated Null Depth=40.33dB [31.61dBA]
Clock drift: 0.23 ppm


Files are NOT a bit-perfect match (match=0.24%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 49.9859% when reduced to 6 bits


---- Phase difference (full bandwidth): 153.693888137683°
0-10kHz: 99.58°
0-20kHz: 135.19°
0-24kHz: 141.09°
Timing error (rms jitter): 104.3μs
PK Metric (step=400ms, overlap=50%):
RMS=-83.3dBr
Median=-83.8
Max=-78.3

99%: -79.08
75%: -81.78
50%: -83.76
25%: -86.68
1%: -103.82

gn=0.999492369120056, dc=0, dr=2.27475495985345E-07, of=20218.9999990423

DONE!

Signature: e2f8ac7cabcd9dd58678c123093d291b

RMS of the difference of spectra: -110.79045175985dB
DF Metric (step=400ms, overlap=0%):
Median=-13.4dB
Max=-1.1dB Min=-43.9dB

1% > -39.88dB
10% > -27.15dB
25% > -19.55dB
50% > -13.38dB
75% > -9.41dB
90% > -6.73dB
99% > -2.15dB

Linearity 2.1bits @ 0.5dB error

3. CD Recording (QRONO d2A) vs. CD source

Spoiler: Deltawave Log

DeltaWave v2.0.13, 2025-06-08T10:05:32.4624128-07:00
Reference: Standard.flac[?] 19424706 samples 88200Hz 24bits, stereo, MD5=00
Comparison: 03 - Rebecca Pidgeon - Spanish Harlem.flac[?] 9596160 samples 44100Hz 16bits, 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
Resampled Reference to 44100Hz
Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.693dB Comparison: -1.551dB
Initial RMS values Reference: -24.892dB Comparison: -21.648dB

Null Depth=17.469dB
Trimming 18 samples at start and 88182 samples at the end that are below -90.31dB level

Cross-correlation found periodic/simple waveform! Trying simple measurement option!
X-Correlation offset: 99268 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Drift computation quality, #1: Good (9.71μs)


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


Final peak values Reference: -4.693dB Comparison: -4.764dB
Final RMS values Reference: -24.76dB Comparison: -24.775dB

Gain= 3.2065dB (1.4465x) DC=0.00088 Phase offset=2250.176359ms (99232.777 samples)
Difference (rms) = -49.36dB [-50.99dBA]
Correlated Null Depth=47.04dB [52.57dBA]
Clock drift: 8.28 ppm


Files are NOT a bit-perfect match (match=0.31%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 49.9995% when reduced to 8.27 bits


---- Phase difference (full bandwidth): 21.4587302818233°
0-10kHz: 15.97°
0-20kHz: 21.06°
0-24kHz: 21.46°
Timing error (rms jitter): 6.7μs
PK Metric (step=400ms, overlap=50%):
RMS=-68.7dBr
Median=-70.9
Max=-61.0

99%: -61.96
75%: -67.06
50%: -70.86
25%: -75.97
1%: -103.51

gn=0.691313630101968, dc=0.000881238148451878, dr=8.27560397260192E-06, of=99232.7774358563

DONE!

Signature: d55d863197aa751f9bb7458fe9513e38

RMS of the difference of spectra: -96.3994677253658dB
DF Metric (step=400ms, overlap=0%):
Median=-29.2dB
Max=-15.9dB Min=-41.6dB

1% > -38.14dB
10% > -34.08dB
25% > -31.85dB
50% > -29.19dB
75% > -25.66dB
90% > -22.29dB
99% > -17.23dB

Linearity 17.9bits @ 0.5dB error

4. MQA-recording vs Undecoded MQA-CD source

Spoiler: Deltawave Log

DeltaWave v2.0.13, 2025-06-08T10:17:58.2878835-07:00
Reference: MQA.flac[?] 19076096 samples 88200Hz 24bits, stereo, MD5=00
Comparison: 03 Spanish Harlem MQA.flac[?] 9507960 samples 44100Hz 16bits, 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
Resampled Reference to 44100Hz
Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.69dB Comparison: -1.562dB
Initial RMS values Reference: -24.81dB Comparison: -21.632dB

Null Depth=13.8dB
Trimming 8415 samples at start and 0 samples at the end that are below -90.31dB level

X-Correlation offset: 9185 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Drift computation quality, #1: Very Good (2.3μs)


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


Final peak values Reference: -4.69dB Comparison: -4.769dB
Final RMS values Reference: -24.796dB Comparison: -24.83dB

Gain= 3.1966dB (1.4449x) DC=0.00088 Phase offset=208.215765ms (9182.315 samples)
Difference (rms) = -45.93dB [-47.81dBA]
Correlated Null Depth=50.88dB [53.66dBA]
Clock drift: 7.47 ppm


Files are NOT a bit-perfect match (match=0.4%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 50.0081% when reduced to 8.68 bits


---- Phase difference (full bandwidth): 30.1368730165672°
0-10kHz: 29.42°
0-20kHz: 30.82°
0-24kHz: 30.14°
Timing error (rms jitter): 121μs
PK Metric (step=400ms, overlap=50%):
RMS=-65.1dBr
Median=-65.9
Max=-58.6

99%: -59.82
75%: -63.25
50%: -65.85
25%: -69.61
1%: -102.49

gn=0.692099737691919, dc=0.000881677640327197, dr=7.46879662431917E-06, of=9182.31521591103

DONE!

Signature: 4015d079385e641ef7709be9ffe808cf

RMS of the difference of spectra: -94.7698049167433dB
DF Metric (step=400ms, overlap=0%):
Median=-32.8dB
Max=-5.8dB Min=-44.4dB

1% > -39.64dB
10% > -37.17dB
25% > -35.74dB
50% > -32.84dB
75% > -27.86dB
90% > -20.85dB
99% > -7.51dB

Linearity 16.6bits @ 0.5dB error

5. CD recording (QRONO d2a) vs. HDTracks Source File

Spoiler: Deltawave Log

DeltaWave v2.0.13, 2025-06-08T10:59:02.7483052-07:00
Reference: Standard.flac[?] 19424706 samples 88200Hz 24bits, stereo, MD5=00
Comparison: 12-Spanish Harlem HDTRACKS.flac[?] 19048448 samples 88200Hz 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

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.685dB Comparison: -0.977dB
Initial RMS values Reference: -24.892dB Comparison: -20.536dB

Null Depth=11.256dB
Trimming 0 samples at start and 98 samples at the end that are below -90.31dB level

Cross-correlation found periodic/simple waveform! Trying simple measurement option!
Phase inverted
X-Correlation offset: 215980 samples
Trimming 2 samples at start and 0 samples at the end that are below -90.31dB level

Residual error too large 505.635348156. Trying alternative drift correction method

Drift computation quality #2: Poor (16.58μs)

Drift found with alternative method: 0.0723ppm


Trimmed 244 samples ( 2.76644ms) front, 2 samples ( 0.022676ms end)


Final peak values Reference: -4.685dB Comparison: -5.316dB
Final RMS values Reference: -24.807dB Comparison: -24.877dB

Gain= 4.341dB (1.6484x) DC=0 Phase offset=2448.752834ms (215980 samples)
Difference (rms) = -22.04dB [-27.51dBA]
Correlated Null Depth=15.43dB [15.97dBA]
Clock drift: 0.07 ppm


Files are NOT a bit-perfect match (match=0.02%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 50.0012% when reduced to 3.81 bits


---- Phase difference (full bandwidth): 118.412949833603°
0-10kHz: 100.50°
0-20kHz: 110.47°
0-24kHz: 112.95°
Timing error (rms jitter): 94μs
PK Metric (step=400ms, overlap=50%):
RMS=-37.7dBr
Median=-38.4
Max=-30.7

99%: -32.46
75%: -36.15
50%: -38.37
25%: -41.53
1%: -59.19

gn=0.60666373053586, dc=3.5485857973899E-06, dr=7.23061591008739E-08, of=215979.999999899

DONE!

Signature: 5699a2b955922b63f7f2895c152c83da

RMS of the difference of spectra: -92.0275867227056dB
DF Metric (step=400ms, overlap=0%):
Median=0.1dB
Max=2.8dB Min=-17.1dB

1% > -9.52dB
10% > -3.64dB
25% > -1.3dB
50% > 0.05dB
75% > 0.95dB
90% > 1.75dB
99% > 2.71dB

Linearity 2.1bits @ 0.5dB error

6. MQA-CD Recording vs. HDTracks Source File

Spoiler: Deltawave

DeltaWave v2.0.13, 2025-06-08T10:29:49.3491515-07:00
Reference: MQA.flac[?] 19076096 samples 88200Hz 24bits, stereo, MD5=00
Comparison: 12-Spanish Harlem HDTRACKS.flac[?] 19048448 samples 88200Hz 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

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.688dB Comparison: -0.977dB
Initial RMS values Reference: -24.81dB Comparison: -20.536dB

Null Depth=11.281dB
Trimming 16786 samples at start and 0 samples at the end that are below -90.31dB level

Cross-correlation found periodic/simple waveform! Trying simple measurement option!
Phase inverted
X-Correlation offset: 43422 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Residual error too large 504.499366042. Trying alternative drift correction method

Drift computation quality #2: Poor (16.93μs)

Drift found with alternative method: 0.3645ppm


Trimmed 240 samples ( 2.721088ms) front, 2 samples ( 0.022676ms end)


Final peak values Reference: -4.688dB Comparison: -5.319dB
Final RMS values Reference: -24.796dB Comparison: -24.866dB

Gain= 4.337dB (1.6476x) DC=0 Phase offset=492.312925ms (43422 samples)
Difference (rms) = -22.01dB [-27.47dBA]
Correlated Null Depth=15.82dB [18.33dBA]
Clock drift: 0.36 ppm


Files are NOT a bit-perfect match (match=0.02%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 49.9961% when reduced to 3.8 bits


---- Phase difference (full bandwidth): 61.409629065583°
0-10kHz: 102.62°
0-20kHz: 83.01°
0-24kHz: 77.06°
Timing error (rms jitter): 592.4μs
PK Metric (step=400ms, overlap=50%):
RMS=-37.8dBr
Median=-38.6
Max=-30.1

99%: -32.6
75%: -36.28
50%: -38.59
25%: -41.8
1%: -58.78

gn=0.60694466088091, dc=3.55795575201226E-06, dr=3.64503765100017E-07, of=43421.9999975449

DONE!

Signature: d48f69a7046b8bd58f0f103096f983c2

RMS of the difference of spectra: -92.0078367600795dB
DF Metric (step=400ms, overlap=0%):
Median=0.1dB
Max=2.8dB Min=-17.4dB

1% > -9.39dB
10% > -3.6dB
25% > -1.15dB
50% > 0.07dB
75% > 0.94dB
90% > 1.76dB
99% > 2.71dB

Linearity 2.1bits @ 0.5dB error

7. CD Source File vs. HD Tracks Source File

Spoiler: Deltawave Log

DeltaWave v2.0.13, 2025-06-08T11:13:12.8084727-07:00
Reference: 03 - Rebecca Pidgeon - Spanish Harlem.flac[?] 9596160 samples 44100Hz 16bits, stereo, MD5=00
Comparison: 12-Spanish Harlem HDTRACKS.flac[?] 19048448 samples 88200Hz 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 Comparison to 44100Hz
Resampled Comparison to 44100Hz
Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -1.551dB Comparison: -0.98dB
Initial RMS values Reference: -21.648dB Comparison: -20.537dB

Null Depth=13.022dB
Trimming 18 samples at start and 88182 samples at the end that are below -90.31dB level

Cross-correlation found periodic/simple waveform! Trying simple measurement option!
Phase inverted
X-Correlation offset: 8519 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Residual error too large 7424.053632669. Trying alternative drift correction method

Drift computation quality #2: Very Poor (23.01μs)


Trimmed 912 samples ( 20.680272ms) front, 208 samples ( 4.716553ms end)


Final peak values Reference: -1.551dB Comparison: -1.972dB
Final RMS values Reference: -21.603dB Comparison: -21.461dB

Gain= 0.9361dB (1.1138x) DC=0 Phase offset=193.174603ms (8519 samples)
Difference (rms) = -18.62dB [-24.05dBA]
Correlated Null Depth=16.24dB [18.93dBA]
Clock drift: -0.01 ppm


Files are NOT a bit-perfect match (match=0.01%) at 16 bits
Files match @ 49.9538% when reduced to 3.33 bits


---- Phase difference (full bandwidth): 99.3147749632823°
0-10kHz: 104.02°
0-20kHz: 100.21°
0-24kHz: 99.31°
Timing error (rms jitter): 188.9μs
PK Metric (step=400ms, overlap=50%):
RMS=-34.9dBr
Median=-35.2
Max=-29.0

99%: -30.31
75%: -33.28
50%: -35.23
25%: -38.48
1%: -55.35

gn=0.897830251253537, dc=1.89824106927886E-06, dr=-1.23588245464565E-08, of=8519.00000000401

DONE!

Signature: 8a98a279b7ef50ed1b9c54419b27b3da

RMS of the difference of spectra: -82.5840441673899dB
DF Metric (step=400ms, overlap=0%):
Median=-0.1dB
Max=2.9dB Min=-14.6dB

1% > -13.16dB
10% > -4.1dB
25% > -1.47dB
50% > -0.13dB
75% > 0.79dB
90% > 1.81dB
99% > 2.63dB

Linearity 2.1bits @ 0.5dB error

Discussion
This is a bit like a Rorschach test. You can look at the same set of data and come to different and completely opposite conclusions, depending on your bias/perspective. A glass of water can scientifically be identical when described as half-full or half-empty, and yet we all appreciate that the two descriptions impart differences.

That said, at least this is how I think you should consider the data for interpretation.
Comparison 1: The Raw Files
This is an unfolded MQA-CD vs regular CD. We know that you lose 1-3 bits of data, theoretically, with MQA encoding. There are also scenarios where different formats have fundamentally different recordings/mixes. The La La Land SACD vs. streaming high-res comparisons is a good example.

Comparison 2: The Recorded Output from each CD
If you own the Bluesound Node Icon, QRONO d2A is being applied to all content, whether MQA-enabled or not. This shows you if QRONO d2A + regular CD is different from a full MQA chain.

Comparison 3: Recording of the CD vs. Digital CD File
This should capture the consequences of QRONO d2a on regular content. This is the impact of Bluesound Node Icon "forcing" QRONO d2a to be enabled all the time.

Comparison 4: Recording of the MQA vs. Unfolded MQA-CD file
To my knowledge, there is no MQA-CD exclusive content -- there are some MQA-exclusive streaming. Suppose there was MQA-exclusive content which means you can play it back on a normal system vs. an MQA system. This comparison may predict the differences.

Comparison 5: CD Recording vs. High Res Source
The high-res source comes from a different album. The mastering may be different.

Comparison 6: MQA-CD Recording vs. High Res Source
The high-res source comes from a different album. The mastering may be different. This is the only way to compare an unfolded MQA CD against a High-Res recording.

Comparison 7: CD Source File vs. High Res Source File
This is the most formal comparison of the mastering between the two files. You can come to your own conclusion, but I would argue that this shows that Comparisons 5 and 6 are of limited use. We don't have a good way to compare the MQA decoded CD chain against the same master if it was distributed high-res.

 

[pjug]

Risking being Captain Obvious, I will comment:

For #1, The CD and MQACD; Iin your comments you said unfolded but I think you meant folded, or not unfolded, since these were loaded directly into DW right? These match well enough that they would seem to come from the same master. The MQA probably has a minimum phase filter applied. Maybe we could attempt to copy the MQA filter to do the same thing. I thnk Paul knows how to undo filters, maybe he could make an MQA undoer for this particular MQACD.

For the 88 KHz recording of the unfolded (fully decoded) MQACD, it would be interesting to look at the spectrum and see if it has real ultrasonic content, or just an image of the audible. Some MQA have been found to have no real ultrasonic content at all.

 

[pkane]

Content: Chesky Ultimate Demonstration Disc, original CD and MQA-CD versions; HDTracks version of The Raven
Source: Bluesound Node Icon
Recorder: E1DA Cosmos ADC, default digital filter, 24/88.1 kHz recording

1. Comparison of Source Files (CD vs. MQA-CD)

Spoiler: DeltaWave Log

DeltaWave v2.0.13, 2025-06-08T09:03:03.5702987-07:00
Reference: 03 - Rebecca Pidgeon - Spanish Harlem.flac[?] 9596160 samples 44100Hz 16bits, stereo, MD5=00
Comparison: 03 Spanish Harlem MQA.flac[?] 9507960 samples 44100Hz 16bits, 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

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -1.551dB Comparison: -1.562dB
Initial RMS values Reference: -21.648dB Comparison: -21.632dB

Null Depth=10.015dB
Trimming 18 samples at start and 88182 samples at the end that are below -90.31dB level

X-Correlation offset: -3749 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Drift computation quality, #1: Excellent (1.91μs)


Trimmed 1811 samples ( 41.06576ms) front, 220 samples ( 4.988662ms end)


Final peak values Reference: -1.551dB Comparison: -1.542dB
Final RMS values Reference: -21.604dB Comparison: -21.61dB

Gain= -0.0186dB (0.9979x) DC=0 Phase offset=-85.017517ms (-3749.273 samples)
Difference (rms) = -50.32dB [-53.87dBA]
Correlated Null Depth=59.91dB [53.44dBA]
Clock drift: 0.02 ppm


Files are NOT a bit-perfect match (match=0.75%) at 16 bits
Files match @ 50.0012% when reduced to 9.37 bits


---- Phase difference (full bandwidth): 62.9998727416267°
0-10kHz: 6.74°
0-20kHz: 58.43°
0-24kHz: 63.00°
Timing error (rms jitter): 14.3μs
PK Metric (step=400ms, overlap=50%):
RMS=-78.8dBr
Median=-79.2
Max=-73.5

99%: -74.29
75%: -77.13
50%: -79.2
25%: -82.13
1%: -103.15

gn=1.00214944387847, dc=-9.3989119362255E-07, dr=2.41113738775602E-08, of=-3749.27250302069

DONE!

Signature: ddc23dc6a809895a2841e2bdd5a7fbe3

RMS of the difference of spectra: -95.5622352690719dB
DF Metric (step=400ms, overlap=0%):
Median=-35.4dB
Max=-14.1dB Min=-51.4dB

1% > -49.71dB
10% > -42.64dB
25% > -39.23dB
50% > -35.39dB
75% > -31.85dB
90% > -27.63dB
99% > -18.55dB

Linearity 18.9bits @ 0.5dB error

View attachment 456443
View attachment 456444
View attachment 456445

2. Comparisons of Analog Recordings (24/88) from Bluesound Node Icon (CD vs MQA-CD)

Spoiler: DeltaWave Log

DeltaWave v2.0.13, 2025-06-08T09:42:59.9203404-07:00
Reference: Standard.flac[?] 19274766 samples 88200Hz 24bits, stereo, MD5=00
Comparison: MQA.flac[?] 19076096 samples 88200Hz 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

Discarding Reference: Start=1.7s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.685dB Comparison: -4.688dB
Initial RMS values Reference: -24.859dB Comparison: -24.81dB

Null Depth=12.912dB
Trimming 16786 samples at start and 1710 samples at the end that are below -90.31dB level

X-Correlation offset: 20219 samples
Trimming 175 samples at start and 0 samples at the end that are below -90.31dB level

Residual error too large 1.8180377. Trying alternative drift correction method

Drift computation quality #2: Poor (16.18μs)

Drift found with alternative method: 0.2275ppm


Trimmed 26 samples ( 0.294785ms) front, 4 samples ( 0.045351ms end)


Final peak values Reference: -4.685dB Comparison: -4.692dB
Final RMS values Reference: -24.806dB Comparison: -24.807dB

Gain= 0.0044dB (1.0005x) DC=0 Phase offset=229.240363ms (20219 samples)
Difference (rms) = -32.31dB [-32.79dBA]
Correlated Null Depth=40.33dB [31.61dBA]
Clock drift: 0.23 ppm


Files are NOT a bit-perfect match (match=0.24%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 49.9859% when reduced to 6 bits


---- Phase difference (full bandwidth): 153.693888137683°
0-10kHz: 99.58°
0-20kHz: 135.19°
0-24kHz: 141.09°
Timing error (rms jitter): 104.3μs
PK Metric (step=400ms, overlap=50%):
RMS=-83.3dBr
Median=-83.8
Max=-78.3

99%: -79.08
75%: -81.78
50%: -83.76
25%: -86.68
1%: -103.82

gn=0.999492369120056, dc=0, dr=2.27475495985345E-07, of=20218.9999990423

DONE!

Signature: e2f8ac7cabcd9dd58678c123093d291b

RMS of the difference of spectra: -110.79045175985dB
DF Metric (step=400ms, overlap=0%):
Median=-13.4dB
Max=-1.1dB Min=-43.9dB

1% > -39.88dB
10% > -27.15dB
25% > -19.55dB
50% > -13.38dB
75% > -9.41dB
90% > -6.73dB
99% > -2.15dB

Linearity 2.1bits @ 0.5dB error

View attachment 456456
View attachment 456457
View attachment 456458

3. CD Recording (QRONO d2A) vs. CD source

Spoiler: Deltawave Log

DeltaWave v2.0.13, 2025-06-08T10:05:32.4624128-07:00
Reference: Standard.flac[?] 19424706 samples 88200Hz 24bits, stereo, MD5=00
Comparison: 03 - Rebecca Pidgeon - Spanish Harlem.flac[?] 9596160 samples 44100Hz 16bits, 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
Resampled Reference to 44100Hz
Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.693dB Comparison: -1.551dB
Initial RMS values Reference: -24.892dB Comparison: -21.648dB

Null Depth=17.469dB
Trimming 18 samples at start and 88182 samples at the end that are below -90.31dB level

Cross-correlation found periodic/simple waveform! Trying simple measurement option!
X-Correlation offset: 99268 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Drift computation quality, #1: Good (9.71μs)


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


Final peak values Reference: -4.693dB Comparison: -4.764dB
Final RMS values Reference: -24.76dB Comparison: -24.775dB

Gain= 3.2065dB (1.4465x) DC=0.00088 Phase offset=2250.176359ms (99232.777 samples)
Difference (rms) = -49.36dB [-50.99dBA]
Correlated Null Depth=47.04dB [52.57dBA]
Clock drift: 8.28 ppm


Files are NOT a bit-perfect match (match=0.31%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 49.9995% when reduced to 8.27 bits


---- Phase difference (full bandwidth): 21.4587302818233°
0-10kHz: 15.97°
0-20kHz: 21.06°
0-24kHz: 21.46°
Timing error (rms jitter): 6.7μs
PK Metric (step=400ms, overlap=50%):
RMS=-68.7dBr
Median=-70.9
Max=-61.0

99%: -61.96
75%: -67.06
50%: -70.86
25%: -75.97
1%: -103.51

gn=0.691313630101968, dc=0.000881238148451878, dr=8.27560397260192E-06, of=99232.7774358563

DONE!

Signature: d55d863197aa751f9bb7458fe9513e38

RMS of the difference of spectra: -96.3994677253658dB
DF Metric (step=400ms, overlap=0%):
Median=-29.2dB
Max=-15.9dB Min=-41.6dB

1% > -38.14dB
10% > -34.08dB
25% > -31.85dB
50% > -29.19dB
75% > -25.66dB
90% > -22.29dB
99% > -17.23dB

Linearity 17.9bits @ 0.5dB error

View attachment 456462
View attachment 456463
View attachment 456464

4. MQA-recording vs Unfolded MQA-CD source

Spoiler: Deltawave Log

DeltaWave v2.0.13, 2025-06-08T10:17:58.2878835-07:00
Reference: MQA.flac[?] 19076096 samples 88200Hz 24bits, stereo, MD5=00
Comparison: 03 Spanish Harlem MQA.flac[?] 9507960 samples 44100Hz 16bits, 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
Resampled Reference to 44100Hz
Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.69dB Comparison: -1.562dB
Initial RMS values Reference: -24.81dB Comparison: -21.632dB

Null Depth=13.8dB
Trimming 8415 samples at start and 0 samples at the end that are below -90.31dB level

X-Correlation offset: 9185 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Drift computation quality, #1: Very Good (2.3μs)


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


Final peak values Reference: -4.69dB Comparison: -4.769dB
Final RMS values Reference: -24.796dB Comparison: -24.83dB

Gain= 3.1966dB (1.4449x) DC=0.00088 Phase offset=208.215765ms (9182.315 samples)
Difference (rms) = -45.93dB [-47.81dBA]
Correlated Null Depth=50.88dB [53.66dBA]
Clock drift: 7.47 ppm


Files are NOT a bit-perfect match (match=0.4%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 50.0081% when reduced to 8.68 bits


---- Phase difference (full bandwidth): 30.1368730165672°
0-10kHz: 29.42°
0-20kHz: 30.82°
0-24kHz: 30.14°
Timing error (rms jitter): 121μs
PK Metric (step=400ms, overlap=50%):
RMS=-65.1dBr
Median=-65.9
Max=-58.6

99%: -59.82
75%: -63.25
50%: -65.85
25%: -69.61
1%: -102.49

gn=0.692099737691919, dc=0.000881677640327197, dr=7.46879662431917E-06, of=9182.31521591103

DONE!

Signature: 4015d079385e641ef7709be9ffe808cf

RMS of the difference of spectra: -94.7698049167433dB
DF Metric (step=400ms, overlap=0%):
Median=-32.8dB
Max=-5.8dB Min=-44.4dB

1% > -39.64dB
10% > -37.17dB
25% > -35.74dB
50% > -32.84dB
75% > -27.86dB
90% > -20.85dB
99% > -7.51dB

Linearity 16.6bits @ 0.5dB error

View attachment 456468
View attachment 456470
View attachment 456471

5. CD recording (QRONO d2a) vs. HDTracks Source File

Spoiler: Deltawave Log

DeltaWave v2.0.13, 2025-06-08T10:59:02.7483052-07:00
Reference: Standard.flac[?] 19424706 samples 88200Hz 24bits, stereo, MD5=00
Comparison: 12-Spanish Harlem HDTRACKS.flac[?] 19048448 samples 88200Hz 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

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.685dB Comparison: -0.977dB
Initial RMS values Reference: -24.892dB Comparison: -20.536dB

Null Depth=11.256dB
Trimming 0 samples at start and 98 samples at the end that are below -90.31dB level

Cross-correlation found periodic/simple waveform! Trying simple measurement option!
Phase inverted
X-Correlation offset: 215980 samples
Trimming 2 samples at start and 0 samples at the end that are below -90.31dB level

Residual error too large 505.635348156. Trying alternative drift correction method

Drift computation quality #2: Poor (16.58μs)

Drift found with alternative method: 0.0723ppm


Trimmed 244 samples ( 2.76644ms) front, 2 samples ( 0.022676ms end)


Final peak values Reference: -4.685dB Comparison: -5.316dB
Final RMS values Reference: -24.807dB Comparison: -24.877dB

Gain= 4.341dB (1.6484x) DC=0 Phase offset=2448.752834ms (215980 samples)
Difference (rms) = -22.04dB [-27.51dBA]
Correlated Null Depth=15.43dB [15.97dBA]
Clock drift: 0.07 ppm


Files are NOT a bit-perfect match (match=0.02%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 50.0012% when reduced to 3.81 bits


---- Phase difference (full bandwidth): 118.412949833603°
0-10kHz: 100.50°
0-20kHz: 110.47°
0-24kHz: 112.95°
Timing error (rms jitter): 94μs
PK Metric (step=400ms, overlap=50%):
RMS=-37.7dBr
Median=-38.4
Max=-30.7

99%: -32.46
75%: -36.15
50%: -38.37
25%: -41.53
1%: -59.19

gn=0.60666373053586, dc=3.5485857973899E-06, dr=7.23061591008739E-08, of=215979.999999899

DONE!

Signature: 5699a2b955922b63f7f2895c152c83da

RMS of the difference of spectra: -92.0275867227056dB
DF Metric (step=400ms, overlap=0%):
Median=0.1dB
Max=2.8dB Min=-17.1dB

1% > -9.52dB
10% > -3.64dB
25% > -1.3dB
50% > 0.05dB
75% > 0.95dB
90% > 1.75dB
99% > 2.71dB

Linearity 2.1bits @ 0.5dB error

View attachment 456476
View attachment 456477
View attachment 456478

6. MQA-CD Recording vs. HDTracks Source File

Spoiler: Deltawave

DeltaWave v2.0.13, 2025-06-08T10:29:49.3491515-07:00
Reference: MQA.flac[?] 19076096 samples 88200Hz 24bits, stereo, MD5=00
Comparison: 12-Spanish Harlem HDTRACKS.flac[?] 19048448 samples 88200Hz 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

Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -4.688dB Comparison: -0.977dB
Initial RMS values Reference: -24.81dB Comparison: -20.536dB

Null Depth=11.281dB
Trimming 16786 samples at start and 0 samples at the end that are below -90.31dB level

Cross-correlation found periodic/simple waveform! Trying simple measurement option!
Phase inverted
X-Correlation offset: 43422 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Residual error too large 504.499366042. Trying alternative drift correction method

Drift computation quality #2: Poor (16.93μs)

Drift found with alternative method: 0.3645ppm


Trimmed 240 samples ( 2.721088ms) front, 2 samples ( 0.022676ms end)


Final peak values Reference: -4.688dB Comparison: -5.319dB
Final RMS values Reference: -24.796dB Comparison: -24.866dB

Gain= 4.337dB (1.6476x) DC=0 Phase offset=492.312925ms (43422 samples)
Difference (rms) = -22.01dB [-27.47dBA]
Correlated Null Depth=15.82dB [18.33dBA]
Clock drift: 0.36 ppm


Files are NOT a bit-perfect match (match=0.02%) at 16 bits
Files are NOT a bit-perfect match (match=0%) at 24 bits
Files match @ 49.9961% when reduced to 3.8 bits


---- Phase difference (full bandwidth): 61.409629065583°
0-10kHz: 102.62°
0-20kHz: 83.01°
0-24kHz: 77.06°
Timing error (rms jitter): 592.4μs
PK Metric (step=400ms, overlap=50%):
RMS=-37.8dBr
Median=-38.6
Max=-30.1

99%: -32.6
75%: -36.28
50%: -38.59
25%: -41.8
1%: -58.78

gn=0.60694466088091, dc=3.55795575201226E-06, dr=3.64503765100017E-07, of=43421.9999975449

DONE!

Signature: d48f69a7046b8bd58f0f103096f983c2

RMS of the difference of spectra: -92.0078367600795dB
DF Metric (step=400ms, overlap=0%):
Median=0.1dB
Max=2.8dB Min=-17.4dB

1% > -9.39dB
10% > -3.6dB
25% > -1.15dB
50% > 0.07dB
75% > 0.94dB
90% > 1.76dB
99% > 2.71dB

Linearity 2.1bits @ 0.5dB error

View attachment 456473
View attachment 456474
View attachment 456475

7. CD Source File vs. HD Tracks Source File

Spoiler: Deltawave Log

DeltaWave v2.0.13, 2025-06-08T11:13:12.8084727-07:00
Reference: 03 - Rebecca Pidgeon - Spanish Harlem.flac[?] 9596160 samples 44100Hz 16bits, stereo, MD5=00
Comparison: 12-Spanish Harlem HDTRACKS.flac[?] 19048448 samples 88200Hz 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 Comparison to 44100Hz
Resampled Comparison to 44100Hz
Discarding Reference: Start=0s, End=0s
Discarding Comparison: Start=0s, End=0s

Initial peak values Reference: -1.551dB Comparison: -0.98dB
Initial RMS values Reference: -21.648dB Comparison: -20.537dB

Null Depth=13.022dB
Trimming 18 samples at start and 88182 samples at the end that are below -90.31dB level

Cross-correlation found periodic/simple waveform! Trying simple measurement option!
Phase inverted
X-Correlation offset: 8519 samples
Trimming 0 samples at start and 0 samples at the end that are below -90.31dB level

Residual error too large 7424.053632669. Trying alternative drift correction method

Drift computation quality #2: Very Poor (23.01μs)


Trimmed 912 samples ( 20.680272ms) front, 208 samples ( 4.716553ms end)


Final peak values Reference: -1.551dB Comparison: -1.972dB
Final RMS values Reference: -21.603dB Comparison: -21.461dB

Gain= 0.9361dB (1.1138x) DC=0 Phase offset=193.174603ms (8519 samples)
Difference (rms) = -18.62dB [-24.05dBA]
Correlated Null Depth=16.24dB [18.93dBA]
Clock drift: -0.01 ppm


Files are NOT a bit-perfect match (match=0.01%) at 16 bits
Files match @ 49.9538% when reduced to 3.33 bits


---- Phase difference (full bandwidth): 99.3147749632823°
0-10kHz: 104.02°
0-20kHz: 100.21°
0-24kHz: 99.31°
Timing error (rms jitter): 188.9μs
PK Metric (step=400ms, overlap=50%):
RMS=-34.9dBr
Median=-35.2
Max=-29.0

99%: -30.31
75%: -33.28
50%: -35.23
25%: -38.48
1%: -55.35

gn=0.897830251253537, dc=1.89824106927886E-06, dr=-1.23588245464565E-08, of=8519.00000000401

DONE!

Signature: 8a98a279b7ef50ed1b9c54419b27b3da

RMS of the difference of spectra: -82.5840441673899dB
DF Metric (step=400ms, overlap=0%):
Median=-0.1dB
Max=2.9dB Min=-14.6dB

1% > -13.16dB
10% > -4.1dB
25% > -1.47dB
50% > -0.13dB
75% > 0.79dB
90% > 1.81dB
99% > 2.63dB

Linearity 2.1bits @ 0.5dB error

View attachment 456482
View attachment 456483
View attachment 456484
View attachment 456485


Discussion
This is a bit like a Rorschach test. You can look at the same set of data and come to different and completely opposite conclusions, depending on your bias/perspective. A glass of water can scientifically be identical when described as half-full or half-empty, and yet we all appreciate that the two descriptions impart differences.

That said, at least this is how I think you should consider the data for interpretation.
Comparison 1: The Raw Files
This is an unfolded MQA-CD vs regular CD. We know that you lose 1-3 bits of data, theoretically, with MQA encoding. There are also scenarios where different formats have fundamentally different recordings/mixes. The La La Land SACD vs. streaming high-res comparisons is a good example.

Comparison 2: The Recorded Output from each CD
If you own the Bluesound Node Icon, QRONO d2A is being applied to all content, whether MQA-enabled or not. This shows you if QRONO d2A + regular CD is different from a full MQA chain.

Comparison 3: Recording of the CD vs. Digital CD File
This should capture the consequences of QRONO d2a on regular content. This is the impact of Bluesound Node Icon "forcing" QRONO d2a to be enabled all the time.

Comparison 4: Recording of the MQA vs. Unfolded MQA-CD file
To my knowledge, there is no MQA-CD exclusive content -- there are some MQA-exclusive streaming. Suppose there was MQA-exclusive content which means you can play it back on a normal system vs. an MQA system. This comparison may predict the differences.

Comparison 5: CD Recording vs. High Res Source
The high-res source comes from a different album. The mastering may be different.

Comparison 6: MQA-CD Recording vs. High Res Source
The high-res source comes from a different album. The mastering may be different. This is the only way to compare an unfolded MQA CD against a High-Res recording.

Comparison 7: CD Source File vs. High Res Source File
This is the most formal comparison of the mastering between the two files. You can come to your own conclusion, but I would argue that this shows that Comparisons 5 and 6 are of limited use. We don't have a good way to compare the MQA decoded CD chain against the same master if it was distributed high-res.

Thanks for taking the time to run these tests! Is there a way for you to share the files used for this analysis? The results look nearly random for the last three comparisons (5, 6, and 7). Seems to me these are largely uncorrelated source files, likely with very different processing and/or mastering -- the differences are too significant to make any conclusions from the results.

 

[GXAlan]

@pjug - thanks for correcting my terminology. Yes, folded or "undecoded" is the right word.

There is no meaningful difference in the high frequency content between MQA-CD and regular CD. There is a difference in the high-res file, but it's a different mastering in my opinion. If you look at the Deltawave logs, the fits are poor but if you look at the correlated waveforms, it's not horrible looking. It could be user error

Now that others have had a chance to comment, here's what I mean by the variable conclusions...
1) MQA is worthless -- the differences are miniscule and the PK Metric shows that the MQA-CD and regular CD are going to be audibly identical whether you unfold or not.
2) QRONO is non-destructive -- the differences of the undecoded content is miniscule and the differences in recordings are slight after decoding
3) MQA is great -- the differences in phase, those subtle, carry down into the audible frequency range and the whole "story" of MQA is something they invented called "temporal blur"

 

[GXAlan]

To add some details, I sent the source files to @pkane who agreed that the “high res” version is not the same master, or if it is processed, processed very differently.

From my own opinion, this injects new debate about high res versus non high res audio. The aligned waveforms are close enough that it doesn’t seem like we are looking at two different takes in the recording studio. However if the recordings that we buy from a high res copy and the recordings that we buy from a CD are different, independent of sampling rate, it potentially leads to audible differences which are attributed to high res but are just different design choices in mastering.

It’s like this comparison of physical media versus streaming…

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

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...

www.audiosciencereview.com

 

[pjug]

If the original point of this is to look at how well MQA nulls with a master, compared to normal downsampling with linear phase filters, I think we know that MQA will be worse because of the effect of phase distortion. I don't think Bob Stuart would argue about that. If you want to see evidence I have some files where @mansr was able to hack a bluesound device and get the MQA high sample rate rendering without recording with a ADC. So these can be used to get very pure results in DW without error contribution from ADC filters.

 

[GXAlan]

If the original point of this is to look at how well MQA nulls with a master, compared to normal downsampling with linear phase filters, I think we know that MQA will be worse because of the effect of phase distortion. I don't think Bob Stuart would argue about that. If you want to see evidence I have some files where @mansr was able to hack a bluesound device and get the MQA high sample rate rendering without recording with a ADC. So these can be used to get very pure results in DW without error contribution from ADC filters.

MQA and QRONO get interchanged a lot. This post also shows you what happens when you apply the QRONO digital filter to NON MQA encoded content.

I think this gets to my point that the Bluesound Node’s “forced” use of QRONO is nothing to worry about, if the benefits of low noise (which matters with my high gain active speakers), Dirac, small footprint, eARC, etc.

Although there is a digital filter from the ADC, recording the output at 88 kHz should minimize the impact of that filter. Likewise, there is something that says that QRONO can be licensed out and the filter is fine tuned for individual DACs.

All that said, with Lenbrook putting new money back into MQA, your digital analysis would be interesting if it sheds any new light to what has been shown before.

 

[pjug]

Although there is a digital filter from the ADC, recording the output at 88 kHz should minimize the impact of that filter.

There is also a HP filter in the ADC for removing DC. We saw in Mani's loopback thread that this messes up nulling quite a bit.

 

[pjug]

Here are comeparisons using @mansr's rendered MQA version of the 2L track.
1st image is DXD vs rendered MQA
2nd image is DXD vs 24/88
3rd image is DXD vs 16/88 (flac is a good deal smaller than the MQA, but it nulls better than MQA)
4th image is DXD with minimum phase filter applied vs rendered MQA, showing that this nulls better in the audible band than in 1st image, and we can do similar as MQA does using just an ordinary filter.

 

[GXAlan]

Here are comeparisons using @mansr's rendered MQA version of the 2L track.
1st image is DXD vs rendered MQA
2nd image is DXD vs 24/88
3rd image is DXD vs 16/88 (flac is a good deal smaller than the MQA, but it nulls better than MQA)
4th image is DXD with minimum phase filter applied vs rendered MQA, showing that this nulls better in the audible band than in 1st image, and we can do similar as MQA does using just an ordinary filter.
View attachment 457153View attachment 457154View attachment 457155View attachment 457156

Everything is in the -100 dB range… and this is all in the inaudible ultrasonic range?

I would be interested to see the phase difference comparison in the audible range.

 

[pjug]

Everything is in the -100 dB range… and this is all in the inaudible ultrasonic range?

I would be interested to see the phase difference comparison in the audible range.

Yes everything is -100 dB in the audible band. I was just using the whole dxd band since that is the reference. I will get a couple phase plots in a little while.

edit: here is the comparison DXD and MQA, and then 2nd image shows simple filtering of the DXD gives similar phase shift so that these files match better.