I published music on Tidal to test MQA - MQA Review

GoldenSound

AI-assisted summary of I published music on Tidal to test MQA - MQA Review — original source. Check the original for full context.

Source published: 2021-04-15

Key ideas

GoldenSound tested MQA by publishing an acoustic track embedded with hidden test signals to Tidal, discovering that the format is measurably lossy, adds high-frequency noise even during silent passages, and causes ultrasonic aliasing down into the audible band. (0:00)
The glowing blue 'MQA Studio' provenance indicator does not guarantee studio origin or bit-perfection; standard 44.1 kHz CD masters trigger the light and upsample to 352.8 kHz, while forum tests showed up to 30% of bits can be truncated without turning the light off. (13:47)
Tidal's HiFi tier delivers undecoded lossy MQA streams rather than true 16-bit/44.1 kHz FLAC on MQA-cataloged tracks, prompting artists like Neil Young to pull their music over unauthorized alterations of original masters. (12:38)
Following early outreach with his findings, GoldenSound's tracks were pulled from Tidal by MQA and distributor access was cut, and MQA's technical rebuttal attributing added noise to omitted dither was refuted by tests showing identical degradation on dithered masters. (33:02)

Detailed notes

## Background Published on YouTube by GoldenSound on April 15, 2021, this video presents an empirical investigation into the claims and technical realities of MQA (Master Quality Authenticated). According to the description and show notes, it is a deep-dive review and measurement analysis exploring the problems and controversies surrounding the format, supported by an extensive list of community references and technical papers.

## TL;DR To test MQA's closed proprietary ecosystem, GoldenSound encoded and published original acoustic tracks containing hidden test signals directly onto Tidal and Deezer. Comparing the resulting streams against the untouched master files revealed that MQA is not lossless, introduces measurable high-frequency noise and non-linear filter ringing, and folds ultrasonic test signals down into the audible spectrum as aliased distortion. GoldenSound further demonstrates that the format often upsamples standard 44.1 kHz Red Book recordings into inflated 352.8 kHz outputs under an unverified 'MQA Studio' authentication light, effectively substituting true lossless FLAC on Tidal with an altered, lossy alternative. Addressing MQA's defensive corporate rebuttal and retaliatory takedowns, the analysis dismantles their arguments regarding dither, filter leakiness, and conversion profiling, urging audiophiles to reject proprietary lock-in in favor of open, verifiable high-resolution audio.

## Atmosphere & style The presentation carries the methodical, unflinching weight of a whistleblower pulling back the curtain on an industry-wide illusion. GoldenSound speaks with calm, articulate precision, avoiding hyperbolic rage in favor of measured scientific scrutiny, clinical oscilloscope charts, and direct spectrum analyses. The pacing conveys mounting incredulity as MQA's shifting marketing claims crumble under empirical testing, punctuated by dry, understated humor regarding corporate defensiveness. Watching the video feels like sitting through a rigorous forensic debriefing—equal parts revelatory, quietly defiant, and deeply validating for an audiophile audience long weary of proprietary obfuscation.

## Bulletin - [00:00](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=0s) GoldenSound tested MQA by publishing an acoustic track embedded with hidden test signals to Tidal, discovering that the format is measurably lossy, adds high-frequency noise even during silent passages, and causes ultrasonic aliasing down into the audible band. - [13:47](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=827s) The glowing blue 'MQA Studio' provenance indicator does not guarantee studio origin or bit-perfection; standard 44.1 kHz CD masters trigger the light and upsample to 352.8 kHz, while forum tests showed up to 30% of bits can be truncated without turning the light off. - [12:38](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=758s) Tidal's HiFi tier delivers undecoded lossy MQA streams rather than true 16-bit/44.1 kHz FLAC on MQA-cataloged tracks, prompting artists like Neil Young to pull their music over unauthorized alterations of original masters. - [33:02](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1982s) Following early outreach with his findings, GoldenSound's tracks were pulled from Tidal by MQA and distributor access was cut, and MQA's technical rebuttal attributing added noise to omitted dither was refuted by tests showing identical degradation on dithered masters.

## Key takeaways - [03:45](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=225s) MQA restricts full unfolding strictly to DACs with analog outputs, preventing independent researchers from capturing a clean digital stream to verify bit-perfection against studio masters. - [05:47](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=347s) Automated ingestion filters initially rejected pure test tones, requiring GoldenSound to embed sweeps, impulses, and white noise into a legitimate acoustic piece to bypass MQA's gatekeeping. - [06:57](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=417s) While Deezer delivered a 100% bit-perfect FLAC match of the original 44.1 kHz master, Tidal's MQA encode suffered from altered waveforms, overshooting transients, and elevated noise. - [08:49](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=529s) DeltaWave null tests on commercial tracks and synthetic masters show that MQA adds background noise across the runtime and alters audible-band frequencies. - [11:11](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=671s) Dynamic range testing using a -60 dBFS tone demonstrated high-frequency noise rising roughly 45 dB below signal level on 44.1 kHz encodes and 20 dB above the test signal on 88.2 kHz-derived files. - [11:11](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=671s) Standard 44.1 kHz master files authenticate as 'MQA Studio' and upsample to 352.8 kHz without recovering any real ultrasonic content, merely applying internal upsampling filters with added noise. - [12:38](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=758s) MQA encoding increased file size from 30 MB to 33 MB on a 44.1 kHz master, disproving claims that the format conserves streaming bandwidth compared to standard FLAC. - [13:47](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=827s) MQA employs an intentionally leaky reconstruction filter that generates mirror images and ultrasonic reflections back into the frequency spectrum above Nyquist. - [16:53](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1013s) Undecoded MQA playback folds genuine ultrasonic tones (such as 35 kHz and 40 kHz) down into the audible spectrum at 9.1 kHz and 4.1 kHz with minimal attenuation. - [18:07](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1087s) MQA's claim that folded ultrasonic data is safely buried below -120 dB is disproven by measurements showing leakage artifacts resting as high as -60 dB, rendering them audible. - [24:23](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1463s) Modern broadband makes proprietary audio compression unnecessary, as native 24-bit/192 kHz lossless FLAC requires less than 5 Mbps—well below 4K video streaming demands. - [27:29](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1649s) iFi's proprietary GTO filter exhibits frequency response curves, leaky roll-off, and ultrasonic imaging identical to MQA unfolding because it was developed with MQA's direct assistance. - [28:57](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1737s) MQA's claim of end-to-end ADC and DAC hardware correction is contradicted by indie ingestion workflows, which collect no data regarding the converters or hardware used. - [35:50](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=2150s) An audience survey revealed widespread consumer confusion, with two-thirds of pro-MQA respondents mistakenly believing MQA matches or outperforms native 24-bit/192 kHz FLAC.

## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=0s) The MQA Experiment and the Backlash

GoldenSound introduces his comprehensive investigation into MQA's marketing claims by explaining a unique experiment: he published his own music on Tidal in MQA format to bypass the lack of public testing tools. In an effort to act in good faith, he reached out to MQA prior to releasing this video to present his findings and give them an opportunity to respond or suggest alternative testing. Instead of a constructive dialogue, his music was swiftly removed from Tidal the very next day, and MQA contacted his publisher to block him and remove all of his content.

GoldenSound reflects on this defensive reaction, noting that “...for a company with nothing to hide they are sure very keen on keeping information like what i'm about to show quiet.” He emphasizes that this video is not an attack on the many excellent hardware manufacturers who have integrated MQA into their products, nor is it meant to dismiss subjective listening preferences. Rather, it is an objective, evidence-based analysis of MQA's specific technical claims. Through his testing, he promises to prove that MQA is not lossless, adds unwanted noise and distortion even in silent portions of tracks, is rarely sourced from true high-resolution masters, and that the "MQA Studio" blue light does not actually authenticate anything.

- **Core Claims to be Proven:** - MQA is not lossless. - MQA adds unwanted noise and distortion, even during silent passages. - MQA is usually not sourced from a high-resolution master. - The "MQA Studio" blue light does not authenticate the file's integrity or origin.

---

[03:45](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=225s) What is MQA and Why is it So Hard to Test?

MQA is marketed as a proprietary music encoding format that promises to deliver the absolute best listening experience by unlocking "every detail of the original master recording." The core marketing claim is that MQA takes high-resolution studio recordings and "folds" the high-frequency information beneath the noise floor of a standard 44.1 kHz file. This file can theoretically be played back on standard, non-MQA DACs as a normal CD-quality track, or it can be "unfolded" back to high resolution. The first unfold is typically performed in software (such as the Tidal app) to output 88.2 kHz or 96 kHz, while subsequent unfolds are handled by hardware renderers or full decoders to reveal the "original" studio resolution.

The primary issue for independent researchers is that MQA is entirely proprietary and closed-source, making it exceptionally difficult to test. Unlike open formats like FLAC, MP3, or DSD, users cannot encode their own files to see how the encoder behaves. Furthermore, MQA licensing rules strictly prohibit any device with a digital output from performing a full MQA unfold. The full unfolding process is restricted entirely to DACs with analog outputs, preventing users from capturing a clean digital stream of the unfolded file to compare it bit-for-bit against the original high-resolution master. GoldenSound explains, “It is only in dax with an analog output meaning it's impossible to test if unfolded mqa really is the same as native hi-res because we cannot get a true digital unfold and check it against the original.”

```mermaid flowchart TD accTitle: Test File Creation and Distribution Flow accDescr: Flowchart showing how GoldenSound created test files disguised as an acoustic track and distributed them to Deezer and Tidal to compare MQA against native FLAC.

A["Original Master (Acoustic Track + Hidden Test Signals)"] --> B["44.1 kHz Version"] A --> C["88.2 kHz 'Deluxe' Version (with 35/40 kHz tones)"]

B --> D["Deezer (Native FLAC)"] B --> E["Tidal (MQA Encoded)"]

C --> F["Deezer (Downsampled to 44.1 kHz)"] C --> G["Tidal (MQA Encoded & Unfolded)"]

D --> H["Result: 100% Bit-Perfect"] E --> I["Result: Overshooting Transients & Added Noise"] F --> J["Result: Clean Downsample, No Added Noise"] G --> K["Result: High-Frequency Noise & Aliasing"] ```

---

[05:47](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=347s) Sneaking Test Signals Past the MQA Gatekeepers

To overcome these testing limitations, GoldenSound decided to publish his own music so he would possess both the exact original master files and the final MQA-encoded distribution files. However, MQA has implemented strict gatekeeping measures to prevent analysis. When he first attempted to submit standard test signals—such as a simple 1 kHz sine wave or an impulse response test—the publisher informed him that the MQA encoder was "unable to encode" the files.

GoldenSound remarks on this convenient limitation: “When i first attempted to do this i was told by the publisher that the mqa encoder was unable to encode the file how convenient...” To bypass these automated checks, he composed a legitimate acoustic track and hid a comprehensive suite of test signals inside it. This included an impulse response, white noise, a square wave, a 32-tone test, and a complete RightMark Audio Analyzer (RMAA) test sequence. He published a standard 44.1 kHz version and an 88.2 kHz "deluxe" version. The deluxe version contained additional hidden high-frequency test tones at 35 kHz and 40 kHz, alongside full-range frequency sweeps from 20 Hz to 44 kHz, allowing him to verify if MQA truly preserves high-resolution data after unfolding.

---

[06:57](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=417s) Analyzing the 44.1 kHz and 88.2 kHz MQA Files

The test results revealed significant differences between the original masters and the MQA-encoded files. When comparing the 44.1 kHz master against the file delivered by Deezer, the FLAC file was 100% bit-perfect. However, the Tidal MQA version was heavily altered by the encoding process. GoldenSound notes, “Immediately the waveform looks different transients are overshooting giving the impression of it being louder at some points and there is more high frequency noise compared to the original even in what should have been silent portions of the track.”

The impulse response test on the MQA file revealed a minimum-phase ringing pattern that did not decay linearly. While minimum-phase filters avoid pre-ringing (which can make transients sound "snappier"), they introduce phase distortion because different frequencies travel through the filter at different speeds. This phase shift can negatively impact musical elements like timbre, which rely on multiple frequencies aligning perfectly in time. Unfolding the file only exacerbated these anomalies. The only way GoldenSound could correct the transient behavior was by running the track through HQPlayer's "poly-sinc-mqa" filter, which is specifically designed to clean up MQA-induced noise and filter errors.

The 88.2 kHz "deluxe" MQA track on Tidal fared even worse. While the Deezer version was cleanly downsampled to 44.1 kHz with no added noise, the Tidal MQA version exhibited high-frequency noise throughout the track—even during passages of absolute silence. Most concerningly, high-frequency content that originally existed above 22 kHz was not preserved cleanly; instead, it was aliased and reflected back down into the audible frequency band as distortion.

- **Key Findings from File Analysis:** - **Deezer FLAC:** 100% bit-perfect match to the original master. - **Tidal MQA (44.1 kHz):** Altered waveform, overshooting transients, and elevated high-frequency noise in silent sections. - **Filter Behavior:** Minimum-phase filter with non-linear decay, causing phase shifts and group delay. - **Tidal MQA (88.2 kHz):** Significant high-frequency noise and severe aliasing/imaging of ultrasonic content back into the audible band.

[08:49](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=529s) The Noise Mystery and Comparing Real-World Masters

GoldenSound notes that MQA encodes are not lossless regardless of original sample rate, and playing them without an MQA DAC or software unfolding creates real playback issues compared to streaming the pure lossless FLAC. Analyzing his test file, he developed a hunch: the elevated high-frequency noise floor centered around 25 kHz seemed proportional to the amount of ultrasonic content MQA was forced to reconstruct. Because his original 88.2 kHz master contained full-range white noise, it pushed that reconstruction mechanism hard. To confirm the hypothesis, he prepared an alternate 88.2 kHz master with the white noise omitted and ultrasonic energy drastically rolled off. Unfortunately, before that revised master went live, MQA intervened and had his tracks pulled from TIDAL entirely, cutting short direct live confirmation.

```mermaid flowchart TD accTitle: Ultrasonic Noise Hypothesis Workflow accDescr: Flowchart contrasting test tracks with and without ultrasonic content to explain noise generation in MQA.

subgraph GoldenSoundTrack["GoldenSound Test Master"] A["88.2 kHz Master (Full White Noise)"] --> B["MQA 44.1 kHz Encode"] B --> C["Heavy 25 kHz Noise Band Across Whole Track"] end

subgraph SamSmithTrack["Commercial Comparison Master"] D["SamSmith Master (No Content >22 kHz)"] --> E["MQA 44.1 kHz Encode"] E --> F["Clean High End (No 25 kHz Noise Band)"] end

C -.->|"Supports Hypothesis"| G["Noise added is proportional to ultrasonic content reconstructed"] F -.->|"Supports Hypothesis"| G ```

He found circumstantial proof by comparing a commercial release that existed both as a native 88.2 kHz file and an MQA file sourced from that same 88.2 kHz master: Sam Smith's track "Too Good at Goodbyes." Examining the native 88.2 kHz file revealed that it actually contained zero musical content above 22 kHz. Consequently, the 44.1 kHz MQA release showed none of the elevated 25 kHz noise band seen in GoldenSound's heavy-ultrasonic test file. Running DeltaWave software comparisons between the native and MQA versions of the Sam Smith track confirmed this behavior: while high-frequency noise was still added by MQA, a strange band of difference or distortion ran through the entire track well within the audible spectrum—an artifact absent from GoldenSound's synthetic files. As GoldenSound sums up the fundamental problem: “MQA or any method aimed at high fidelity shouldn't be adding noise.”

- MQA files played back without decoding add background noise that is absent in native lossless streaming. - Files with natural bandwidth roll-offs around 22 kHz create far less ultrasonic noise than tracks with rich high-frequency spectra extending up to 44 kHz. - DeltaWave null testing of real-world commercial releases exposes audible-band differences and alterations introduced by the MQA process across the entire runtime.

[11:11](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=671s) Low-Level Linearity, Dynamic Range Loss, and Upsampling 44.1 kHz Tracks

Analyzing low-level performance reveals significant signal degradation across both 44.1 kHz and 88.2 kHz MQA encodes. GoldenSound included a standard dynamic range test signal: a 1 kHz sine wave at -60 dBFS. On the original unencoded master, dynamic range was pristine, limited only by the FFT gain of the measurement analyzer. In the 44.1 kHz MQA file, high-frequency noise appeared roughly 45 dB below the signal level, accompanied by broadband noise across the entire spectrum sitting about 65 dB below the tone. In the MQA file generated from the 88.2 kHz master, this high-frequency noise rose to roughly 20 dB *above* the -60 dB test signal.

```mermaid flowchart LR accTitle: 44.1 kHz Master Upsampling Chain accDescr: Flow diagram showing how a standard 44.1 kHz file is labeled MQA Studio and upsampled to 352.8 kHz.

A["Original Master: 44.1 kHz / 16-24 bit"] --> B["TIDAL Ingestion"] B --> C["Authenticated as 'MQA Studio'"] C --> D["DAC Unfolding / Rendering"] D --> E["Output Display: 352.8 kHz Sample Rate"] E -.-> F["No extra data retrieved: pure upsampling with added noise"] ```

Investigating the unfolding process yields even more troubling discrepancies. When playing GoldenSound's standard 44.1 kHz test track, the hardware DAC authenticated it as "MQA Studio" and unfolded it to a massive 352.8 kHz sample rate. Because the source file was never a high-resolution recording, no ultrasonic information existed to unfold; the DAC simply applied an internal upsampling filter. TIDAL offers no interface indicator to tell consumers whether an MQA file contains genuine high-resolution material or represents an upsampled standard-resolution master. When GoldenSound asks why any listener would accept this, he points out the obvious absurdity: “And if it is a 44.1 kilohertz original, why would you want the MQA version when all it's doing is adding noise and other unwanted effects compared to playing lossless?”

- The -60 dBFS 1 kHz test tone revealed high-frequency noise sitting 45 dB below the signal level on 44.1 kHz encodes, and 20 dB *above* the signal level on 88.2 kHz-derived encodes. - A standard 44.1 kHz master file triggers full "MQA Studio" authentication, upsampling all the way to 352.8 kHz despite containing zero ultrasonic content. - Consumers cannot differentiate between an upsampled Red Book release and a genuine high-res master within the TIDAL interface.

[12:38](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=758s) TIDAL's Forced MQA Delivery and Neil Young's Defection

Examining stream delivery directly in the TIDAL desktop application exposed a major bait-and-switch for subscribers paying for uncompressed audio. Selecting the "HiFi" quality tier instead of "Master" did not serve the original lossless CD-quality FLAC; TIDAL merely streamed the lossy, bit-mangled MQA file without running the unfolding code. GoldenSound argues that paying customers are being deceived, declaring that if lossless streaming is eliminated on titles labeled "Master," users should cancel their accounts immediately: “Even if MQA is a benefit in some cases, it should be a choice; it should not be mandatory. Speak with your wallet and demand lossless.” Furthermore, MQA fails to fulfill its primary pitch of bandwidth savings: GoldenSound's 44.1 kHz original master was 30 megabytes, while the resulting MQA file swelled to 33 megabytes.

This practice provoked industry pushback from prominent recording artists, most notably Neil Young, who removed his catalog from TIDAL after discovering the platform converted his 44.1 kHz recordings into MQA "Master" tracks without consent. GoldenSound quotes Young's scathing assessment of TIDAL's format conversion directly: “In his own words: 'TIDAL's master is a degradation of the original to make it fit in a box that collects royalties. That money is ultimately paid by the listeners. I am not behind it. I am out of there, gone. My masters are the original.'”

- Selecting "HiFi" on an MQA-encoded album on TIDAL delivers an undecoded MQA stream rather than a bit-perfect 16-bit / 44.1 kHz lossless FLAC. - File size increased from 30 MB (original FLAC) to 33 MB (MQA encode), contradicting marketing claims of efficient bandwidth compression. - Neil Young pulled his entire catalog from TIDAL to protest the unauthorized alteration and monetization of his CD-quality masters.

[13:47](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=827s) The Illusory Blue Light and Leaky Filter Artifacts

The MQA "Studio" provenance indicator (the glowing blue light on certified DACs) provides zero guarantee of audio integrity or bit accuracy. GoldenSound highlights tests by user FrederikV on the Audiophile Style forums, who demonstrated that up to 30 percent of the bits in an MQA file could be completely truncated without extinguishing the blue light. GoldenSound's own 44.1 kHz file triggered the same authentication indicator despite having its master data altered and degraded by MQA's pipeline. Summarizing the indicator's real value, GoldenSound states plainly: “This track has been altered, and so I, the artist, am telling you this is not the version that would have been created or heard in the studio, nor is it truly high resolution. This blue light is MQA's marketing, nothing more.”

```mermaid flowchart TD accTitle: MQA Leaky Reconstruction Filter Imaging accDescr: Diagram demonstrating how MQA leaky filtering creates ultrasonic mirror images and time-domain distortion.

subgraph SoftwareUnfold["First Unfold (TIDAL / Roon)"] A["MQA 44.1 kHz Source"] --> B["Upsampled to 88.2 kHz"] B --> C["Leaky Reconstruction Filter"] C --> D["Upward Mirror Images of Audible Content"] end

subgraph TimeDomainIssues["Impulse & Waveform Degradation"] D --> E["Square Wave Notch Distortion"] D --> F["Excessive Overshoot & Non-Linear Pre/Post Ringing"] end ```

When software like TIDAL or Roon unfolds that 44.1 kHz master to 88.2 kHz, spectrum analysis shows mirrored frequency reflections bouncing back into the ultrasonic band above 22.05 kHz. This imaging artifact is directly caused by MQA's unusually slow, "leaky" upsampling filter, an architectural flaw detailed by blogger Archimago and presented by Chris Connaker at the 2018 Rocky Mountain Audio Fest (RMAF). At that talk, MQA representatives in attendance offered no technical explanations for the leakiness, choosing instead to repeatedly interrupt Connaker, demand to know Archimago's real identity, and deflect by insisting he ask dCS why they support the format. Using the white noise section of his test track, GoldenSound mapped MQA's unfolding filter curve against HQPlayer's sinc-L filter (a pristine 2-million-tap reference similar to the Chord Electronics M Scaler). MQA's filter exhibited atrocious ultrasonic attenuation and odd frequency irregularities. In the time domain, reconstructed square waves suffered from severe overshoot, non-linear ringing, and unnatural notches at each transition point. Performing a full hardware decode only multiplied these upward images, extending the notched distortion further down the waveform.

- Forum member FrederikV proved MQA files retain hardware authentication even after truncating over 30% of their data bits. - MQA's upsampling filter is intentionally leaky, creating mirror images of musical content above the Nyquist cutoff (22.05 kHz). - MQA representatives at RMAF 2018 disrupted Chris Connaker's presentation rather than addressing the filter's measurable imaging artifacts. - Synthetic square-wave tests reveal aggressive waveform overshoot, non-linear ringing, and notch artifacts introduced during unfolding.

[16:53](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1013s) Ultrasonic Aliasing in Hi-Res Master Files

To evaluate how genuine high-resolution material behaves under MQA encoding, GoldenSound embedded 35 kHz and 40 kHz sine tones, along with RightMark Audio Analyzer (RMAA) IMD tests and frequency sweeps extending to 44 kHz, inside his original 88.2 kHz master. Because a standard 44.1 kHz container has a Nyquist limit of 22.05 kHz, these tones cannot exist natively within that sample rate. Examining the undecoded 44.1 kHz MQA file revealed severe aliasing: instead of safely stowing the tones away or filtering them out, the encoder folded the 35 kHz and 40 kHz tones back down into the audible band at 9.1 kHz and 4.1 kHz, respectively.

These aliased tones sat well above the baseline noise floor with minimal attenuation, meaning any listener playing MQA tracks on standard non-MQA DACs will hear loud aliased distortion inside the audible band. The continuous RMAA sweeps similarly hit the 22.05 kHz ceiling and bounced backward down through the audible spectrum. When GoldenSound enabled the first unfold, the 35 kHz and 40 kHz sine tones shifted back to their proper ultrasonic positions, and the sweeping tones unfolded back toward 44 kHz. While this confirmed MQA does restore some high-frequency content as advertised, the decoded signal remained heavily degraded by a massive band of noise centered around 25 kHz.

- Ultrasonic sine waves at 35 kHz and 40 kHz alias directly down to 9.1 kHz and 4.1 kHz in the audible band when played without MQA decoding. - RMAA test sweeps bouncing off the 22.05 kHz Nyquist barrier generate audible down-sweeps on standard DAC playback equipment. - First unfolding successfully restores ultrasonic frequencies to their proper positions, but leaves an elevated band of noise centered near 25 kHz.

[18:07](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1087s) Debunking MQA's Folding Claims and Dynamic Range Loss

Analyzing the decoded output reveals that the unfolding process leaves severe artifacts lingering across the lower frequencies, pushing the effective noise floor all the way up to roughly -60 dB. This directly contradicts the core technical marketing published on Bob Stuart's *Bob Talks* blog, which asserts that folded high-frequency content is tucked safely into the dynamic range window below -120 dB where it would remain completely inaudible. GoldenSound's test file demonstrates that this assertion is demonstrably false: the content was barely attenuated at all, meaning higher-level musical signals folded into this space would be undeniably audible. To verify this beyond the graphs, GoldenSound conducted a simple listening check and was able to hear the folded test tones playing out loud quite clearly.

```mermaid flowchart TD accTitle: Comparison of Claimed vs Actual MQA Folding accDescr: Diagram contrasting MQA marketing claims of burying folded content below minus 120 dB against test measurements revealing audible leakage at minus 60 dB.

subgraph MQA_Claim["MQA Marketing Claim"] direction TB C1["Ultrasonic Content Folded"] --> C2["Buried Below -120 dB"] C2 --> C3["Completely Inaudible"] end

subgraph Actual_Result["Measured Reality"] direction TB R1["Ultrasonic Content Folded"] --> R2["Attenuated Only to ~ -60 dB Noise Floor"] R2 --> R3["Clearly Audible Artifacts in Low Frequencies"] end ```

The consequence is a severe degradation of dynamic range across both the audible and ultrasonic spectra whenever MQA unfolds high-frequency material. Performing a full hardware decode fails to resolve the issue; instead, it repeatedly mirrors and reflects the file's spectral content upward throughout the ultrasonic band. In evaluating whether MQA successfully reconstructs ultrasonic information, GoldenSound concludes that while it technically returns high frequencies, it does so at the cost of heavily compromised accuracy, poor dynamic range, persistent ultrasonic noise, and audible-band distortion left in the wake of the unfolding process. As GoldenSound summarizes the stark divergence between marketing and measurement: “It does not by any stretch retain 100% of the original recording, and this claim is false.”

- MQA's *Bob Talks* blog claims folded audio rests below -120 dB, but test measurements demonstrated artifacts sitting around -60 dB. - High-frequency test tones folded into the audible band were loud enough to be heard directly by ear. - The 44.1 kHz base files suffer from substantial aliasing reflected down into the audible range at high levels. - Full decoding creates repeated upward reflections into the ultrasonic spectrum rather than clean reconstruction.

---

[19:54](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1194s) Quickfire Technical Summary of MQA's Flaws

To distill the technical findings, GoldenSound presents a rapid-fire rundown of the established evidence. First and foremost, MQA is provably lossy whether played back in its folded stock form or through an unfolding DAC. Furthermore, a substantial portion of the MQA catalogue on Tidal consists merely of upsampled standard-resolution files rather than true high-resolution masters. When an MQA file contains genuine high-frequency content, it aliases those frequencies downward into the audible range with minimal attenuation. When unfolded, the decoder restores the upper frequencies and removes the majority of the downward aliasing, but leaves behind severe background noise and spurious harmonic artifacts.

```mermaid flowchart LR accTitle: MQA Signal Processing Workflow and Identified Faults accDescr: Pipeline showing original masters entering MQA encoding, revealing upsampling, aliasing leakage, and leftover distortion after decoding.

subgraph Encoding["Tidal Ingestion & Encoding"] Master["Original Track<br>(Often 44.1 kHz 16-bit)"] --> Upsample["Hidden Upsampling<br>to MQA Container"] end

subgraph BasePlayback["Folded Base File Playback"] Upsample --> Folded["High-Frequency Aliasing<br>Leaked into Audible Band"] end

subgraph DecodedPlayback["Decoder Hardware / Software"] Folded --> Decoded["Unfolded Output:<br>Restores Highs, Leaves -60 dB Noise & Distortion"] end ```

Equally troubling is the impact on streaming platforms: Tidal no longer delivers genuine lossless audio for tracks flagged as "Master." Listeners are restricted either to lossy MQA or compressed streams, completely locking out native bit-perfect FLAC. GoldenSound points out that listeners seeking bit-perfect lossless must turn to rival services like Qobuz or Deezer. The vaunted "authentication" blue light provides zero verification of the original source's sample rate, file integrity, or fidelity to the studio session. However, Roon users can uncover the truth: looking at the final sample rate listed in the track description exposes the actual source sample rate of the master file, revealing how many tracks are merely 44.1 kHz recordings disguised as high-res. Finally, like-for-like file size comparisons demonstrate that MQA has no bandwidth advantage over a standard 44.1 kHz FLAC (MQA is actually larger), though it remains smaller than an uncompressed 24-bit/192 kHz FLAC.

---

[21:18](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1278s) Transparency vs. Obfuscation: Rob Watts, DSD, and the Snake Oil Divide

Stepping back from pure data analysis, GoldenSound offers personal commentary as an audiophile passionate about gear and legitimate technical innovation. The audio industry has always hosted its fair share of snake oil, but products traditionally split into two camps: those whose claims can be tested and independently verified by third parties, and those that deliberately avoid scrutiny. Designers like Rob Watts of Chord Electronics openly champion bespoke designs, such as the WTA (Watts Transient Aligned) filter, claiming it represents the ideal method for 44.1 kHz audio playback. While one may agree or disagree with Watts's conclusions, his philosophy allows open evaluation: any user with a Chord DAC or M Scaler can run test signals through the hardware, analyze the output, and engage with Watts directly on forums like Head-Fi where he explains his engineering rationale in public.

A similar transparency defines formats like DSD, which is an open ecosystem where anyone can encode, decode, test, and critique the technology. Even audiophile cables—frequently ridiculed as pure snake oil—can be tested for empirical differences; GoldenSound notes that measuring a generic BNC cable against an AudioQuest Carbon BNC revealed measurable jitter differences. Whether those variations justify the cost or are audible is up for debate, but the crucial factor is testability. MQA, by contrast, operates on complete obfuscation. The company neither publishes rigorous proof of its claims nor permits the public to test its ecosystem, forbidding consumer-level encoding tools and locking down the digital output of full decoders. “It just operates on a ‘trust us, it's better’ approach,” GoldenSound notes, while its marketing claims remain elusive and continually shift whenever third-party engineers poke holes in the narrative.

- Rob Watts (Chord Electronics) openly discusses filter design on forums like Head-Fi and permits users to measure hardware like the Hugo DAC and M Scaler. - DSD remains an open format accessible to any engineer or hobbyist wishing to evaluate its strengths and weaknesses. - Independent jitter measurements proved an AudioQuest Carbon BNC cable measurably outperformed a generic BNC, highlighting that real physical claims can always be tested. - MQA prohibits user encoding, locks downstream digital decoders to block digital capture, and offers marketing claims that shift once disproved. - While MQA's custom "white glove" analog-to-digital correction work is fascinating (as documented in their "Fairy Tales" article), it represents bespoke remastering, not the reality of the mass automated encoding applied across Tidal. - Stuart’s theoretical folding model falls apart on 16-bit files, which lack the dynamic range required to bury folded ultrasonic data inaudibly.

---

[24:23](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1463s) Monopolies, Bandwidth Realities, and Industry Pushback

GoldenSound emphasizes that even if MQA were mathematically flawless and completely lossless, it should never be permitted to become an inescapable monopoly. Consumers must retain the freedom to select native, uncompressed high-resolution PCM over proprietary alternatives. The argument that MQA is essential to conserve bandwidth in the modern era is completely obsolete: an uncompressed 24-bit/192 kHz stream requires under 5 megabits per second. Any household capable of streaming 4K video on YouTube or Netflix possesses more than enough internet speed to handle native high-resolution audio without compression. Instead, consumers are being forced into a closed ecosystem where MQA replaces native masters, leaving listeners with a lossy approximation as their only avenue to high-resolution content.

Manufacturers are not embracing MQA out of an organic belief in its sonic superiority, but because widespread consumer demand leaves them with little choice. High-end manufacturers like Linn have publicly opposed the format, while PS Audio's Paul McGowan has openly discussed their reluctance. McGowan confirmed that PS Audio integrated MQA compatibility solely because the market demanded it, stating candidly that he still dislikes the technology. Hardware vendors face the grim reality that failing to pay MQA licensing fees risks losing retail sales. GoldenSound calls on audiophiles to reject the format, refuse to pay the hardware licensing premiums baked into equipment costs, cancel subscriptions to platforms that replace lossless archives with MQA, and demand total transparency from manufacturers. GoldenSound delivers a clear warning: “If a company can't provide its own evidence and goes to great lengths to prevent anyone else from testing it, that should probably raise quite a few red flags.”

- A native 24-bit/192 kHz FLAC stream requires under 5 Mbps of bandwidth, well within the capability of standard modern broadband. - High-end audio companies such as Linn have published scathing critiques of MQA's proprietary model. - Paul McGowan of PS Audio acknowledged that MQA was added to PS Audio products strictly under market pressure, not engineering preference. - Hardware manufacturers are pressured to license MQA to avoid losing sales, passing the royalty costs directly to consumers. - As a closing technical recommendation, GoldenSound notes that if one must interact with MQA files, performing the first unfold in software is far preferable to running the hardware's subsequent unfolding filters.

[27:29](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1649s) DAC Filter Recommendations and the iFi GTO Filter Discovery

GoldenSound advises listeners on the optimal playback chain: perform the initial decode and then upsample or oversample using the DAC's internal digital filter rather than engaging any downstream MQA rendering. Doing so eliminates most audible-band aliasing while suppressing unwanted ultrasonic images and noise.

While testing, GoldenSound used the iFi micro iDSD Diablo, a full MQA decoder described as a fantastic product. However, evaluating the unit revealed a striking technical detail regarding iFi's proprietary GTO (Gibbs Transient Optimized) filter: its measured frequency response curve is virtually indistinguishable from an MQA renderer filter. Testing showed that the GTO filter displays the exact same "leaky" roll-off characteristics and generates ultrasonic imaging identical to MQA unfolding. Because it is a minimum-phase filter, phase-dependent sonic attributes like instrumental timbre can suffer. Reverting to iFi's firmware version 7.0 and selecting the standard stock filter resolved these issues entirely. The similarity proved unsurprising upon reviewing iFi's technical paper, which acknowledged that the GTO filter was developed with direct assistance from MQA. While GoldenSound maintains that hardware supporting MQA is not inherently bad, consumers should voice their reluctance to pay licensing premiums for features they do not want.

```mermaid flowchart TD accTitle: Recommended MQA DAC Decoding Flow vs GTO Filter Issues accDescr: Flowchart contrasting standard DAC upsampling against MQA rendering and iFi GTO leaky filtering.

subgraph PreferredPlayback["Recommended Signal Path"] A["MQA Core Decode"] --> B["Standard DAC Internal Upsampling"] B --> C["Suppressed Aliasing, Reduced Imaging & Lower Noise"] end

subgraph MQAandGTO["MQA Rendering / iFi GTO Filter"] D["MQA Render / iFi GTO Filter"] --> E["Slow Leaky Minimum-Phase Roll-Off"] E --> F["Ultrasonic Imaging & Degraded Phase/Timbre"] end ```

- GoldenSound urges consumers to tell equipment manufacturers: “Tell manufacturers you don't want to pay extra for it.” - iFi firmware 7.0 with the stock filter avoids the leaky minimum-phase imaging observed in the GTO filter. - The iFi GTO filter design paper explicitly notes development assistance from MQA.

[28:57](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1737s) MQA's Defense: Production Chain Assumptions and Lack of Transparency

Before the video went live, MQA provided a formal response. MQA opened by claiming their format is substantially more advanced than traditional digital audio because it compensates for ADC and DAC conversion processes across the full hardware chain. GoldenSound dismantles this claim: when submitting tracks via MQA's recommended publishing pipeline for independent artists, the encoder never requested any information regarding the production equipment, converters, or monitoring hardware used. Furthermore, modern recordings rarely feature an all-analog signal digitized through a single ADC model; they routinely blend multiple converters, synthesizers, sample libraries, and in-production digital processing steps within a single project.

GoldenSound addresses MQA's dismissive reaction to third-party testing: > “When someone like Chris Connaker provides evidence the question gets dodged and you try to change the subject.”

The absence of open analysis tools stands in direct contrast to virtually every standard audio format. Users can play MQA files but have no means of inspecting how masters were manipulated, validating claims against placebo, or verifying test signals without laborious workarounds. Rather than providing technical clarity, MQA responded to direct inquiries by having the author's tracks purged from publishing platforms.

- MQA claims end-to-end hardware profiling, but indie upload workflows collect zero converter or hardware metadata. - Contemporary music production uses diverse ADC models, plugins, and synthesized sources, making global ADC compensation impossible. - Independent verification is blocked by a closed ecosystem that lacks objective evaluation tools.

[30:42](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1842s) Encoder Limitations, Studio Tool Obscurity, and File Authentication

MQA argued in their statement that their indie encoder is not engineered to handle content where statistical audio profiles shift mid-track or where material deviates from "natural" acoustic sound. GoldenSound points out the absurd practical consequence: if synthetic textures break the encoder, vast swathes of modern genres—from electronic music to hybrid pop—are unsupported. Notably, eight out of the top ten streamed tracks on TIDAL are predominantly digital or synthesized productions, yet they are served as MQA.

While MQA points to proprietary "studio tools" that purportedly solve these issues, their existence remains entirely opaque: the public does not know what they do, who holds licenses, which commercial releases utilized them, or how they perform under scrutiny. GoldenSound suggests MQA adopt an evaluation model similar to Signalyst's HQPlayer, which lets users test its professional engine by restricting playback to 60-second intervals (or even 10 seconds for MQA) without causing commercial harm.

GoldenSound also rejects MQA's assertion that the artist or label bears sole responsibility for confirming the sound on TIDAL to validate the blue MQA authentication indicator. The encoder never communicated this policy, nor does official marketing explain that an MQA file can have a third of its data discarded while still illuminating the authentication light. Even the test file submitted at standard 44.1 kHz sounded different and measured differently from the uncompressed master, contradicting claims of lossless delivery.

- Current TIDAL top-ten metrics show that non-acoustic, synthetic music dominates the platform's MQA catalog despite MQA's caveats. - GoldenSound questions MQA's definitions: “Saying that MQA is lossless but only within vague unspecified criteria and lossless only using non-conventional evaluation methods isn't exactly reassuring.” - If an encoding algorithm changes its behavior based on musical content, it is adaptive DSP rather than bit-perfect lossless transmission.

[33:02](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1982s) Debunking MQA's Excuses: Dither Claims, Leaky Filters, and High-Frequency Noise

GoldenSound reviews MQA's specific technical rebuttals regarding the submitted audio files. MQA alleged that several files triggered server-side encoding errors; however, GoldenSound only ever received a single automated error notification, which occurred on isolated artificial test bursts like bare impulses and square waves (a known behavior also seen on hardware such as the Chord Electronics Hugo M Scaler).

MQA further claimed the measured noise floor degradation was a "naive mistake" resulting from GoldenSound failing to apply dither. GoldenSound refutes this outright: > “The first track I submitted called Try Again did have dither and it showed all of the issues mentioned in this video in fact some to a worse extent.”

Dither was intentionally omitted only on the subsequent two tracks to test whether external dithering interacted poorly with MQA's encoder.

When MQA objected to the term "leaky" for their slow upsampling filters—labeling the term "derogatory"—GoldenSound dryly jokes about whether the filter's feelings were hurt, before explaining why the engineering critique stands. Genuine high-resolution audio (such as native 96 kHz) provides over 13 times more bandwidth between the 20 kHz audio ceiling and the Nyquist limit than 44.1 kHz, making steep brick-wall filtering unnecessary without resorting to MQA's lossy folding tricks.

MQA claimed the measured ultrasonic aliasing occurred solely because artificial test tone amplitudes were excessively hot. While GoldenSound had uploaded a secondary track with reduced ultrasonic levels to measure down-aliasing thresholds, MQA's takedown of his catalog prevented TIDAL from publishing it; he offers to publish an immediate correction at the top of the video description if MQA provides the processed file proving otherwise.

Finally, MQA asserted that the added noise was merely inaudible ultrasonic dither concentrated at 25 kHz. Measurements disprove this: on both 44.1 kHz and fully unfolded outputs, the noise curve encroaches deep into the audible audio band, reaching an audible −43 dB at 20 kHz.

[35:50](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=2150s) Misleading Master Sample Rate Labels and Consumer Confusion

GoldenSound addresses the metadata surrounding sample rates. Although an MQA file contains tags indicating the original master sample rate, TIDAL’s native interface obscures this information entirely, leaving subscribers with the impression that every "Master" badge represents a full 192 kHz delivery unless they run secondary inspection software like Roon.

This UI opacity actively misinforms consumers about what they are streaming. GoldenSound conducted an audience survey of more than 240 respondents to quantify consumer perception: > “Of the people that said MQA was a good thing, two-thirds of them believed that MQA releases were the same or better than 24-bit 192 kilohertz FLAC.”

The gulf between MQA's technical realities and consumer marketing claims represents the core problem with the format's commercial adoption.

- 46% of survey respondents believed MQA was beneficial to audio reproduction. - Among pro-MQA respondents, fully two-thirds mistakenly believed MQA matches or outperforms genuine 24-bit/192 kHz FLAC files. - TIDAL lacks native indicators displaying true source sample rates, requiring external programs like Roon for verification.

[36:38](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=2198s) The Core Issue of Transparency and Final Acknowledgments

GoldenSound brings his investigation to a close by emphasizing why the entire ecosystem surrounding MQA desperately requires transparency. He argues that even if one assumes the best-case scenario—that a 24-bit, 192 kHz recording folded down into a 44.1 kHz container could unfold into something completely indistinguishable from an original native 192 kHz recording—that still would not excuse the broader reality of the format. A substantial portion of the MQA catalog available on streaming services consists merely of 44.1 kHz files generated from standard 44.1 kHz CD-quality masters. These files simply cannot match or exceed the quality of native, uncompressed high-resolution FLAC, and marketing them under the guise of an upgraded audiophile experience creates a deeply misleading impression across the industry. As GoldenSound stresses regarding this persistent marketing narrative, “The vast number of MQA releases that are 44.1 kilohertz from 44.1 kilohertz masters are not the same or better than native high-res FLAC, and this misconception is concerning.”

Wrapping up the presentation, GoldenSound hopes the technical breakdown has proven both informative and engaging for the audience. He extends his personal gratitude to Daniel Mellinger and Amos (widely known in the community as Currawong) for their hands-on assistance during the production of the video. Amos in particular curated an extensive compendium of articles, technical analysis, and documentary evidence detailing the history and behavior of MQA—much of which extends beyond what could fit into the video and is provided in the description for viewers wishing to investigate further. GoldenSound also credits the broader technical community whose foundational research, open tools, and testing methodologies informed his experiments, giving credit to Archimago, John Atkinson, Miska, Mans Rullgard (mansr), and Frederik V. Finally, he thanks the 250 individuals who took part in his community MQA survey, gives a special shout-out to his top-tier Patreon supporters, and invites viewers to continue the conversation in his Discord server or the patron Telegram group.

- **Mastering realities:** A massive portion of MQA's streaming catalog originates from basic 44.1 kHz source masters rather than genuine high-resolution material, nullifying claims of superior high-frequency reconstruction. - **Key research contributors:** Currawong (Amos) compiled extensive reference documentation; technical testing and analytical tools drew on prior work by Archimago, John Atkinson, Miska, Mans Rullgard, and Frederik V. - **Top Patreon supporters recognized:** Diamond and Legend tier patrons Chris Ross, Kyle Gravitas, Crack, Daniel Mellinger, King Jong-un, and Lana Bennett.

## Notable quotes > [12:38](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=758s) Even if MQA is a benefit in some cases, it should be a choice; it should not be mandatory. Speak with your wallet and demand lossless. — GoldenSound

> [12:38](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=758s) TIDAL's master is a degradation of the original to make it fit in a box that collects royalties. That money is ultimately paid by the listeners. — Neil Young

> [24:23](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1463s) If a company can't provide its own evidence and goes to great lengths to prevent anyone else from testing it, that should probably raise quite a few red flags. — GoldenSound

> [08:49](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=529s) MQA or any method aimed at high fidelity shouldn't be adding noise. — GoldenSound

> [36:38](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=2198s) The vast number of MQA releases that are 44.1 kilohertz from 44.1 kilohertz masters are not the same or better than native high-res FLAC, and this misconception is concerning. — GoldenSound

## Mentioned GoldenSound, MQA, Tidal, Deezer, Qobuz, FLAC, HQPlayer, Sam Smith, DeltaWave, Neil Young, FrederikV, Audiophile Style, Archimago, Chris Connaker, Rocky Mountain Audio Fest (RMAF), dCS, Chord Electronics, M Scaler, Bob Stuart, Bob Talks, Rob Watts, Watts Transient Aligned (WTA), DSD, AudioQuest, Linn, PS Audio, Paul McGowan, iFi audio, iFi micro iDSD Diablo, Gibbs Transient Optimized (GTO), Signalyst, RightMark Audio Analyzer (RMAA), Roon, Daniel Mellinger, Currawong (Amos), John Atkinson, Miska, Mans Rullgard (mansr), Chris Ross, Kyle Gravitas, Crack, King Jong-un, Lana Bennett

## Video description A deep dive into MQA, and its problems. Testing starts at 5:45

Come chat on discord: https://www.thegoldenone.co.uk Support me on patreon and join the private telegram chat: https://www.patreon.com/goldensound (or subscribestar) https://www.subscribestar.com/goldensound

How to test MQA's leaky filter: https://audiophilestyle.com/forums/topic/62669-how-to-demonstrate-mqas-leaky-filter/ 2018 RMAF MQA talk by Chris Connaker: https://www.youtube.com/watch?v=NSv0lcHlawk Archimago's MQA article: https://audiophilestyle.com/ca/reviews/mqa-a-review-of-controversies-concerns-and-cautions-r701/ PS Audio vid on MQA: https://www.youtube.com/watch?v=lPfmWKjiccA Schiit on MQA: https://www.schiit.com/news/news/why-we-wont-be-supporting-mqa#:~:text=Today%2C%20Schiit%20Audio%20announced%20that,so%20that%20they%20may%20choose Linn on MQA: https://web.archive.org/web/20201111211105/https://www.linn.co.uk/blog/mqa-is-bad-for-music Neil Young removes albums from TIDAL: https://neilyoungarchives.com/news/1/article?id=Tidal-Misleading-Listeners

中文简报

我在 Tidal 上发布音乐以测试 MQA——MQA 评测为测试 MQA 的封闭专有生态系统，GoldenSound 将含有隐藏测试信号的原创原声曲目进行编码，并直接发布到 Tidal 和 Deezer 上。将生成的流媒体音频与未经处理的母带文件进行对比后发现，MQA 并非无损，它会引入可测量的搞频噪声和非线性滤波器振铃，并将超声波测试信号向下折叠至可听频谱内形成混叠失真。GoldenSound 进一步证明，该格式常常在未经验证的“MQA Studio”认证指示灯下，将标准的 44.1 kHz 红皮书 CD 录音升频为虚高的 352.8 kHz 输出，实际上在 Tidal 上用一种经过篡改的有损格式替代了真正的无损 FLAC。针对 MQA 官方带有防御性质的反驳和报复性下架举措，该分析驳斥了关于抖动、滤波器泄漏及转换配置文件的说辞，并呼吁音响发烧友拒绝专有锁定，转向开放且可验证的高解析度音频。 GoldenSound 通过在 Tidal 上发布嵌入了隐藏测试信号的原声曲目来测试 MQA，结果发现该格式在测量上是有损的，即使在静音段也会增加高频噪声，并会导致超声波混叠向下折叠至可听频段。 (0:00) 亮蓝色的“MQA Studio”来源指示灯并不能保证录音室源头或位完美；标准的 44.1 kHz CD 母带也会触发该指示灯并升频至 352.8 kHz，而论坛测试显示，高达 30% 的位即使被截断，指示灯也不会熄灭。 (13:47) 对于编入 MQA 目录的曲目，Tidal 的 HiFi 层级提供的是未解码的有损 MQA 流，而非真正的 16-bit/44.1 kHz FLAC，这促使 Neil Young 等艺术家因原始母带遭未经授权的篡改而撤下自己的音乐。 (12:38) 在就其发现进行早期沟通后，GoldenSound 的曲目被 MQA 从 Tidal 下架，其分发商权限也被切断；MQA 将新增噪声归咎于缺少抖动的技术反驳，也被后续测试推翻，测试显示加入抖动的母带存在完全相同的劣化。 (33:02)

Reading summary

GoldenSound conducted an independent empirical analysis of MQA by distributing an original acoustic track embedded with hidden test signals to Tidal, concluding that the proprietary format is measurably lossy, adds ultrasonic and audible noise, causes aliasing, and frequently upsamples standard-resolution masters under misleading marketing claims. Bypassing Closed Encoders for Independent Audio Analysis (5:47) Because MQA strictly prohibits third-party encoding tools and prevents DACs with digital outputs from performing a full unfold, independent evaluation of the format against original studio masters has historically been blocked. To overcome this limitation, GoldenSound authored an acoustic track embedded with a suite of analytical signals—including impulses, square waves, white noise, multi-tone tests, and high-frequency sweeps up to 44 kHz. When initial submissions containing bare test tones were rejected by MQA's automated encoder, embedding them within a legitimate acoustic track bypassed the gatekeeping. GoldenSound submitted the files to both Deezer and Tidal at 44.1 kHz and 88.2 kHz sample rates. When he shared his findings privately with MQA before publication to offer an opportunity for clarification, MQA contacted the distributor, resulting in the immediate deletion of his tracks from Tidal. GoldenSound notes that this aggressive removal and refusal to supply verification tools contradicts MQA's public positioning of operating with technical transparency. Measurable Lossiness, Added Noise, and Aliasing Distortion (6:57) File comparisons between the original masters and streaming outputs revealed significant signal degradation on MQA encodes, whereas Deezer delivered bit-perfect FLAC files. In standard 44.1 kHz MQA streams, GoldenSound observed overshooting transients, minimum-phase ringing that introduces phase distortion, and an elevated noise floor even across completely silent passages. Synthetic dynamic range tests using a 1 kHz sine wave at -60 dBFS showed high-frequency noise sitting roughly 45 dB below signal level on 44.1 kHz encodes, and rising 20 dB above the test tone on 88.2 kHz-derived encodes. Undecoded playback on standard DACs produced severe downward aliasing: ultrasonic 35 kHz and 40 kHz tones embedded in the 88.2 kHz master were reflected back into the audible spectrum at 9.1 kHz and 4.1 kHz, respectively. While software unfolding restored these tones to their ultrasonic frequencies, the decoded signal exhibited a persistent band of reconstructed noise centered around 25 kHz. Contrary to claims on Bob Stuart's blog that folded data remains inaudible below -120 dB, test artifacts appeared around -60 dB and were clearly audible during listening checks. Leaky Reconstruction Filters and Hardware Artifacts (13:47) GoldenSound evaluated MQA's upsampling behavior against reference linear filters, finding that MQA utilizes an unusually slow, "leaky" minimum-phase reconstruction filter. During software unfolding to 88.2 kHz, this filter allows mirror images of musical content to reflect upward past the 22.05 kHz Nyquist cutoff into ultrasonic frequencies. Time-domain analysis of square waves demonstrated severe overshoot, non-linear pre- and post-ringing, and sharp notch distortion at transition points that worsen when passed to hardware renderers. This behavior is not restricted to MQA-branded decoding alone. In evaluating the iFi micro iDSD Diablo, GoldenSound discovered that iFi's proprietary GTO (Gibbs Transient Optimized) filter generated the identical leaky roll-off curve and ultrasonic imaging artifacts seen in MQA unfolding. A review of iFi's technical white paper revealed that the filter had been developed with direct engineering assistance from MQA. GoldenSound recommends that users wishing to play MQA files perform only the first unfold in software and rely on a conventional DAC filter for oversampling rather than using hardware MQA rendering. Deceptive Provenance Indicators and Ecosystem Monopolization (12:38) The investigation revealed that MQA's "Studio" blue-light authentication offers no guarantee of studio fidelity or original high-resolution provenance. Forum testing by user FrederikV established that up to 30 percent of an MQA file's data bits could be stripped away without extinguishing the blue indicator light. Furthermore, GoldenSound's standard 44.1 kHz master file triggered the blue light and unfolded to 352.8 kHz on a certified DAC despite possessing zero ultrasonic data, demonstrating that MQA routinely applies internal upsampling while masquerading as genuine high-resolution material. On Tidal, subscribers selecting the "HiFi" tier were served undecoded, lossy MQA streams instead of true 16-bit/44.1 kHz lossless FLAC whenever an MQA version existed in the catalog. File size comparisons also refuted MQA's bandwidth-saving claims, as the 44.1 kHz test track swelled from 30 MB in standard FLAC to 33 MB in MQA. These practices prompted major industry backlash: artist Neil Young pulled his catalog from Tidal to protest unauthorized format conversions, and manufacturers like Linn and PS Audio's Paul McGowan have publicly criticized the commercial pressure to pay MQA licensing fees. GoldenSound concludes that MQA functions as an adaptive, lossy digital signal processing format rather than a transparent, bit-perfect transmission standard. With modern broadband capable of streaming uncompressed 24-bit/192 kHz audio at under 5 Mbps, he argues that proprietary compression algorithms are obsolete and urges consumers and hardware manufacturers to resist paying licensing premiums for closed ecosystems.

精炼阅读摘要

评测机构GoldenSound通过向流媒体平台发布内嵌测试信号的原声音轨，对MQA专有音频编码展开逆向实测，证实该格式不仅属于有损压缩，还会引入可测量的背景噪声与高频混叠失真；调查同时指出，MQA的“Studio”蓝灯认证无法验证母带真实性，其封闭的授权商业模式正在以牺牲原生无损音频为代价误导消费者。实验设计与突破封闭门控 (5:47) 由于MQA属于闭源专有格式，且其授权协议严格禁止任何具备数字输出的设备进行完整硬件展开，独立研究者长期无法直接获取展开后的纯数字流来与高解析原始母带进行位比对。为绕过这一门控机制，GoldenSound自行创作了一首原声吉他音轨，并在音频中隐蔽嵌入了脉冲响应、白噪声、方波以及RMAA测试序列，分别制作了44.1 kHz标准版与包含35 kHz和40 kHz超声正弦波的88.2 kHz版本，分发至Deezer和Tidal平台。比对测试显示，Deezer分发的文件与原始44.1 kHz母带实现了100%完全一致的位完美还原，证明常规无损分发完全可行；而Tidal分发的MQA版本波形被严重修改，瞬态响应出现过冲，并在本应静音的片段中引入了高频噪声。在GoldenSound本着善意将测试数据提交给MQA寻求技术回应后，MQA并未提供合理解释，反而在次日联络发行商强行下架了其全部测试曲目并封禁了其发布权限。信号失真与高频混叠实测 (16:53) 测量数据表明MQA严重违背了高保真原则。在针对低电平线性度进行的-60 dBFS 1 kHz测试中，44.1 kHz编码文件在全频段滋生了显著底噪，高频噪声仅低于测试信号45 dB；而在88.2 kHz生成的MQA文件中，高频噪声甚至高出-60 dB测试信号约20 dB。针对商业母带（如Sam Smith的曲目）进行的DeltaWave差值比对同样证实，MQA编码会在全曲可闻频段内留下无法消除的失真痕迹。在超声保留能力测试中，未解码的44.1 kHz MQA文件未能平滑衰减超声内容，而是将88.2 kHz母带中原本位于35 kHz和40 kHz的超声测试音混叠折叠回可闻频段，分别在9.1 kHz与4.1 kHz处以高电平呈现，且人耳清晰可闻。MQA官方博客曾宣称折叠信息安全隐匿在-120 dB以下的动态范围中，但实测底噪与伪影抬升到了约-60 dB；虽然第一层软件展开能将超声频段复位，但在25 kHz附近依然残留了浓密的噪声带，并导致严重的相位失真。蓝灯认证幻象与假高解析升频 (13:47) 硬件DAC上点亮的“MQA Studio”蓝灯常被营销为“录音室母带真实性”的权威认证，但各项证据证明该指标仅为营销包装。独立论坛测试此前已证实，即便将MQA文件超过30%的有效数据位直接截断，硬件解码器依然会亮起蓝灯；GoldenSound提交的纯44.1 kHz测试文件在播放时不仅触发了蓝灯认证，更被解码器强制升频至352.8 kHz，而该文件原本根本不包含任何超声频段信息。由于Tidal原生界面隐瞒了曲目的真实母带采样率，普通用户极易将常规CD规格升频曲目误认为高解析录音。这一做法曾引发著名音乐人Neil Young的公开抗议，其因发现Tidal未经授权将原本的44.1 kHz母带转化为MQA格式而撤下了全部作品目录。此外，Tidal的HiFi订阅层对标记为Master的专辑仅提供未解码的有损MQA数据流，彻底剥夺了用户的真无损选择；原版30 MB的FLAC文件在经MQA封装后体积膨胀至33 MB，完全打破了其节省带宽的核心宣传。封闭生态垄断与官方辩解破产 (28:57) 针对测试结果，MQA官方辩称其编码器仅针对自然声学音乐设计，无法处理合成测试波形，并声称高频噪声升高是由于测试者未施加抖动（dither）导致的失误。GoldenSound反驳指出，首版提交的音轨完全包含了抖动且同样呈现出严重的噪声与过冲；同时Tidal热播前十的曲目大多包含合成器与现代数字音色，若编码器无法处理人工波形，则说明该技术本质是具有破坏性的自适应数字信号处理（DSP），而非位完美传输。针对MQA辩护其滤波特性的说法，实测表明其所谓的缓慢平滑滚降属于典型的“泄漏”（leaky）滤波器，导致音频频段向上镜像反射并产生方波严重切角与过冲，测试中发现iFi由MQA协助开发的GTO滤波器同样存在完全相同的泄漏缺陷。GoldenSound强调，现代宽带串流24-bit/192 kHz原生FLAC所需的码率不足5 Mbps，带宽节约说早已过时；包括Linn与PS Audio在内的多家高端音频制造商均对MQA的封闭授权模式表达了反感或抗拒，消费者不应为这种劣质的有损专有格式支付硬件溢价。 GoldenSound认为，音频行业的技术进步应建立在像DSD或Chord滤波算法那样可被第三方独立验证的透明基础之上，而非依靠信息封锁与不可检验的商业话术；MQA大量将标准44.1 kHz母带包装为高解析音频并强制替代真无损流，不仅未能兑现还原录音室细节的承诺，反而以垄断授权损害了创作者与消费者的权益。

Detailed notes — zh

## Background 本内容由音频评测频道 GoldenSound 于 2021 年 4 月 15 日发布在 YouTube 平台上。视频简介指出，这是一次针对 MQA（Master Quality Authenticated）专有格式及其严重缺陷的深度技术调查，核心客观测试部分从 5 分 45 秒正式展开。 ## TL;DR 为了穿透 MQA 严格闭源且限制数字输出的黑盒生态，GoldenSound 以独立音乐人身份亲自制作并向 TIDAL 发布了嵌入 RMAA、单频正弦波及全频扫频等测量信号的母带，随后通过比对流媒体返回文件揭露了其技术真相。实测数据表明，MQA 绝非无损格式，其高频折叠不仅会引入严重的底噪抬升与谐波失真，还会造成剧烈的高频混叠，且所谓的 MQA Studio 硬件蓝灯认证即便在丢失 30% 数据时依然照常点亮。更严重的是，TIDAL 在用户选择 HiFi 档位时依然强行分发未展开的 MQA 有损文件，且绝大多数 MQA 曲目仅是由 44.1 kHz 普通 CD 母带虚假升采样而来。GoldenSound 逐一戳穿了 MQA 官方事后给出的回避性公关辩解与封禁审查手段，呼吁发烧友拒绝为昂贵的专利溢价买单，共同捍卫原生无损高解析音频与开放的行业透明度。 ## Atmosphere & style 全片充斥着严密、硬核且冷静的调查记者式质感，主讲人 GoldenSound 操着沉稳从容的英音，面对封号与下架打压不仅毫无惧色，反而以条理清晰的技术图表逐个反击。叙事节奏层层递进，从自制母带遭遇阻挠的悬疑感，逐步演变为示波器与频谱仪前无可辩驳的法庭公诉，其间偶尔流露出的冷幽默更反衬出官方借口的荒谬。整段观看体验宛如见证一位技术硬核的发烧友以一人之力击碎行业巨头的商业神话，既有抽丝剥茧的理智爽感，又激荡着对音频真实保真理念的坚定捍卫。 ## Bulletin - [00:00](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=0s) 经过对亲自上传至 TIDAL 的测试曲目进行声学测量，实测证实 MQA 无论是在未展开状态还是经硬件完全解码后均非无损格式，不仅无法保留 100% 原始录音，还会引入显著的高频底噪与谐波失真。 - [16:53](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1013s) MQA 的编码与滤波机制存在严重缺陷：其泄漏滤波器会在 22 kHz 以上产生镜像反射，且未解码播放高解析曲目时会将 35 kHz 和 40 kHz 超声信号分别混叠成 9.1 kHz 与 4.1 kHz 的可闻高频杂音。 - [11:48](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=708s) DAC 设备上的“MQA Studio”蓝灯认证并不代表音频完整性或母带保真度，即使人为丢弃超过 30% 的音频数据，甚至对纯 44.1 kHz 母带进行虚假升频，该指示灯依然会照常亮起。 - [35:50](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=2150s) TIDAL 平台在曲目标记为 Master 时完全取消了真正的无损 FLAC 串流，即使用户选择 HiFi 档位，接收到的依然是包含编码杂质的未展开 MQA 文件；独立民调显示 46% 认可 MQA 的受访者中有三分之二被误导认为其优于 24-bit/192kHz 原生 FLAC。 ## Key takeaways - [00:00](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=0s) MQA 并非无损格式，无论在未解码状态还是软硬件展开之后，均无法做到与原始母带完全吻合的位完美传输。 - [03:45](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=225s) MQA 专有且闭源的商业生态严格限制授权，禁止硬件输出完整展开后的数字信号，阻碍了第三方通过纯数字比对验证其准确性。 - [05:47](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=347s) MQA 编码器内设了滤波与检测机制拦截标准测试信号，迫使测试者必须将 RMAA、方波、冲激等信号嵌入吉他原声音轨中进行实测。 - [08:49](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=529s) MQA 引入的高频噪声电平与其试图重建的超声频内容体量正相关，在 88.2 kHz 母带编码下，其超声频噪声甚至倒挂高出 -60 dB 测试信号达 20 dB。 - [11:48](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=708s) 流媒体平台上绝大多数 MQA 音源并非来自高解析度母带，而是将普通的 44.1 kHz CD 规格母带直接升采样包装而成。 - [11:48](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=708s) TIDAL 平台涉嫌剥夺用户的无损选择权：当曲目存在 Master 版本时，用户即刻丧失获取纯正 16-bit/44.1 kHz FLAC 的途径，HiFi 档位下发的也是未展开的 MQA 文件。 - [11:48](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=708s) 硬件设备上的“MQA Studio”蓝灯认证机制名存实亡，实测表明丢弃文件超过 30% 的数据位仍能正常点亮蓝灯，无法作为母带未被篡改的凭据。 - [14:31](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=871s) MQA 采用的升频数字滤波器高频滚降极其缓慢且严重“泄漏”，在软件第一层展开时便会在可闻频段上方产生明显的向上镜像伪影与非线性振铃。 - [16:53](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1013s) 高解析母带折叠机制存在灾难性混叠，在未解码设备上播放时，35 kHz 和 40 kHz 的超声频信号会直接折叠下移为 9.1 kHz 与 4.1 kHz 的刺耳可闻杂音。 - [18:07](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1087s) 在展开超声内容时，测试文件的本底噪声恶化到 -60 dB 左右，高频噪声在 20 kHz 处高达 -43 dB，完全打破了官方宣传的“隐藏在 -120 dB 人耳不可闻区间”的谎言。 - [19:54](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1194s) 在相同音频规格下，MQA 相比普通无损格式完全没有带宽与体积优势，其 44.1 kHz 版本的体积反而大于原版的无损 FLAC。 - [21:18](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1278s) 与 Chord 或 DSD 等开放并接受客观测量的音频技术不同，MQA 采取严密封闭策略，缺乏客观证据并千方百计阻挠外界独立取证。 - [23:25](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1405s) 在当今家庭宽带下流式传输 24-bit/192 kHz 原生无损所需的带宽不足 5 Mbps，MQA 以节约带宽为由推行的有损压缩在现实中毫无必要。 - [27:29](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1649s) iFi DAC 内置的 GTO 滤波器在实测中表现出与 MQA 渲染滤波器完全一致的泄漏特性与最小相位失真，技术白皮书证实其确为双方联合设计。 - [33:02](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1982s) MQA 官方将测试暴露的底噪增加归咎于用户未加抖动（Dither），但对比测试表明添加了 Dither 的版本问题同样存在甚至更加严重。 - [35:50](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=2150s) 民意调查显示 46% 认可 MQA 的用户中有高达三分之二坚信其音质优于 24-bit/192kHz 原生 FLAC，反映出虚假商业宣传对公众认知造成的严重误导。 ## Detailed walkthrough ### [00:00](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=0s) 调查缘起与发布测试曲目的波折 GoldenSound 收集了所有能够获取的证据，旨在全面检验 MQA（Master Quality Authenticated）的市场宣传，探究该格式究竟是发烧友应当追求的音质福音，还是应当极力规避的技术陷阱。为了彻底验证 MQA 的实际表现，他甚至亲自制作音乐并发布到了流媒体平台 TIDAL 上进行 MQA 编码，结果在测试过程中发现了比外界预想中更多的严重问题。在视频公开发布之前，GoldenSound 本着诚意与公正的原则主动联系了 MQA 官方，提交了自己收集到的所有测试证据，希望在视频上线前给对方一个回应与澄清的机会。然而，他等来的第一个“官方回应”并非邮件交流，而是其发布在 TIDAL 上的所有测试曲目被以极快速度全网下架；次日，MQA 方面还联系了他所使用的音乐发行商，直接封禁了他的账户并删除了全部内容。随后，他收到了 MQA 的回复邮件。正如 GoldenSound 直言： > “不幸的是，对于一家坚称自己问心无愧、毫无隐瞒的公司来说，他们却异乎寻常地急于封杀我即将展示的这类信息。” 所幸 GoldenSound 仍妥善保存着原始母带与编码后的 MQA 副本文件，测试工作并未受阻。他进一步明确了三点立场：第一，本期视频绝非针对采用 MQA 解码芯片的硬件厂商，业内有大量优秀的音频设备支持 MQA，这属于商业环境所致；第二，主观听感因人而异，部分发烧友原本就偏好失真更高的器材，客观测量数据不代表主观喜好的绝对标准；第三，在面对技术争议时应当保持理性与客观，而不是诉诸网络谩骂，MQA 官方长期以来缺乏公开透明的数据，理应由事实与测量来发声。 - **核心论证要点预告**： - MQA 并非无损格式（Not lossless）。 - MQA 会引入不必要的底噪与谐波失真，甚至在音轨完全静音的片段中同样存在。 - MQA 极少使用真正的高解析度母带作为音源。 - 所谓的“MQA Studio”蓝灯认证机制名存实亡，并未真正验证音质完整性。 --- ### [03:45](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=225s) MQA 的官方承诺与封闭黑盒生态在官方宣传中，MQA 被描绘为一种革命性的音频编码技术，号称能够“解锁原始录音母带的每一个细节”，在发烧友订阅 TIDAL 且购买支持 MQA 的 DAC 设备的前提下，提供极致的听音体验。官方声称它能保留 100% 的原始录音信息，并通过精确的时间轴特性帮助听众定位乐器与演奏者，构建出逼真的 3D 声音结像。其核心技术构想被称为“音频折叠”（Audio Origami）：将原本的高解析度录音信息折叠并打包进标准的 44.1 kHz 文件中，将原本的高频信号编码隐藏在可听频段的本底噪声之下。在标准 DAC 上播放时，它向下兼容输出常规的 20 Hz 至 20 kHz 频段；而在 MQA 解码设备上则能实现“展开”（Unfold）： ```mermaid flowchart LR accTitle: MQA 三层展开流程 accDescr: 从44.1kHz原始折叠文件通过核心解码展开到88.2kHz，再由渲染器或完整解码器展开至原始母带采样率。 subgraph S1["第一层展开（Core Decode）"] A["44.1 kHz MQA 文件"] -->|"软件解码（TIDAL / Roon 等）"| B["88.2 / 96 kHz 信号"] end subgraph S2["后续硬件展开（Rendering）"] B -->|"MQA 渲染器（Renderer）"| C["第二/三层硬件展开"] A -->|"全解码器（Full Decoder）"| C C --> D["原始高采样率输出（最高至 352.8 kHz）"] end ``` 然而，验证这套理论面临着极高的技术壁垒。MQA 属于完全专有且闭源的商业格式，不同于 MP3、DSD 或 FLAC，外界无法自行利用编码器制作测试文件。更关键的是，MQA 严格限制授权，禁止任何设备在数字接口输出完整展开后的数字音频流，只允许通过 DAC 的模拟接口输出。这种软硬件层面的层层限制，使得利用纯数字信号比对（Bit-perfect Diffing）核验其展开准确度变得异常艰难。 --- ### [05:47](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=347s) 植入测试信号与母带对比实测为了绕开 MQA 的重重封锁，GoldenSound 必须获得一组能够同时拥有“原始未编码数字母带”与“经 MQA 编码后音轨”的样本。最万无一失的途径就是以音乐人身份亲自发行专辑。最初，GoldenSound 尝试直接上传纯粹的测试信号（如 1 kHz 单音正弦波和冲激响应测试信号），但发行商立刻通知 MQA 编码器出现故障报错，无法对该文件进行编码。这证实了 MQA 在编码流程中内设了滤波或检测算法，专门拦截标准的音频测试信号。为打破限制，GoldenSound 巧妙地制作了一段轻柔的原声木吉他演奏音轨，并将丰富的声学测量测试信号嵌入其中： - 完整的 RightMark Audio Analyzer（RMAA）测试序列 - 冲激响应脉冲信号（Impulse Response） - 白噪声（White Noise） - 方波信号（Square Wave） - 32 音多音频测试（32-tone test） - 制作了标准 44.1 kHz 版本以及 88.2 kHz 高解析度“豪华版”（Deluxe），并在高频隐藏了 35 kHz 与 40 kHz 纯音，以及 20 Hz 至 44 kHz 的全频段扫频信号。随后，他将这些曲目分发至各大平台，并将各平台的流媒体输出与原始本地母带进行了严格比对。 | 音频版本与平台 | 对比结果 / 表现特征 | |:--- |:--- | | **Deezer（44.1 kHz 母带）** | 100% 位完美（Bit-perfect），波形与本地原始母带完全吻合，无任何篡改。 | | **Deezer（88.2 kHz 母带）** | 正常下采样至 44.1 kHz，切除 22 kHz 以上高频，相位保留完好，未引入多余底噪。 | | **TIDAL MQA（44.1 kHz 母带）** | 波形发生明显突变，瞬态出现过冲（Overshooting），引入高频噪声，静音段同样有底噪。 | | **TIDAL MQA（88.2 kHz 母带）** | 频谱极其混乱，静音段充斥大量高频杂音，22 kHz 以上高频信号发生严重混叠（Aliasing）。 | 在观察 44.1 kHz 的 MQA 音轨时，冲激响应测试显示出一种典型的最小相位（Minimum Phase）振铃图案，且其能量衰减并不呈线性衰减。虽然最小相位滤波器能避免前振铃（Pre-ringing），让部分瞬态听起来更加干脆利落，但其代价是破坏了相位线性——不同频率穿过滤波器的群延迟并不一致，从而破坏了依赖多重谐波同相呈现的音色（Timbre）准确性。该音轨在展开后问题更加严重，GoldenSound 只能借助 HQPlayer 内置的专用“Poly-sinc-mqa”滤波器才能勉强滤除部分噪声。而在 88.2 kHz 的高解析度测试中，MQA 更是将原本高于 22 kHz 的超声频段信号直接折叠镜像到了人耳可闻频段之内，形成了严重的混叠失真。针对实测中暴露出的波形污染与高频互调杂音，GoldenSound 断言 MQA 根本没有实现真正的无损传输。 ### [08:49](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=529s) 超声频噪声与 MQA 的重建缺陷无论原始采样率如何，MQA 文件都不是无损的；在没有使用 MQA DAC 或展开（unfolding）功能的情况下直接播放，相比流式传输纯正的无损 FLAC，会产生诸多问题。GoldenSound 提出一个假设：MQA 在 25 kHz 频段引入的额外噪声，其强度与 MQA 所需重建的超声频（ultrasonic）信号量成正比。由于测试文件中包含了覆盖至 88.2 kHz 母带全频段的白噪声，该频段的噪声极其显著。为了验证这一假设，GoldenSound 特意制作了第二版 88.2 kHz 母带——去除了全频段白噪声并大幅切除了超声频内容，然而在曲目上线前，MQA 施压下架并审查删除了他在 TIDAL 上的全部内容，导致该项直接对比未能完成。随后，GoldenSound 通过市售商业曲目验证了这一推论。他找到了 Sam Smith 的《Too Good at Goodbyes》，该曲目同时存在原生 88.2 kHz 版本以及源自 88.2 kHz 母带的 44.1 kHz MQA 版本。分析发现，44.1 kHz MQA 版本中并没有出现测试文件那样庞大的高频噪声带。然而在检查原生 88.2 kHz 母带后发现，该音轨在 22 kHz 以上原本就没有任何实际声学内容。这充分表明：MQA 引入的高频噪声水平，与其试图“折叠”和重建的超声频内容体量直接挂钩。更为诡异的是，利用 DeltaWave 软件对 Sam Smith 的两个版本进行差异比对时，除了证实 MQA 确实残留了高频噪声并非无损之外，整个音轨在清晰的人类可闻频段内，居然横贯着一条不明的失真差值带。对于高频早已被切除的 44.1 kHz 母带或“伪高解析”母带，额外引入的噪声相对较小；但对于高频真正延伸到 44 kHz 的高解析母带，MQA 编码不仅在超声区域，而是在“整首曲目的全频段”都粗暴地叠加上了一整层噪声。正如 GoldenSound 指出的那样：“MQA 或任何旨在实现高保真的方法都不应该添加噪声。”这种需要重建的高频信息越多、文件底噪就越恶化的现象，完全背离了高保真的初衷。 - 低电平信号处理能力极其糟糕：原始测试母带包含一段作为行业动态范围标准测试信号的 -60 dB 1 kHz 正弦波，原始文件没有任何多余噪声，仅受 FFT 处理增益限制。 - 动态范围大幅缩水：在 44.1 kHz MQA 文件中，高频噪声仅比 -60 dB 测试信号低 45 dB，全频谱底噪则仅低 65 dB 左右。 - 噪声倒挂现象：在基于 88.2 kHz 高解析母带编码生成的 MQA 文件中，其引入的高频噪声电平竟然比 -60 dB 的测试信号本身还要高出整整 20 dB。 --- ### [11:48](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=708s) 虚假的展开、格式强制与 Neil Young 的抗议 MQA 展开是否如其宣称的那样“100% 还原原始录音”？测试揭示了一个极其荒谬的事实：GoldenSound 提交的纯 44.1 kHz 母带，在播放时同样点亮了“MQA Studio”认证灯，并且 DAC 直接将其展开至 352.8 kHz 的采样率。这个文件从来就没有所谓的高解析母带，根本不存在任何可供“解折叠”提取的高频信息，DAC 仅仅是在进行常规的升频（up-sampling）。更严重的是，TIDAL 平台对用户完全隐藏了底层文件的真实采样率，消费者根本无法甄别哪些 MQA 音乐是真正的高解析转录，哪些只是被强行拉伸升频的普通 CD 级别文件。对于 44.1 kHz 母带而言，MQA 版本的体积甚至从原始无损 FLAC 的 30 MB 膨胀到了 33 MB，不仅牺牲了带宽，还附送了不可消除的噪音。 ```mermaid flowchart TD accTitle: TIDAL平台MQA文件分发与展开欺骗流程 accDescr: 展示用户即使选择HiFi无损音质，TIDAL依然下发体积更大的未展开MQA有损文件，并在播放时虚假升频点亮Studio蓝灯。 subgraph Source["原始母带输入"] A["44.1 kHz / 16-bit 真实母带 (30 MB)"] end subgraph TIDAL["TIDAL 服务端下发机制"] B["强行编码为 MQA 格式 (33 MB)"] C{"用户选择播放品质"} D["Master 品质: 下发 MQA 文件"] E["HiFi 品质: 依然下发未展开 MQA 文件 (非真无损 FLAC)"] end subgraph Playback["终端回放与虚假认证"] F["DAC 强行升频至 352.8 kHz"] G["点亮 'MQA Studio' 蓝灯 (虚假母带保真)"] end A --> B B --> C C -->|"选择 Master"| D C -->|"选择 HiFi"| E D --> F E --> F F --> G ``` 更恶劣的商业行为体现在流媒体的分发逻辑上。当用户在 TIDAL 客户端主动勾选“HiFi”（通常代表 16-bit/44.1 kHz 纯无损 FLAC）档位时，TIDAL 并没有提供原本纯净的无损文件，而是将包含编码杂质但未展开的 MQA 文件直接喂给了用户。GoldenSound 愤怒地直言：“我个人付费订阅 TIDAL 就是为了听无损，如果 TIDAL 连标有 Master 的曲目都不再提供真无损，那我将取消订阅，我也呼吁其他人一起退订。”他强调，即便某些人认为 MQA 有益，它也绝不应成为剥夺无损选择权的强制霸王条款。传奇摇滚音乐家 Neil Young 曾因类似事件愤而将自身全部作品从 TIDAL 下架。TIDAL 在未经 Neil Young 授权且从未获得高解析母带的前提下，擅自将其原本仅为 44.1 kHz 的母带包装成 MQA“高解析 Master”格式上架。Neil Young 在声明中对其进行了辛辣的抨击：“TIDAL 的 Master 是对原作的降级篡改，只是为了把它塞进一个收取专利版税的盒子里；而这笔钱最终全是由听众买单。我绝不支持这种行径，我要彻底退出那里，我的母带才是真正的原版。” - 认证蓝灯毫无安全性可言：Audiophile Style 论坛用户 FrederickV 进行的逆向测试表明，即使人为截断并丢弃文件超过 30% 的音频数据位（bit），DAC 上的 MQA 蓝灯依然照常亮起。 - 伪造的“录音室母带真实性”：GoldenSound 以创作者身份证明，自己提交的文件已被 MQA 改动，DAC 亮起的蓝灯根本不代表这是录音室听到或期望的声音，“那盏蓝灯不过是 MQA 的营销把戏，仅此而已。” --- ### [14:31](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=871s) 泄露的数字滤波器与高频镜像走样在深入探究 MQA 展开的核心算法时，严重的数字信号处理缺陷暴露无遗。当原始 44.1 kHz 音轨通过 TIDAL 桌面端或 Roon 软件进行第一层展开（Core Decode）至 88.2 kHz 时，频谱图上赫然出现了一道将音乐高频部分向上镜像反弹（upwards reflection / imaging）的镜像伪影。这段原本在 22 kHz 以上绝对纯净平坦的音频，硬生生被无中生有地填满了镜像高频杂质。这一现象本质上是 MQA 采用的数字升频滤波器极度“泄漏”（leaky）所致。早在 2018 年的 Rocky Mountain Audio Fest（RMAF）音响展研讨会上，Audiophile Style 创始人 Chris Connaker 就公开披露过 Archimago 绘制的这一滤波器泄漏证据。当时在场的 MQA 官方员工不仅未能给出合理解释，反而多次粗鲁打断演讲，顾左右而言他地质问 Chris“为什么不先去问问顶级 DAC 厂商 dCS 为何支持 MQA”，并试图以 Chris 未公开“Archimago”真实姓名来全盘否定测试的可信度。Chris 当场严正回击：测试者的真实姓名毫不重要，因为这个严重的镜像失真现象在任何人的家里都能够轻松被复现验证。 GoldenSound 随后利用测试曲目中注入的纯白噪声，逆向剖析了 MQA 展开滤波器的频响与阻带衰减特性，并与发烧界公认的高水准基准——HQPlayer 拥有 200 万抽头（taps）、类似 Chord M Scaler 设计的 sinc-L 线性相位滤波器进行了直观对比。对比显示，MQA 的展开滤波器高频滚降极其缓慢软弱，超声阻带衰减能力极差，甚至伴随着怪异的起伏纹理。这正是不可闻高频镜像大量漏进可闻频段的物理根源。 | 评估项目 | 原始纯净 44.1 kHz FLAC | MQA 软件第一层展开 (88.2 kHz) | MQA DAC 完全硬件解码 (Full Decode) | |:--- |:--- |:--- |:--- | | **超声镜像伪影** | 完全无 (22.05 kHz 以上严格截止) | 出现明显的向上镜像反射杂波 | 产生更多阶次的高频重影翻折 | | **方波瞬态响应** | 正常吉布斯现象/平直响应 | 出现异常严重的过冲与非线性振铃 | 过冲持续，边沿跌落切口进一步向低频扩散 | | **方波跳变沿波形** | 阶跃边缘过渡平顺连续 | 边沿过渡区出现反常的“凹槽缺口 (Notch)” | 凹槽形态更加恶化并向更深频段侵蚀 | | **可闻频带噪声** | 纯净底噪，仅受量化限制 | 持续存在抬升的高频背景噪声 | 背景高频噪声贯穿全曲且无法消除 | --- ### [16:53](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1013s) 高解析母带折叠测试：严重的频谱混叠为了验证 MQA 是否真能如宣传所言、在 88.2 kHz 真正高解析母带中精确保留并“解折叠”还原超声波信息，GoldenSound 在原始 88.2 kHz 母带中人工埋入了 35 kHz 和 40 kHz 两个高频测试纯音。这两个频点远超常规 44.1 kHz 音频格式所能容纳的 22.05 kHz 奈奎斯特极限；同时，母带中还嵌入了 RightMark Audio Analyzer（RMAA）的互调失真（IMD）和全频正弦扫频信号，频率最高覆盖至 44 kHz。当测试母带被 MQA 编码压缩成 44.1 kHz 基础文件、且在没有执行任何 MQA 展开解码的环境下播放时，灾难性的混叠（aliasing）出现了。根据奈奎斯特采样定律，未加恰当低通抗混叠滤波的 35 kHz 和 40 kHz 超高频信号，直接被镜像混叠翻折回了人耳极度敏感的可闻频带： - 35 kHz 正弦信号下移混叠成 **9.1 kHz** 刺耳纯音； - 40 kHz 正弦信号下移混叠成 **4.1 kHz** 的清晰可闻纯音。这些混叠杂音并不是藏在听觉底噪之下的微弱信号，而是电平极高、几乎没有任何衰减的真实存在。与此同时，RMAA 的连续扫频信号在越过 22 kHz 节点后，不仅没有被滤除，反而径直“掉头向下”，倒灌扫描回 20 Hz 到 20 kHz 的可闻声学空间。这意味着，任何不支持 MQA 硬件解码的用户在播放这类“折叠”高解析曲目时，耳膜都在承受着严重的可闻频段互调与混叠失真。在开启第一层软件展开后，这两个被扭曲的 35 kHz 和 40 kHz 纯音确实被翻转搬移回了超声频原始位置，RMAA 倒灌的扫频信号也完成了回正。这证实了 MQA 所宣称的“高频还原”原理在形式上确实部分运作。然而正如 GoldenSound 明确总结指出的那样，整个系统的实现水准依然破绽百出：“这表明 MQA 关于恢复高频内容的说法至少在部分上是真实的；然而，它远非完美——在围绕 25 kHz 的核心区域，我们仍然面临着一条庞大密集的噪声带。” ### [18:07](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1087s) MQA 展开后的伪影、底噪与失真问题在对 MQA 音频进行展开（Unfold）测试时，测试文件的低频区域留下了大量额外的伪影，导致本底噪声直接飙升到约 -60 dB。这一实测结果与 MQA 官方的宣传大相径庭。在 MQA 的“Bob Talks”官方博客中，作者声称高频内容会被折叠并隐藏在 -120 dB 以下的区域，因而绝对不会被人耳听到。然而实测文件直接推翻了这一说法：高频信号几乎完全没有经过实质性衰减，较强电平的信号是完全可以被人耳感知到的。GoldenSound 亲自进行了听音验证，明确指出自己能够非常清晰地听到这些原本不应出现的测试信号音。这意味着无论是在人耳可闻的音频频段，还是在超声波频段，当 MQA 尝试展开超声波内容时，其实际动态范围都变得极度糟糕。即使完成完整的全解码（Full Decode），这些由于混叠产生的失真和缺陷依然没有消失；文件中的信号只是被多次向上镜像反射到了超声波频段。针对最初提出的测试核心问题，结论是清晰的：在技术层面，MQA 展开后确实“还原”了高频内容，但它是以急剧恶化的精度、严重受损的动态范围以及不可忽视的噪声为代价的。 > “It does not by any stretch retain 100% of the original recording, and this claim is false.”（无论从任何角度看，它都绝没有保留 100% 的原始录音，这项官方宣传完全是虚假的。）这种做法在技术上很难称得上是重大成就。对于 44.1 kHz 采样率的文件而言，由于超声波内容只是简单地以高电平被混叠并向下反射回可闻频段，处理后的音频几乎失去了实用价值。它根本没有像官方宣称的那样被隐藏在人耳不可闻的极低电平区域；只要原始录音中含有丰富的高频成分，MQA 的这种折叠机制就会在听感上产生极其明显且严重的听觉缺陷。 ### [19:54](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1194s) 测试结论速览：MQA 的实测真相完成一系列客观技术测试后，GoldenSound 快速梳理并总结了本次实测确立的几项硬核结论： - **非无损格式**：无论是在未经解码的原始状态，还是在展开解码之后，MQA 都不是无损音频。 - **大量假高解析度**：绝大多数 MQA 发行版与原始母带采样率无关，它们仅仅是将 44.1 kHz 的文件进行了升采样（Up-sampling），并非源自真正的高解析度母带。 - **高频严重混叠**：MQA 会将高频信号在极少衰减的情况下，直接混叠混入可闻音频范围。 - **展开后遗留噪声**：展开过程虽然补回了高频并移除了大部分混叠信号，但却在音频中留下了可观的底噪与谐波失真。 - **Tidal 抛弃真无损**：在 Tidal 平台上，只要曲目标记为“Master”，就再也找不到真正的无损串流，用户只能得到有损压缩或有损的 MQA。要听真正的无损，只能转向 Qobuz 或 Deezer 等竞争平台。 - **认证蓝灯毫无意义**：MQA 的认证机制和硬件上的“蓝灯”并不保证源文件的采样率、文件完整性或未被篡改，更不代表听到的声音与录音室现场一致。 - **体积毫无优势**：在相同规格下，MQA 相较于普通无损没有任何文件体积优势；MQA 文件不仅比 44.1 kHz FLAC 更大，仅仅比原生高解析度 FLAC 稍小一些。 ```mermaid flowchart TD accTitle: MQA 实测处理流程与缺陷 accDescr: 展示高频音频信号经过 MQA 折叠、下采样、传输至最终展开后产生的混叠与底噪问题 subgraph Fold ["编码与折叠阶段"] A["高频/超声波原始信号"] --> B["高频混叠并折叠至可闻频段"] B --> C["高电平伪影与噪声 (-60 dB 底噪)"] end subgraph Dist ["平台分发 (如 Tidal)"] C --> D["封装为 MQA 容器 (非真正无损 FLAC)"] D --> E["点亮硬件认证蓝灯 (无真实保真度担保)"] end subgraph Unfold ["终端解码与展开"] E --> F["软件展开 (第一层展开)"] F --> G["全硬件解码 (镜像反射至超声频段)"] G --> H["动态范围劣化 / 遗留不可逆失真"] end ``` 如果使用 Roon 播放器，用户可以清晰识别 MQA 文件的原始母带采样率：曲目描述中显示的 MQA 采样率最后一段即为原始母带规格。只要该数值显示为 44.1 kHz，就证明该专辑完全只是在进行升采样处理。 ### [21:18](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1278s) 音响发烧友的审视：开放测试与“蛇油”之别在阐明客观技术事实后，GoldenSound 分享了自己作为一名纯粹音响发烧友的主观立场与评价。他热爱音频硬件与音乐，也乐于见到能真正提升听音体验的技术突破。然而音响圈向来充斥着各种“蛇油”（虚假宣传）产品。在这个领域中，产品通常可以划分为截然不同的两类：一类是提供客观证据或允许第三方独立验证的，另一类则是严密封闭、拒绝接受检验的。例如 Chord Electronics 的 DAC 设计师 Rob Watts，他高调宣扬自己的 WTA 滤波器是重放 44.1 kHz 音频的最佳方案。即便有人不同意其观点，但任何拥有 Chord DAC 或 M Scaler 的玩家都可以自由播放、录音并测量任何测试信号。Rob Watts 本人也活跃在 Head-Fi 等发烧友论坛，详细撰写技术文章解释其架构。他虽然不会公开滤波器的核心底层系数，但给出了充足的数据让消费者做出明智的购买判断。同样的道理适用于 DSD 格式——它是开放的，任何人都可以编码、播放并深入剖析其优缺点。甚至在最容易被斥为“蛇油”的线材领域，客观测量依然可行。例如对比普通 BNC 线与 AudioQuest Carbon BNC 线时，能通过仪器测出更低的时基抖动（Jitter）。无论这种差异在主观上是否能听出来、是否物有所值，至少其营销主张是可以被实测验证或证伪的。然而 MQA 采取了完全相反的姿态： > “Not only has MQA not provided their own evidence, they prevent you from gathering your own.”（MQA 不仅从未拿出他们自己的证据，还千方百计阻止你自己去获取证据。）普通用户无法自由编码 MQA 文件，硬件也禁止录制完整解码后的最终纯数字输出。MQA 的商业模式完全建立在“相信我们，这声音更好”的话术之上。一旦旧的宣传说辞被技术社区拆穿，他们的公关口径就会立刻变得模糊并不断调整。 ### [23:25](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1405s) MQA 的公关迷局与行业垄断危机不可否认，MQA 曾展示过一些引人入胜的概念，例如针对老旧录音设备的缺陷进行针对性校正的“白手套”（White Glove）处理，这类技术本身非常惊艳（推荐阅读相关的《Fairy Tales》技术文章）。然而，MQA 官方却通过精心包装的宣传话术，诱导大众误以为所有 MQA 核心产品都是这样运作的。他们向公众暗示所有的 MQA 发行版都超越原生 FLAC，但现实中绝大多数曲目只是由普通 44.1 kHz 文件升采样包装而成。在 Bob Stuart 的技术博客中，他解释高频折叠机制的方式在 16-bit 音频中根本无法实现，因为 16-bit 根本没有足够的动态范围来容纳那些折叠信息。每当技术圈提出具体质疑时，MQA 总是顾左右而言他，回避实质问题。在技术硬伤不断暴露的同时，MQA 却在产业端大肆扩张，全面蚕食流媒体目录与硬件生态。即便退一步假设 MQA 真的完美无缺、做到了 100% 无损，它也绝不应该成为市场上唯一的音频选项。消费者理应拥有自主选择权，能够自由在原生高解析度文件、MQA 以及其他格式之间做出抉择。目前令人担忧的现实是，某些真正拥有高采样率母带的唱片，在流媒体上只提供 MQA 版本，彻底剥夺了原生高解析格式的存在空间。这种对高解析音频市场的垄断趋势对消费者极为不利。在网络传输带宽方面，传输 24-bit/192 kHz 的无损音频所需的带宽不足 5 Mbps。在发达国家绝大多数家庭宽带环境下，只要能够流畅观看 YouTube 或 Netflix 视频，就绝对有能力毫无压力地串流 24/192 原生无损音频。在带宽充裕的今天，MQA 本质上是一种被强加给用户的昂贵有损压缩机制。硬件制造商之所以纷纷支持 MQA，绝非因为工程师们真心认为它优于原生高解析音频，而完全是因为消费者被营销驱动产生了购买需求。包括 Linn 以及自身 DAC 产品支持 MQA 的 PS Audio 在内的诸多厂商，都曾公开表达对 MQA 的反对。PS Audio 的创始人 Paul 在视频中坦言，加入 MQA 完全是因为大量客户的指名要求；这纯粹是一个商业妥协，否则公司就会流失订单。撇开技术层面的恶劣表现不谈，MQA 在营销传播层面无疑展示了颠倒黑白的操盘范例。面对这种剥夺选择权且破坏音质的格式，GoldenSound 呼吁发烧友用钱包投票： - 拒绝购买捆绑了 MQA 高昂授权费的音频硬件。 - 拒绝订阅像 Tidal 这样下架真实无损音源、强制向用户推送 MQA 的平台。 - 坚决要求音频企业保持透明与诚信，允许第三方进行独立客观测试。最后，GoldenSound 补充了一个实用的技术建议：MQA 的第一层展开（First Unfold，通常展开至 88.2/96 kHz）机制不同于硬件执行的第二层与第三层渲染展开。如果发烧友在某些场景下别无选择、必须播放 MQA 文件，折中的最佳方案是在播放软件端（如 Roon 或 Audirvana）执行第一层核心展开，而避免使用全硬件渲染，以尽可能减少后续滤波带来的更严重干扰。 ### [27:29](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1649s) 绕过 MQA 渲染器与 iFi GTO 滤波器的惊人内幕针对 MQA 解码造成的混叠与高频镜像问题，GoldenSound 给出了一个折中的播放建议：如果 DAC 允许，可以在完成第一层核心解码展开（Core Decode）后，关闭 DAC 的后续 MQA 渲染（Rendering）功能，改用 DAC 自带的常规内部滤波器进行过采样或升频。这种做法不仅能够消除绝大部分可闻频段内的混叠失真，还可以滤除不需要的高频镜像以及一部分伴随的高频噪声。在本次视频测试中，GoldenSound 使用的硬件是具备完整 MQA 解码能力的 iFi iDSD Diablo。他高度赞扬了这款产品本身的硬件素质，但在分析其内置的 GTO（Gibbs Transient Optimised）滤波器响应时，意外发现了令人玩味的细节： ```mermaid flowchart TD accTitle: iFi GTO 滤波器与 MQA 渲染滤波器的关联 accDescr: 展示 iFi GTO 滤波器与 MQA 滤波特性高度一致，及其引发的音质缺陷与恢复建议。 subgraph Comparison["滤波器频响与特性比对"] MQA["MQA 渲染滤波器 (Renderer Filter)"] GTO["iFi GTO 滤波器"] MQA <-->|"曲线与 leaky 特性完全一致"| GTO end subgraph Defects["GTO 产生的负面影响"] GTO --> Leak["严重泄漏与高频镜像 (Imaging)"] GTO --> MinPhase["最小相位失真 (影响音色等相位相关特性)"] end subgraph Solution["建议应对策略"] Stock["降级至固件 7.0 / 切换回原生标准滤波器"] Advise["向厂商反馈：拒绝为 MQA 捆绑授权买单"] MinPhase -.-> Stock end ``` 通过实测比对，iFi GTO 滤波器的频响曲线与 MQA 渲染滤波器的曲线近乎完全吻合。两者均表现出相同的“泄漏（leaky）”特性并引发高频镜像，且同属最小相位（Minimum Phase）设计，会破坏音色等与相位敏感的声音特质。随后查阅 iFi 官方关于 GTO 滤波器的技术白皮书更坐实了这一点——该滤波器正是在 MQA 的协助下联合设计的。GoldenSound 强烈建议 iFi DAC 用户刷回 7.0 固件并改用默认的 stock 滤波器。他借此重申立场：产品支持 MQA 并不代表硬件本身糟糕，但发烧友应当主动发声，告诉设备制造商用户不需要捆绑 MQA，也不想为此支付额外的授权溢价。 - “I would strongly recommend that if you have an iFi DAC you use the stock filter, not GTO.”（GoldenSound 强烈建议拥有 iFi DAC 的用户使用出厂默认滤波器，而非 GTO。） - “Tell manufacturers you don't want to pay extra for it; tell them that you want their products without MQA added on.”（GoldenSound 呼吁用户告诉厂商不想为 MQA 多付钱，希望购买不加装 MQA 的纯粹产品。） --- ### [28:57](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1737s) MQA 官方回应的逻辑漏洞：制作端补偿与缺乏透明度在视频上线前，MQA 官方针对 GoldenSound 的测试发来了正式回复。MQA 在回应开篇大谈其技术比传统方法先进，声称该算法能对信号链两端（录音端与播放端）的转换失真进行补偿修正。然而 GoldenSound 指出这一说法在现实中根本站不住脚：他是完全按照 MQA 推荐给独立音乐人的标准流程上传音频并完成 MQA 转码的，整个流程中没有任何人询问过他录音或混音时使用了哪款硬件或转换器；此外，测试全程均在官方认证的 MQA 软硬件环境下完成。更深层的破绽在于现代音乐制作的工作流机制：现今极少有曲目是纯模拟录制后仅经由单一型号 ADC 数字化完成的。一首歌往往混杂了来自不同声卡、合成器、采样器以及各种数字音频工作站处理手段的声音，根本无法用单一的“端到端硬件补偿”来囊括。面对 MQA 声称“无法理解用户为何会对评估 MQA 感到沮丧”，GoldenSound 直接驳斥了这种避重就轻的态度： - **端到端分析工具的缺失**：不同于市面上几乎所有音频格式，用户根本拿不到官方分析工具来客观检验 MQA 到底对音频做了何种修改，无法排除安慰剂效应，也无法验证它是否与原始母带一致。 - **面对质疑的敌对姿态**：当业界专家 Chris Connaker 提供确凿测试证据时，MQA 顾左右而言他并转移话题；当独立评测人联系他们时，甚至试图通过下架其作品来施行审查与噤声。 - “That is my frustration: it isn't transparent in the slightest, and MQA responds in a hostile manner whenever criticized.”（GoldenSound 直言：“这就是我的挫败感所在：它毫无透明度可言，而且每当面对批评时，MQA 总是一副敌对姿态。”） --- ### [30:42](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1842s) 不自然的音频限制、黑盒工作室工具与可笑的蓝灯认证 MQA 在回应中辩解称，其编码器并未设计用于处理“曲中统计特性剧烈变化”或“听起来不像自然声学声音（does not resemble natural sound）”的内容。GoldenSound 质问，这是否变相承认了 MQA 根本无法正确处理电子音乐，或是结合了真实乐器与合成音色的人声/配乐曲目？现实是 Tidal 排行榜前十名的热门曲目全非纯模拟原生录音，其中却有 8 首打着 MQA 标识。 ```mermaid flowchart LR accTitle: MQA 黑盒环境与专业透明工具对比 accDescr: 对比 MQA 缺乏透明度的黑盒工具与 HQPlayer 开放评估版本的行业实践。 subgraph MQA["MQA 现状：完全黑盒"] T1["声称有未公开的 Studio Tools"] --> T2["外界不知谁能使用、何种曲目使用"] T2 --> T3["声称有条件无损，拒绝给出严格定义"] end subgraph Industry["行业透明范例：Signalyst HQPlayer"] H1["提供专业完整功能评估版本"] --> H2["通过限制输出时长为 60 秒防商业滥用"] H2 --> H3["公众可自由分析算法真实表现"] end ``` 针对 MQA 所谓的“工作室专属调校工具”，GoldenSound 建议官方不妨借鉴 Signalyst HQPlayer 的做法——哪怕推出一个限制音频长度为 10 秒的评估版本供公众与分析师实测，既不损害商业利益，又能证明清白。在 DAC 上的“蓝灯（Blue Light）”最高级别母带认证问题上，MQA 在回复中竟将责任完全甩给上传者，声称“责任在于提交者在音乐上线 Tidal 后去自行监听并确认声音”。这种推脱此前从未在任何推广材料或上架流程中提及。更荒谬的是，实测早已证明就算把 MQA 文件强行丢弃三分之一的数据，硬件上的蓝灯依然会照常亮起；甚至连提交的 44.1kHz 文件在经过 MQA 打包后，无论是主观听感还是客观数据，都与 GoldenSound 提交的原版母带完全不同。 - “Does that then imply that electronic music won't work with MQA, or that tracks with a combination of natural recording and synthesized parts won't work?”（GoldenSound 质疑道：“这是否意味着电子音乐在 MQA 下根本行不通，或者结合了自然录音与合成器音色的曲目也无法正常工作？”） - “The onus is on the submitter to check the content when it arrives on Tidal and confirm the sound... and it also doesn't address the fact that you can make an MQA file, throw away a third of it, and still have that light show up.”（GoldenSound 反驳道：“在文件上线 Tidal 后去确认声音的责任全在提交者……但这根本无法掩盖事实：你制作一个 MQA 文件，扔掉三分之一的内容，那盏蓝灯却依然会亮起。”） - “It might sound more natural to some people, but sounding natural and being lossless to the original master are not the same thing.”（GoldenSound 强调：“在某些人听来它也许更自然，但‘听起来自然’和‘对原始母带保持无损’根本就不是一回事。”） --- ### [33:02](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1982s) 逐条戳穿官方借口：抖动、泄漏滤波与高频噪声的可闻性在技术细节层面，MQA 对 GoldenSound 指出的各项缺陷进行了逐一抗辩，但均被 GoldenSound 依据严格的控制变量实验逐条击破： ```mermaid flowchart TD accTitle: MQA 抗辩借口与实测事实反驳 accDescr: 对比 MQA 提出的多项抗辩理由与 GoldenSound 实际测试中记录的客观技术事实。 subgraph DitherDebate["争议点 1：底噪增加归咎于未加抖动？"] MQA_D["MQA：用户未做 Dither 属于低级错误"] GS_D["实测：第一版专门加了 Dither，问题更严重；后续为排除变量才发无 Dither 版"] MQA_D <--> GS_D end subgraph FilterDebate["争议点 2：泄漏滤波器与高频混叠"] MQA_F["MQA：Leaky 是贬义词，除了陡峭滤波别无选择"] GS_F["实测：真正的高解析（如 96kHz）频宽达 13 倍裕量，根本无需泄露滤波妥协"] MQA_F <--> GS_F end subgraph NoiseDebate["争议点 3：失真与噪声是不可闻的高频抖动？"] MQA_N["MQA：无额外失真，噪声仅集中在 25kHz 超声区"] GS_N["实测：噪声直接侵入 20kHz 可闻带，高达 -43dB，完全处于人类可闻范围"] MQA_N <--> GS_N end ``` 针对 MQA 声称的“系统报错拦截”，GoldenSound 证实除了纯冲激响应或方波等极端合成测试信号外，他从未收到过其他警告信息。针对 MQA 称其“未添加抖动（Dither）是低级错误导致底噪”的指责，GoldenSound 反击称自己第一首提交的音轨《Try Again》正是严格添加了 Dither 的，结果所有缺陷不仅存在甚至更恶化；为了控制变量、防止抖动算法干扰，他才在后续曲目中去除了 Dither，犯下盲目推断错误的恰恰是未仔细核对三套文件的 MQA 官方。对于 MQA 抱怨“泄漏（Leaky）”一词带有贬损性且声称陡峭砖墙滤波（Brick-wall Filter）是唯一替代品，GoldenSound 调侃道“滤波器本身可没有感情，如果伤到了它的感情我深表歉意”。但从数字信号处理角度看，原生 96kHz 高解析音频在可闻上限与奈奎斯特频率（Nyquist frequency）之间留出了比 44.1kHz 大 13 倍以上的充裕过渡带，完全允许采用温和且无振铃的线性滤波，根本无需引入 MQA 这种引发镜像混叠的残缺滤波架构。关于高频噪声，MQA 坚称那是集中在 25kHz 的不可闻超声抖动，但测试数据明确显示其噪声裙边已经一路侵入到 20kHz 可闻频带内，并在 20kHz 处达到了高达 -43 dB 的惊人电平——这无论在理论还是实际听感上都是完全不可忽视的杂音。 - “Firstly, I don't think the filter itself has feelings, but if I have hurt them, then I apologize.”（GoldenSound 幽默地反唇相讥：“首先，我不认为滤波器本身会有感情，但如果我真的伤害到了它的感情，那我道歉。”） - “Going from 44.1 kHz to 96 kHz gives you over 13 times more distance between the audible band and the Nyquist frequency... the solution isn't MQA, it's to use native high-res.”（GoldenSound 论证道：“从 44.1kHz 提升到 96kHz，在可闻频带与奈奎斯特频率之间留出了超过 13 倍的缓冲间隔……解决方案绝非 MQA，而是使用真正的原生高解析音频。”） - “It extends down into the audible band and is at -43 dB at 20 kHz. That is audible.”（GoldenSound 明确指认：“它延伸进了可闻频段，并且在 20kHz 处达到了 -43dB。这绝对是听得见的。”） --- ### [35:50](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=2150s) 缺乏透明度的曲目信息与被误导的消费者民调在信息呈现方面，MQA 宣称其编码在回放时会明示原始母带的采样率。GoldenSound 指出，普通消费者在最主要的载体 Tidal 官方应用中根本看不到任何母带采样率的具体参数，除非借助像 Roon 这样昂贵的专业第三方播放软件。这种遮遮掩掩的信息展示配合铺天盖地的商业宣传，直接让绝大多数终端消费者误以为只要戴上“MQA”光环，音质就等同甚至优于原生的 24-bit/192kHz 无损 FLAC。为了验证公众究竟被误导到了何种程度，GoldenSound 专门进行了一项收集了超过 240 份样本的独立民意调查。调查数据揭示了惊人的认知偏差： - 在所有受访者中，有 **46%** 的人认为 MQA 是一项优秀的正面技术。 - 在这部分认可 MQA 的人群中，竟有整整 **三分之二（约 66.7%）** 的人坚信“MQA 发行版本的音质等同于或超越了 24-bit 192kHz 的无损 FLAC”。这种基于虚假营销所建立起来的庞大认知错觉，正是 GoldenSound 认为 MQA 对整个音频行业造成严重危害的核心症结所在。 - “Tidal shows no indication at all of original sample rate; you have to use a program like Roon to show this information.”（GoldenSound 披露道：“Tidal 上根本没有任何关于原始母带采样率的标识；你必须借助于 Roon 这样的独立软件才能查到这些数据。”） - “Of the people that said MQA was a good thing, two-thirds of them believed that MQA releases were the same or better than 24-bit 192 kHz FLAC. This is why I think it's a problem.”（GoldenSound 揭示调查数据并总结道：“在那些认为 MQA 是好东西的人里，有三分之二的人坚信 MQA 发行版与 24-bit 192kHz FLAC 相同甚至更好。这就是为什么我认为它是一个严重的问题。”） ### [36:38](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=2198s) 总结 MQA 的核心误区与行业透明度缺失 GoldenSound 在全片结尾针对 MQA 的核心宣传逻辑进行了严厉的总结，强调行业必须追求更高的透明度。即使在假设的情境下——哪怕一段 24-bit/192kHz 的录音在折叠为 44.1kHz 之后，听感上真的能与原生 192kHz 录音完全一致，MQA 的商业现实也依然站不住脚。现实中 Tidal 等平台上绝大多数的 MQA 唱片根本不是用高规格母带折叠而来，而是直接来自 44.1kHz 的 CD 规格母带；这些文件无论如何都不可能等同于、更不可能优于原生的高解析度 FLAC。GoldenSound 强调这种由宣传制造出来的广泛认知偏差是非常令人担忧的： > “Even if a 24-bit 192 kHz recording folded down to 44.1 kHz was absolutely identical to a native 192 kHz recording, the vast number of MQA releases that are 44.1 kHz from 44.1 kHz masters are not the same or better than native hi-res FLAC, and this misconception is concerning.” - **折叠假设的破灭**：即便“音频折叠”能做到无损还原高频，现实中大量打着 MQA 标签的音源实际只是普通 44.1kHz CD 母带套上 MQA 编码，根本无法无中生有产生真正的高解析度细节。 - **市场误导性**：流媒体宣传让大量用户误以为只要看到 MQA 认证或展开标志，就获得了超越原生无损 FLAC 的听觉体验，这严重混淆了音频质量与专有编码包装的区别。 ### [37:02](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=2222s) 致谢贡献者、社区测试者与支持者在完成整套测试与结论陈述后，GoldenSound 向协助完成本次调查与视频制作的社区技术先驱和朋友致以谢意。他特别感谢了 Daniel Mellinger 以及 Amos（即音频发烧圈内熟知的 Currawong）。Amos 整理了一份涵盖大量针对 MQA 争议的深度文章与技术证据清单，其中许多无法在视频中完全展开的素材都已被收录在视频简介的参考链接中。此外，GoldenSound 表达了对音频技术社区资深研究者的敬意，感谢 Archimago、John Atkinson、Miska、Måns Rullgård 以及 Frederik V 所提供的测试工具、测量数据与公开技术资料。同时，他也向参与 MQA 调查问卷的 250 位受访者表达了感谢。在视频的最后，GoldenSound 感谢了所有赞助频道的 Patreon 会员，特别是处于最高支持梯队的 Diamond 与 Legend 级别赞助者，包括 Chris Ross、Kyle Gravitas Crack、Daniel Mellinger、King Jong-un 以及 Lana Bennett，并欢迎感兴趣的观众加入他的公开 Discord 社区或 Patreon 专属 Telegram 群组参与进一步的交流探讨。 ## Notable quotes > [11:48](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=708s) TIDAL 的 Master 是对原作的降级篡改，只是为了把它塞进一个收取专利版税的盒子里；而这笔钱最终全是由听众买单。我绝不支持这种行径，我要彻底退出那里，我的母带才是真正的原版。 — Neil Young > [21:18](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1278s) Not only has MQA not provided their own evidence, they prevent you from gathering your own. — GoldenSound > [30:42](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1842s) It might sound more natural to some people, but sounding natural and being lossless to the original master are not the same thing. — GoldenSound > [33:02](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1982s) Going from 44.1 kHz to 96 kHz gives you over 13 times more distance between the audible band and the Nyquist frequency... the solution isn't MQA, it's to use native high-res. — GoldenSound > [27:29](https://www.youtube.com/watch?v=pRjsu9-Vznc&t=1649s) Tell manufacturers you don't want to pay extra for it; tell them that you want their products without MQA added on. — GoldenSound ## Mentioned GoldenSound, MQA, TIDAL, Deezer, Qobuz, Roon, Neil Young, Chris Connaker, Archimago, Rob Watts, Chord Electronics, Chord M Scaler, iFi iDSD Diablo, iFi GTO Filter, HQPlayer, DeltaWave, RightMark Audio Analyzer (RMAA), Sam Smith, Too Good at Goodbyes, PS Audio, Linn, dCS, AudioQuest, Daniel Mellinger, Amos (Currawong), John Atkinson, Miska, Måns Rullgård, Frederik V ## Video description A deep dive into MQA, and its problems. Testing starts at 5:45 Come chat on discord: https://www.thegoldenone.co.uk Support me on patreon and join the private telegram chat: https://www.patreon.com/goldensound (or subscribestar) https://www.subscribestar.com/goldensound How to test MQA's leaky filter: https://audiophilestyle.com/forums/topic/62669-how-to-demonstrate-mqas-leaky-filter/ 2018 RMAF MQA talk by Chris Connaker: https://www.youtube.com/watch?v=NSv0lcHlawk Archimago's MQA article: https://audiophilestyle.com/ca/reviews/mqa-a-review-of-controversies-concerns-and-cautions-r701/ PS Audio vid on MQA: https://www.youtube.com/watch?v=lPfmWKjiccA Schiit on MQA: https://www.schiit.com/news/news/why-we-wont-be-supporting-mqa#:~:text=Today%2C%20Schiit%20Audio%20announced%20that,so%20that%20they%20may%20choose Linn on MQA: https://web.archive.org/web/20201111211105/https://www.linn.co.uk/blog/mqa-is-bad-for-music Neil Young removes albums from TIDAL: https://neilyoungarchives.com/news/1/article?id=Tidal-Misleading-Listeners Further reading on MQA: GTO filter analysis: https://audiophilestyle.com/forums/topic/55951-ifi-foe-fum/#comments Benchmark Audio's statement about MQA transient shifting: https://www.theabsolutesound.com/articles/benchmark-media-systems-la4-line-amplifier-and-dac3-b-digital-to-analog-converter RealHD stops supporting MQA: https://www.realhd-audio.com/?p=7218 Archimago on transient smearing https://archimago.blogspot.com/2018/02/musingsmeasurements-on-blurring-and-why.html Rob Watts on digital filters: https://www.head-fi.org/threads/watts-up.800264/page-134#post-15968970 Benchmark on MQA: https://benchmarkmedia.com/blogs/application_notes/163302855-is-mqa-doa Meridian has doctored tests of hi-res in the past: https://audioinvestigations.blogspot.com/2015/07/the-great-debate-amir-vs-arny.html Filter ringing (with a link to Mans/troll audio) https://audiophilestyle.com/forums/topic/30381-mqa-is-vaporware/page/699/?tab=comments#comment-983918 Aliasing vs. imaging: https://audiophilestyle.com/forums/topic/30381-mqa-is-vaporware/page/683/?tab=comments#comment-980885 Music aliasing distortion example: https://audiophilestyle.com/forums/topic/30381-mqa-is-vaporware/page/683/?tab=comments#comment-980857 Linear phase vs. Minimum phase: https://troll-audio.com/articles/linear-and-minimum-phase/ Sox settings to duplicate an MQA filter: https://audiophilestyle.com/forums/topic/37821-stereophile-series-on-mqa-technology/page/7/?tab=comments#comment-754014 JA's post showing signal aliasing: https://audiophilestyle.com/forums/topic/30381-mqa-is-vaporware/?do=findComment&comment=980401 mansr's original Dragonfly DAC analysis: https://audiophilestyle.com/forums/topic/30572-mqa-technical-analysis/page/20/?tab=comments#comment-690307 Why MQA's digital filters blur the music: https://audiophilestyle.com/forums/topic/30381-mqa-is-vaporware/page/470/?tab=comments#comment-936690 A Survey of Musical Instrument Spectra to 102.4 KHz https://www.cco.caltech.edu/~boyk/spectra/spectra.htm RealHD Audio on MQA: https://www.realhd-audio.com/?p=6517 Bob Stuart wanted to censor AS: https://audiophilestyle.com/forums/topic/30381-mqa-is-vaporware/page/584/?tab=comments#comment-958011 References in MQA's response: [1] 'A Hierarchical Approach for Audio Archive and Distribution', JAES Volume 67 Issue 5 pp. 258-277; May 2019. http://www.aes.org/e-lib/browse.cfm?elib=20456 [2] 'The Gentle Art of Dithering', JAES Volume 67 Issue 5 pp. 278-299; May 2019. http://www.aes.org/e‐lib/browse.cfm?elib=20457 [3] MQA Q&A Sidebar 1: https://www.stereophile.com/content/mqa‐questions‐and‐answers‐sidebar‐1‐ example‐nielsen‐2l‐120‐track‐1 [4] MQA Q&A Sidebar 2: https://www.stereophile.com/content/mqa‐questions‐and‐answers‐sidebar‐2‐ example‐portland‐state‐chamber‐choir [5] MQA Sidebar 3: https://www.stereophile.com/content/mqa‐questions‐and‐answers‐sidebar‐3‐example‐ portland‐state‐amazing‐grace‐audibility‐analysis [6] Bobtalks: MQA CD https://bobtalks.co.uk/blog/mqaplayback/origami‐and‐the‐last‐mile/ 0:00 - Intro 1:15 - Summary of topics 1:47 - Pre-video clarifications 3:44 - What is MQA? 5:41 - How I got test files encoded in MQA 6:57 - MQA 44.1khz vs Native 44.1khz 8:14 - MQA HiRes vs Native 88.2khz 11:11 - Reduced Dynamic Range 11:47 - Does unfolding work? (Many MQA releases are just upsampled 44.1khz) 12:41 - No Lossless on tidal if track is MQA 13:47 - MQA "Authentication" means nothing 14:30 - MQA Upsampling filter is leaky 15:33 - Filter attenuation analysis 16:54 - HiRes unfolding analysis 18:53 - Full decode analysis 19:54 - Conclusion 21:19 - My Opinion, and testing cables 25:46 - Why are manufacturers supporting MQA? 27:13 - Best way to play MQA files 27:44 - ifi GTO filter is MQA 28:56 - MQA's response 37:02 - Credits/Thank you The GoldenSound youtube channel, and all content is created and owned by GoldenSound LTD [Source](https://www.youtube.com/watch?v=pRjsu9-Vznc)

Source quotes

[11:48] Neil Young: TIDAL 的 Master 是对原作的降级篡改，只是为了把它塞进一个收取专利版税的盒子里；而这笔钱最终全是由听众买单。我绝不支持这种行径，我要彻底退出那里，我的母带才是真正的原版。 [21:18] GoldenSound: Not only has MQA not provided their own evidence, they prevent you from gathering your own. [30:42] GoldenSound: It might sound more natural to some people, but sounding natural and being lossless to the original master are not the same thing. [33:02] GoldenSound: Going from 44.1 kHz to 96 kHz gives you over 13 times more distance between the audible band and the Nyquist frequency... the solution isn't MQA, it's to use native high-res. [27:29] GoldenSound: Tell manufacturers you don't want to pay extra for it; tell them that you want their products without MQA added on.

Source quotes

[12:38] GoldenSound: Even if MQA is a benefit in some cases, it should be a choice; it should not be mandatory. Speak with your wallet and demand lossless. [12:38] Neil Young: TIDAL's master is a degradation of the original to make it fit in a box that collects royalties. That money is ultimately paid by the listeners. [24:23] GoldenSound: If a company can't provide its own evidence and goes to great lengths to prevent anyone else from testing it, that should probably raise quite a few red flags. [8:49] GoldenSound: MQA or any method aimed at high fidelity shouldn't be adding noise. [36:38] GoldenSound: The vast number of MQA releases that are 44.1 kilohertz from 44.1 kilohertz masters are not the same or better than native high-res FLAC, and this misconception is concerning.

Original transcript

[0:00] in this video i've put together every [0:01] piece of evidence i could to test many [0:03] of mqa's marketing claims and see if [0:05] it's something that audiophiles should [0:06] want [0:07] or avoid i even got my own music [0:09] published on title in mqa [0:11] just to prove to you how troublesome it [0:13] really is and i've discovered a few [0:15] things along the way that indicate mqa [0:16] has more problems than we originally [0:18] thought [0:19] i did reach out to mqa whilst making [0:20] this video in order to present the [0:22] evidence that i'd collected and give [0:23] them a chance to respond before the [0:25] video went live [0:26] the first response i got was actually [0:28] having all of my music removed from [0:30] tidal the next day i received an email [0:32] back from mqa [0:33] and i'll discuss their response at the [0:34] end of the video mqa's marketing has [0:37] been [0:37] vague at best and it's changed over the [0:39] years there's a lack of definitive [0:40] evidence about it [0:41] additionally due to some only partially [0:43] related information being released such [0:45] as the white glove service correcting [0:46] for recording gear imperfections [0:48] which has nothing to do with mqa's [0:50] normal releases lots of people have [0:52] formed their own different ideas as to [0:53] what mqa is [0:55] what it does what it improves and how it [0:57] works i will be sticking to mqa's own [0:59] claims the things they themselves say it [1:01] does [1:02] and discussing obvious issues found in [1:03] my testing that would relate to real [1:05] world use [1:06] though if you do have any specific [1:07] questions after feel free to join my [1:09] discord [1:09] or my private patreon chat if you're [1:11] feeling supportive and i'll do my best [1:13] to answer any questions [1:14] if you're short on time and want a quick [1:16] summary this video will prove the [1:18] following points [1:19] mqa is not lossless mqa adds unwanted [1:22] noise and distortion [1:23] even in silent portions of tracks mqa is [1:26] not usually sourced from a high [1:28] resolution master [1:29] mqa and its mqa studio blue light [1:32] doesn't actually authenticate anything [1:35] and then i'll be discussing the various [1:36] other issues discovered throughout my [1:38] testing [1:38] if you want to skip to a particular [1:39] section there's timestamps on the screen [1:41] but if you want the full story then grab [1:43] a cup of tea [1:44] sit back and enjoy so before we start i [1:47] want to make a few things clear [1:48] one all of the evidence i gathered and [1:50] all of the issues that i found [1:51] were sent to mqa prior to this video [1:53] going live so that they could address [1:55] the concerns put forward [1:57] after i sent my email they did make sure [1:58] to swiftly get my tracks removed from [2:00] tidal [2:00] and also spoke to the publisher i used [2:02] to have them block me and remove all of [2:04] my content [2:05] i acted in good faith to make sure that [2:07] this video would be as fair as possible [2:09] and give mqa a fair chance to respond to [2:12] talk to me [2:13] and suggest alternative testing or [2:15] explain why some of these issues are [2:17] present [2:17] i didn't want this to be one-sided i [2:19] wanted to get their view [2:21] but unfortunately for a company with [2:23] nothing to hide they are sure [2:24] very keen on keeping information like [2:26] what i'm about to show quiet [2:28] i still have the original and mqa files [2:31] so it doesn't stop me or others from [2:32] testing [2:33] but i think it says a lot about nqa as a [2:35] company [2:37] two this video is not an attack on any [2:39] manufacturer that has incorporated mqa [2:41] into their products for reasons which [2:42] will be discussed at the end of the [2:43] video [2:44] this video is intended to inform you the [2:46] viewer about the issues and false claims [2:48] surrounding mqa [2:49] nothing else there are a huge number of [2:51] fantastic products which happen to have [2:53] mqa support [2:54] this is about mqa specifically not the [2:56] products that support it [2:58] three obviously this is going to be a [3:00] controversial topic and even if after [3:02] watching this video you still think mqa [3:04] is beneficial [3:04] or preferable to native flak then that's [3:06] entirely up to you subjective preference [3:08] is well subjective and i'm not going to [3:10] stop you liking what you like [3:12] some people prefer dacs amps headphones [3:14] and speakers that provably distort more [3:16] and that's okay objective performance [3:18] doesn't always align with subjective [3:20] preference [3:20] but please do not leave abuse towards me [3:23] or others [3:24] simply because your opinion differs i'm [3:26] providing evidence that mqa's claims are [3:28] not [3:28] true and if you disagree with the [3:29] evidence provided i would strongly [3:31] encourage [3:32] you to gather your own the more [3:33] information is available the better [3:35] and so far mqa themselves have provided [3:37] almost none [3:38] we should be able to discuss these [3:39] things in a level-headed and [3:41] constructive manner [3:42] don't resort to internet bickering so [3:45] what is mqa and what does it claim to do [3:48] mqa is a music encoding format that [3:50] promises to unlock [3:51] every detail of the original master [3:53] recording it will deliver the absolute [3:55] best possible listing experience that is [3:57] provided you have a title subscription [3:59] and pay for a dac which has mqa support [4:02] this they claim allows you to unfold [4:04] additional information hidden within mqa [4:06] files [4:07] it enables the listener to position the [4:08] instruments and performers to build a 3d [4:10] sonic picture [4:11] and it retains 100 of the original [4:14] recording [4:15] the claim is that it takes original [4:17] high-resolution studio recordings [4:19] and folds the information into a 44.1 [4:21] kilohertz file [4:22] by encoding what was the high frequency [4:24] information [4:25] down under the noise floor of the [4:27] audible band theoretically this can then [4:29] be played back natively on a standard [4:31] dac with just the 20 hertz to 20 [4:33] kilohertz information available [4:35] or it can be unfolded back to high [4:37] resolution [4:38] the first unfold can be done in software [4:40] taking the 44.1 kilohertz file to 88.2 [4:43] some devices called renderers can do the [4:46] second and third unfolds [4:47] once they've been fed information that [4:49] has already been unfolded once by tidal [4:51] for example [4:53] and some devices called full decoders [4:55] can do all three unfolds internally and [4:57] mqa claims that this then reveals the [4:59] original resolution [5:01] the original studio master the problem [5:04] is [5:04] it's nearly impossible to test whether [5:06] any of this is true [5:07] mqa is proprietary and closed source [5:10] unlike other formats such as mp3 dsd or [5:12] even flac [5:13] we can't encode music in mqa ourselves [5:16] we cannot use test files to find out [5:17] what's really going on [5:18] and as they go to great length to ensure [5:20] that this is not possible it's [5:22] exceptionally difficult to test anything [5:24] no device with a digital output is [5:26] allowed to have full mqa unfolding [5:27] capability [5:28] it is only in dax with an analog output [5:31] meaning it's impossible to test if [5:32] unfolded mqa really is the same as [5:34] native hi-res because we cannot get a [5:36] true digital unfold [5:38] and check it against the original [5:41] there are a lot of factors that mqa have [5:42] put in place to make it very difficult [5:45] to test claims and figure out what's [5:46] going on [5:47] so in order to look deep at what mqa is [5:49] doing in various situations [5:51] i needed a track where both the real [5:52] original master and the mqa versions [5:55] were available [5:56] and so what better way to do that and be [5:57] 100 certain than to get something [6:00] published myself [6:01] i wanted not just to publish music [6:02] though i wanted to get some test files [6:04] encoded such as a one kilohertz tone or [6:06] an impulse response test [6:08] when i first attempted to do this i was [6:10] told by the publisher that the mqa [6:12] encoder was [6:13] unable to encode the file how convenient [6:16] it seems that mqa's got checks in place [6:17] to prevent anyone from getting test [6:19] signals encoded [6:20] so i put together a quick acoustic track [6:23] and [6:23] just so happened to put an impulse [6:25] response white noise square wave 32 tone [6:27] test [6:28] and the entire rightmark audio analyzer [6:30] test sequence inside of it [6:31] for good measure not only did i publish [6:33] a 44.1 kilohertz version [6:35] i also put together and published an [6:37] 88.2 kilohertz high-res version [6:39] with a few extra tricks hidden inside [6:41] including some test tones at 35 [6:43] kilohertz and 40 kilohertz [6:45] and full range sweeps going from 20 [6:47] hertz to 44 kilohertz to test if mqa [6:49] truly does retain all hi-res information [6:52] from the original master once it's [6:53] unfolded [6:54] or if it gets removed in the compression [6:56] process [6:57] first let's check to see if the mqa [6:59] encoded 44.1 kilohertz file is the same [7:02] as a native 44.1 kilohertz flack [7:04] i checked my master against the file on [7:06] deezer and it was absolutely [7:08] 100 bit perfect it was the same file [7:12] on title though my mqa version was not [7:15] the same the mqa encoding process had [7:17] altered it [7:18] immediately the waveform looks different [7:19] transients are overshooting [7:21] giving the impression of it being louder [7:23] at some points and there is more high [7:25] frequency noise compared to the original [7:27] even in what should have been silent [7:28] portions of the track [7:30] looking in at the impulse response [7:32] inside the file this showed a minimum [7:34] phase ringing pattern which [7:35] also did not decay linearly minimum [7:37] phase filters have the advantage of not [7:39] having any pre-ringing which means that [7:41] transients can sound snappier or faster [7:44] however it has the disadvantage that [7:45] different frequencies [7:47] move through the filter at different [7:48] speeds and this can throw off musical [7:50] elements that rely on multiple [7:52] frequencies at the same time [7:54] such as timbre once unfolded this [7:57] problem only got worse [7:59] the only way i was able to correct it [8:00] was by running the track through hq [8:02] players [8:02] polysync mqa filter which is designed to [8:05] remove some of the anomalies and noise [8:07] caused by mqa [8:08] and it did a pretty good job here quite [8:10] white impulses and transients are [8:11] behaving like this [8:13] i don't know for my 88.2 kilohertz [8:15] tracks labeled as [8:16] deluxe the deezer version was just down [8:18] sampled to 44.1 kilohertz and [8:20] otherwise looked good any high frequency [8:22] content was just cut off [8:24] the impulse response preserved phase and [8:26] there was no added noise [8:27] fantastic however the tidal mqa version [8:30] looked very messy [8:31] there was a lot of high frequency noise [8:33] again even in parts of the track that [8:35] was supposed to be completely silent [8:38] also content that was previously in the [8:40] above 22 kilohertz range [8:42] had been aliased or reflected back down [8:44] into the audible band and [8:45] more on this later so no mqa files are [8:49] not lossless regardless of the original [8:51] sample rate [8:52] and if no mqa dac or unfolding is used [8:54] they will present issues compared to [8:56] streaming the lossless flag [8:58] i had a hunch that the amount of noise [9:00] in this 25 kilohertz band that was added [9:02] was proportional to the amount of [9:03] ultrasonic content that mqa had to [9:05] reconstruct [9:06] my file has quite a lot of noise in this [9:08] 25 kilohertz band [9:09] but then my file also does have full [9:11] range white noise in the original 88.2 [9:14] kilohertz master [9:15] to test this i did prepare another 88.2 [9:17] kilohertz master [9:18] without the white noise and with the [9:20] ultrasonic content drastically cut down [9:22] i'd submitted this to the publisher but [9:24] unfortunately before it went live [9:26] mqa had censored and removed all of my [9:28] content from tidal [9:29] so i wasn't able to confirm this the [9:31] reason that i suspected this though was [9:33] because there was one track which i [9:34] found where there was a native 88.2 [9:36] kilohertz version available [9:38] and an mqa version which was sourced [9:40] from an 88.2 kilohertz master [9:43] sam smith too good at goodbyes in the [9:45] 44.1 kilohertz mqa version of the file [9:48] which was made from an 88.2 kilohertz [9:50] master [9:50] we don't see the same band of high [9:52] frequency noise like we did in my file [9:54] however looking at the original native [9:57] 88.2 kilohertz version [9:58] there was never any content above 22 [10:00] kilohertz anyway so this then suggests [10:03] that the amount of noise [10:04] that mqa adds is proportional to the [10:06] amount of ultrasonic content that mqa [10:08] has to reconstruct [10:10] additionally one interesting thing was [10:12] that when these two files are compared [10:13] using delta wave [10:15] firstly we can see that there is still [10:16] some high frequency noise that mqa has [10:18] added [10:18] it's not lossless at all but there is [10:21] also this strange [10:22] band of distortion or difference across [10:24] the entire track [10:25] well within the audible band this wasn't [10:27] present on my file [10:29] i don't know what's causing this but i [10:30] felt that it was important to mention [10:33] so for 44.1 kilohertz masters or high [10:36] res masters where the ultrasonic content [10:37] was cut anyway [10:39] there won't be that much noise but still [10:40] some however for a track like this where [10:43] there is content going all the way up to [10:45] 44 kilohertz [10:46] there is a huge band of noise across the [10:48] entire track [10:50] not just where that ultrasonic content [10:51] was this is still a big problem [10:54] mqa or any method aimed at high fidelity [10:56] shouldn't be adding noise [10:57] the fact that this noise is present and [11:00] especially the fact that it's present [11:01] across the entire [11:02] file is worrying and it seems to be that [11:05] mqa gets more noisy the more high [11:07] frequency content mqa tries to [11:09] reconstruct [11:11] other issues were present in both the [11:12] 88.2 and 44.1 khz versions [11:15] low level signals were handled very [11:17] poorly and dynamic range is reduced [11:19] my original included a minus 60 db one [11:22] kilohertz sign [11:23] which is the standard test signal for [11:25] dynamic range the original had no noise [11:27] whatsoever [11:28] and was restricted only by the fft gain [11:31] in the mqa file there was high frequency [11:32] noise about 45 db below the signal level [11:35] and noise throughout the spectrum about [11:37] 65 db below [11:39] in the mqa file produced from our hi-res [11:41] master this high frequency noise was [11:43] actually 20 db [11:45] above the signal level next let's look [11:48] at exactly what's going on when [11:49] unfolding [11:50] does unfolding fix the issues and truly [11:52] reveal the original master [11:54] as mqa claims does it retain 100 of the [11:57] original recording [11:59] the first thing i found was curious was [12:01] that even my 44.1 kilohertz track [12:03] authenticated as mqa studio [12:05] and the dac unfolded it to a 352.8 [12:08] kilohertz sample rate [12:10] there was never a high resolution master [12:12] there is no extra information being [12:14] retrieved here [12:15] it is just being up-sampled there are [12:17] also no indicators on title that my [12:19] tracks came from different sample rates [12:21] how are customers supposed to know which [12:24] mqa releases are actually high [12:25] resolution [12:26] and which are just being up-sampled and [12:28] if it is a 44.1 kilohertz original [12:31] why would you want the mqa version when [12:33] all it's doing is adding noise and other [12:35] unwanted effects [12:36] compared to playing lossless more [12:38] concerning still [12:40] tidal actually didn't serve the lossless [12:42] file [12:43] when i selected the hifi quality setting [12:45] it just streamed the mqa file [12:47] without unfolding it i personally paid [12:49] for titles so that i could stream [12:50] lossless and if tidal no longer offers [12:53] that [12:53] at least for tracks labeled as master [12:55] i'm going to cancel my subscription and [12:57] i would encourage others to do the same [12:59] even if mqa is a benefit in some cases [13:02] it should be a choice [13:03] it should not be mandatory speak with [13:05] your wallet and demand lossless [13:08] a further note on this my original 44.1 [13:10] kilohertz master [13:11] was 30 megabytes and the mqa version is [13:14] 33. [13:15] it is bigger and does not offer a size [13:17] advantage [13:18] the artist neil young actually removed [13:20] all of his music at one point from title [13:22] in protest [13:23] after they provided mqa hi-res releases [13:26] despite the fact that he had never given [13:27] them permission to do so [13:29] and had never provided them with [13:30] anything other than a 44.1 kilohertz [13:32] master [13:33] in his own words tidal's master is a [13:35] degradation of the original to make it [13:37] fit in a box that collects royalties [13:39] that money is ultimately paid by the [13:41] listeners i am not behind it [13:43] i am out of there gone my masters are [13:46] the original [13:47] now then let's talk about authentication [13:49] that little blue light [13:51] it means nothing and mqa's provenance [13:53] and authentication is not secure [13:55] user frederick v on audio file style was [13:58] able to get rid of over 30 [14:00] of the file truncating 30 of the bits [14:03] and that blue light [14:04] still showed up and here with my tracks [14:07] even though i only ever provided them a [14:09] 44.1 kilohertz master [14:11] and the release on title is not the same [14:13] as my original master [14:15] this light is still showing up this [14:17] track has been altered [14:18] and so i the artist i'm telling you this [14:21] is not the version that would have been [14:22] created or heard in the studio [14:24] nor is it truly high resolution this [14:26] blue light [14:27] is mqa's marketing nothing more so then [14:31] let's have a look at unfolding [14:32] once the track that was originally a [14:34] 44.1 kilohertz master [14:35] is unfolded through tidal or roon an [14:38] 88.2 kilohertz file is produced [14:40] looking at the spectrum we can see that [14:41] there is an upwards reflection of the [14:43] music [14:44] when looking at the 44.1 kilohertz [14:46] original this was never there in the [14:48] original master [14:49] it is an unwanted artifact this is due [14:52] to the upsampling filter that mqa uses [14:54] being [14:54] leaky an issue which has actually been [14:56] discussed by ashimago in an article [14:58] linked in the description [14:59] and in a 2018 armath talk by chris [15:02] konica [15:03] there were nqa staff present for this by [15:05] the way but they didn't offer any [15:06] explanation [15:07] instead they chose to interrupt chris [15:09] quite regularly to make arguments that [15:10] he should ask dac manufacturer dcs why [15:12] they support mqa [15:14] and that the information couldn't be [15:15] trusted given as he wouldn't share the [15:17] name of [15:17] archimago who had provided the graph [15:21] to address that second point chris quite [15:23] rightly said that it didn't matter who [15:24] ashumanga was [15:25] anyone can quite easily show this [15:26] imaging issue at home and if you'd like [15:28] to replicate this issue yourself [15:30] i've made a post showing quite easily [15:32] how to do just that [15:35] using the white noise portion of my [15:36] track i was able to look at the [15:38] ultrasonic attenuation performance and [15:39] filter design of mqa's unfolding filter [15:42] the white line here shows mqa's filter [15:44] and the blue line shows hq player's sync [15:46] l filter as an ideal reference this is a [15:49] 2 million tap filter [15:50] similar in designed to that of the chord [15:52] m scaler if you're familiar with that [15:54] mqa's filter has really poor attenuation [15:56] as well as strange patterns in the [15:58] ultrasonic range this is why ultrasonic [16:00] content is being added [16:01] in the unfolded 44.1 kilohertz version [16:04] not only is the unwanted ultrasonic [16:05] content added [16:06] audible band content hasn't been fixed [16:08] either there's still persistent high [16:10] frequency noise [16:11] the square wave has excessive overshoot [16:13] non-linear ringing [16:14] and a strange notch to the transitions [16:16] there's absolutely nothing i can see [16:18] here that is in any way beneficial to [16:20] the listener [16:20] or the integrity of the song and plenty [16:22] that would be considered problematic [16:25] if we do a full decode these problems [16:27] are effectively repeated [16:28] with more upward images created this [16:31] part for example is a 20 hertz to 20 [16:33] kilohertz sweep [16:34] and this musical content stopped at 22 [16:36] kilohertz [16:37] there should be absolutely nothing above [16:39] that but the leaky up sampling filter [16:41] design [16:42] is introducing artifacts that strange [16:45] notch on the square wave is now [16:46] extending lower [16:47] the 44.1 kilohertz master has gained [16:49] nothing from this unfolding process [16:51] and it's had all sorts of problems added [16:53] so what about the versions that did come [16:55] from a high-res release the ones that [16:57] were 88.2 kilohertz [16:59] in the original file for the 88.2 [17:01] kilohertz release i added a 35 kilohertz [17:04] and 40 kilohertz test tone [17:05] these are frequencies which are too high [17:07] to be stored in a 44.1 gigahertz file [17:10] and so this was designed to test whether [17:11] mqa would retrieve this information [17:14] once it was unfolded the writemak audio [17:16] analyzer test also has sweeps and imd [17:19] tests which go all the way up to 44 [17:21] kilohertz [17:23] in the mqa encoded 44.1 kilohertz file [17:26] these signs have been aliased down into [17:28] the audible band at 9.1 and [17:30] 4.1 kilohertz these are not below the [17:33] noise floor these are very audible [17:35] and have barely been attenuated at all [17:37] the same has happened for the sign [17:39] sweeps in the right mark audio analyzer [17:40] test [17:41] once the frequency reaches 22 kilohertz [17:43] it starts moving back down and aliases [17:45] into the audible [17:46] band again once the first unfold was [17:48] done however these 35 [17:49] and 40 kilohertz signs were back in [17:51] their correct place [17:52] and the writemak audio analyzer sweeps [17:54] had been flipped back [17:56] indicating that mqa's claim of restoring [17:58] high frequency content is at least [18:00] partially true however it is far from [18:03] perfect [18:03] we still have a massive band of noise [18:05] centered around 25 kilohertz [18:07] and the unfolding has left further [18:09] artifacting in lower frequencies [18:11] resulting in a noise floor of about -60 [18:13] db [18:14] in mqa's bob talks blog they claim that [18:16] content is folded back into the region [18:18] below -120 db [18:20] and should therefore be inaudible we can [18:22] see from my file that this claim is just [18:24] not true this has hardly been attenuated [18:27] at all and higher level signals would [18:29] absolutely be audible [18:30] i checked this by just listening myself [18:33] and i could quite clearly hear these [18:34] tones being played [18:43] so this means that we've got a very poor [18:45] dynamic range in both the audible [18:47] and the ultrasonic bands when mqa is [18:49] unfolding ultrasonic content [18:52] once a full decode was done again these [18:54] problems did not go away [18:56] and we simply see the content of the [18:57] file reflected upwards multiple times [18:59] into the ultrasonic range [19:01] so to answer question number two yes [19:04] technically mqa does put back high [19:06] frequency content once unfolded [19:08] however it does so with vastly reduced [19:10] accuracy and dynamic range [19:11] introduces concerning levels of noise [19:13] that do not go away once unfolded and it [19:16] leaves fairly high level noise and [19:17] artifacts in the audible band when [19:19] previously folded content is removed [19:22] it does not by any stretch retain 100 [19:25] of the original recording and this claim [19:27] is false [19:29] it is also in my own opinion not exactly [19:32] a huge achievement given as the 44.1 [19:34] kilohertz file [19:35] could be barely usable all of this [19:37] content has simply been aliased and [19:39] reflected back down into the audible [19:40] frequency range [19:42] at high levels it is not folded down at [19:44] inaudible levels as claimed by mqa [19:47] and depending on how much high frequency [19:48] content was originally present [19:50] it could be very audible and very [19:52] problematic [19:54] so then with the testing out of the way [19:55] let's quickfire some points that we now [19:57] know [19:57] mqa is not lossless in either stock or [20:00] unfolded form [20:01] regardless of the original file sample [20:03] rate most mqa releases are just being [20:06] up-sampled and are not sourced from [20:07] high-resolution masters [20:09] mqa aliases high frequency content down [20:12] into the audible range with minimal [20:13] attenuation [20:15] mqa does put back high frequency content [20:17] and removes most of the previously alias [20:19] content when unfolded [20:20] but it leaves significant noise and [20:22] distortion behind [20:24] tidal no longer offers true lossless [20:26] streaming for any track marked as master [20:28] that i could find you either have [20:30] compressed or lossy mqa [20:32] the lossless flag is not available and [20:35] you have to use an alternative service [20:36] such as cobuzz [20:37] or deezer mqa authentication and the [20:41] blue light gives absolutely no guarantee [20:44] of source sample rate file integrity or [20:46] lack of alteration [20:48] and it does not guarantee that the sound [20:50] is the same as was in the studio [20:52] if you are a rune user you can actually [20:54] see what the original sample rate of an [20:56] mqa release was [20:58] the last portion of the track [20:59] description where it gives the mqa [21:01] sample rate is the original mass to [21:02] sample rate [21:03] so anything labeled as 44.1 kilohertz [21:06] here is just being upsampled [21:09] mqa like for like offers no file size [21:12] advantage [21:13] and is actually bigger than 44.1 [21:15] kilohertz flack though it is smaller [21:17] than high-res flack [21:18] now with evidence and testing of the way [21:20] i want to give my own opinions about the [21:22] situation [21:23] mqa in general and why i did this [21:24] testing the following is my own review [21:26] not necessarily fact i'm an audiophile i [21:29] love the gear i love the music and i [21:31] love when a technological advancement [21:33] brings an improvement to our listening [21:34] experience [21:35] but anyone in this hobby will happily [21:37] tell you that there is a lot of snake [21:39] oil about [21:40] there are plenty of products that [21:41] promise a better experience and claim to [21:43] be the best way of doing whatever it is [21:45] they're doing [21:46] but there are two key categories those [21:48] that have evidence or can be tested by a [21:50] third party [21:51] and those that cannot there are people [21:53] like rob watts from cord [21:54] who is quite vocal about the fact that [21:56] he believes his wta filter design [21:58] is the optimal method for playing back [21:59] 44.1 kilohertz audio [22:01] and not only is evidence given for this [22:03] you can test it yourself [22:05] i don't necessarily agree but me or [22:08] anyone with a chord dac or chord m [22:09] scalar is free to play and record any [22:11] music we like through them [22:13] test any signals we wish and rob watts [22:15] is also very responsive on places such [22:17] as headphy [22:18] writing up detailed posts about how and [22:20] why his design works [22:21] he's not going to hand over the filter [22:23] coefficients but he'll answer enough [22:24] questions and provide evidence [22:26] for you to make an informed decision as [22:28] to whether you agree [22:29] and would like to purchase his product [22:31] it's a similar story with dsd [22:33] it's open anyone can encode and playback [22:36] files using it [22:37] and it's completely able to be poked [22:38] prodded tested explored [22:40] and have the pros and cons discussed [22:42] even the market segment most commonly [22:44] regarded as being snake oil [22:45] cables can be tested when comparing the [22:48] jitter performance between a decent [22:50] generic bnc cable [22:51] and the audioquest carbon bnc cable i [22:54] was able to get measurably lower jitter [22:56] with the audioquest cable [22:58] now whether this is worth the cost or [23:00] audible to you is a different question [23:02] but the point is i can test their [23:04] marketing claims and i can prove or [23:06] disprove them [23:07] but then there are companies like mqa [23:09] not only has mqa not provided their own [23:11] evidence [23:12] they prevent you from gathering your own [23:14] you can't encode files in mqa [23:16] or record the final output of a full [23:18] decoder it just operates on a trust us [23:20] it's better approach their marketing [23:22] claims are vague [23:23] and change when they get disproven [23:25] they've released information and done [23:26] some things that are [23:27] genuinely fascinating mqa's white glove [23:30] treatment correcting for problems in [23:31] recording equipment is [23:33] truly fantastic and i would strongly [23:35] recommend giving the fairy tales article [23:37] a read [23:37] it's highly recommended but the way that [23:40] it's presented [23:41] makes people believe that this is how [23:42] mqa's core product works [23:44] which it isn't the implication is that [23:47] all mqa releases are better than flac [23:49] when the majority of it is just [23:51] upsampled 44.1 [23:53] bob stewart's blog provides an [23:54] explanation for how mqa folds high [23:56] frequency info [23:57] which could never work with a 16-bit [23:59] file as it wouldn't have enough dynamic [24:01] range [24:02] when issues are presented to them they [24:03] dodge the question and try to divert the [24:05] conversation [24:06] and all the while more and more music [24:08] content is being replaced with mqa [24:10] more and more hardware is supporting mqa [24:12] and it continues to spread [24:14] with the supposed benefits being [24:15] completely unproven and its problems [24:17] steadily being exposed as people find [24:19] ways around the testing roadblocks [24:21] such as what i've done today even if we [24:24] were to assume that mqa is absolutely [24:26] perfect that it is 100 [24:28] lossless and it does everything as [24:29] described in the marketing material it [24:32] should not be the only option [24:33] you should still have a choice you [24:35] should still be able to choose between [24:37] high resolution native sample rate audio [24:39] and mqa as well as any other option that [24:42] chooses to come along [24:43] now as i've shown in this video most mqa [24:45] releases are just being up-sampled from [24:47] 44.1 kilohertz [24:49] they are not sourced from a [24:50] high-resolution master but there are [24:52] some releases where it is genuinely from [24:54] a high sample rate master [24:56] and there is no native high sample rate [24:58] release [24:59] this is a problem you as a consumer [25:01] should not be comfortable [25:02] with one company growing a monopoly on [25:05] the high resolution audio market [25:07] 24 bit 192 kilohertz audio uses less [25:10] than 5 megabits per second [25:12] the vast majority of internet [25:13] connections in the developed world [25:15] can absolutely accommodate that if you [25:17] can watch youtube [25:18] or netflix you can stream lossless in 24 [25:21] 192 [25:22] without any concern whatsoever mqa [25:25] is nothing but an expensive lossy [25:28] compression mechanism [25:29] being forced on you and provided as the [25:31] only option if you want to access some [25:33] of this high-resolution content [25:35] and even then due to the fault shown in [25:37] this video it is not the same [25:39] as the native high-res content that [25:41] ideal option [25:42] of true native audio has been taken away [25:45] from you [25:46] the reason hardware manufacturers are [25:48] supporting mqa isn't because they [25:49] genuinely think it works and that it's [25:51] better than native high-res [25:52] it's because you the customer are [25:54] demanding it [25:56] various companies such as lin and [25:58] even ps audio whose dax do support mqa [26:01] have all spoken out against mqa links [26:03] again are in the description [26:05] paul from ps audio talks in a video [26:07] about this he says that mqa was added [26:09] because so [26:10] many people were demanding it but he [26:11] still doesn't like it [26:13] it's a business decision it's being done [26:16] because if they do not implement it they [26:18] will lose [26:18] customers and regardless of mqa's [26:20] technical performance [26:22] no one can dispute that it is an [26:23] absolutely fantastic example [26:25] of the power of marketing i like [26:28] progress i like advancements in allowing [26:30] us to enjoy music all the more [26:31] but mqa has various demonstratable [26:34] problems [26:35] none of which exist if you just stream [26:37] lossless [26:38] flak even at 44.1 i can't recommend that [26:41] anyone uses mqa in any situation [26:44] and you should vote with your wallet [26:45] don't buy products that have cost the [26:46] mqa licensing fees tacked on [26:48] don't subscribe to platforms like tidal [26:50] that are removing lossless content [26:52] and forcing you to use mqa for a lot of [26:54] tracks and demand that companies be [26:56] honest and transparent about the [26:57] products they sell [26:58] demand that they allow third-party [27:00] testing to confirm their marketing [27:02] claims [27:03] if a company can't provide its own [27:05] evidence and goes to great lengths to [27:06] prevent [27:07] anyone else from testing it that should [27:09] probably raise quite a few red flags [27:13] one last thing i'd like to talk about [27:15] the first unfold of mqa is different to [27:17] the second and third which is why some [27:18] devices can only do the second and third [27:21] and require the player software to do [27:22] the first if you have [27:24] no choice but to use an mqa file my [27:26] recommendation would be to do the first [27:28] unfold in software [27:29] and then up sample or oversample [27:31] normally with your internal dac filter [27:33] don't do any further mqa rendering [27:36] this will remove most of the audible [27:37] band aliasing but it will also remove [27:39] unwanted high frequency images and [27:41] some of the noise [27:44] during this video i use the ifi idsd [27:46] diablo for testing [27:48] it is a full mqa decoder and i have to [27:50] say it's a fantastic product i love it [27:52] but i did find out something kind of [27:54] interesting about ifi's gto filter [27:56] whilst i was making this [27:59] this is the filter response of ifi's gto [28:01] filter [28:02] look familiar that's probably because [28:04] this is the filter response of mqa's [28:06] renderer filter [28:08] they are basically identical a quick [28:10] test showed that ifi's gto filter [28:12] exhibited the same leaky characteristics [28:14] and caused imaging in the same way as [28:16] the mqa unfolds did [28:17] and it's also minimum phase meaning that [28:19] phase-dependent sound traits such as [28:21] timbre [28:22] can be degraded by using it however [28:24] swapping back to the 7.0 firmware and [28:26] using the stock filter was absolutely [28:28] fine [28:28] i would strongly recommend that if you [28:30] have an ifi dac you use the stock filter [28:33] not gto i guess i shouldn't have been [28:35] surprised when i've read the i5 paper [28:37] about the gto filter [28:38] and saw that it was designed with [28:40] assistance from mqa [28:42] the diablo is still a truly fantastic [28:44] product and again just because something [28:46] supports mqa doesn't mean that the [28:48] product is bad [28:48] but tell manufacturers you don't want to [28:51] pay extra for it [28:52] tell them that you want their products [28:54] without mqa added on [28:57] mqa did respond to me prior to this [28:59] video going live i'm not gonna address [29:00] every single point because a lot of it [29:02] is just marketing speak [29:03] but the full response is on screen now [29:05] so you can read through [29:06] and the references are linked in the [29:08] description the first part mostly talks [29:10] about how [29:11] mqa is more advanced than conventional [29:13] approaches and intends to account for [29:14] the conversion process at each end of [29:16] the chain [29:17] this makes no sense given as i got my [29:19] files encoded in mqa [29:21] and wasn't even asked for information [29:23] whatsoever about the production hardware [29:25] i used this was all done using mqa's [29:27] recommended method for indie artists by [29:29] the way [29:30] and all of the issues in this video were [29:32] captured using fully certified mqa [29:34] hardware and software [29:36] more to the point even if you did have [29:38] information about the source and dac [29:39] hardware components [29:41] how would you account for the behavior [29:43] with tracks where multiple adcs [29:45] synths recording methods and [29:46] in-production editing methods were used [29:48] hardly any music nowadays is produced in [29:51] analog and then digitized or digitally [29:53] recorded [29:54] using only one adc model we don't [29:56] understand your frustration about [29:58] evaluating mqa [30:00] my frustration is that no end-to-end [30:02] analysis tools are available [30:03] like they are for nearly every other [30:05] format [30:06] yes we can listen to any mqa song but we [30:09] have no information about how it's been [30:11] altered [30:12] we can't check for placebo we don't know [30:14] if it's the same master [30:15] and we have no way to objectively test [30:17] with test tones or signals [30:19] other than jumping through a lot of [30:20] hoops like i had to do for this video [30:23] additionally when someone like chris [30:24] conica provides evidence the question [30:26] gets dodged and you try to change the [30:28] subject [30:29] when i reach out to you you attempt to [30:31] censor me by removing all of my content [30:33] that is my frustration it isn't [30:35] transparent in the slightest [30:37] and mqa responds in a hostile manner [30:39] whenever criticized [30:42] in regards to the tracks i had used for [30:44] testing in this video mqa says [30:46] this encoder is not configured to deal [30:48] with content where for example [30:50] the statistics change mid-song or where [30:52] the audio does not resemble natural [30:53] sound [30:54] does that then imply that electronic [30:56] music won't work with mqa or that tracks [30:57] with a combination of natural recording [30:59] and synthesized parts won't work [31:02] right now none of the tracks in titles [31:04] top 10 are true analog recordings and [31:06] yet 8 out of 10 of them [31:08] are in mqa maybe the [31:11] studio tools that are referenced in your [31:13] response and on your website [31:14] do help to overcome these issues but [31:16] again there's no [31:18] transparency we have no idea what these [31:20] tools are [31:21] who has access to them which releases [31:23] were made using them [31:24] and we have no way to evaluate them [31:26] would you ever consider making a [31:28] limited version of these available [31:29] perhaps releasing a version that only [31:31] allows for a max 10 second audio length [31:33] this would allow people to test properly [31:35] without having any commercial impact [31:37] and is similar to what signalist does [31:38] with their hq player software [31:40] where you can use the hq player [31:42] professional software but it caps the [31:44] output to 60 seconds of audio a little [31:46] transparency would go [31:47] a long way and saying that mqa is [31:50] lossless but only within vague [31:52] unspecified criteria and lossless only [31:54] using non-conventional evaluation [31:56] methods [31:57] isn't exactly reassuring in regards to [32:00] my mention of the blue mqa [32:02] authentication indicator they said the [32:04] onus is on the submitter to check the [32:06] content when it arrives on tidal [32:07] and confirm the sound again this was [32:10] never mentioned during the publishing [32:12] process it is not mentioned [32:14] in your marketing material and it also [32:15] doesn't address the fact that you can [32:17] make an mqa file [32:18] throw away a third of it and still have [32:21] that light show up [32:23] plus even the 44.1 kilohertz file that i [32:26] submitted [32:26] both sounds different and is objectively [32:29] different to the master that i published [32:31] and again this is the same for any other [32:33] track [32:34] where a high sample rate version and mqa [32:36] version are available [32:37] they're not the same the next part of [32:40] their response goes on to explain how [32:41] mqa adaptively identifies and responds [32:44] to content in a track which again [32:45] implies that it can't be lossless if [32:48] it's behaving differently to different [32:49] types of recording and music [32:51] then the result won't be consistent it [32:53] might sound more natural to some people [32:55] but sounding natural and being lossless [32:57] to the original master [32:58] are not the same thing now they also [33:02] discussed the files that i had published [33:03] i should firstly mention that the [33:05] information and status messages they're [33:07] providing here [33:08] i only ever received one of these the [33:10] one saying that the encoder was unable [33:12] to encode the file [33:13] and that was only for the files where it [33:15] was literally just a test file [33:16] an impulse response or square wave [33:18] without any other content [33:19] which likely wouldn't pass mqa's checks [33:21] for test files [33:23] other products like the called m scalar [33:24] have impulse test detection as well [33:27] all of the other error messages i never [33:29] received i don't know if these are [33:30] genuine [33:31] and given as mqa had the publisher [33:32] remove all of my content [33:34] i can't go back and check in response to [33:37] my claims about added noise [33:39] mqa claims that this was my fault as i [33:41] didn't dither the files [33:42] claiming this was a naive mistake no mqa [33:46] it wasn't in fact this was something i [33:47] had explicitly tested [33:49] the first track i submitted called try [33:51] again did have dither [33:52] and it showed all of the issues [33:53] mentioned in this video in fact some to [33:55] a worse extent [33:57] to give mqa the benefit of the doubt and [33:58] check that the dithering wasn't [34:00] adversely affecting anything [34:01] i published the next two tracks without [34:03] dither so this isn't my fault at all and [34:06] the only naive mistake made was mqa not [34:08] looking at all three files i had [34:09] published [34:09] dither was something i considered and it [34:11] was something that i checked if it [34:12] altered mqa's performance [34:15] in response to others and my description [34:17] of the mqa upsampling filters as [34:19] leaky they claim this term is derogatory [34:21] and inappropriate [34:22] and that the only alternative would be [34:24] brick wall band limiting [34:25] firstly i don't think the filter itself [34:27] has feelings but if i have hurt them [34:28] then i apologize [34:30] secondly perhaps the biggest advantage [34:32] of genuine native high-res audio [34:34] is just this it allows much more [34:36] flexibility with the filter design [34:38] going from 44.1 kilohertz to 96 [34:40] kilohertz [34:41] gives you over 13 times more distance [34:44] between the audible band [34:45] and the nyquist frequency so if the [34:47] argument is that brick wall filters are [34:49] bad which to be clear in many aspects i [34:51] agree with [34:51] the solution isn't mqa is to use native [34:54] high-res in regards to aliasing they [34:57] claim that this is simply because the [34:58] levels of the signal were too high and [35:00] it wouldn't occur in real music [35:02] for this point i'm quite happy to accept [35:04] this might be the case in fact as i [35:06] mentioned earlier in the video [35:07] i prepared another track with ultrasonic [35:09] content at a much lower level to test [35:11] what level this was then alias down at [35:13] unfortunately you had my tracks removed [35:15] and this never made it to tidal [35:17] if this claim is true then send me the [35:19] mqa encoded version of that file [35:21] and i will put at the top of the [35:22] description of this video that mqa does [35:24] not actually have any issues with [35:25] aliasing [35:27] lastly they say that despite what's [35:29] shown here mqa doesn't add distortion [35:31] and the noise added is inaudible high [35:33] frequency dither noise [35:35] it might be centered on 25 kilohertz in [35:37] the unfolded version [35:38] but in both the 44.1 and unfolded [35:41] versions [35:42] it extends down into the audible band [35:44] and is at minus 43 db [35:46] at 20 kilohertz that is audible [35:50] they also say that every mqa file will [35:52] tell you the original sample rate of the [35:53] master [35:54] and this is true but my complaint wasn't [35:56] that it didn't tell you anywhere [35:58] it was that it wasn't clear in the [36:00] slightest tidal [36:01] shows no indication at all of original [36:03] sample rate [36:04] you have to use a program like roon to [36:06] show this information [36:08] this means that mqa's marketing is [36:10] misleading consumers into thinking that [36:12] mqa releases aren't the same or better [36:15] than 24 bit 192 kilohertz flack [36:17] i did a survey which got over 240 [36:20] responses [36:20] of those people 46 said that they [36:23] thought mqa was a good thing [36:25] the people that said mqa was a good [36:27] thing two-thirds of them [36:28] believed that mqa releases were the same [36:31] or better [36:32] than 24-bit 192 kilohertz flack [36:36] this is why i think it's a problem this [36:38] is why it needs to be more transparent [36:40] even if a 24 bit 192 kilohertz recording [36:43] folded down to 44.1 kilohertz was [36:45] absolutely identical to a native 192 [36:48] kilohertz recording [36:49] the vast number of mqa releases that are [36:51] 44.1 kilohertz from 44.1 kilohertz [36:54] masters [36:55] are not the same or better than native [36:57] high-res flack [36:58] and this misconception is concerning [37:02] hopefully you found this video [37:03] informative and enjoyable i'd like to [37:05] thank daniel mellinger and amos aka [37:07] kurawong for helping me out with the [37:09] making of this video [37:10] amos has compiled a long list of [37:12] interesting articles and evidence [37:13] surrounding mqa [37:14] a lot of which was not touched on this [37:16] video and all of them are in the [37:18] description below [37:19] i'd also like to thank ashimago john [37:21] atkinson miska mans roll guard and [37:23] frederick v [37:24] for the information tools and testing [37:25] that they have made available [37:27] thank you as well to all of the 250 [37:29] people that responded to my mqa survey [37:31] additionally a huge thank you to all of [37:33] my patrons especially my diamond and [37:35] legend here patrons [37:36] chris ross kyle gravitas crack daniel [37:39] mellinger [37:39] king jong-un and lana bennett you guys [37:41] are fantastic [37:42] if you have any questions do feel free [37:44] to join my discord or my private [37:45] telegram chat for patrons if you're [37:47] feeling supportive [37:48] have a great day