Digital Done Wrong
Installment 1: Introductory Overview
Playback Reconstruction Accuracy
by Digital Claiming Superior Time Domain Response
This article will directly affect and influence millions of you around the world - virtually all of you who have a digital signal anywhere in your home entertainment system (or professional A/V equipment). To be honestly candid, this is probably the most influential article ever published on digital (especially when all its installments are released and assembled together).
This article will shock the 600,000 among you who read Stereophile, since it shows that (and how and why) our esteemed colleagues at this widely read publication (and at other publications) have been wrong in virtually everything they have said about digital - indeed, their pronouncements are so far wrong that they are backwards, the very opposite of what digital really is and how digital actually works. Just one example: Stereophile consistently always says (even in their brand new January 2018 issue) that a shorter impulse response in digital playback products provides better time domain transient response performance - but the truth is the complete opposite, and shorter impulse response actually provides worse time domain accuracy, and more time domain distortion, of the reconstructed signal waveform that is played back from digital.
This article reviews 86 of the world's most advanced digital audio products, using the world's most advanced digital system, MQA. But MQA itself is merely the most advanced spearhead of a very large and important modern revisionist digital engineering movement worldwide, that nobly seeks to provide better time domain transient response from digital systems, e.g. with filters optimized for better time domain performance. This article will completely overthrow that entire worldwide digital engineering movement. This article will show and prove that the million digital products sold to consumers, designed for this allegedly superior time domain performance, in point of fact are fatally defective designs - and actually accomplish the complete opposite of the time domain accuracy they purport to honor, instead actually butchering the time domain signal waveform they try to reconstruct, in playing back your digital recordings.
Thus, this article will also shock thousands of digital engineers around the world, into realizing that most of the digital products they have designed for the past 34 years are fatally defective, actually butchering the signal, rather than correctly reconstructing the original signal in the time domain from the digital samples. This article will show them that, despite their noble intent and brilliant brains and PhD digital education, they actually have such utter lack of comprehension, about how the basics of digital actually work, that they commit 100 huge blunders in their designs of the digital products that millions of you have purchased.
These 100-blunder digital product designs butcher the signal (rather than correctly reconstructing it), because they violate the way digital actually works. These 100 blunders, committed by digital engineers worldwide (and by publications that support their views) not only are high in number, but also are each severe in degree, with many being completely backwards, the very opposite of the truth about the way digital actually works. The high number and severity of these 100 blunders proves that, regrettably, digital engineers worldwide (and their supporting media) don't have the slightest clue in comprehending how digital actually works.
This article will shock these thousands of engineers into scrapping their 100-blunder digital models and theories, and scrapping their product designs based on these 100-blunder models. This article will hopefully spur them into creating fresh new digital product designs that can and do obey the way digital actually works.
And, to close the circle, this article will then directly affect and influence millions of you to chuck your millions of fatally defectively designed digital products into the trash, and purchase new designs that actually deliver on the promise of digital.
Sonic Summary of 86 Consumer Electronics Products Using MQA
We'll begin this article, to be published in serial installments, with a review of MQA and the 86 consumer electronics products (plus professional products) using MQA.
Logically, MQA belongs somewhere in the middle of our overall serial installments, since the overall article's umbrella title is Digital Done Wrong. But we're prioritizing MQA up front, for several reasons. MQA is the latest and most advanced spearhead of an important and widespread revolutionary revisionist digital engineering movement, which nobly tries to provide better time domain transient response performance from digital filters and digital in general. MQA provides the best (quickest) time domain impulse response of any digital system to date. Stereophile's brand new January 2018 issue has two of our esteemed colleagues (the Editor and a science/technical writer) endorsing this quick impulse response by MQA - and (more crucially) agreeing that this portends superior time domain transient response from MQA in reproducing digital signals. And, finally, MQA is an au courant hot topic, having recently engendered a lot of comment and controversy, much of it misinformed and utterly wrong.
The 86 consumer electronics products, that today incorporate MQA playback, sound similar to one another in MQA mode, which of course makes sense since the algorithms provided by MQA to all OEMs are basically the same. So the real review question for these 86 digital products boils down to asking, and analyzing, how MQA itself sounds (and then of course technically explaining why it sounds this way).
We listened to a variety of MQA playback products, and to a variety of MQA encoded recordings. The most revealing and definitive sonic test was a simple recording of Bob Dylan and his acoustic guitar, provided by MQA themselves as a demo, with both the PCM version and the MQA version sourced from the same original Columbia tapes. By simply flicking a switch at moments of our own choosing, we could A-B the PCM version against the MQA version (the playback electronics for this demo was a Mytek Brooklyn, which has been universally praised, driving very good headphones).
There were huge and startling sonic differences between the two renditions, differences which also turned out to be very informative of the technical differences between conventional PCM vs. MQA. These specific sonic differences we heard were also subsequently confirmed and corroborated, by a later technical analysis, explaining the reasons for these sonic differences.
MQA's chief technical claim to fame is its superior time domain transient response performance, so naturally we first paid attention to the one featured and repeated time domain transient on this recording, the attack transient of each guitar plck. This time domain transient attack sounded dramatically different through the two competing digital systems.
As you know, the picking of a guitar string is a mechanical process, that produces a pretty hard sounding momentary transient attack, which is subsequently followed by the smooth sounding sustain, of the just-picked string resonating freely on its own. We focused our attention on the time domain momentary transient attack of the picking itself, since this is supposedly MQA's claimed (and acclaimed) forte. The PCM version rendered this picking sound correctly, with excellent hardness and speed and coherent focus and lack of overhang, all of which superbly conveyed the strong and tactile quality of each pick's time domain transient. In short, the PCM guitar picks sounded physically real, as though we could reach out and touch them.
What about the time domain performance of the MQA guitar picks? Were they also strong and tactile and real, hopefully even more so (as theory, and MQA's claims of superior time domain transient response, led us to expect)? We were shocked by what we heard. Not at all what we expected. The guitar pick attack transients from MQA were not at all strong, tactile, coherent, or real. Instead, these naturally strong and hard, tactile and coherently focused time domain transients were literally eviscerated by MQA, transformed to merely a weak shadow of the correct attack transient sound. Indeed, these naturally strong and hard attack transients were so badly eviscerated that they virtually disappeared entirely, from MQA's reconstructed time domain signal waveform and the sound of this MQA waveform.
The sustain portion of each note's string resonance seemed to begin, without the sound of that string ever having been picked. With MQA obliterating and erasing the sound of the guitar pick transient from its time domain reconstruction of the signal waveform, the physical effect of the resonating string's sustain seemed sonically to begin and emerge without ever having had any physical cause (the picking of that string). Thus, MQA was sonically committing the ultimate non-causal filter heresy, making the bell ring without ever having been struck.
Similarly, MQA softens, eviscerates, or completely erases virtually all transient attacks, which is especially pernicious and noticeable when the attack transient is naturally hard, e.g. the "t" of a triangle ting, or the "t" sound formed by a singer's tongue and teeth.
Some na´ve listeners might subjectively prefer all their sounds to be softened, or even have the naturally hard transient attacks be completely erased. Such na´ve listeners might well prefer the sound of MQA to other competing systems that reveal the truth about real sounds, so they might well erroneously pronounce MQA's sound to be superior precisely because of what is in fact MQA's gross distortion and reconstruction failure here, with all brief transient attacks.
The second sonic contrast we noted was that MQA's high frequencies had a much more airy, open quality, whereas PCM's high frequencies sounded more closed in. They were surely different, but that does not answer the crucial questions of which is better, and why.
One sonic clue was that MQA's high frequencies were also soft, defocused, diffused, phasey, and fuzzy smeared (not articulate, individuated, or coherently focused). This suggests that something was wrong with these high frequencies (perhaps the same factor producing their airier, more open quality).
On the other side of the coin, we happen to know that, for technical reasons (to be explained in another installment of this article), low sample rate digital in general and PCM in particular do have problems that lead to their sounding too dull and closed in at high frequencies.
Thus, this particular sonic difference could objectively well be a matter of both systems being somewhat wrong, but in different ways, and on opposite sides of the ideal accurate middle (as sonically revealed by high sample rate PCM).
Which did we subjectively prefer? Neither. Our job is to objectively observe and analyze. We can well imagine many listeners subjectively preferring the more airy, open, softer and more relaxed presentation of high frequencies by MQA. But it would be an objective mistake to pronounce MQA better. Indeed, in point of objective fact, it might well turn out that MQA's different sonic presentation here, however subjectively pleasing to some listeners, is actually caused by an objective factual technical defect in MQA, in which case we can assuredly state that MQA's sound is factually and objectively wrong here, despite it being subjectively pleasing to many.
The third sonic contrast we noted was that the MQA rendition had a much larger surrounding space, with rich ambience. Since the advent of stereo all of us have strived for better reproduction of large acoustic spaces, and their richer ambience, to frame the direct sound of the performing musicians. So all of us are naturally biased toward thinking that more is always better. And many na´ve listeners will be persuaded by MQA's larger spatial portrayal to believe that this MQA rendition must be better, and must be yet another way in which MQA is superior to other digital systems. But, as Goldilocks taught us, more is not always better, and a "just right" middle ground might well be the best of all, especially if it represents accuracy. We are seasoned professional and analytical listeners, so we objectively probed deeper when we heard MQA's richer rendition of ambience and a large space. Our initial subjective enthusiasm was quickly tempered by 3 objective scientific considerations.
First, the large space itself was portrayed as a large amorphous bloated vague halo somewhere beyond the performers - but not as a well defined hall or room, with well defined walls at a well defined distance, where the original recorded performance actually took place.
So there was clearly something wrong with MQA's enhanced spatial imaging. This inferentially scientifically suggests that MQA's whole enhanced imaging phenomenon might well be due to an inaccurate mistake MQA was making, which is the complete opposite of MQA's own claim that MQA does something better than competing digital systems, hence provides superior and more accurate spatial imaging.
Many na´ve listeners have joined MQA's own implied claim that bigger, more bloated spatial imaging is always better. And they also have joined MQA in jumping to the totally unfounded inferential assumption that this bigger hence 'better' spatial imaging must be due to MQA's very short (hence more 'ideal') time domain impulse response - which they (erroneously) assume must be providing more accurate time domain transient response reconstruction of the original signal waveform, hence more accurate reproduction of the original spatial venue imaging as the recording mikes heard it.
Second, we have heard this vague kind of ambience and bloated spatial enrichment before, and it was always due to an error in the system, or deliberate signal manipulation, both of which are by definition erroneous inaccuracies.
Science knows that human perception of space and ambience can be artificially boosted, beyond an accurate portrayal, by mixing certain types of deliberate inaccuracies (i.e. time domain distortions) into a music recording. For example, the mere addition of random incoherent phase versions of a recording's signal, to the original accurate recording, boosts the human ear/brain's perception of rich ambience and space (the well-known Damaske effect). This artificial effect can be boosted even further by cross feeding 2 stereo channels into each other, especially with polarity inversion, which artificially creates yet more random incoherent phase patterns due the interference being cross channel, not just from the same channel.
And the artificial boosting of space and ambience can be magnified yet even further by the addition of reverberation, which introduces not only time delay, but also introduces yet more complex random incoherent phase interference patterns. Indeed, reverb boxes have been used for decades to artificially create a seeming large room or hall around performers who in fact were close miked - a huge time domain distortion of the signal, and of its space and ambience from the actual recording signal as the mike heard it. This deliberate signal manipulation and distortion crudely mimics what a concert hall does acoustically, creating lots of random incoherent phase interference with long delay paths (hence down to low spectral frequencies), coming at you from all directions, due to multiple reflections from the distant plural hall walls. This is what tells the listener's ear/brain that the concert hall is a large space and is rich in ambience, both live and also when playing back a recording.
It's worth noting that, for the 60 year life of the stereo vinyl disc since 1958, all stereo phono cartridges have committed this same kind of time domain error (to a milder degree), artificially boosting their portrayal of a recording's space and ambience beyond the accurate signal on the master tape. How and why do they do this? All stereo phono cartridges have pretty bad inter-channel crosstalk, which begins in the midranges and gets very bad in the trebles. We discovered many years ago that this bad crosstalk had a peculiar property that no one else seems to have noticed (even to this resurgent heyday of vinyl): the crossfeed signal that channel A badly leaks to channel B is actually an inverted polarity version of channel A's signal (and of course vice versa). Thus, all stereo cartridges actually act as a Carver holography processor, crossfeeding inverted polarity signals, which also artificially creates complex random incoherent phase interference patterns, hence more artificially boosted space and ambience.
We designed a circuit to cancel this cartridge crosstalk, and the sonic results were both stunning and educational. With the cartridge crosstalk canceled, the cartridge was now delivering spatial imaging much closer to our copy of the master tape (which has excellent channel separation, hence by definition is the reference standard for what accurate spatial imaging should sound like from this recording). True, the corrected phono cartridge's portrayal of space and ambience was not as boosted, thus not as large and rich. But the portrayal was now more accurate.
And the nicest surprise was that, in this more accurate portrayal of space and ambience, the vague bloat was gone, and in its place was much better localization specificity. We could now hear where the hall walls were, and how far away they were, and even better hear where the performers were, in relationship to each and to the stage and the hall. In short, the entire portrayed spatial image jelled into reality, instead of being a vague amorphous bloated blob.
That's the contrast we heard in the MQA portrayal of space and ambience, vs. an accurate portrayal.
Third, we know that Columbia had a penchant for very close miking, especially of intimate musical performances such as Bob Dylan's voice and acoustic guitar. It's inconceivable that such a large space with such rich ambience would have been captured by Columbia as the actual mike signal. And this rich spatial bloat is not heard at all from the PCM rendition of this very same master tape.