A Psychoacoustic Rationale for the Subjective Evaluation of Stereophonic Sound Systems by Reviewers and Audiophiles (Ralph Glasgal)

With the discussion of ambiophonics ("ambio") and crosstalk cancellation (XTC) last time, I came across a number of fascinating writings from Ralph Glasgal in the Ambiophonics Facebook group. The one below is one of them I think worth considering that I had not seen before, explicitly written as copyright-free. I've made some mild formatting edits and wording substitutions/additions for clarity as well as emphasis in areas I felt of importance. I trust this mildly edited version retains the full meaning and intent of the original found at the group above. A copy of this has been sent to Mr. Glasgal but I'm not sure if he's reviewing his E-mails or active in audio discussions these days in his 90s.

A Psychoacoustic Rationale for the Subjective Evaluation of Stereophonic Sound Systems by Reviewers and Audiophiles

By Ralph Glasgal, BEP, MSEE, AES, IEEE

glasgal@ambiophonics.org

(original text posted on Facebook - October 15, 2015)

The oft repeated assertion that high end components of stereo systems can only be truly evaluated by listening rather than by measurement or blind A/B testing by listening panels has been argued for decades now. The following text will, I believe, provide a psychoacoustic explanation of why this position is largely correct and not a question of practicing voodoo audio.

Very few audiophiles will want to read what follows since it is rather long, too technical, and not what they are used to reading in Stereophile or The Absolute Sound. But this article is for the record, for discussion forums, and to have it available for future debates. Anybody is free to use or forward this text. It is not copyrighted. Supporting technical references by a variety of authors are available at ambiophonics.org (mirrored).

While analogies are always subject to dispute, refutation, and even ridicule, I will start with an analogy that discusses similar visual subjectivity issues in order to temporarily avoid stirring up deeply held sonic beliefs.

Imagine that a talented still scenic photographer has produced a very high resolution color negative with an aspect ratio that provides a full height 180° half circle if printed or projected. Imagine also that there is a second negative that produces the rear half circle so that if you move your head or turn around you still see a great close-to-perfect 360° panoramic image. There is also a 3D version of this picture for those into hardcore realism. Now a frontal only, 2D color negative, extracted from this original master, is sold to a member of a photography club who wants to reproduce a positive version of this large pixel data file (call it 2.0) to enjoy and show his friends. By preference, he only uses a black and white enlarger or high pixel printer and only ever makes 8 ½” by 11” black and white prints. So when he loads this 2.0 super fine color negative into his enlarger or digital picture program, the color pixels are deliberately eliminated and the aspect ratio is cropped to fit the printed paper size. Some clouds that would be too far to the side and up above to be included are compressed and shifted to fit on the page. Of course, the 3D and rear parts of the original image file (call it 4.0) never enter into the equation.

When he looks at the print sitting in his living room, he decides that if he tweaks the gamma, the contrast, the brightness, gets a higher-resolution developer, and glossy print paper, all rave reviewed by B&W Photo Magazine, he can get more of an illusion of depth and even implied color from this master color negative. So he does this and he indeed sees an improvement. But when he shows it to a friend, the friend opines that the first print really had more depth and implied color and had better resolution because the newer print paper was too reflective and the new developer/toner induced a bit of grey.

Our hero then looks the next day at both prints under his recreation room lighting and decides that neither print really has as much depth or suggestion of color as he first thought. In the end he finds a combination of tweaks that satisfies him and is resigned to the fact that none of his friends agree with his opinion as to the best way to make a now data-poor print, even though of high-resolution, extracted from the original data-rich negative, and still looking like a version of reality or artistic perfection in his mind. 

I hope you get the idea. The print lacked many of the basic elements of visual realism including 3D, color, and an uncompressed aspect ratio. It did have the high fidelity physical elements of resolution, low noise, low pixel distortion, and potentially perfect greyscale. Since any black and white reduction from this relatively perfect negative is not going to provide a normal everyday visual impression to any normal brain, each individual will react to this lack of visual realism in a different way. Also, a slight imperfection in the original full color wide-angle negative like a few missing pixels or a gamma error in one area, will never be noticed in a full color print especially if it is fully panoramic and 3D. But in the black and white print version, the slightest aberration will likely stand out and be quite noticeable. Even without considering the relative absolute realism of the original photograph (nothing is perfect), there is no such thing as a black and white photography system that everyone will agree is the best possible way to do partial justice to such an original high resolution, surround, color, and, likely digital, picture file.

This essentially means that a reviewer of high quality black and white film/printer/projector/print paper/lens/etc. must resort to a subjective opinion on how close a device or tweak comes to producing either the optical illusion equivalent of normal 3D color vision, or more likely some other arbitrary standard of visual perfection once the basic black and white parameters of pixel resolution, contrast, etc. are determined to be excellent. Again, any artifice usually requires subjectivism to judge its success relative to some generally agreed upon standard be that realism or an abstract artistic concept like Film Noir in movies where color is not welcome.

Now in the audiophile world, we have many great 2.0 recordings (LPs, CDs, etc.) that have indeed captured a full 180° wide stage for orchestras, opera, oratorios, movies, etc. Also, we have mostly scorned 4.0 (5.1+) media like DVDs, SACDs, BDs, that have both front and rear very wide sound stages as found in some movies. For music, in both theory and now indeed in practice, these existing media can provide quite a real 3D concert hall acoustic experience. To satisfy the brain that a sound field is real, you need more than just flat frequency response, proper level, low distortion, and perfect resolution 2.0 files to play. In say a concert hall with a wide stage, you have some time differences between head spaced mics of up to 700 microseconds, some level differences up to 10 dB, early reflections coming from behind and all around that change with the stage location of an instrument or voice, and later diffuse but directional hall reverberation also coming from the sides, rear, and overhead (the latter the least desirable because it is mono).

Despite the many microphone arrangements used, all of this was and is now actually captured to a large extent on LPs, CDs, SACDs, DVDs, BDs, etc. So, many of these music files are like the wide angle 3D color negatives that cry out to be fully reproduced. Even recording/mastering engineers have never actually heard all the localization data that is included in their mixes. 

So let us assume we have a great 4.0 recording which has the same TD (time difference), LD (level difference), and early/late reflections that you would have heard had you been at the mic position hopefully at the best seat in the hall. Now we are going to arbitrarily and stubbornly reproduce this file using just two speakers at 60°. We just buy the 2.0 version of this original 4.0 file so it actually has rear and other hall reverb mixed into the front channels so that is what we have to work with. Now we don’t know it, but using two speakers this way changes the data in the file that reaches the ears in a variety of ways. It reduces the original 700μs on the disc to a maximum of about 220 depending on your head size and exact angle to the speakers and also similarly reduces the maximum level difference at the ears in the listening room to less than 5dB even if LD is 10dB on the 2.0 version (i.e. LP or CD) we bought. So now instead of the 180° stage that is on the disc, we hear only a compressed 60° version of it. Also, the early reflection listening room pattern remains the same whether a source is coming from the left speaker, the right speaker, both, or in between. This is sort of a "flat Earth" effect if I have to subjectively describe what it sounds like.

Note that now all or most of the recorded hall ambience is coming from the front rather than from all around as in a real hall. Also, each human pinna acts like a direction finder. So in stereo, the pinnae detect that sounds are coming from both speakers +/-30° regardless of where an instrument is panned to on the recording or where the mid frequency part of the stereo sonic illusion seems to place it. This, with the other errors above, is an impossible artifact for the brain to resolve so it decides that what it is hearing is a sonic illusion, a recording, and a sound field that is not real. But like the black and white print, which the brain knows is not normal, the brain can accept the stereo illusion for what it is and like it or not.

It gets worse! The use of speakers this way causes a series of large but narrow dips and peaks starting at around 1500Hz. These peaks and dips resemble pinna direction-finding patterns and so tend to confuse the brain with contradictory localization cues that are high frequency content dependent and unnaturally static for all instrument positions.

Finally, at a live concert, a cough or a program rattle is hardly noticeable and off in left field somewhere. In a stereo system playing an LP, every tick and pop is front and center. Also front and center now are small amounts of harmonic distortion, and relatively otherwise inaudible minor changes in frequency response. This sensitivity to otherwise minor aberrations is also why some tweaks are inordinately audible. This is not theory. This, and all the above impairments can be demonstrated to be audible and what is better, eliminated. 

So what has this got to do with rating stereo components and tweaks? Well, like in the optical illusion or black-and-white print case, the stereo sonic illusion, because of all its faults, will be heard differently by different listeners. This individualism is amplified in the case of sound because a pinna is like a fingerprint, very individual. Head sizes also vary. The term for this is HRTF - Head-Related Transfer Function. Despite many attempts to produce average functional HRTFs, everyone has a different HRTF. So each one of us responds differently to a particular listening angle, to the mic arrangement or panning used in making the recording, to speaker type, room treatments, cables, etc. Even the same individual will find that a given adjustment sounds one way today and differently next week because the chair has moved, his head is now up, down, rotated, leaning, or a different recording has LD and TD values that after being acoustically processed through the stereo loudspeaker equilateral triangle now have ITD and ILD (I = Interaural) values that are less than satisfying compared to that other recording played last week.

Just as you don’t need to have a "golden eye" to see that a black-and-white print lacks color, so you don't need a "golden ear" to detect the lack of concert hall or other forms of sonic realism in any stereo loudspeaker system. If you rate stereo components and tweaks via 60° loudspeakers based on their approximation to binaural hearing or concert hall realism, this rating must be subjective, and usually such an opinion can only be valid for one individual in a particular listening situation. So, like black-and-white photography, 60° stereo reproduction is inherently a subjective art form. A "golden ear" may have a 'golden' HRTF, but your 'iron' HRTF is what you have to listen with. "Golden ears" are just ears that are able to decide for their own brains how a given tweak balances the tradeoffs in the reality deficit inherent in a 60° stereophonic sound field. Again, a reviewer's HRTF is quite unlikely to be like yours so their observations are unlikely to apply to your own pinnae, head size, listening angle, or just plain sonic preferences.

So for all the above technical reasons, the only logical way to discuss the merits of stereo components, tweaks, and gizmos is indeed subjectively but on a strictly individual basis. It is likely that HRTFs can be grouped into classes where a particular tweak will sound the same to all members of the same class but the number of such HRTF classes seems to be in the hundreds. So again, any suggested tweak to a physiologically unnatural system such as this can only be evaluated subjectively and on an individual basis. So, a subjective comment about the realism or other sonic attribute of a product or a recording in a magazine or blog should not be taken seriously if it was listened to using speakers at 60°. Likewise, double blind tests involving a stereo system for things like realism, depth, clarity, stage width, localization, etc. can never be objective, except for the one individual involved who may then use subjective criterion relatively unbiased by extraneous non-sonic influences.

--------------------

Time for a serious discussion.

IMO, regardless of whether one agrees with Glasgal around the ambio (as opposed to standard 2-channel stereo) approach to speaker playback, I think we can appreciate that this article is at the level of discussions we, as serious audiophiles, need to have when thinking about high-fidelity reproduction. It's obviously very different than the level of Industry-sponsored articles we see in the glossy pages of Stereophile and TAS these days.

It's great that Glasgal framed the subjective experience of stereo listening with an emphasis on the brain being asked to assess a fundamentally unreal-sounding 2-channel stereo playback system. Regardless of how fancy or expensive our speakers, or amplifiers, or source, stereo playback can only be a "data-poor" mimic of the real experience. No doubt, the mind cannot ultimately judge this experience to be fully convincing compared to real world voices, instruments, and other sounds. Nobody accepts that a B&W 2D image is any more than just a shadow of seeing the scene in real life even if there is artistic merit in the artificial reproduction.

B&W classic! Moonrise, Hernandez, New Mexico - Ansel Adams, 1941.
Regardless of how it's printed and on what paper, nobody insists that this is an accurate image of what was actually seen by Adams' eyes that afternoon along the highway, right? Yet we often seem to compare our 2-channel speaker playback with some ideal "absolute sound" of reality.

For the most part, I agree with Glasgal in the article. Everyone has their own subjective preferences and opinions. I certainly don't doubt that idiosyncratic HRTF will have an impact with some preferring the frequency peaks and valleys of one speaker versus another for example. Perhaps the only way to ensure some semblance of a standard is to insist that all reviewers undergo training and certification in a treated room with controlled characteristics like reverb time. This is never going to happen among the audiophile press for all kinds of reasons.

Although I try to be accepting of fellow audiophiles' observations and believe that we're generally honest, I would add to Glasgal's observations with a reminder that we cannot be naïve when we read subjective-only reviews in magazines or watch YouTube videos. It would be foolish of us to ignore the impact of friendships, advertising dollars, financial incentives, and "in kind" benefits on the convenient, unprovable, "subjective" opinions of reviewers in publications or monetized video channels!

Furthermore, I believe it's important to recognize that there are certain classes of tweaks and devices that basically make no difference to sound quality based on rational, sober evaluation. These are the ones that audiophiles will find most controversial - for good reasons! For example, decent quality cables "sound" the same regardless of MSRP; suggesting that there are "Best Interconnects: $10,000 and up" is frankly ludicrous. This stuff IMO is not "audiophile hobby" material, but rather simply targeted at the "high end", low insight, undiscerning listeners who want to show off their frivolous consumerism regardless of astronomical prices.

High-resolution DACs and amplifiers likewise under controlled listening tests more than likely will sound the same once we reach some key resolution parameters (including things like output impedance for amp-speaker matching, which could make a difference). I believe digital "Bits Are Bits" based on experience, so it makes no sense to go nuts into expensive streamers / computers / servers, or bizarre "audiophile" ethernet routers unless one can demonstrate meaningful changes in measurements. Yes, as suggested by Glasgal, I think there are tweaks that can make a difference (mainly in the vinyl/LP/turntable world), just not the ones above.

IMO, "subjective" differences claimed to be significant between listening sessions when no objective changes can be demonstrated in the equipment are more than likely just a reflection of mood/stress, ear wax, head tilt, amount of sinus congestion, psychological brand recognition, etc. rather than actual sonic change that we might want to pursue spending much money on for lack of value.

I'm not sure if Glasgal's last sentence is clear to me though:

"Likewise, double blind tests involving a stereo system for things like realism, depth, clarity, stage width, localization, etc. can never be objective, except for the one individual involved who may then use subjective criterion relatively unbiased by extraneous non-sonic influences."

Indeed, I'm not sure if any of us can be 100% objective about anything especially when judging an ultimately unreal experience. However, if there are significant differences to be heard, I think blind testing will detect those differences among sensitive trained listeners. Is there anyone in this world who can reliably "then use subjective criterion relatively unbiased by extraneous non-sonic influences"? I'm not sure how certain any of us can be unbiased, or not have our subjective criteria molded, when exposed to non-sonic influences! Sometimes, these influences act on the subconscious, tugging at inner ego-driven needs we don't normally express or even acknowledge (like being treated by a respected audio designer or company and made to feel special). There can be very powerful psychological elements at work within an audiophile which we might even be ashamed to talk about honestly due to their primitive nature. We can be sure that advertising dollars seek to influence, even exploit these internal dynamics though. This is why it's important to be mindful about the potential "subjectivisms" out there.

Let's end by pulling back a bit and remembering that this is not to say we don't have the right to buy whatever we want based on our heart's desires, including for those "non-utilitarian"/luxury benefits of owning nice-looking sound systems. An amp with higher distortion and elevated noise floor (like this) will have some audiophiles subjectively enjoying the sound and are willing to pay whatever asking price. That's just fine because it can be subjectively "euphonic" for the individual, even if we would likely not call such a device particularly "high fidelity" based on modern objective standards.

Thanks for the fascinating ideas and writings over the decades, Mr. Glasgal.

Remember audiophiles, if you're using a 2-channel stereo system, have the ability to rearrange your speakers' listening angle, plus have DSP capabilities, I highly encourage you to try out ambio playback as discussed last week. Make sure to give Queen's "Bohemian Rhapsody" a try, tracks like Pink Floyd's "Time", or more contemporary material like say Billie Eilish's "bad guy" (she's whispering in your ears), "Colombia, Mi Encanto" off Encanto, Jean-Michel Jarre's Oxymore, or Shpongle's stuff, to hear the space open up and speakers melt away even more than they might be already in your sound room!

Are You Shpongled? (1999, DR8) is a pretty wild ride with XTC/ambio - check out "Monster Hit".

I hope you're enjoying the music, dear audiophiles.

Addendum: October 18, 2023

Found this other post from the Ambiophonics Facebook Group from Glasgal - written April 4, 2020; again with some light editing. I trust he wouldn't mind I repost here:

"The two-channel cul de sac" - This is the title given to the final paragraph of an article by J Gordon Holt entitled "Space... The Final Frontier" Stereophile Magazine, March 1994:

"As long as we remain stubbornly committed to two-channel stereo, further advancement in reproduced realism just won’t happen. Sure we can continue indefinitely to tweak what we have now, for a minuscule improvement here and a subtle improvement there. And while such endeavors are worth-while, it’s time we faced the fact that trying to reproduce 3-D space from a 2-D system is ultimately futile. Surround-sound is the only way to do it. Many audiophiles will prefer the familiar to the challenge of the new. But the world will pass by those audiophiles who insist that two-channels up front are all they will ever need to reproduce acoustical space."

After moving to The Absolute Sound magazine, Gordon wrote in issue 133:

"I’m not saying Ambiophonics can't make stereo material sound magnificently rich and spacious. In fact, Mr. Glasgal's demonstrations of Ambiophonics almost never fail to wow even the most loyal two-channel partisan, although this has a much to do with the criminal ineptitude of most surround demos as it does with the excellence of Ambiophonics. Glasgal's surround demos allow wrap-around spaciousness of surround to speak for itself: most demos hit you over the head with it, to make sure you notice it's there."

In a personal meeting with Gordon, he stated that he believed the right technology for music surround was matrixing à la Dr. David Griesinger of Lexicon. So Gordon wrote:

"My main reservations about Ambiophonics are that (1) it seems an absurdly complicated way of doing what other systems designed specifically for stereo recordings (like Citation's 6-axis, Lexicon's Logic 7, and Meridian's Trifield processors) do with disarming directness, (2) its surround field is rarely likely to mesh with the front soundstage ambience, and (3) it adds reverb that never existed during the recording session, to recordings that probably already have more than enough reverb of the own recorded on them."

Needless to say, matrixing schemes faded gracefully as SACD, and other 5.1 discreet channel media have become standard. It surprised me that Gordon could so steadfastly believe that ambience extraction from encoded 2.0 media would work any better that the earlier similar Quadraphonic matrixing scheme for direct sound. It is also not true that the front stereo pair has valid ambience that makes sense coming from a frontal direction. Also such frontal ambience will sound like an echo if present at concert hall levels. I would also maintain that now it is a lot easier to convolve a real hall impulse response then extract ambience enmeshed within a direct sound media channel. But, to coin a few phrases, "hindsight is always 20/20", and "time marches on". The picture shows the domestic concert hall as it was when Holt wrote his comments.