A new theoretical model for stereo imaging

[Keith_W]

AFAIK there is no thread on ASR to discuss Gary Eickmeier's @geickmei article in the Sep 2025 edition of AudioXPress so I thought I would summarize it. Gary posted a brief summary of his article here which includes a link to purchase the magazine - it is USD$15 (EDIT: AUD$15, USD$9!). IMO, if the reasoning in this article is sound, then it has major implications for speaker design and listening room setup. I will generate my own images to avoid copyright issues.

And another thing: I neither agree nor disagree with the author. I haven't made up my mind! I made this thread because it might generate interesting discussion. I strongly encourage interested readers to purchase the AudioXPress magazine and read it yourself. Don't worry, AudioXPress is not a subjectivist rag, a few of their writers are here on ASR. Including @SIY (speaking of which, where the heck is that grumpy dog?!? I haven't seen him for months!).

The article starts off by discussing current audio paradigms. He points out that loudspeakers are positioned with respect to our ears, not the room - the "stereo triangle". In this model, the axis of the loudspeaker is what we want to hear, and the reflections are a "nuisance variable".

He also makes a distinction between "field type" systems and binaural systems. A "field type" system has to be specifically recorded for playback on stereo speakers or multichannel surround - this is the vast majority of available recordings. Likewise, a binaural system needs to be recorded using a dummy head and played back on a binaural system - headphones, or speakers set up for binaural reproduction with crosstalk cancellation and suppression of reflections. He points out that in a "field type" setup, we do not need to worry about the mechanism of human hearing (HRTF, ILD, ITD, etc) because these are strictly binaural concerns. The two are not the same, he points out that it is erroneous to think that crosstalk cancellation is the key to fixing imaging problems which arise in field type recordings.

He then proposes a new model for field type setups:

In this model, speakers in a room can be thought of as a lamp in a room where every wall is covered in a full length mirror. Whilst a hall of mirrors will generate an almost infinite number of reflections, it is different with speakers. Only the first and second order reflections are important for imaging, because the other reflections die off and become more diffuse in time. It is argued that the position of these phantom speakers (my term, not his) is responsible for apparent soundstage width (ASW) - again my term and not his. All imaging takes place within the area bound by the two actual speakers and the six phantom speakers as shown.

It can be seen that the radiating pattern of the speaker, and the position of the speakers relative to each other and to the room, has a major influence on the strength of the phantom speakers.

- a highly directive speaker will generate weaker phantom speaker images and cause the ASW to narrow and localize between the two front speakers,
- a speaker deliberately designed to radiate most of its energy rearwards (author cites Bose 901) will generate a strong reflected sound, causing the ASW to localize further away from the front wall,
- placing the speakers closer to the front wall will flatten the depth of the soundstage by decreasing the distance between the actual speaker and its mirror image,
- placing the speakers further apart will increase the "hole" in the centre of speakers.

The author points out that this has major implications for loudspeaker design. Speakers now become "image model projectors" - in his own words, "we can easily see how loudspeaker design is mostly about radiating pattern, frequency response becomes power response, and the "axis" becomes a thing of the past".

He then proposes a new type of "variable directivity speaker" - a speaker with drivers front and rear. The gain of the rear drivers should be adjusted so that it is 6dB louder than the front driver, and spaced appropriately from the front and side walls.

My comments:

- the article does not comment on other phantom speakers, for example from the rear wall, ceiling, and floor. One would imagine that these would be ruinous to stereo imaging so should we attenuate them?
- I am not so sure about the "axis becom[ing] a thing of the past". There is still the Law of the First Wavefront, so the axis is still important.
- there is no mention in the article about how this changes our strategy for deploying room treatment like diffusers and absorbers. I would imagine that if all this is correct, then we do not want any sound absorption because this will attenuate the position of the phantom speakers.

I hope I have done a good job of summarizing the article and not made any mistakes or misrepresented the author. By necessity I have left a lot out. I found the article thought provoking and well worth my money.

 

[Flaesh]

Bipolar speakers are rare, and additional rear drivers are occasionally used. Omni speakers aren't very popular also. Is 6dB of rear gain necessary and sufficient to produce wonderful, marvelous, miraculous, magical sound?

 

[ZolaIII]

Back to front refractions ratio, ISO 3382-2 ordinary rooms or ISO 3382-3 open offices... and when you can't close to the wall to minimise timing difference. Really nothing new, just wrapped up nicely to take you 15$.
Its not really about perfect triangle nor it should be but same distance (additionally compensated) between speakers and listening position. In my experience double chamber absorbers across the wall and with sizable depth improve ISO 3382-1 to a point you can't achieve even pulling them out a lot in large room. Length of the room is still imperative so I will critique your change of orientation as long as I live (until you change it anyway).

 

[Pareto Pragmatic]

He points out that loudspeakers are positioned with respect to our ears, not the room - the "stereo triangle".

So I must have imagined all that stuff about speakers not being too close to walls or corners. And I must have dreamed something about a "rule of thirds" I guess, which I am sure no one else has heard of, it being from dreamland and all.

Sorry, but a false premise right out of the box makes me snarky.

Room --> speaker placement --> seating position. That's how it goes in my world.

 

[EERecordist]

Recording remix rooms are often designed with a live end / dead end philosophy. The front of the speaker cabinet is a hard surface, the glass to the studio is a hard reflective surface, and you will often find hard and articulated surfaces in the front of the room, including the console itself. Then the sides, back, ceiling, and floor have absorbing surfaces. If you have the money you will have bass traps.

The mix engineer lays out the stereo image alternating between the near field and far field monitors, including adding echo.

The mastering engineer for stereo applies slight adjustments to EQ and compression for the publication medium: CD, vinyl, tape, and the many variations on streaming. The mastering engineer does not alter the layout of instruments and microphones across the speakers.

Remixing for spatial would affect the spatial image, and spatial remixing often has access to the stems/tracks or may use AI instrument demixing. From there echo is applied in the spatial array.

This setup is similar to a classic concert hall with an orchestra shell to send the music, instruments or voices, toward the audience and not up into the fly space or stage wings, Then the front of the hall is ornamented by plaster curly cues which are the live end and produce a complex sound field from the individual instruments. At least the floor of the hall is somewhat dead, and sometimes the sides and back with loges and balconies because the chairs and audience are absorbers. The golden era of concert halls still remaining was the mid1700s to the 1930s. The latter period includes movie theaters, many retain the proscenium and front of hall complex ornamental hard diffusers but make other audio compromises. To the early hall designers it looked good, and it sounded good, perhaps by accident at the beginning.

In my opinion, the mystique of the "stereo image" or "sound stage" is defined by the degree to which the listening room is similar or different from the mixing room where the image was made, by combining instrument spot mics, and distance mics, including mixing distance mic arrays to 2 channels.

From reading the summary only, I disagree with Mr Eickmeier. He has been at this theory for over 35 years and has written extensively on many platforms, including his original 1989 paper at the AES you can find on his blog.

 

[krabapple]

So I must have imagined all that stuff about speakers not being too close to walls or corners. And I must have dreamed something about a "rule of thirds" I guess, which I am sure no one else has heard of, it being from dreamland and all.

Sorry, but a false premise right out of the box makes me snarky.

Room --> speaker placement --> seating position. That's how it goes in my world.

Gary Eickmeier is no hack. He''s been experimenting with 'bipolar' speaker designs for decades...and has produced some surprising results

He does address speakers being too close to walls -- it's literally discussed in Keith W's summary:

- placing the speakers closer to the front wall will flatten the depth of the soundstage by decreasing the distance between the actual speaker and its mirror image,
- placing the speakers further apart will increase the "hole" in the centre of speakers.

Did you even read the whole thing?

 

[Juhazi]

Dipoles work that way... But utilizing reflections like Eickmeyer is very delicate, practically unpractical . Fortunately we have multichannel systems and processors for that, and automated setup for spl and latency.

Google scholar found only two citations of the original AES paper.... https://scholar.google.com/scholar?hl=fi&as_sdt=0,5&q=gary+eickmeier&btnG=

 

[Pareto Pragmatic]

Did you even read the whole thing?

While I recognize that a summary of what someone said is not what they say... no. I stopped at the point where what I read was way off.

I sometimes miss good material with such snap judgements, true. But the time saved not wading through crap more than offsets it.

In this case, I was way off base. Thanks for taking the time to point that out to me.

 

[Duke]

Here's my analysis of Gary Eickmeier's approach (I'm leaving a few things out for the sake of focusing on what I think the main points are):

Let's imagine that you care a lot about spatial quality, and in particular you really like it when the soundstage width extends beyond the speakers. It's like the music is detached from the speakers, and the illusion is more convincing and enjoyable without the distraction of the soundstage stopping at the speakers. And suppose you read Floyd Toole's book and learn that this happens because strong early same-side-wall reflections expand the Apparent Source Width in the direction of those reflections. You even read where this is an attribute that most listeners enjoy and prefer to have.

Suppose you start to wonder whether this same principle could be used to expand the soundstage in other directions. If so, obviously the speakers would need to create strong first reflections in those other directions, and you'd want those reflections to have the correct spectral balance. Maybe you experiment with a dipole or bipole speaker and find that, yes, the soundstage DOES expand in the directions of the additional spectrally-correct and sufficiently-loud reflections.

So now you want to optimize. This includes 1) figuring out what the preferred radiation pattern should be; 2; figuring out how loud the sound going into those reflections should be; and 3) figuring out what the best speaker locations would be.

Well, Gary has already gone through this process. If I understand correctly, he finds that a bipolar or dipolar pattern works best, but with the rear-firing radiation about 6dB louder than the front-firing radiation. This 6 dB compensates for the fall-off with distance due to the reflections having a significantly longer path length than the direct sound. And the preferred positioning is shown in @Keith_W's opening post in this thread.

Imo Gary's approach is innovative, valid, well thought-out, and based on the same well-known principle that expands the Apparent Source Width in the direction of the first strong sidewall reflections.

 

[krabapple]

Well, Gary has already gone through this process. If I understand correctly, he finds that a bipolar or dipolar pattern works best, but with the rear-firing radiation about 6dB louder than the front-firing radiation.

Hmm, I wonder if there is a well-known brand (but anathema among audiophiles) that does something like this in its flagship model?


Note too that ASW is one of factors people tend to overlook as foundational to Toole's conception of good home audio sound*. Like 'envelopment', it is something most tested auditioners want, ....but squirrely types like, say, online audio forum readers, might be outliers.


* the most common being his disrecommendation of room eq etc above Schroeder if you are using speaker that measures great in a Klippel/anaechoic chamber -- people forget those italicized conditions.

 

[Bwaslo]

I have not read the paper, only the summary that starts this thread. But one thing was mentioned that makes me reply, because it relates to something I've set up in my system (adjustments still progressing).

A speaker with additional driver(s) on the back to bounce sound off boundaries?? Is this the 1970s or what? If this kind of approach is worth doing, why not use modern technology - the reverse side or an otherwise (preferably directional) driver can be driven by electronically delayed signal, even also convolved with impulse responses of a larger more diffuse room. Somewhat related to Duke's "Late Ceiling Splash" developments but with benefit of (now economical) additional amplifiers and DSP, and "splashing" off other surfaces besides ceiling. Because of sufficient delay and hall convolution, it can be done without changing the general tone of the sound or losing sharp imaging, particularly with constant directivity speakers such as Ascilab C6B (what I use). The advantage is that it can put the listener sonically in an environment similar to live music where sound comes at him from all directions (rather than like just through an open window in front).

(Of course, this is all assuming stereo. More advanced would be multichannel with Atmos and actual recorded directed signals. But for many, myself included, stereo is the primary interest because of established music collections, and because we don't feel comfortable in a room with speakers or other audio gear everywhere -- the last thing I want is to have it so evident that it is all contrived and pointing at me!).

The processed splash scheme is probably considered anti-purist if not heresy by most (but isn't all recorded and presented audio very contrived anyway?). I don't mean something like a 21st century Bose 901 (strong, vague unfocused effect that can't be turned off), but more subtle, adjustable and configurable, set to be just perceptible enough to allow a listener to feel he's in a room the music is played in. Not musicians in his room, nor him in the recoding venue, but at least both in some contrived room together.

 

[sfogg]

A speaker with additional driver(s) on the back to bounce sound off boundaries?? Is this the 1970s or what? If this kind of approach is worth doing, why not use modern technology - the reverse side or an otherwise (preferably directional) driver can be driven by electronically delayed signal, even also convolved with impulse responses of a larger more diffuse room. Somewhat related to Duke's "Late Ceiling Splash" developments but with benefit of (now economical) additional amplifiers and DSP, and "splashing" off other surfaces besides ceiling. Because of sufficient delay and hall convolution, it can be done without changing the general tone of the sound or losing sharp imaging, particularly with constant directivity speakers such as Ascilab C6B (what I use). The advantage is that it can put the listener sonically in an environment similar to live music where sound comes at him from all directions (rather than like just through an open window in front).

(Of course, this is all assuming stereo. More advanced would be multichannel with Atmos and actual recorded directed signals. But for many, myself included, stereo is the primary interest because of established music collections, and because we don't feel comfortable in a room with speakers or other audio gear everywhere -- the last thing I want is to have it so evident that it is all contrived and pointing at me!).

The processed splash scheme is probably considered anti-purist if not heresy by most (but isn't all recorded and presented audio very contrived anyway?). I don't mean something like a 21st century Bose 901 (strong, vague unfocused effect that can't be turned off), but more subtle, adjustable and configurable, set to be just perceptible enough to allow a listener to feel he's in a room the music is played in. Not musicians in his room, nor him in the recoding venue, but at least both in some contrived room together.

I wouldn't think you would need to use ambiance generation, 2 channel music has plenty of recorded info about the space already in it. Just need to break it down to source, first reflections and reverb. QLI does this but needs a room full of speakers to reproduce the hall sound in the recording. Pairing that processing with some form of beamforming soundbar to add all the additional 'channels' would likely make the in room speaker requirements more manageable.

 

[RexrothPigeon]

Well, Gary has already gone through this process. If I understand correctly, he finds that a bipolar or dipolar pattern works best, but with the rear-firing radiation about 6dB louder than the front-firing radiation.

Did I miss the controlled testing that led to this finding? (genuine question)

 

[Duke]

Did I miss the controlled testing that led to this finding? (genuine question)

I don't know what Gary's investigative process was, but his ideas have been subjected to controlled blind testing evaluation.

Gary submitted a DIY design of his as one of three speakers evaluated for spatial quality in a controlled blind test. The test was in response to this challenge issued by Siegfried Linkwitz:

"The challenge [is] to find the optimum radiation pattern and placement of stereo loudspeakers in a room for the creation of phantom sources and simultaneous masking of real sources."

Gary's entry won. It beat the dipolar Linkwitz Orion and a monopolar Behringer speaker that stood out for its surprisingly good measured response.

You can read an overview of the test here. I've only read the first post in that thread.

The test itself is described in detail here.

 

[Duke]

Somewhat related to Duke's "Late Ceiling Splash" developments but with benefit of (now economical) additional amplifiers and DSP, and "splashing" off other surfaces besides ceiling...

Giving credit where credit's due, it was my colleague Jim Romeyn who first thought to take advantage of the vertical dimension, which increased the path-length-delay for our rear-directed driver(s). Our latest iteration is independently aim-able which indeed enables "splashing" off other surfaces besides the ceiling. But it sounds like you already figured out that, and more, on your own!

 

[Flaesh]

One local audiophile (who recently died suddenly, relatively early) installed his Lowthers in his attic, rear-facing.
Another local audiophile had been ranting online for years about the benefits of reflected sound, I advised him to install his speakers rear-facing; this guy also recently died of old age .

 

[RexrothPigeon]

I don't know what Gary's investigative process was, but his ideas have been subjected to controlled blind testing evaluation.

ok so not exactly
a speaker he made according to his theories beat out (A) a linkwitz orion with no sub or EQ (lol) and (B) a terrible Beringer speaker in a test with 13 listeners. I'm no expert, but it seems to me they did a commendable job trying to produce a controlled test, so that's good.

The test unequivocally can't support this claim, however. And that's what I was asking about.

he finds that a bipolar or dipolar pattern works best, but with the rear-firing radiation about 6dB louder than the front-firing radiation

 

[Flaesh]

I agree with Earl, the test procedure and criteria are not very convincing.

 

[kyuu]

The notion that playing around with reflections somehow negates the importance of the on-axis response doesn't jive for me. Unless the point really is to just replicate the Bose 901 experience of a very diffuse mess.

Also given that most people's rooms are not rectangles with perfectly and evenly reflective walls, and the notion of pulling the speaker out from the wall by some feet isn't going to fly in most setups (hell, mine are inside the wall) makes this difficult to translate even if we take it at face value.

 

[Duke]

The test unequivocally can't support this claim, however. And that's what I was asking about.

It sounds to me like you have objections.

Are you objecting to Gary's approach because it has not been proven to your satisfaction, or are you objecting because you have reasons for believing that it will not give good results?

Or, have I misunderstood you?