Small Room Acoustics: Example of impact with (ultra) high-end speakers in Stereophile. (On AI audio writers, and Asia vacation pictures. 😊)

Last time, we discussed small room acoustics and audiophile listening rooms. In that article, I used an idealized loudspeaker with flat frequency response from 40Hz up. As I'm still traveling, away from my gear, I figured for this article, let's look around published data in audio articles to examine room effects and consider what this means.

One widely available and consistent source of objective data over the decades comes from Stereophile. While there are clear examples in the magazine where they perpetuate the "woo" in audiophilia, I appreciate that they do provide measurements for audiophiles who can read graphs and consider the technical performance of devices beyond subjective claims of how "great" something sounds, at some point in time, to someone, in some nebulous-quality sound room.

So let's take a look at a very recent Stereophile article - this one reviewing the Marten Mingus Septet Statement Edition loudspeakers - and consider the effects from room acoustics. Clearly, at an asking price of US$200k/pair, we're in the zone of "ultra high-end" luxury audio hardware. These speakers would be way above and beyond what the vast majority of audiophiles (never mind "music lovers" at large!) would ever spend.

From that article, and for ease of comparison, let's examine a couple of graphs together to see the change between the pseudo-anechoic measurement and what the frequency response looks like in the reviewer's (Jim Austin's) room for these speakers: 

If you haven't looked at such graphs in the Stereophile Measurements section of reviews in detail before, I hope you find the striking difference in shape of the curves interesting since that top graph and the bottom red curve are the same speaker but under different conditions/environments.

The upper graph shows pseudo-anechoic frequency response (gated sweep plus nearfield drivers and ports); how the Marten Mingus Septet speakers perform within the 30° horizontal listening window when you take the room out of the equation. Notice that despite all the positive comments stated by Austin's subjective review about the crossovers, time-domain performance and fancy drivers, there are quite a few irregularities around the crossover points (midrange said to cover between 800Hz and 6kHz).

We can compare this pseudo-anechoic measurement with other recently-reviewed floorstanding loudspeakers like the Magico S2 ("only" ~US$38k/pair), and EgglestonWorks Andra 5 (also ~US$40k/pair), both achieving smoother frequency response. We use these (pseudo-) anechoic measurements to provide a reasonable "apples to apples" comparison of speaker tonality with room effects taken out.

[A comment on Jim Austin's article "Time Is On My Side". Of course time performance is important for high-fidelity playback. Temporal precision can affect things like inter-aural delay between channels to give us more precise 3D soundstaging. Likewise, timing is essential to convey precise attacks and decays.

BUT. Physiologically, human hearing is mostly judged on frequency response because the cochlea has a tonotopic organization. The ear acts like an FFT diagram where frequencies will specifically stimulate hair cell populations in the ear based on frequency. High frequencies stimulate the stiffer and narrower base of the cochlea, whereas low frequencies affect the apex of the cochlea where it's wider and more flexible.

While it's ideal to achieve precise frequency and temporal performance, since the frequency response is significantly more important for human hearing, this means it should not be sacrificed. Even back in 1998, John Atkinson commented on the lack of subjective preference for loudspeakers with accurate "time-coincident" impulse response. Fast forward to 2026; if time-coincident behaviour was essential, it would have been a standard design character of any high-fidelity loudspeaker over $10,000! This is obviously not the case.

Simply put, human ears are not very sensitive to time-domain performance of loudspeaker crossovers where the impulse response delay between drivers are typically an invariant <5ms.]

The lower graph is the frequency response in Jim Austin's room. Notice how different this looks once you actually place the Marten loudspeakers in an actual room!

Whereas the Marten Mingus Septet are already far from perfect in the quasi-anechoic frequency response (with ~10dB range below 4kHz), look at the irregularities now in his room with >15dB bass fluctuation from 20Hz to 200Hz. While we don't know the sound room's dimensions, we can clearly see the effects in the "Modal Zone" and irregularities in the "Transition Zone" likely a combination of  SBIR, LBIR, surface bounce and the presence of other furniture. Higher up in the quasi-diffusion frequency zone, it's good to see the steady-state frequency response averages out the pseudo-anechoic anomalies with some expected roll-off by 20kHz.

Conveniently, that in-room graph also shows the Wilson Alexx V's (MSRP ~US$140k/pair in 2021) response in the same room although position and toe-in would certainly be different to some extent. Comparing these graphs at the listening position, clearly we would expect the tonality to be quite different between the two when listening. The Martens having a more recessed mid-range and more energy above 5kHz, whereas the Wilsons have more midrange (vocal) presence with about a -6dB/octave tilt from 4kHz, potentially perceived as more "polite" when fed harsher and sibilant recordings. Hopefully, the Wilson Alexx V still maintained enough "brilliance" so the tonality isn't too bass/mid-heavy given that -5dB by 10kHz to -10dB at 20kHz dip. Much of this must be judged based on subjective listener preferences of course.

The power of the room cannot be understated when we see both the Wilson Audio and Martens have very similar bass response with strong peaks at 35 and 65Hz. Clearly, if one spent $140k-200k on speakers like these, it would be imperative to get some bass traps as a minimum to better control this coloration.

Over the years, I have suggested that if an audiophile is to seriously review products purely based on subjective listening, at least a couple of factors should be accounted for, and better yet, disclosed for readers:

1. Confirm that their hearing acuity is good - a simple audiogram would be nice. I would suggest that this is especially important if the reviewer is over 65 years old - I think I'm being very liberal already with that age.

2. Understand the room that the reviewer is listening in. It's not unreasonable to let readers know to what extent the space has been optimized enough such that the loudspeakers have been allowed to show their potential, right?

I find it disturbing that in these magazine reviews, the "Associated Equipment" supplementary material typically include all kinds of stuff that really have minimal, if even any, audible relevance (assuming competent set-up) - like the Innuos server-streamer (see streamer bit-perfection). Furthermore, is it not silly to be listing stuff like the AudioQuest power conditioner, Innuos ethernet switch, and the mind-numbing string of cables used?! We can understand why this stuff is listed within the cultural context of audiophilia and the consumerist drive, right?

But where in all of these reviews over the years do we get a sense of the reviewer's ears as being adequate for the task of evaluating loudspeakers "worth" $200k in sound quality? And where in all of that does he mention the nature of his room, its size, whether it's treated, how the speakers are oriented, and where he's sitting in that space?

Interestingly, there are these nuggets in the text to the Marten review that help us learn a little about the room Jim Austin listened in:

I set up the speakers using Marten's fairly specific guidance—a roughly equilateral setup with at least 10' or so between the speakers and the listener, with the speakers toed in. I made sure they were more than 32.5" from the front wall and "not ... close" to the sidewalls, as advised. In-room symmetry was the one part of Marten's setup directions I was unable to comply with. As always in my room, the setup was asymmetric, with the right speaker some 3' from a record shelf (a decent diffuser at high and high-mid frequencies) and the left speaker very far from any wall: In fact, there's a dining area between that speaker and the nearest wall. I toed the Septets in just a little at first, but once I learned about the first-order crossovers, I aimed them straight at the listening seat so that most of the sound I was hearing was direct and on axis.

And later:

<While at Rockport Technologies' demo room> I noticed the soundstage right way. It extended a few feet beyond the sidewalls and very far back—it was deep—but it ended precisely at the front wall. I already knew that moving speakers out from the wall would expand image depth, but the precise correlation with the position of the front wall was new.

Inspired, back in New York, I moved the Martens a couple of feet farther out from the front wall. I moved the listening seat back by about the same amount. After some fine-tuning, the speakers ended up just under 11' apart and 10.5' from my ears, precisely level, and toed in toward my ears. I'd never listened with speakers so far out into the room before, with any speakers—mainly because it was now impossible to sit at one of the places at the dining table.

So, the reviewer is listening to these US$200k/pair speakers in a room that is:

1. Not a dedicated enclosed sound room but is shared with the dining area.

2. Highly asymmetrical. Admittedly doesn't conform to manufacturer recommendations.

3. His revised listening position involved moving speakers further away from the front wall (I'm guessing around 5 feet?), but this meant pushing his listening seat closer to the back wall. How far back is the sitting position to the back wall?

4. We don't know about the state of the room treatment, but by the look of the in-room response with the strength of those bass peaks, it doesn't look like he used many bass traps nor DSP correction to tame the bass. Also, if this room is in New York, what's the ambient noise level?

Well, as luck would have it, we actually do have a glimpse of this room from back in 2021 published in this interview Austin did for the Polish magazine HighFidelity:

I took the liberty to apply wide-angle lens correction to the original image. Wilson Alexx V loudspeakers are 161cm tall. Looks like 10' ceilings which is good. And significantly more spacious than Michael Fremer's basement with his Chronosonic XVX, check out the in-room measurements in that space.

In discussion forums, it's common for audiophiles to share pictures of the rack, components and speakers but many will not show the room size, seating position or the location of the back wall in their photos. There has always been an inherent focus on the "stuff" rather than the environment which plays such a profound effect on the sound.

As you can see, in 2021, Austin was using the Wilson Alexx V speakers. While things might have changed since the picture, there is no sign of audio panels used to absorb reflections. The hardwood floor has a rug over it that doesn't appear particularly thick. I presume there are no ceiling absorbers. The speakers are pushed quite close into the corners, basically against a partial wall (looks like kitchen counter height?) on the left, and there's a bit more space behind the right speaker but there's a reflective window back there. Indeed, it's quite an asymmetric layout already, made even more uneven with reflective/diffusive LPs on the right side with speakers a few feet away, but on the left, I presume there's a comparatively deep dining area that would act absorptive. One also wonders about the RT₆₀ in such a room. What's the R-L balance like in such an asymmetrical room?

If this was just another audiophile hobbyist's room, I obviously would not care to drag all this stuff up with the measurements in that space and concerns about what they show. However, this is Stereophile, they published the measurements for readers to consider, and even Austin said in the article "Like most audiophile folk, reviewers are always in pursuit of better sound, for our pleasure and to make our systems better reviewing tools." - what kind of "reviewing tool" is that room? At least for us in North America, I consider Stereophile the best audiophile magazine so I'm a bit disappointed that this is the sound room of the Editor-In-Chief who has been charged since 2019 to not just overseeing operations, but also review some of the most expensive gear out there including 6-figure loudspeakers.

Imagine a top-tier supercar reviewer writing pages about a multi-hundred-thousand dollar vehicle and telling us how amazing it is but he only drives along city streets with no access to a proper race track (or maybe access to the unrestricted portions of the Autobahn). In the same way, assuming that these ultra-high-end speakers are capable of a special level of sound quality (not necessarily true), without a treated, dedicated, confirmed quiet room, how can the reviewer claim he's hearing anything close to their potential? At the very least, what do we make of any comments he might have about the bass quality considering the low-end coloration?

When I look at the picture above with the Wilson Audio Alexx V, I can appreciate the "cool" looking pair of speakers in the same way that a Ferrari might attract my attention when driving by. However, very quickly I'll get distracted by the poorly matching environment, thinking... "that's a shame". 😕

I encourage you to make sure to optimize your sound room, dear audiophiles. As I've suggested over the years, high-fidelity and high-end isn't about putting thousands and thousands of dollars into all kinds of fancy boxes, or cables. The vast majority of the money dedicated to "hardware" needs to be placed into speakers and sound room. Megabuck speakers demand a partnering room of the same class; anything else would be unevenly yoked.

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

BTW... About using AI writers.

I found it interesting in the HighFidelity 2021 article when they asked Jim Austin a question about AI, and his response (WP is the interviewer Wojciech Pacuła, emphasis Austin's):

WP Do you think that we, audio writers, are going to be replaced with an AI algorithm in the future?

JA I'm quite familiar with the AI world, as my son deals with it (at Google).

I get the impression that we are secured against artificial intelligence, as it is a very human undertaking. It is all about human experience. Inner experience. And if there is something that we know about artificial intelligence is that it is inscrutable. It often gives us the right answers, but we almost never know why. Artificial intelligence may have its models in the human brain, but it is alien to us.

I suppose it is possible that artificial intelligence might convincingly forge an audio review one day, but you need a cynical approach to believe that this may really happen – and I am not cynical.

Hmmm, in 2026 now, 5 years later, the idea of an AI writing purely subjective audio hardware reviews is clearly not far fetched at all! As discussed a few years ago already, "creative" writing is something AI can excel at. Let's write a 3-paragraph review of the Marten Mingus Septet in less than 1 minute [NOTE: What follows is AI-generated for illustration and may not be completely accurate]:

The Marten Mingus Septet: Transcendence at Six Figures

There comes a point in every serious audiophile's journey when one encounters a loudspeaker that simply is music — and the Marten Mingus Septet, at $200,000 per pair, is unquestionably such a transducer. Standing 126 cm tall and tipping the scales at 95kg per cabinet, the Septet employs a four-way configuration: dual 8" aluminum honeycomb woofers, a 7" ceramic mid-bass driver, a 3" beryllium midrange, a 1" diamond tweeter extending response beyond 50 kHz. The hand-rubbed piano-black lacquer (seven coats, in case you were wondering) catches light like a still pond at dusk. Set up in my room with 1.8m behind them and toed in approximately 12 degrees toward the listening seat, driven by a pair of 200-watt monoblocks of suitable pedigree, the Septets vanish into the room with an effortlessness that borders on the supernatural. The soundstage materializes not so much between the speakers as through them — extending well beyond the outer baffles and reaching backward into a holographic depth that makes one reconsider what stereo reproduction is actually capable of.

Spinning Bill Evans's Sunday at the Village Vanguard (Riverside, 1961) was the moment I knew I was hearing something genuinely different. Scott LaFaro's bass — that impossibly fluid, conversational instrument frozen in time just ten days before his death — emerged with a tactile presence that revealed individual finger movements on the fingerboard, the subtle string buzz against the wood, the way the room itself responded to each note. On my reference $75k floorstanders, LaFaro is wonderful; on the Septets, he is present, occupying a specific cubic volume of air with weight and intention. The audience ambience — the clinking glasses, the half-suppressed coughs, the room tone of that long-vanished Greenwich Village basement — formed a believable acoustic envelope around the trio rather than the vague "atmosphere" lesser speakers produce. Paul Motian's brushwork on the snare drum had genuine textural information: I could hear the wire bristles against the head, not merely the impact, but a silky, textural sweep that cuts through the silence.

Switching to Herbert von Karajan's 1962 Berlin Philharmonic recording of Beethoven's Symphony No. 9 (DG) confirmed the speakers' command of large-scale orchestral material. The opening pianissimo tremolando of the strings — that primordial mist from which the first movement gradually coalesces — was rendered with an almost unsettling delicacy, individual desks of violins discernible within the collective shimmer. When the full orchestra and chorus erupted in the finale, the Septets handled the dynamic transition without the slightest trace of compression or driver strain, and crucially, without the bass overhang that plagues even excellent speakers at this scale. Whether these improvements scale linearly with the $125,000 price differential over my merely excellent reference speaker is, of course, a separate question — one each prospective owner must answer for themselves, ideally with their wallet at a safe distance from the showroom.

A reasonable first draft for a minute's work without even "persona prompting" the AI or applying specific constraints. The audiophile editor can then review and confirm the details, check the musical passages it speaks of, enhance the language. IMO, the writing as an intended vehicle to increase awareness of this product to the reader and how great it sounds would be the same whether human vs. AI-generated given how superfluous subjective reviews have become at this point.

While it's still early in the adoption phase of embedding AI into the workflow of technical work (like in actual science and medicine - see 2026 Stanford AI Report), I don't for a minute think that using AI to generate subjective audiophile reviews which are low risk if inaccurate, based on minimal objective grounding, with acceptance of hype, has not already seeped into much of what we read. Without a measurements section, magazines like The Absolute Sound and HiFi+ could be written very efficiently with mostly AI-generated text.

Who knows what the percentage is currently, but it's probably not unreasonable in 2026 to say that about half of superficial, uncritical, consumer content out there used AI at some level. I also wonder whether the publications have clear guidelines about the limits of generative AI used in their articles.

[To be clear, I don't think there's anything wrong with using AI in the writing process. Depending on the article, I'll sometimes create an outline with the help of AI to make sure I convey the most important ideas. The actual articles have been created the old fashioned way though. 😄 I've been using AI image generators for years now. This is mainly as visual guides like the title image and to help space the text into the various sections.

In the short term, my main concerns with using AI are inaccuracy and repetitive slop. Obviously one should never use AI to generate subjective listening impressions (at most, use it to refine the text). In the long term, by becoming dependent on AI, a certain portion of the population will lose their written communication skills. A terrible idea to allow AI use during the foundational phases of educational programs (not just in childhood).] 

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

Greetings from S. Korea as I drop this post! 🙂

It has been awhile since I've been in this country. Having already passed through Singapore and Malaysia, it's amazing to me how quickly things change compared to the relative reticence to development and slow pace of construction we see in North America. It will be interesting to see how things have changed in China given their pace of adoption for renewable energy, an essentially cashless urban society, and technological growth.

Of course, Asia has its problems. But as North Americans, I think we have a thing or two to learn from the industriousness, perseverance and discipline out East.

A few vacation pictures as per my usual practice here on the blog.

Singapore - Marina Bay Sands in the distance. Also notice the number of tankers and cargo ships in the water. They're always out there but maybe a few more recently with the Iran war. Gasoline prices high in S.E. Asia but comparatively inexpensive in S. Korea.

Singapore Changi Airport - Jewel Rain Vortex - tallest indoor waterfall.

Busan, S. Korea - Songdo Cable Car.

Busan, S. Korea - At Jagalchi Fish Market (largest in Korea). Live baby octopus sashimi anyone? The delicacy is called San-nakji. Poor baby octopus chopped up live and served still wriggling. If you've ever had raw squid, you have an idea of the taste. Just make sure to chew well so the tentacles don't get stuck.
Pick the fresh seafood downstairs among the fishmongers and their "goods". They'll set you up upstairs with you selection ready to consume. Beer and soju often helps. Prices still reasonable given the freshness compared to what I usually pay in Vancouver. As always in Asia when prices not set, haggle.

Busan, S. Korea - Gamcheon Cultural Village. This place has been called the "Santorini of Korea". Lots of Antoine de Saint-Exupéry's The Little Prince references, art, and characters throughout this place.

Seoul, S. Korea - Gyeongbokgung Palace. I think it's great how they encourage visitors to wear the traditional hanbok with free admission.

I hope you're enjoying some awesome music through your hi-fi systems, dear audiophiles!