Interview about Measurements with Pearl Sibelius Speaker Designer

[MattHooper]

I finally took a look at the interview posted in the OP.

My skeptical hackles go up when anyone in high end audio starts to talk about "things we can't measure" or how measurements don't help with X or Y.

Nonetheless, the Pearl speaker designer did have some interesting points. When answering the question how it seems un-intuitive that such a small driver could produce the sensation of deep bass, I really liked his analogy to holding up a teeny reed used by a Bassoon, pointing out just how small the actual vibrating source is, yet can be heard easily through a hall in the context of an entire symphony playing...because of the effects of marrying that vibration to an amplifying pipe system.

 

[312elements]

The Harmon curve is derived from work done by Dr. Floyd Toole. Dr. Toole is a member of this forum; you might do a search for his posts; he's been active recently.



As far as I know, there are only two rules here.
1) Be civil. That means that you get the respect that you give. And if you don't get that respect, then be civil anyway.
2) This site belongs to @amirm . He owns it, not us. He pays for it out of his own pocket. He sets the rules. If there's any question, he will clarify. That is his right.

Jim

I think you took my post the wrong way. I’m sorry. It’s likely that I wasn’t clear. I’ll ask it again differently. I have no argument with the Harmon curve or any other curve. None at all. I’m saying if we define high fidelity by accurately reproducing sound why is the Harmon curve ok, but an alternate curve not high-fidelity? It doesn’t make sense to me. If you were to say that high fidelity required a flat speaker response, I could totally accept that. If you were to tell me that most people prefer a downward tilted frequency response, the data supports that, that works. To say that one house curve is high fidelity and another house curve is not high fidelity. That’s my issue. That’s what I would like explained.

 

[NTK]

I think you took my post the wrong way. I’m sorry. It’s likely that I wasn’t clear. I’ll ask it again differently. I have no argument with the Harmon curve or any other curve. None at all. I’m saying if we define high fidelity by accurately reproducing sound why is the Harmon curve ok, but an alternate curve not high-fidelity? It doesn’t make sense to me. If you were to say that high fidelity required a flat speaker response, I could totally accept that. If you were to tell me that most people prefer a downward tilted frequency response, the data supports that, that works. To say that one house curve is high fidelity and another house curve is not high fidelity. That’s my issue. That’s what I would like explained.

If you are referring to the downward sloping predicted in-room response (PIR) curve, it is not the "Harman house curve" nor the "Harman target curve". It is simply what one would likely measure for the steady state in-room response with good forward firing speakers in a typical room. The design target should always be anechoic flat on-axis (or listening window) response, and anechoic smooth, gradually changing off-axis response.

See Dr Toole's posts.

I cannot trust the Harman speaker preference score

Some bookshelf speakers get midrange distortion from the bass freqencies hitting the single midrange driver at high volumes, even with with an 80 Hz crossover. Depends on if your speakers exhibit this, or if you even notice. Mine did, and I noticed, and I hated it. I don't know how high of a...

www.audiosciencereview.com

Blind Listening Test 2: Neumann KH 80 vs JBL 305p MkII vs Edifier R1280T vs RCF Arya Pro5

My take as I've done many times, speakers that exaggerate the pattern on/off-axis may sound harsh while speakers with inverted pattern sound smoother. How can a speaker exaggerate the pattern on-axis?

www.audiosciencereview.com

 

[312elements]

Thank you @Jim Taylor and @NTK . I definitely misunderstood that point. I think I’m up to speed now.

 

[MattHooper]

Thank you @Jim Taylor and @NTK . I definitely misunderstood that point. I think I’m up to speed now.

Another way of putting it, as I understand it: The research showed that people prefer a flat, neutral sound from a speaker. So one seeks a flat frequency response on axis.
The problem is that in many rooms many speakers that are flat on axis will produce an overall room response that will deviate from neutral, so too bright or whatever.
The idea is that a gradually downward sloping, even off-axis response will help the overall response "stay neutral" in terms of our perception. So the downward slope isn't abandoning neutral: it's a way of preserving it, once rooms become involved.

 

[ahofer]

such as a narrow-radiating horn for Elsa's ice palace.

I got a good laugh out of that one.

 

[Mart68]

An example that quickly pops to mind is the last time I auditioned the Kii Audio Three speakers, which had been carefully set up at a dealer and dialed in. They sounded boxless, almost "invisible," virtually full range, precise and revealing. I can absolutely see anyone saying "these bring out all the differences in recordings." And yet, to my ear, though recordings sounded very different *within the general timbre of that speaker,* absolutely nothing sounded "right" or quite natural. I never got that "it could be a real acoustic guitar playing" sensation.

Yet in the same demo room they had some Martin Logan hybrid speakers set up. I'm not a total fan of ML, but out of curiosity I played a bunch of the same tracks and...by gum!...it sounded so much more convincing. There was that warm, woody sparkle to acoustic guitar that I'm so familiar with. Drum sticks hitting snares, rims had that "right there" immediacy and recognizable character, etc. Each recording sounded different. And I'm sure that system would not measure as neutral and low distortion (in some aspects) as the Kii speakers.

But my basic point remains that what you may have found "uncoloured" I may have have found obviously homogenized, relative to what I'm judging things against, and in terms of my own taste.

Cheers.

I've had this discussion before with someone else and in the end we concluded that the difference between us is that I don't listen to any classical music and not much in the way of acoustic instruments recorded without any effects. So 'The sound of real instruments' isn't my benchmark; it's the difference in production styles.

I don't think there is any way to square that circle, nor do I think we need to. But it's interesting and explains a lot.

 

[SSS]

Good video explaining acoustics and physics. The interviewer could have read it already in the appropriate books. There is no mystery behind the construction of the speaker. Interesting for non-professionals in this area.

 

[Floyd Toole]

I think you took my post the wrong way. I’m sorry. It’s likely that I wasn’t clear. I’ll ask it again differently. I have no argument with the Harmon curve or any other curve. None at all. I’m saying if we define high fidelity by accurately reproducing sound why is the Harmon curve ok, but an alternate curve not high-fidelity? It doesn’t make sense to me. If you were to say that high fidelity required a flat speaker response, I could totally accept that. If you were to tell me that most people prefer a downward tilted frequency response, the data supports that, that works. To say that one house curve is high fidelity and another house curve is not high fidelity. That’s my issue. That’s what I would like explained.

If you dig through my posts you will see several occasions on which I say something like: "A room curve is a result, not a target". The so-called Harman (with an "a") is what one might measure, above 400-500 Hz in a typical listening room from a loudspeaker that is highly rated in double-blind tests. Such a loudspeaker will almost certainly have flattish and smooth on-axis response and gradually changing and smooth off axis response. These characteristics indicate (a) a lack of resonances that color the sound and (b) that reflected sounds in a room are likely to have a timbral signature that is similar to the dominantly important direct sound. However, it is possible for compensating errors (flaws) in a loudspeaker to result in a similar curve, so room curves are not reliable evidence of sound quality. Likewise equalizing a flawed loudspeaker to match that curve does not guarantee good sound. For that one need comprehensive anechoic data, the "spinorama" is an example. Stop trusting room curves, ignore the single number "Harman" rating. Learn how to interpret anechoic data. Spinoramas are becoming more common. They contain the "secrets"..

 

[MattHooper]

Such a loudspeaker will almost certainly have flattish and smooth on-axis response and gradually changing and smooth off axis response. These characteristics indicate (a) a lack of resonances that color the sound and (b) that reflected sounds in a room are likely to have a timbral signature that is similar to the dominantly important direct sound. However, it is possible for compensating errors (flaws) in a loudspeaker to result in a similar curve, so room curves are not reliable evidence of sound quality.

Dr. Toole,

You've written about how a speaker with less resonance coloring the sound will "disappear" better even in blinded mono tests. (And also that mono tests will actually give an even better indication of a speaker disappearing - that is getting out of the way of the sound, so it doesn't sound like it's "coming from a speaker.")

Also, you've said that insofar as a speaker disappears this way in mono, that will tend to translate to similar "disappearing" low coloration performance in stereo.

Of course, having hard spinorama data to look at will always be the most reliable indictors. But lacking that, do you think that we can make any similar inferences from how much speakers seem to "disappear" from stereo listening tests? (Since that's how most of us tend to audition speakers under our consideration). In other words: if one pair of speakers seem to disappear better in stereo, even hard panned sonic images apparently floating free of the cabinets, will that suggest low coloration/low resonances in of itself?

Or are there too many confounding factors once we are doing stereo listening? (Perhaps frequency response sculpting can result in one speaker disappearing more than another).

So for instance, I have two floor standing speakers - Thiel 2.7 and Joseph Audio Perspective, both floor standing speakers rated down to 35Hz. The Thiels cast an enormous soundstage, with nice precise imaging, though sonic images more toward the sides seem just a bit more gloming in to the speaker (so a bit more of a "U-shaped" soundstage). Whereas the Joseph speakers seem to utterly "disappear" as apparent sound sources, no "U-shape" to the soundstage, and all sonic images even those to the left or right, don't seem to be coming from the speakers. Might one infer the Josephs are lower in some form of coloration from this?

Thanks.

 

[312elements]

If you dig through my posts you will see several occasions on which I say something like: "A room curve is a result, not a target". The so-called Harman (with an "a") is what one might measure, above 400-500 Hz in a typical listening room from a loudspeaker that is highly rated in double-blind tests. Such a loudspeaker will almost certainly have flattish and smooth on-axis response and gradually changing and smooth off axis response. These characteristics indicate (a) a lack of resonances that color the sound and (b) that reflected sounds in a room are likely to have a timbral signature that is similar to the dominantly important direct sound. However, it is possible for compensating errors (flaws) in a loudspeaker to result in a similar curve, so room curves are not reliable evidence of sound quality. Likewise equalizing a flawed loudspeaker to match that curve does not guarantee good sound. For that one need comprehensive anechoic data, the "spinorama" is an example. Stop trusting room curves, ignore the single number "Harman" rating. Learn how to interpret anechoic data. Spinoramas are becoming more common. They contain the "secrets"..

Why are loudspeakers not designed for a flat PIR? Is that a design limitation or a design choice?

 

[thecheapseats]

..."A room curve is a result, not a target"... and.... it is possible for compensating errors (flaws) in a loudspeaker to result in a similar curve, so room curves are not reliable evidence of sound quality. Likewise equalizing a flawed loudspeaker to match that curve does not guarantee good sound. For that one need comprehensive anechoic data...

that ^^^ a perfect mouthful of reality....

 

[tuga]

Stop trusting room curves, ignore the single number "Harman" rating. Learn how to interpret anechoic data. Spinoramas are becoming more common. They contain the "secrets"..

Thank for bringing some sanity to the discussion.

 

[tuga]

Another way of putting it, as I understand it: The research showed that people prefer a flat, neutral sound from a speaker. So one seeks a flat frequency response on axis.
The problem is that in many rooms many speakers that are flat on axis will produce an overall room response that will deviate from neutral, so too bright or whatever.
The idea is that a gradually downward sloping, even off-axis response will help the overall response "stay neutral" in terms of our perception. So the downward slope isn't abandoning neutral: it's a way of preserving it, once rooms become involved.

My next question is what angle should the PIR show?
And to my understanding it depends on the directivity characteristics: as you move from narrowing to omni the pitch/angle should increase (the latter only really suitable for positioning far from boundaries.
But what if one’s listening near-field or the early-reflections are treated?
And then there’s constant directivity speakers.

As I’ve said earlier, let’s not oversimplify a complex matter…
I find that the Spinorama and the Preference ratings are making lazy people just be lazy. No free lunches and all that.
Even PIR is a rough prediction and really only valid above Schroeder if your speakers are positioned symmetrically in a symmetrical room and with symmetrical furnishings (a dedicated theatre or studio).

 

[YSC]

My next question is what angle should the PIR show?
And to my understanding it depends on the directivity characteristics: as you move from narrowing to omni the pitch/angle should increase (the latter only really suitable for positioning far from boundaries.
But what if one’s listening near-field or the early-reflections are treated?
And then there’s constant directivity speakers.

As I’ve said earlier, let’s not oversimplify a complex matter…
I find that the Spinorama and the Preference ratings are making lazy people just be lazy. No free lunches and all that.
Even PIR is a rough prediction and really only valid above Schroeder if your speakers are positioned symmetrically in a symmetrical room and with symmetrical furnishings (a dedicated theatre or studio).

I will put it this way as in logical explaination/understanding of the preference.

IMO, the results of most people prefer a on axis neutral, and off axis downward sloping with good directivity, is that it ended up upon various music, no tones will stand out being missing or overly boosted, hence "natural" and "Hi-fi". that is the big, first requirement for most people to prefer that kind of sound: on axis and reflected sounded similar in tonality, hence at least not artificial or "something weird", passing that mark, we likely percieve that the sound is natural.

After that basic desire to get a natural sound, the PIR slope and beam width, room treatment could vary a lot due to personal preference on tonal balance of the music type we choose, it's like how pin point you want the image is, or how apparent sound stage is varies from individual to individual, but most likely ppl would like something not having parts "this sounded weird" as our brain compare on and off axis response to do a lot of background processing. so the poorer the extension, the worse you feel as more are "missing completely", then the more irregular the on axis response is, you feel the dominant sound is weird, followed by with poorer the directivity and hence the off axis reflected sound is out of balance, our brain would feel something is not right, if horizontal directivity is great, but vertical directivity is off, the roof or floor reflection would also sound not sso right also, maybe that affects our ability to pin point the stereo image, and that's why coaxials sound more precise.

 

[thewas]

PIR on its own doesn't tell much as someone could take a loudspeaker with non-smooth directivity and equalise the on-axis response for a smooth PIR, but it wouldn't sound the same good as a loudspeaker with flat direct sound and smooth directivity (have even tested that myself more than a dozen of times). Like Dr. Toole writes humans perceive sound differently than a microphone at the listening position making a temporal average of direct and reflected sounds.

 

[SSS]

If you dig through my posts you will see several occasions on which I say something like: "A room curve is a result, not a target". The so-called Harman (with an "a") is what one might measure, above 400-500 Hz in a typical listening room from a loudspeaker that is highly rated in double-blind tests. Such a loudspeaker will almost certainly have flattish and smooth on-axis response and gradually changing and smooth off axis response. These characteristics indicate (a) a lack of resonances that color the sound and (b) that reflected sounds in a room are likely to have a timbral signature that is similar to the dominantly important direct sound. However, it is possible for compensating errors (flaws) in a loudspeaker to result in a similar curve, so room curves are not reliable evidence of sound quality. Likewise equalizing a flawed loudspeaker to match that curve does not guarantee good sound. For that one need comprehensive anechoic data, the "spinorama" is an example. Stop trusting room curves, ignore the single number "Harman" rating. Learn how to interpret anechoic data. Spinoramas are becoming more common. They contain the "secrets"..

Read your book "Sound Reproduction" from 2008. Since physics and acoustics did not change since then it is still a valuable read and it describes all the mechanisms of sound distribution. Therefore to my opinion there is and will be no loudspeaker which is perfect for every kind of room. It all depends on size, dampening and recflectiions. Lisened recently to very good speakers which were Wilson Audio and Geithain in a so called hi-fi studio I got several impressions. Soundstage was very dependent on the kind of mastering mix and if there was reverbation in the mix. If the listening room has almost no reflections the sound is just only between the two speakers and it sounds not natural. My brain is trained that every room has reflections and an own sound. Therefore for me it is not desirable to hear the same what the mixing engineer heard since as I saw often the mixing room is sound dead and nearfield. Lack of room impression happens to me also when listening with earphones. Good for analytics but no fun.

 

[ctrl]

Why are loudspeakers not designed for a flat PIR? Is that a design limitation or a design choice?

For standard loudspeaker designs, a PIR that slopes slightly toward high frequencies is a "natural" result.
The resulting slope angle of the PIR depends on the LS design, whether the LS radiation narrows more or less toward high frequencies.

There is one exception, a constant directivity (CD) design produces both a flat on-axis frequency response (and/or flat LW) and a flat PIR - it's easy to recognize, since the DI curve is also flat. But such designs are extremely difficult to realize and can sound too "bright" in poorly treated listening rooms.

For consumer products, a PIR that drops uniformly to high frequencies is most likely a deliberate design choice to prevent a too bright sound in modern listening rooms. Since a flat PIR design is more expensive to design and implement (requires e.g. depending on the design, large cabinets) which is also a design limit for many low-cost LS products.

Of course, it is also possible to force a flat PIR with a corresponding on-axis FR rising to high frequencies by crossover tuning or EQ - but this will definitely sound too "bright" in a normal listening room.

A few examples

Coax design with very even radiation. PIR downward slope controlled by waveguide.
First hor deg0-90, then CTA-2034, last PIR


wide hor and narrow ver radiation with small tweeter, woofer for low frequencies is not shown. PIR "naturally" slight downward slope.
First hor deg0-90, then CTA-2034 with PIR


constant directivity horn above 1kHz for CD radiation - the DI above 1kHz is nearly constant, so is PIR.
First hor deg0-90, then CTA-2034, last PIR


waveguide and cardioid slots for CD radiation, woofer for low frequencies is not shown.
First hor deg0-90, then CTA-2034 with PIR

 

[312elements]

For standard loudspeaker designs, a PIR that slopes slightly toward high frequencies is a "natural" result.
The resulting slope angle of the PIR depends on the LS design, whether the LS radiation narrows more or less toward high frequencies.

There is one exception, a constant directivity (CD) design produces both a flat on-axis frequency response (and/or flat LW) and a flat PIR - it's easy to recognize, since the DI curve is also flat. But such designs are extremely difficult to realize and can sound too "bright" in poorly treated listening rooms.

For consumer products, a PIR that drops uniformly to high frequencies is most likely a deliberate design choice to prevent a too bright sound in modern listening rooms. Since a flat PIR design is more expensive to design and implement (requires e.g. depending on the design, large cabinets) which is also a design limit for many low-cost LS products.

Of course, it is also possible to force a flat PIR with a corresponding on-axis FR rising to high frequencies by crossover tuning or EQ - but this will definitely sound too "bright" in a normal listening room.

A few examples

Coax design with very even radiation. PIR downward slope controlled by waveguide.
First hor deg0-90, then CTA-2034, last PIR
View attachment 281320View attachment 281318 View attachment 281319

wide hor and narrow ver radiation with small tweeter, woofer for low frequencies is not shown. PIR "naturally" slight downward slope.
First hor deg0-90, then CTA-2034 with PIR
View attachment 281332 View attachment 281333

constant directivity horn above 1kHz for CD radiation - the DI above 1kHz is nearly constant, so is PIR.
First hor deg0-90, then CTA-2034, last PIR
View attachment 281315View attachment 281313View attachment 281314

waveguide and cardioid slots for CD radiation, woofer for low frequencies is not shown.
First hor deg0-90, then CTA-2034 with PIR
View attachment 281344 View attachment 281343

That is a solid answer! Thanks so much for taking the time. Why is "too bright" not considered subjective here? I get that these companies are in the business of selling speakers and from a marketing and customer satisfaction standpoint, adapting to the desires of your customers is an important part of staying in business. I don't question that logic at all. But in terms of a faithful reproduction of the original content, it would seem that this logic deviates from that. I often read, on this forum, that anything that deviates from a faithful reproduction of the original content is broken. Why is this the exception to the rule?

 

[Floyd Toole]

Why are loudspeakers not designed for a flat PIR? Is that a design limitation or a design choice?

Humans can separate the direct - first arrival - sound from later arrivals - reflections. In terms of sound quality and image localization the direct sound is dominant. That is why it is important to have flat and smooth - i.e. neutral timbre - on-axis and listening window frequency responses. Two ears and a brain are much smarter than a 1/4-inch microphone that simply adds things together.