Why Synergy horns?

[Keith_W]

Should probably give you a few assumptions. I assumed that the T30 was 0.3sec, and the ceiling height was standard Australian 2.7m.

Anyway, want to guess why the two speakers are so similar below 400Hz? This is the published FR of the Yamaha NS1000:



And the Monitor Silver-8:



How come they match up almost exactly?

 

[hopkins]

Since the room has such a profound impact on the sound of a loudspeaker and it is impossible to listen in a blind test at an audio store, there is little that an audiophile can do to make a rational decision. Fortunately, science has come to the rescue with a set of measurements, that have been proven to demonstrate an extremely close correlation with sound quality, as based on carefully controlled double-blind listening tests. This group of measurements have been adopted as the industry standard for measuring loudspeakers.

I could not access the paper you linked, but I am wondering what type of speakers were used to conduct these tests and reach these conclusions...

Does this article relate to the same study?

Loudspeaker Measurements Standard: Our Procedure for Objectively Analyzing Speaker Performance

The Audioholics Loudspeaker Measurements Standard objectively measures performance of speaker systems. Frequency response, distortion, sensitivity, polar & listening window response, impedance, etc.

www.audioholics.com

The variety of speakers tested seem very limited, and in that context it would not be surprising that a limited set of basic measurements could correlate with listening preferences!

If only audio were so simple...

 

[schlager]

Just put the links in google search and not in the top search bar. Some are probably behind a paywall.

 

[schlager]

More here... I have other files, but are too large even when zipped.

 

[schlager]

hopkins, it seems that the audioholic paper, is pure technical/theoretical. Harman Group has conducted many blind listening test, with professional audio people and layman.

If we have to sum up the listening test and what different people statistically prefers, then the two most important things that a loudspeakers should have, is even on-axis FR and a gradually rolled off off-axis FR, resembling the on-axis FR.

 

[RCanelas]

A couple of ideas around interpreting Toole's preference results:

- obviously, correlation doesn't imply causation
- sample size is too small and with a large bias towards experienced listeners
- used speaker topologies was fairly banal and limited. No Synergy horns there. No dipoles, electrostats, isodynamic or others

Now, these are not shots at the work. I agree with his ideas, I use them in my own work, because they seem sound. But I always approach the results critically and try to compare with other ideas.

This is a navigational warning reminding us of our natural and unavoidable tendency to extrapolate something beyond its validity bounds. The 0.85 max correlation coefficient means very little to me if the question it answers to is not a good one, i.e. if the variables in question are not independent. These studies represent the currently best possible, but they still fall very short of guaranteeing statistical hygiene. There is no unified conceptual model for audio perception. We have pieces and bits, and some of them contradict, so my guess is that we're still far from it, and making blanket categorical statements incurs a large risk of being inaccurate.

 

[Solypsa]

A couple of ideas around interpreting Toole's preference results:

- obviously, correlation doesn't imply causation
- sample size is too small and with a large bias towards experienced listeners
- used speaker topologies was fairly banal and limited. No Synergy horns there. No dipoles, electrostats, isodynamic or others

Now, these are not shots at the work. I agree with his ideas, I use them in my own work, because they seem sound. But I always approach the results critically and try to compare with other ideas.

This is a navigational warning reminding us of our natural and unavoidable tendency to extrapolate something beyond its validity bounds. The 0.85 max correlation coefficient means very little to me if the question it answers to is not a good one, i.e. if the variables in question are not independent. These studies represent the currently best possible, but they still fall very short of guaranteeing statistical hygiene. There is no unified conceptual model for audio perception. We have pieces and bits, and some of them contradict, so my guess is that we're still far from it, and making blanket categorical statements incurs a large risk of being inaccurate.

If I have not said so before, please allow me to mention how much I enjoy your ability to write about technical subjects. Thank you

So far this thread is good fun.

 

[schlager]

- sample size is too small and with a large bias towards experienced listeners

After Toole left Harman Group, Sean Olive kept these speaker test going. As I recall they also had a panel speaker in the test (Martin Logan).

There is no unified conceptual model for audio perception.

I still believe that we have a very good understanding of room- and psychoacoustics and that the two most important things that a loudspeakers should have, is even on-axis FR and a gradually rolled off off-axis FR, resembling the on-axis FR. It sounds like a reasonable and logic argument, no?

 

[RCanelas]

I still believe that we have a very good understanding of room- and psychoacoustics and that the two most important things that a loudspeakers should have, is even on-axis FR and a gradually rolled off off-axis FR, resembling the on-axis FR. It sounds like a reasonable and logic argument, no?

I agree, but we probably don’t align in the ‘how much’ of that importance is to the overall. There are a number of effects, nonlinear with spl, that I consider almost as relevant and are just as likely to make or break a system imo.
Just think distortion spectra, snr etc. We have some indication that the sweet spots on these are not intuitive. Zero is unlikely the correct answer here.
To make things even morel interesting, we’re all in Toole’s circle of confusion, so as soon as we move preference estimates outside of a controlled environment, all bets are off

tldr: I agree with your assessment, I just can’t extrapolate to very universal stances from there, not enough information.

 

[schlager]

Just think distortion spectra, snr etc. We have some indication that the sweet spots on these are not intuitive. Zero is unlikely the correct answer here.

Some probably prefer some distortion, but I like to keep things a bit more grounded, so strictly speaking, adding components to the original signal, does not make the signal more correct and then we move in the wrong direction, away from high fidelity.

Toole’s circle of confusion, will be a problem as long as there is no recognised standard for loudspeakers. The result is markedly different "house sound" from different studios, especially in older recordings. But I see light at the end of the tunnel. Today, music is mixed on slightly better speakers combined with headphones, but we still have no established norms for that.

Film sound is based on calibration against standards, so that the sound you hear in your local cinema roughly matches what was approved in the studio. The same applies to image reproduction in cinemas and TV. There is a standard for color reproduction, and then you can measure, calibrate and assess against this. I find it strange that a common standard can't be incorporated in music studios.

The players in the reproduction chain for recorded music apparently consider themselves to be above such trivialities. They like to add an "artistic expression" where it should just be a pure transport stage. In that regard they team up with the subjective audiophile camp.

 

[hopkins]

Some probably prefer some distortion, but I like to keep things a bit more grounded, so strictly speaking, adding components to the original signal, does not make the signal more correct and then we move in the wrong direction, away from high fidelity.

Exactly why DSP is a terrible idea. Sorry, I could not help myself

 

[schlager]

I might be flogging a dead horse here, but it is essentially frequency response that we react to when we judge sound. We can train our ears and then open our eyes, to how it sounds. Often, too high/too low level in parts of the spectrum will be described with incredibly many nice words like "warm", "cold", "clinical", "comfortable", "enveloping", "articulate" etc. but they are really just very poor substitutes for simpler explanations like "low level between 100-300 hz" or "too high level above 5 khz".

Here it gets really interesting, what the heck is it that Herb has concluded here?

https://www.stereophile.com/content/manger-p1-loudspeaker-page-2

"The first thing I noticed when I installed the 50Wpc Pass Labs XA25 amplifier ($4900) was how quiet the Manger p1s sounded. This quiet I experienced was not about the signal-to-noise ratio of the Pass Labs amp, nor was it about the Mangers' lack of cabinet vibrations or port noise. It was about the Manger p1's conspicuous absence of that scratchy, tinnitus-like "cone noise" that shadows the upper-midrange and presence region of most two-way dynamic speakers. This lack of cone noise is more of a normal state in planar magnetic speakers. But, to my surprise, the Manger Sound Transducer was even quieter than my Magnepan .7s, which are my in-house reference for quiet.

When scratched or struck by a drumstick, every material generates its own resonant sound. (Think cast bronze bell vs molded clay bell.) Therefore it stands to reason: The material composition of the Manger Sound Transducer is likely the prime cause of its conspicuous quiet.

The MST's plastic sandwich does not ring or resonate like hard paper, Kevlar, or aluminum."

Here we go straight from an observation to a conclusion in no time, without any proof to support this. What is he talking about? None of this makes sense in terms of listening experience from a speaker, but maybe the next part does?

"consequently, it does not impart a false liveliness. Instead, in my room, it did the inverse: The Manger driver contributed a barely noticeable dull or damped-sounding undertone that imparted a sense of restrained refinement to every recording I played"

Hey, a reference to frequency response! Can we look at a measurement and find out what can give an indication of what the fellow is experiencing? Of course, a quick look at the measurements of the speakers clearly shows what he has picked up by the ears.




This useless speaker has barely any sound off-axis above 2000 hz, no wonder it sounds "quiet" and "boring" or "muffled". For my part, there is no relevant information to learn from the many flowery words and gross assumptions without some kind of reality check. In this case, a quick look at the measured response of the speakers, was enough to tell me that this speaker will sound boring and dull and free of unpleasant ear piercing sounds.

If I had to do a report to such a speaker, it would preferably have been to the police. I the same time I could have saved you guys quite a few minutes of wasted reading. Does that count for something?

 

[schlager]

Now there is a slight possibility that someone in the hifi community may have worked with a scientific approach, read about the research in the area and, in addition, experimented like any other enthusiast for many years.

Joe D'Appolito tireless work through 30 years led him to provide the strongest predictors of loudspeaker preferences.

• On-axis frequency response
• Impulse response
• Cumulative spectral decay
• Polar response
• Step response
• Impedance
• Efficiency/Sensitivity
• Distortion
• Dynamics

This lines up very well with what Harman also found out in their many blind listening tests, namely that on axis FR and polar response (dispersion pattern) are the two most critical parts of sound quality.

Many loudspeakers have obvious defects in the dispersion pattern, so they will never be really good, but it gets an awful lot better when you can adjust the direct sound, with EQ, to have a fairly neutral sound balance.

The paradox is that many people who prefer a certain "coloring" of the sound stubbornly claim that they are looking for the most "transparent" sound possible, and then the argument collapse and the misunderstanding become complete.

What I would like to see is a reckoning with classic hifi thinking where, among other things, trial and error with components for "matching" is an important part. And then turn the focus to more, let’s say, reality based solutions with roots in what we actually know.


A quick and dirty list of important points will then be:

- Active speakers that measure well in the essential areas (good power response and dispersion pattern that suits the area)

- Optimizing the placement of both the seating position and speakers

- Use a separate subwoofer/bass solution with time adjustment.

- Treat the room acoustically as much as possible. Broadband treatment and preferably use some diffusion if the dimensions of the room allows it.

- Use EQ to fine-tune the frequency response and the tonal balance to the listening position.

- Use good electronics that measures well and have impedance matching and sufficient power reserves.

Such an approach gives a completely different result and is much closer to reality IMO, than the classic approach. Moreover, you do not get random results that vary; you have a completely different control over the whole setup. However, it is obvious that not everyone can follow all of these points, but as a general rule, you can do some. It is, for example, difficult to see why there should be any reasonable reason to choose something other, than a loudspeaker with good dispersion pattern.

Classic hifi thinking has very little to do with high-end sound IMO. Good PA, for example expensive JBL solutions, will knock your socks off in comparison. Others probably think otherwise.

 

[RCanelas]

This is the interesting part, where the rubber meets the road.

Just came back from Munich, where you could try to do basic comparison exercises of general presentation styles. Apart from brutal directivity and overall quality problems with some designs, you could taste it all in a relatively short time, in obviously less then ideal conditions.

I won't name things, but there are clear outliers in the overall correlation between the principles you enumerated and my perceived quality score. I heard tear inducing reproductions from systems with obvious power response defects, electronics that are nowhere near sota as far as SINAD goes. Simultaneously I heard coma inducing presentations from systems I bet are amazingly linear in all dimensions. I also heard the exact opposite, flawed systems sounding flawed where I would expect and more textbook systems sounding pretty good.

So I'll repeat myself. These are obviously good ideas, they are internally consistent and smell of a constitutive model. Unfortunately when you throw the listener in the mix, some very relevant parts seem to be missing. They are not black magic, they're just not well represented I guess.

 

[schlager]

Unfortunately when you throw the listener in the mix, some very relevant parts seem to be missing.

Yep, science probably doesn't help if one have a skewed preference I'll probably look into that in a future post. Unfortunately, I missed High End Munich this year, hopefully I make it next year.
Thanks for your contribution.

 

[hopkins]

I do not have the technical competence that you both have. I am curious about the topic, and not just when it comes to speakers, but with any "objective" approach to audio. Concerning speakers, my questions would be:

- are there aspects of speaker performance that are difficult to assess with basic measurements? I read, for example, about the Doppler effect and its induced distortions. How can you apprehend that with a basic frequency sweep? Obviously, some have said that this effect is negligeable, but it is only an example. Are frequency sweeps really adequate to understand the effect of speaker cabinet resonance? Etc...

- more generally, to the extent that no speaker is perfect for all measurements/indicators, how do you assess the relative significance of these performance indicators? The more parameters there are, the more difficult it becomes to assess their relative significance.

- combining the two previous points, we could perhaps see that samples of speakers used in tests may never be comprehensive enough to be conclusive about listener preferences.

- the mechanisms of our hearing are not fully understood. What are the implications for all this? Also, do measurements with microphones adequately simulate our hearing mechanism?

 

[schlager]

I read, for example, about the Doppler effect and its induced distortions.

That could be a problem for a full range driver. By dividing the audio band between two or more drivers, we can circumvent that issue. The Doppler modulation is also a function of displacement and thus driver area. Big drivers don't have to move much, compared to smaller drivers. Finally, by horn loading the driver it will barely move at all.

Are frequency sweeps really adequate to understand the effect of speaker cabinet resonance?

No, we would need an accelerometer. Stereophile measures cabinet resonances. Unless they are severe they will be masked by the audio signal.

When building cabinets, thick walls are often used as an assurance, that things are in place, and then you use slightly smaller braces because you have so much faith in the thick walls.

Here the size of the cabinet, the size of the panels, whether you have several surfaces that sing at the same frequency, whether you have any degree of tension and whether you have any form of extra absorption in the walls, all are things you should have a fairly clear plan for, especially in slightly larger cabinets.

This is how I did my JBL 5644 copies.



Walls are 33 mm thick and well braced. Then stuffed with thick absorption material.

 

[hopkins]

No, we would need an accelerometer. Stereophile measures cabinet resonances. Unless they are severe they will be masked by the audio signal.

How do you determine the level at which this "masking" occurs? There is a whole "genre" of speakers designed to specifically address this (ex: granite speakers!). Are the perceived improvements biased? Innocent question..

 

[RCanelas]

@hopkins

allow me to be absolutely categorical in this, for once

There is absolutely nothing about a loudspeaker that can't be measured and modelled to an effectively arbitrary precision. These are macro scale electro-mechanical devices, of the simple kind. In my PhD I routinely measured transient states of fluids in scales of the micro meter, with an energy spectrum with a lot more decades than the audio one. Every non-linear and hysteretic effect that might show up is covered in first year material science, vibration propagation in structural dynamics and electric effect almost by Gauss's and Faraday's original writings. And so on. It is all just a matter of putting it together.

This is not to say it is simple, far from it, but is easily done. F1 teams where I consulted did it. Landing a Falcon rocket means we do it or it fails catastrophically. This is also not to say we 'can' do it, just that it is easily possible from the technical POV. There is no incentive to do it beyond a certain point. No money on the table. Hence we don't do it.

Now comes the second part: even if you fully characterize a loudspeaker and it's environment, you'd still need to do the same for the listener, to be sure you hit the target (maximizing preference). It is not just the mechanical side of the auditory mechanism, also the perception side. Full blown individual (and time variant) psychoacoustics characterization. This is the truly complicated part IMO, and where I think that as laudable as the efforts of Toole et al are, they are very very low order approximations that tend to fail on extreme/edge cases. All of audiophilia is an extreme case as far as I'm concerned.

Putting it differently, these preference studies and concerns are very useful to characterize a population at large, so you know what type of product you can sell by the hundreds of thousands, not the dozen. That is not to say they are remotely invalid, but I can recognize their limitations, context and notable empirical exceptions.

 

[hopkins]

Thank you for this very clear reply.