Some speakers slant, some don't. Why?

[semente]

This is from Siegfried Linkwitz's website:

 

Design of Loudspeakers

Some general observations

 

The typical loudspeaker product is designed to make money and not necessarily to provide accurate sound reproduction.

Since customers prefer small, unobtrusive speakers and judge sound quality by the amount of bass that they hear and by high frequencies they had not noticed before, there is a staggering number of essentially identical designs on the market that meet these requirements at different price points.

No wonder then that there is a generic loudspeaker sound and that you can always tell whether something that you hear originates from a speaker and not from a live source.

The marketing departments of the different speaker manufacturers are busy to point out differentiating features and breakthrough inventions when it comes to the highest price points, but in reality box loudspeaker design has come to a the end of a road and all you will hear are slight variations on the same theme.

The fundamental problems of box re-radiation and non-uniform power response in a room are at best only partially solved by these conventional designs

 

Sound reproduction is about creating an auditory illusion.

When the recorded sound is of real instruments or voices there is a familiar, live reference in our auditory memory. The illusion of hearing a realistic reproduction is destroyed by distortion that is added anywhere in the signal chain from microphone to loudspeaker, but the speaker is by far the biggest culprit.

Every designer focuses on the on-axis frequency response as if it were the all determining distortion parameter.

Sometimes great attention is paid to the phase response in an attempt to preserve waveform fidelity, which at best can only be achieved for a single listening point in space.

Ignored usually, though of much greater importance, is resonance in drivers and cabinets and the slow release of stored energy that goes with it. 

Furthermore, the uniformity and flatness of the off-axis frequency response which we hear via room reverberation and reflections is rarely a design goal.

You can check the naturalness of the timbre by listening from another room.

Does it sound like a loudspeaker is playing?

The imbalance in the speaker's power response between low and high frequencies destroys the illusion.

 

And then there are the non-linear distortions, the ones that add sounds that were not present in the original.

They are easily measurable in the form of harmonic and intermodulation distortion products.

Rarely do non-linear distortion considerations enter into the design of speakers.

Otherwise, consumer stores and recording studios would not abound with 2-way designs - usually a 6.5" woofer/midrange and 1" dome tweeter in a ported box - that are physically incapable of the sound levels claimed for them and which by distortion often create a unique box loudspeaker bass experience, variously described with slam and speed.

Some design attempts for reducing non-linear distortion lead to line source speakers.

While successful at this they introduce phantom image distortions and, coupled with a conventional vented woofer of sorts, they suffer the same uneven power response and rich excitation of room resonances as the typical box speaker design.

 

(...)

 

http://www.linkwitzlab.com/design_of_loudspeakers.htm

 

 

 

[semente]

8 hours ago, esldude said:

Those were all from stereophile measurements, which I should have credited.  The old software would do that for links from Stereophile automatically.  They don't do impulse or squarewave more recently.  The Treo does show a nice step response which is near textbook.

 

Of course how important is that?  Here is the step response of some Revel Ultima Salon 2 speakers which sound quite fantastic by most accounts.  This step response is similar to that of all the Revel speakers.  Most of which get near universal acclaim for sounding good.

 

I believe the ESL63 had a resonance between the panel and metal cover around 13 khz which is why it rings on.  The later near identical design of the 989 did reduce that problem. 

 

Two (negative) aspects one should mention are the significant driver overlap and the ineffective control of the mid/mid-woofer resonance (plots from Sphile's measurements of the 2CeSigII):

 

 

The 2CeSigII also has a very bright in-room tonal balance, compared here to the PSB Synchrony One (which measures very flat in anechoic conditions) at JA's - bottom graph, though tis may or may not be related with the crossover design...

 

 

 

 

 

The Revel Ultima Salon2s also measure very flat in anechoic and in-room at JA's (blue):

 

 

[semente]

13 hours ago, Ralf11 said:

 

Jud, look at the foothills are the foot of the mountains - some stick out a lot more than others, indicating a room mode resonance that should be treated

 

may also be from the speaker itself, but I don't think you will find much of that mechanism on a Vandersteen

 

The waterfall can be used for different purposes such as determining room resonances or if the speaker is measured in free space (or nearfield with a narrow time-window) speaker resonances.

 

Speaker resonances, or better yet driver resonances are what we should be looking for in Sphile's measurements.

 

Here is a good example where you can perfectly identify the breakup resonances produced by the aluminium cone (blue, -5dB @ 5.8kHz) and dome (red, 20dB @ 25kHz) of a 2-way Monitor Audio loudspeaker with a :

 

Here you can see how it affects the frequency response.

The woofer appears to be low-passed using a first-order, 6dB/octave filter.

 

Monitor Audio Studio 15, cumulative spectral-decay plot of woofer at 44" (0.15ms risetime).

 

Monitor Audio Studio 15, cumulative spectral-decay plot at 44" (0.15ms risetime).

 

Monitor Audio Studio 15, cumulative spectral-decay plot calculated from the output of an accelerometer fastened to the center of the cabinet's sidewall (MLS driving voltage to speaker, 7.55V; measurement bandwidth, 2kHz).

 

Read more at http://www.stereophile.com/content/monitor-audio-studio-15-loudspeaker-measurements

[semente]

Here's another example, BnW's 683 S2, this time a kevlar midrange cone breaking up at 3.6kHz:

 

B&W 683 S2, cumulative spectral-decay plot on tweeter axis at 50" (0.15ms risetime).

 

B&W 683 S2, anechoic response on tweeter axis at 50", averaged across 30° horizontal window and corrected for microphone response, with complex sum of nearfield responses plotted below 300Hz

Read more at http://www.stereophile.com/content/bowers-wilkins-683-s2-loudspeaker-measurements

[semente]

Here's another way to measure decay (lower colour temperature equals lower sound level, signal output at top horizontal side of the plot, time extends from top to bottom, frequency increases from left to right):

 

 

Graham Audio LS5/9

http://www.hi-fiworld.co.uk/index.php/loudspeakers/65-reviews/726-graham-audio-bbc-ls59.html

[semente]

4 hours ago, semente said:

Here's another way to measure decay (lower colour temperature equals lower sound level, signal output at top horizontal side of the plot, time extends from top to bottom, frequency increases from left to right):

 

It's actually not another way to measure but to display.

[semente]

14 minutes ago, Jud said:

 

Richard Vandersteen's interview indicates a deliberate decision to achieve time/phase coherence at the cost of flatter frequency response across the band.

 

Ricardo has mentioned before this bothers him.  It didn't bother me with the 2Ci and 2Ce, and assuming the 3A Sig is similar, there either.  One way to look at it is that I learned to hear this sound signature as "correct," or at least what I was looking for in audio reproduction.  Another viewpoint is that since people do hear differently, perhaps I was among a sub-population predisposed to pay greater attention to time/phase coherence in evaluating the realism of reproduced sound.

 

I also find this a plausible explanation and one's favourite musical genres also play a part here.

 

Sphile's 3A measurements show a more even tonal-balance although the upper mids and treble are shelved up by a few dB, maybe the 3A Sig is even better in this respect.

There seems to be quite a bit of resonance coming from the midrange driver (kevlar?):

 

 

Vandersteen 3A, cumulative spectral-decay plot at 50" (0.15ms risetime)

 

Read more at http://www.stereophile.com/content/vandersteen-audio-3a-loudspeaker-sidebar-3-measurements-page-4

[semente]

1 minute ago, Jud said:

 

I'm pretty sure Vandersteen has never gone in for Kevlar.  Carbon fiber or fiber/balsa sandwich, maybe.

 

The reviews is pretty worthless in that respect.

The specs page mentions 4.5" midrange unit with patented, low-diffraction magnet, curvilinear polycone, and ferrofluid cooling

[semente]

7 hours ago, Ralf11 said:

I'd heard they sound really good, but...

 

...but it depends on who's judging? 

 

R

[semente]

Speaking about decay, I wonder how much it affects resolution / low-level detail retrieval...

 

Here are a few examples:

 

Eminent LFT-16

planar-magnetic tweeter and midrange, cone woofer below 

 

Martin Logan Electromotion

electrostatic high/mid panel, cone woofer

 

Amphion Argon 3L

dome tweeter, cone midwoofer

 

Graham Audio LS5/9

dome tweeter, cone midwoofer

 

Tannoy Kensington

horn tweeter (compression driver), cone woofer

 

 

 

[semente]

3 hours ago, Nikhil said:

 

Semente,

 

How do you interpret those graphs?  I can't make out the axis parameters.

Of the lot I really like those Graham Audio LS5/9 s - very engaging sound.

 

Regards.

 

  

 

 

3 hours ago, esldude said:

Frequency along the x axis 20 hz to 20 khz on the right.

 

Time along y axis with 0 at the top and a bit over 200 milliseconds along the bottom.  The right hand side indicates decibel level with color.

 

Normally I see those rotated 90 degrees from this orientation, but there is no standard for that.

 

Ideal would be a narrow band of red and yellow very near the top dropping to black quickly.  A 2D variation on the info contained in 3 D waterfall plots. 

 

Hi Nikhil,

 

@elsdude has anticipated himself and I have nothing to add to his reply, except I think that the graph might have been made clearer if time ("Y" axis) were to extend from bottom to top.

 

R

[semente]

On 28 March 2017 at 2:37 AM, botrytis said:

You really can't judge a speaker by the printed graphs. Your room, your equipment, YOUR ROOM TREATMENTS all interact.

 

Just go listen and make the decision for yourself.

 

I agree with the underlined bit but if you are implying that a speaker with better decay characteristics with not have better decay in any room I would have to disagree.

But, of course, decay isn't everything.

 

Measurements focus on one particular aspect of performance and logically they do not characterise the whole sound of a speaker.

But some, like frequency response, have good correlation with listening and a comprehensive set, such as those provided by Sphile and Soundstage, is very informative.

 

Measurements are complementary to listening, they'll help you make a more informed (or learned) decision.

 

R