Wilson Audio Specialties Alexx loudspeaker Measurements

Sidebar 3: Measurements

Because this speaker's size and 452-lb weight made shipping one of them to me for measurement impossible, I drove my test gear to Michael Fremer's place. I used DRA Labs' MLSSA system and a calibrated DPA 4006 microphone to measure the Wilson Audio Alexx's frequency response in the farfield, and an Earthworks QTC-40 for the nearfield and in-room responses.

The first dilemma was on what axis to examine the speaker's quasi-anechoic behavior. Wilson's Peter McGrath had adjusted the Alexx's tweeter and midrange modules so that their axes converged on the positions of Michael's ears in his listening room, which were 36" from the floor and 85" from the tweeter modules. The tweeters themselves were 54" from the floor, so after we'd wheeled one of the Wilsons (serial number 0145) into Michael's driveway—easier to say than to do—I set up the microphone at my usual 50" distance on the line connecting its tweeter to a position 85" away and 36" from the ground. I ran a set of tests, then moved the mike to where Michael's ears would have been and repeated the tests. The problem, however, was that at the greater distance, with the mike 36" above the ground, early reflections of the speaker's output from the ground drastically reduced the measurements' resolution in the midrange.

Looking first at the Alexx's voltage sensitivity, my estimate of 91.3dB(B) agreed with the specified 91dB/2.83V/m. This speaker will play loudly with relatively few amplifier watts—or it would if its impedance were not very low, meaning that the standard 2.83V drive signal used for specifying sensitivity is equivalent to much more than 1W. The impedance is specified as 4 ohms nominal, with a minimum magnitude of 1.5 ohms at 2.85kHz. My measurement is shown in fig.1. Although Wilson's specification is correct, in that the minimum magnitude (solid trace) is 1.5 ohms just below 3kHz—I measured 1.44 ohms at 2.9kHz—the impedance remains at or below 3 ohms from 55Hz to 6kHz, and the electrical phase angle (dotted trace) exacerbated the drive difficulty in some regions of the audioband. The combination of 3 ohms magnitude and a phase angle of –44° between 56 and 59Hz will stress amplifiers, as will the combination of 2.2–2.6 ohms and a phase angle of +40° between 4 and 5kHz. Alexx owners need to match the speakers with amplifiers unfazed by very low impedances.

417wilsonal.Walexfig1.jpg

Fig.1 Wilson Audio Alexx, electrical impedance (solid) and phase (dashed) (2 ohms/vertical div.).

Wilson's heroic cabinet construction meant that I could find no panel resonances. The impedance graph suggests that the large port is tuned very low in frequency, and nearfield measurements of the woofers' outputs revealed that they each had a minimum-motion notch between 21 and 24Hz. The port's output, again measured in the nearfield, peaked between 10 and 30Hz, but didn't begin its upper-frequency rolloff until 65Hz. The port's output in the midrange was well suppressed, however. Though the upper woofer extends very slightly higher in frequency than the lower woofer, they otherwise have similar responses, and are crossed over to the lower-midrange unit at around 150Hz. That unit, in turn, seems to be crossed over to the upper midrange unit at about 800Hz.

Below 300Hz, the blue trace in fig.2 shows the complex sum of the nearfield port, woofer, and midrange responses, taking into account their radiating areas, amplitudes, and acoustic phase angles. The peak in the midbass is an artifact of the nearfield measurement technique; the Alexx's low frequencies don't quite extend to the port tuning frequency, which suggests that the woofer alignment is somewhat overdamped. Higher in frequency, the blue trace shows the Wilson's farfield response at 50" averaged across a 30° horizontal window centered on the tweeter axis. There is a small peak centered on 1.6kHz as well as a suckout an octave higher, but the suckout fills in at the 85" microphone distance (red trace). Other than the upper-midrange peak, the speaker's balance is relatively even between 100Hz and 20kHz, with small peaks balanced by small dips.

417wilsonal.Walexfig2.jpg

Fig.2 Wilson Audio Alexx, anechoic response on tweeter axis at 50" (blue) and 85" (red), both averaged across 30° horizontal window and corrected for microphone response, with complex sum of nearfield midrange, woofer, and port responses plotted below 300Hz.

Because of the Alexx's bulk, I wasn't able to examine its horizontal and vertical dispersions in any detail. However, looking at the individual responses that I averaged to produce the response in Michael's listening room (fig.3), it looks as if the Alexx maintains its spectral balance below 20kHz over windows of ±5° vertical and ±30° horizontal, both centered on the tweeter axis.

417wilsonal.Walexfig3.jpg

Fig.3 Wilson Audio Alexx, spatially averaged, 1/6-octave response (red), and of Wilson Alexandria XLF (blue), in MF's listening room.

The spatially averaged response (fig.3, red trace) was produced by taking 20 1/6-octave–smoothed spectra individually for the left and right speakers using SMUGSoftware's FuzzMeasure 3.0 program and a 96kHz sample rate, in a rectangular grid 36" wide by 18" high and centered on the positions of MF's ears. For reference, the blue trace shows the response taken under identical conditions of the Wilson Alexandria XLF, which Michael reviewed in January 2013. The two speakers behave similarly above 200Hz, though the Alexx is more even overall. The slight peak in the upper midrange visible in fig.2 is evident here as well, but I suspect it emphasizes the presentation of recorded detail rather than being perceived as a coloration. The dip between 100 and 200Hz and the boosted output between 20 and 100Hz for both speakers will be respectively due to destructive and constructive interference between the direct sound and the reflections from the nearby boundaries. The Alexx produced more low frequencies in-room than did the Alexandria; while MF was impressed by the new speaker's bass, I found it a bit ripe for my taste.

In the time domain, the step response on the tweeter axis (fig.4) reveals that the tweeter and midrange units are connected in positive acoustic polarity (footnote 1), the woofers in negative polarity (footnote 2). More important, the output of each drive-unit can be seen to blend smoothly with that of the next lower in frequency, suggesting good crossover design. The cumulative spectral-decay or waterfall plot on the tweeter axis (fig.5) shows a clean initial decay, but with then some low-level hash throughout the midrange and treble. I suspect that this behavior is due in part to early reflections from the complicated, multi-part enclosure, but the tweeter does have a high-Q resonance apparent at 15kHz. (This should not be confused with the interference from the computer monitor's line-scan frequency just below 17kHz.)

417wilsonal.Walexfig4.jpg

Fig.4 Wilson Audio Alexx, step response on tweeter axis at 50" (5ms time window, 30kHz bandwidth).

417wilsonal.Walexfig5.jpg

Fig.5 Wilson Audio Alexx, cumulative spectral-decay plot on tweeter axis at 50" (0.15ms risetime).

Measuring a loudspeaker as large as Wilson Audio's Alexx without access to an anechoic chamber is problematic, and involves what speaker measurement maven Floyd Toole once referred to in an AES lecture I attended as "hand waving." But there's a lot to admire in the Alexx's measured performance—as well as some things that raised my eyebrows, such as its waterfall plot and amplifier-punishing impedance. But when Michael asked me if the measurements confirmed that he should sell his Alexandrias and buy the Alexxes, I'm afraid I punted. As much as I agree with Michael about the superiority of the Alexx's midrange and highs, I missed the Alexandria's lows.—John Atkinson



Footnote 1: Following the review's publication, I reexamined the step responses of the 2 midrange units and have posted them here. You can see that while the output of the lower midrange unit is in positive acoustic polarity, that of the upper midrange unit is in negative polarity. However, in fig.4, it can be seen that the step response of the upper midrange unit still blends smoothly with the step responses of the tweeter and lower midrange unit to give a time-coherent output.

Footnote 2: In this respect, the Alexx differs from the Alexandria XLF, which has the tweeter and woofers in positive acoustic polarity, the midrange units in inverted polarity. See fig.8 here.

COMPANY INFO
Wilson Audio Specialties
2233 Mountain Vista Lane
Provo, UT 84606
(801) 377-2233
ARTICLE CONTENTS

COMMENTS
SpeakerScott's picture

"The Alexx shares with the Alexandria XLF Wilson's Aspherical Group Delay technology, whereby the positions of its individually enclosed midrange and treble drivers can be adjusted, precisely, to recreate a time-correct waveform at the listening position."

Would it be possible, at some point in the future reviewing a Wilson Audio speaker to measure the spot at the listening position where this is actually true? I've never once seen a review with anything remotely indicating a time domain accurate speaker. JA has accomplished it in the past with Dunlavvy and others...but never Wilson.

-Scott

John Atkinson's picture
SpeakerScott wrote:
Would it be possible, at some point in the future reviewing a Wilson Audio speaker to measure the spot at the listening position where [recreating a time-correct waveform] is actually true?

The step response calculated from the in-room measurement of the Alexx at the position of Mikey's ears that I made with the Fuzzmeasure system is not in conflict with the quasi-anechoic step response shown in fig.4, ie, tweeter and midrange units are connected in positive acoustic polarity, the woofers in negative polarity. And as I said in the review, the output of each drive-unit blends smoothly with that of the next lower in frequency. So while the Alexx's output is not time-coincident, it is time-coherent, which I feel more important.

John Atkinson
Editor, Stereophile

SpeakerScott's picture

"The step respons calculated from the in-room measurement of the Alexx at the position of Mikey's ears that I made with the Fuzzmeasure system is not in conflict with the quasi-anechoic step response shown in fig.4, ie, tweeter and midrange units are connected in positive acoustic polarity, the woofers in negative polarity."

Mr. Atkinson, this statement is a bit confusing...is there a second unpublished step response measurement in addition to figure 4? My issue is not with the performance of the system in the time domain itself, for a large format multi-driver system the measurement in figure 4 is entirely reasonable. I've measured way worse...as have you, and we've both measured better.

My issue comes from the manufacturers claim that adjusted, precisely, to recreate a time-correct waveform at the listening position..

In this case your measurements show, that after adjustment the response at the measurement location is not a correct representation of a step response.

In fact after studying the step response I may disagree with your assertion that the tweeter and midrange are in positive acoustic polarity. Notice the very short positive spike at approximately 3.75mS, in the positive direction, followed very quickly by a negative going "spike" just before 4mS. The negative going one I believe is the tweeter. As you study it, if that negative going spike wasn't there the remaining positive triangle would be the approximate shape from a 6-7" midrange driver operating over a relatively wide bandwidth. Put another way if the woofer(s), midranges and tweeter are operating over three frequency bands, how can you have two negative going spikes if two of the driver bands are supposed to be positive.

My interpretation is that the midrange is the only driver operating in positive acoustic polarity, and the woofer(s) and tweeter inverted. It is unlikely that either woofer would show the negative going risetime between 3.75 and ~3.9mS due to the low pass filters.

The fact that the midrange has such a sharp risetime and long tail indicate they are operating over widebandwidth, possibly with large amounts of overlap both acoustically and electrically which would explain the low impedance in the 500Hz to 3kHz range and the difference in on axis response 50" and 85". I would hazard a guess that had you been able to run vertical polar response measurements there would have been relatively large nulls and peeks across that bandwidth......

Again, my issue is not with the measurements themselves, but the disparity between marketing statements and measured performance.

a.wayne's picture

Never happen in the Time Domain with a reverse phase tweeter and JA 36 " height measuring technique.

John i know its 450 lbs , but I'm sure Between you, Mickey and Darrell you guys could pop this thing on its side and do a couple GP measurements on tweeter axis , no gating or splicing necessary ... :)

Want a better idea ? You could fly out to WA and do your measurements there , it would make for a great story , crap ! i can remember when SP used to do great measurement stories like that ..

Regards

John Atkinson's picture
a.wayne wrote:
Never happen in the Time Domain with a reverse phase tweeter...

The Alexx's tweeter is connected in positive acoustic polarity - see fig.4 in the review's Measurements.

John Atkinson
Editor, Stereophile

SpeakerScott's picture

I think the tweeter and woofer are negative...see my rather long winded response to Mr. Atkinson...

-Scott H.

John Atkinson's picture
SpeakerScott wrote:
I think the tweeter and woofer are negative...see my rather long winded response to Mr. Atkinson...

I am sorry but you are both wrong. You can find the impulse response of the Wilson Alexx, taken in Michael Fremer's listening room, at www.stereophile.com/content/wilson-alexx-inroom-impulse-response. You can see that the tweeter's output - the sharp spike at the beginning of the impulse - is in positive acoustic polarity.

And to confirm what I described in the text associated with fig.4 in the review, you can find the step responses of one of the midrange units and one of the woofers at www.stereophile.com/content/wilson-alexx-midwoofer-steps. The midrange unit's output is in positive polarity (like those of the tweeter and the other midrange unit), the woofer's in negative polarity.

John Atkinson
Editor, Stereophile

SpeakerScott's picture

Mr. Atkinson,

Thank you for the response and additional measurements. I think I understand better now....and now I'm even more perplexed about the time domain performance marketing claims of Wilson Audio.

The additional midrange/woofer step response measurements you provided do not adequately explain the sharp negative spike between roughly 3.5mS and 4mS in Fig. 4 of the review measurements. It's a bit difficult to see given the difference in time scale resolution (5 total mS with .5mS minor divisions vs. 50mS with 2mS minor divisions) but there is nothing in the dotted line that shows the sharp negative spike as it appears in Fig 4.

That now leads me to believe that the tweeter and one midrange are positive polarity while the woofer(s) and other midrange are negative unless you have additional measurements that would indicate the root cause.

-Scott

John Atkinson's picture
SpeakerScott wrote:
Thank you for the response and additional measurements. I think I understand better now....and now I'm even more perplexed about the time domain performance marketing claims of Wilson Audio.

You're welcome, Scott. But I owe you and a.wayne an apology, as while you were incorrect about the polarity of the Wilson's tweeter, I was also incorrect when I said that both of the midrange units were in the same polarity.

SpeakerScott wrote:
The additional midrange/woofer step response measurements you provided do not adequately explain the sharp negative spike between roughly 3.5mS and 4mS in Fig. 4 of the review measurements.

I reexamined the step responses of the 2 midrange units and have posted them at www.stereophile.com/content/wilson-alex-midrange-steps. You can see that while the output of the lower midrange unit is in positive acoustic polarity, that of the upper midrange unit is in negative polarity. So the negative-going spike that puzzles you is actually the step response of the upper midrange unit. It blends smoothly with the step responses of the tweeter and lower midrange unit to give a time-coherent output.

SpeakerScott wrote:
That now leads me to believe that the tweeter and one midrange are positive polarity while the woofer(s) and other midrange are negative unless you have additional measurements that would indicate the root cause.

You are correct and I will amend the review text accordingly.

John Atkinson
Editor, Stereophile

a.wayne's picture

Hello John,

Yes, Lazy typing on my part , I did say reverse phase , instead of out of phase, as all the drivers dont share the same phase, , I did not specify as i should if positive or negative phase and yes you are correct the tweeter is of positive polarity, but that was not what i was addressing.

BTW, Is there any reason why you still favor using gated measurements with such short reflective paths ?

With such a large complex speaker , why not measure it at WA..? we would all like to see what it looks like in the time domain.....

Regards...

John Atkinson's picture
a.wayne wrote:
yes you are correct the tweeter is of positive polarity, but that was not what i was addressing.

See my recent response to SpeakerScott, where I confirm that the two midrange units are not in the same polarity.

a.wayne wrote:
Is there any reason why you still favor using gated measurements with such short reflective paths?

Other than the in-room averaged-response measurements, I use a gated measurement because I want to examine the speaker's anechoic behavior, ie, the direct sound that reaches the ear first.

a.wayne wrote:
With such a large complex speaker , why not measure it at WA?

I have only once in 30 years of measuring speakers traveled to a manufacturer to measure a speaker. One problem I encountered was that having the designer on hand interferes with the measurements because if I find a problem, the designer wants then to change something with the speaker to address the problem, which means the sample is no longer representative of what our reviewer auditioned.

John Atkinson
Editor, Stereophile

a.wayne's picture

"I have only once in 30 years of measuring speakers traveled to a manufacturer to measure a speaker. One problem I encountered was that having the designer on hand interferes with the measurements because if I find a problem, the designer wants then to change something with the speaker to address the problem, which means the sample is no longer representative of what our reviewer auditioned.

John Atkinson"

Ahhh,

Thanks for the response and explanation John ......

Regards ..

John Atkinson's picture
a.wayne wrote:
Thanks for the response and explanation John ......

You're welcome. What I find interesting about this speaker is that while the adjustments Wilson offers don't result in a time-coincident speaker, they do fine-tune the arrival times, taking the different driver polarities into account, to give a smooth blend of their time-domain outputs hence optimal integration in the crossover region. But using the phrase "Aspherical Group Delay" for this facility, no matter how effective, seems misleading.

John Atkinson
Editor, Stereophile

SNI's picture

I wonder when somebody will say thew words that really should go with a speaker like this.
The time domain is really terrable, and the impedance is even worse.
How can this ever be called high end?
To me this looks like someone without any knowledge of speaker design ran into a huge stack of components, drivers, MDF and veneer, and had a lot of fun.
If I did this speakers measured performance, I would start all over again.

The sure to come comment asking "Did you listen to it?" will be answered no, I did not, and I do not see any reason for it.
Good measurements will not promise anything when it comes to listening experiments, but bad measurements will always do so. I´ve never heard any speaker or other component perform well if the measurements were lousy.
And this set of measurements is way in the dessert.

Michael Fremer's picture

Do not listen. You are too foolish a person to let your ears tell you anything.

a.wayne's picture

In the time domain , it's hard to tell based on the way John is measuring the speaker, but i fail to see what's wrong with the impedance graph.

Could you be more specific ...?

SNI's picture

The impedance from ie. 60-200Hz is around 3 Ohms or lower, this is the area with the most enrgy in music.
This lowish impedance will make the life difficult for about almost any poweramplifier, causing the output stage to operate in it´s less linear area for no reason at all.
Also class A amplifiers will quickly be forced away from their most linear mode of operation.
And of all the worst @ 60Hz you have 3 Ohms in conjunction with around 45 dgrs. of phase shift. This is horrible.
To me this looks like very bad speakerdesign, which only quality seems to be it´s size.
And we should not forget the measured time domain behavior, which looks even worse. JA has a lot of times been able to measure TD, with much better results than this. Even with speakers with just as large distance from top to buttom as this one.
So I see no excuse for the poor Time Domain measurements.

Should I say somtheng positive, then it would be, that this speaker is so expensive, that it almost certainly will be paired with very powerfull poweramplifiers, such as Krell, Boulder or some class D powerhouse maybe @ a similar pricetag.
BTW class D will probably be the only amplifier type, which can function as "almost" linear into loads like this. Class A will definately need "sauna" bias to perform at its best.

a.wayne's picture

SNI,

Thanks for the response and I do agree with your concerns , but do oppose your conclusions.

1. Firstly it's nearly impossible to design a multi driver system and maintain an 8 ohm ( where class-A ratings are valid) zmin, also 60-200 is not the bulk of most music , the difficult area for most music reproduction is in the 400- 3K Region and while 3 ohm accompanied with a -45deg phase angle could represent some difficulty to most dinky toy amps , this speaker has reasonably high sensitivity , this does go along way in alleviating drive difficulty and should not be an issue with any competent 2 ohm rated amplifier. I can imagine most listening will be done in the 1-2 watts RMS range with 200 watts or so on dynamic peaks. This would mean 2-5 watts of class-A power at 3 ohms should suffice, if this is your concern ( Only 20 watts@ 8 ohm of class-A power would be necessary to achieve this ).

2. TD, agree, but , I would have to see the anechoic/GP measurements ( with phase) on tweeter axis with the alignments done for proper TD and not setup for subjective listening as tested .

3. Actually IMO, class-D amplifiers are best suited for 8 ohm operation and not for low Z loads.

Regards ..

SNI's picture

Mr Wayne

I wish it only was flimsy power amps, that will be stressed with impedances like this speaker. But I´m afraid that will be wishfull thinking.
Even the most stable and strong amps performs better with a more human load.
But even if that is left aside, the TD is still horrible, and I´m flabbergasted that someone making speakers this expensive, can get the idea of reversing the phase of one or more drive units.

I´m sorry, but to me this looks like carpentry more than speakerbuilding, and for the latter only for one self.
As a commercial product at the asking price, we have one of these high price audio products so frowned upon by engineers.

Class D will remain linear far longer than class A will @ lower frequencies, but at higher ones class A will be able to stay linear into more difficult loads than class D.
Class A/B is not linear at anytime, but the harder the load, the more distortion you´ll get especially at low volume.

Michael Fremer's picture

There's no wood involved. Please never listen to these speakers. Look at the measurements and decide. That's your best bet here.

monetschemist's picture

Mr. Atkinson,

I wonder if it would be possible, and worthwhile, to do an article (maybe a series) that more fully explain what we see in your graphs? Perhaps take a couple of speakers that you have recently tested, do some more annotated graphs ("this point here shows the effect of an out of phase frammistat or perhaps a busted knobflicker"). It would be especially cool if you are able to correlate the measurement with what weirdness it creates in the listening experience...

Perhaps this is not feasible, but it seems to be an interesting idea.

John Atkinson's picture
monetschemist wrote:
I wonder if it would be possible, and worthwhile, to do an article (maybe a series) that more fully explain what we see in your graphs?

I have written 3 articles on this subject, based on a paper I presented to the Audio Engineering Society 20 years ago:

www.stereophile.com/features/99/index.html.

www.stereophile.com/features/100/index.html.

www.stereophile.com/features/103/index.html.

And in 2011 I gave an illustrated lecture on the subject at the RMAF:

www.youtube.com/watch?v=j77VKw9Kx6U.

John Atkinson
Editor, Stereophile

monetschemist's picture

I look forward to reading these! Thanks very very much.

es347's picture

..and was mildly impressed. The fit and finish on any Wilson speaker is top shelf but the sonics have never been my cup of tea..

Michael Fremer's picture

judging how a speaker sounds at a hi-fi show hotel room is a fool's errand. It's funny how people who like a certain brand will say "oh it didn't sound good, but it's a hotel room". And for brands they don't like, they leave out the "but"....and just say what you said.

JimboJumbo's picture

Another unimpressive test result for a very expensive passive speaker.

My God, no wonder some of these manufacturers ask us to listen rather than measure and/or be guided by tests.

To the discussion about the step response; as it was clear to me what was initially written about was wrong.

My interpretation is that the upper mid and woofers are negative polarity and the rest positive.

In some of the comments John says; “It [the negative going transitions of the upper mid, bass, and overall step response] blends smoothly with the step responses of the tweeter and lower midrange unit to give a time-coherent output”, then he later says “What I find interesting about this speaker is that while the adjustments Wilson offers don't result in a time-coincident speaker, they do fine-tune the arrival times, taking the different driver polarities into account, to give a smooth blend of their time-domain outputs hence optimal integration in the crossover region. But using the phrase "Aspherical Group Delay" for this facility, no matter how effective, seems misleading”.

With all due respect, that step response is not symbolic of a time coherent frequency divided 4 way system.

Such disparity between marketing statements and measured performances.

The impedance, CSD, and step/frequency response should be way better for a speaker of this name/cost.

Michael Fremer's picture

The comments here from people who have never heard these speakers and from one who heard them in a hotel room at a hi-fi show. I have never met any of these people but they are all really butt ugly.

es347's picture
es347's picture

..are you and Peter Breuninger still BFFs?

X