Magnepan LRS Speaker Review

[tuga]

Simple - if it is flat at 300 Hz, one octave below at 150Hz it must be -6 dB, and one octave below 150 Hz it is 75 Hz with another -6 dB, for a total of -12 dB at 75 Hz.

Can you use these LW plots of the PSB T3 to illustrate your point?

 

[tuga]

Well OK, but if the PIR calculation is off it's nothing to do with how the measurements were taken. As far as I understand, the PIR is a weighted average of listening window, early reflections and sound power. Its calculation doesn't even involve any of the specialized Klippel math.

If someone thinks the PIR is wrong they're going to need to collect some inconsistent measurements(the JA one doesn't seem inconsistent to me, and I frankly trust the mess Stereophile makes of measurement below 300hz much, much less than the Klippel anyways.) and potentially show the correct weights for a dipole if they differ for whatever reason.

I agree, I later corrected to interpretation of the Klippel data.

 

[Blumlein 88]

The Klippel measures (ultra-)nearfield, unlike anechoic.
Panels have a very large radiating surface.
There's a slight chance (ultra-)nearfield might not the best way to go at it...

This was @John Atkinson on measuring the ESL-2912:

It's always difficult to assess the acoustic behavior of a large panel speaker like the ESL-989.
First, the nearfield measurement of the low frequencies will not show the effect of the dipole cancellation, as the antiphase backwave increasingly wraps around to cancel the speaker's direct output with decreasing frequency.
The peaked-up bass shown in fig.2 will therefore tend to be more flat in the farfield than it appears in this graph. In addition, when measuring a loudspeaker, it is assumed that the microphone is in the speaker's farfield; ie, several times the largest dimension of the diaphragm.
With the largest dimension of the ESL-2912 being its panel height of 54" and my routine distance from the microphone being 50", this assumption is no longer true.
As a result, due to the proximity effect, there will be a slight downward trend with frequency in the measured response, as fig.2 also shows.

Fig.2 Quad ESL-2912, anechoic response on center-panel axis at 50",
averaged across 30° horizontal window and corrected for microphone response,
with nearfield response of woofer plotted below 300Hz.

https://www.stereophile.com/content/quad-esl-2912-loudspeaker-measurements

Simply no. Though done ultra-nearfield, the Klippel results can be extrapolated to any distance. You are wasting your time arguing with the Klippel results because it has been shown to equal anechoic results to a high degree of agreement. And the manner in which works is as valid for panels as for box and cone speakers.

Now interpretation is another issue. JA mentions he is measuring near field and proximity effect corrupts results. The Klippel device has no such issues.

Again would be nice if some LRS owners would do an REW measurement for us.

 

[Vladimir Filevski]

Can you use these LW plots of the PSB T3 to illustrate your point?

Yes, I can. PSB T3 is a classic monopole speaker, so you should subtract gradually up to 6 dB at the peak at 75 Hz (Stereophile measurements) and - voila! It is the same as the true (within chamber error) anechoic measurements from Soundstage.
For a dipole loudspeaker you need additional -6 dB/octave slope.

 

[DSJR]

I've been reading much of this thread with something akin to horror!

My experience of Maggies as a breed, has been that the one thing that really helps 'sort them out' is POWER. I remember a particular dem of a lower mid size model some years back now (1,3's?) where we compared two early Levinson power amps (black fronted and I'm sorry but I forget the numbers). The larger more powerful model brought these panels to life, the sound-field expanding in all areas and the speakers themselves 'disappearing' more from the scene. Really nice with the bigger amp and with the smaller amp, not so different from some of the less positive comments here. I heard some 1.7's a couple of years back and driven with I think, a Naim amp (totally inappropriate on almost all levels ) and they were totally 'meh' and dismissable frankly. Remembering their ancestor's dem, I suspect a larger more capable amp would have helped no end...

Thirty five years ago, a once well known 'tweaky reviewer' who lives in a fascinating split-level appartment in London's Barbican (older UK based readers will guess who) bought a pair of IIIa's. As the listening seat was a few feet above the level the speakers were sited, he mounted them on the side walls off the ground, turned them round with magnets behind and on the occasion I heard them (could have been Krell KSA50 driven?), the sound was most impressive (on my then numerous visits, we always tended to concentrate on the music a lot more than the sound of the stereo gear he had set up). He went away from them subsequently as in his words, it was like watching a tacky soap opera in full high-def IMAX or similar all the time (hope non UK readers understand what I'm trying to say). No doubt a better room environment and large enough amp would have done wonders for them and for me, this kind of presentation is far preferable to omni-directional types which 'play the room' even more, further reducing direct soundwaves I think..

I also heard a lovely sounding set of SMGa's years ago when visiting clients. For small scale orchestral and chamber music at sensible volumes, they were delightful in this client's room - amp was an integrated Revox I think (100WPC?) with matching tuner set to UK radio 3 or 4!

As for the Quads, the 63 was measured more conventionally in the 80's (1m and 2m distance I remember in the small-form HiFi Choice 'Speaker' group test books) and the response was pretty flat from mid bass to lower hf, the only subjective rider on that being in my experience, the domestic version needed lifting off the floor and tilting back slightly to get over the substantial mesh protection grilles under the cloth covers, to prevent a rather dead tonal balance (get 'em right and they were magical in the right setting at not silly rock levels). I've only fleetingly heard the later post tale-over models, small and large, but these don't seem to deviate too much from the original 63 version, although all of them have a very slight 'sparkle' at hf which can aid string tone no end (I think it was measured in the 63).

Many of you do seem to have larger rooms and use subs to get the bottom octaves re-enforced. We tend not to do this on the UK and as I said in the review (and subsequent annihilation here) of the ATC 19 early model, a rolled away bass isn't necessarily a bad thing in smaller rooms...

 

[Frank Dernie]

The Klippel measures (ultra-)nearfield, unlike anechoic.
Panels have a very large radiating surface.
There's a slight chance (ultra-)nearfield might not the best way to go at it...

I would expect this only to be a problem if the measurement point cloud has too few locations.
As long as there are sufficient measurement locations the anechoic soundfield can be calculated accurately any distance from the speaker.
I don't see any problem with the measurements, only the in room estimation which, IME, is very, very sensitive to room position. I have no idea how the PIR calculation relates to position and orientation wrt walls but I suspect it is only here where a poor estimation may be made (if one has been, which is not guaranteed).
IMO

 

[tuga]

Yes, I can. PSB T3 is a classic monopole speaker, so you should subtract gradually up to 6 dB at the peak at 75 Hz (Stereophile measurements) and - voila! It is the same as the true (within chamber error) anechoic measurements from Soundstage.
For a dipole loudspeaker you need additional -6 dB/octave slope.

Thanks, that makes sense.

 

[tuga]

I would expect this only to be a problem if the measurement point cloud has too few locations.
As long as there are sufficient measurement locations the anechoic soundfield can be calculated accurately any distance from the speaker.
I don't see any problem with the measurements, only the in room estimation which, IME, is very, very sensitive to room position. I have no idea how the PIR calculation relates to position and orientation wrt walls but I suspect it is only here where a poor estimation may be made (if one has been, which is not guaranteed).
IMO

It is clear to me now that my comments should have been directed to the interpretation of the Klippel data, namely the PIR.

 

[Thomas savage]

Sorry - it invalidates EVERYTHING you've said. I've owned many pairs of Magnepans for many years. I attend 30-40 live concerts every year. I listen to a lot of music through my Magnepans, KEFs, Totems and KLH's. Passing judgement on speakers you have never heard is simply nonsensical. Revelling in one's ignorance is an interesting way to make a point. To quote Benjamin Franklin, "We are all born ignorant, but one must work hard to remain stupid."

This sort of thing is very unhelpful and betrays the spirit of debate we want to host here .

 

[Vladimir Filevski]

As for the Quads, the 63 was measured more conventionally in the 80's (1m and 2m distance I remember in the small-form HiFi Choice 'Speaker' group test books) and the response was pretty flat from mid bass to lower hf, ...

I do remember vintage "booklet" style HiFi Choice magazine. All loudspeakers were measured 8 m off ground, so for dipole it was essentially anechoic measurement, better than in an anechoic chamber. Later they were using anechoic chamber.
Yes, Quad 63 was very flat anechoically all way down - in contrast to Magnepan LRS. But we have to look at the price difference...

 

[Juhazi]

Still guessing, but this is how I see the "poor" PIR estimate...
NFS sotware calculates Estimated In-room Response (called PIR in this thread) from nearfield scans. Most speakers are monopole with a box and radiate onmispherically in bass, and that give the typical PIR with sloping response in high freq. But a dipole has much less spl at angles 60-120deg horizontally and as well vertically in bass. Typical monopoles on the other hand have less spl at wide angles/high freq! This will result in less than expected PIR spl of low freq for dipoles like Magnepan LRS! Add to this line radiator vertical interferences at high spl

This should be really simple to verify with a multipoint or MMM room response measurement of LRS and another boxed 3-way speaker and then comparing the scans, PIR and measured room response of both. But still were are just arguing and nitpicking with details and midsunderstanding each others...

So, who will add LRS's in-room response here?

Actual in-room response with long gating or RTA, or averaged multipoint response
(Missing)


Amir has only published only some measured room responses of Klippel-tested speakers, I don't remember any now.

 

[Vasr]

We'll hear something much like it measures, but it will almost always be more uneven and elevated vs anechoic measures.

So in essence there is no measurement that matches what we hear the way you are asking of it to do. We know enough to piece it together pretty well.

So the measurement that tells us what we'll hear listening to speakers in a typical setting is the Klippel measurements down to 500 hz or a bit lower depending upon room size. That lower end part is very important, but the in room version of it is something of a kludge, an add on. You probably should rely on in room measures for the low end. But there is no universal room.

Good. Like I said a trick question.

You have agreed in this thread many times that in-room measurement should have been taken. So we are not disagreeing on that front. Especially when an inference is made on the behavior of the DUT in that lower region. Let us ignore for now whether the Klippel measurements taken when the unit is NOT placed in its listening position is a good indicator of observed behavior because there is a more fundamental problem even if it was.

Your characterization of the in-room measurement gets to a fundamental dilemma of experimental science.

Do you use a system that is unable to model the observed phenomena sufficiently (by eliminating some contributing parameters) or do you use a system that has variances due to those parameters to provide a precise answer? When the goal is to make an inference on what the observed behavior is likely to be in general.

It is a trick question again. The answer is it depends on which one gets closer to the observed behavior in expected scenarios. If the observer subjectively "heard" the results of the former, did he hear the sound or did he hear the measurement? Hence, the need for corroborating evidence before drawing conclusions.

Take as a rough conceptual analogy of weather forecasting - the models for forecasting beyond a day or two gets way too complicated and depends on far too many variables. So, you use a simpler model for short-term forecasting for which precise data is available and you use a more complex model but with a subset of contributing data to make long-term (3-5 days) prediction. If the latter is good 70% of the time, you will take that over a simplified model that is correct only 20% of the time.

I see a similarity here in the prediction of low-frequency behavior.

For anyone that has stared at enough REW measurements, it is fairly obvious what is a specific room mode speaker placement behavior (typically nulls and peaks) and what is speaker behavior in most rooms (roll-offs). The latter is seldom affected much by room geometry (I am leaving out specific room treatments, etc). So, if you were to make an inference on likely bass extension, you would want a measurement that more accurately predicts the roll-off even if the peaks and troughs may be affected by the room. So, I wouldn't dismiss it off as a kludge. Don't confuse the precision of measurement with adequacy of the model.

The Klippel measurement interpretation here suggests a 300hz roll-off. In-room measurement seems to suggest a 50hz roll-off. This is a fundamental and discordant result between the two that makes a significant difference to the evaluation of the speaker. Before you accept the former as the "truth" for what one can experience in ANY room, you would need to corroborate with actual in-room measurements even if it is not a "perfect" measure for EVERY room.

Until then, it is bad science - an inappropriate use of a measurement/model.

 

[Vasr]

Regardless from my perspective the only real way to move forward for those insisting that something is wrong with the measurements would be to acquire some REW measurements of the LRS in various places and see whether they actually do have more bass than the PIR calculation is expecting them to have.

Or to put it another way, the only way to say that the Klippel measurements were correct in predicting in-room behavior is to acquire some REW measurements of the LRS in various places and see whether they actually do have such a "broken" bass deficiency the PIR calculation is expecting them to have. It could be, I don't know. But this is what you do in experimental science when a model predicts a behavior.

As an owner I honestly don't see how a panel the size of the LRS could possibly produce sufficient bass when some of the significantly larger Magnepans are already somewhat disappointing in that area, until you get up to the 3.7 or 20.7 at least. Basically I don't think these results are nearly as surprising as so many people think.

This is a bit misleading. Because it is using an ambiguous/imprecise term "sufficient bass".

Magnepan models differ in their bass extension (the point at which the response starts to roll-off) just like for any speaker of any design. The earlier smaller Maggies (not including the wall-mounted tiny MMG-W which rolled off at 500hz - see REW I posted earlier) rolled off around 100hz and so would be heard as weak in bass. As you went up the chain, the bass extension would go lower fairly flat all the way down to say 30hz with bigger and bigger panels BUT the problem was that for the price at each point, there were other speakers which had better bass extension or you could get better bass extension at lower prices. This is what is behind the reputation of Maggies as not having sufficient bass.

The predicted measurements here suggests a 300hz roll-off for the smaller LRS Maggie significantly worse than any of the decades-old similar size Maggies of the past. This to me is very surprising. It says either the LRS is really broken and should be burnt OR the measurement does not capture reality OR the interpretation of the measurement is incorrect.

 

[RayDunzl]

Experiment out of curiosity here with my non-magnepanels and little JBLs.

Set 75dB pink level at the listening position.

Measure level "at the drivers" - microphone a few millimeters away.

MartinLogan reQuest - 80dB
JBL LSR 308 - 110dB

 

[Sancus]

Experiment out of curiosity here with my non-magnepanels and little JBLs.

Set 75dB pink level at the listening position.

Measure level "at the drivers" - microphone a few millimeters away.

MartinLogan reQuest - 80dB
JBL LSR 308 - 110dB

I'm really confused how that would be relevant to the discussion at all, honestly.

 

[Sancus]

The predicted measurements here suggests a 300hz roll-off for the smaller LRS Maggie significantly worse than any of the decades-old similar size Maggies of the past. This to me is very surprising. It says either the LRS is really broken and should be burnt OR the measurement does not capture reality OR the interpretation of the measurement is incorrect.

Well, the panel area of the LRS is only about 80% larger than the MMG-W. The 1.7i is 3.25x(!) as large as the MMG-W, and I always thought of it as the smallest panel size where they legitimately start to sound decent. Smaller ones I've heard(MMG-W, center channel variants) sounded pretty bad to me. So I still don't think it is that surprising. I don't really believe that Magnepan finally found some magic that lets them make ever-smaller panels work. I think they are just making products to a price point.

 

[NTK]

Or to put it another way, the only way to say that the Klippel measurements were correct in predicting in-room behavior is to acquire some REW measurements of the LRS in various places and see whether they actually do have such a "broken" bass deficiency the PIR calculation is expecting them to have. It could be, I don't know. But this is what you do in experimental science when a model predicts a behavior.
...

I don't know how aware people are on the origin of the "Harman target curve", which is often used as the reference which the estimated in-room response (aka PIR) curve is compared to.

@thewas_ has quoted Dr Toole in this post in another thread many months ago, which I'll again quote here [see note below for the link to Dr Toole's original post]:

Floyd Toole said:​

...​

As the "creator" of the "new Harman target curve" I can clear up some misunderstandings. Those who have the 3rd edition of my book can see where the curve came from - Figure 12.4. It is nothing more than the steady-state room curve that results from measuring any of several forward-firing loudspeakers that have been awarded very high ratings in double-blind listening tests. These steady-state room curves are substantially predictable from the "early reflections" curve in the spinoramas, as is illustrated.​

​

Now, if you measure such a curve or something very close to it, and your speakers are conventional forward firing designs, it means that you probably have chosen well. ...​

Therefore, IMO the "standard" interpretation of the PIR curve may indeed not be as applicable to panel type (dipole) or omni speakers.

[Note] Dr Toole's original post in AVS Fourm: https://www.avsforum.com/threads/of...vel-home-theater-thread.2515137/post-57820394

 

[RayDunzl]

I'm really confused how that would be relevant to the discussion at all, honestly.

Something is relevant, I'm not sure it has been discovered as of yet.

I'd find this to be a fundamental difference between a panel and a point source, if extrapolating soundfields.

Oh well.

Carry on!

 

[Blumlein 88]

Something is relevant, I'm not sure it has been discovered as of yet.

I'd find this to be a fundamental difference between a panel and a point source, if extrapolating soundfields.

Oh well.

Carry on!

I've seen it referred to as surface loudness. Like in lighting surface brightness or intensity. You could have the same lumens from a single bulb filament which might be blinding. Or a large panel with low level non blinding surface intensity.

 

[Blumlein 88]

Okay, not as good as someone who would measure their LRS for us.

I did an REW room simulation for a room 18 ft x 14 ft x 10 ft. Put subs 4 ft from rear wall, 3 ft from side wall. Two of them. Only 4 inches apart from each other. Set the crossover at 400 hz. Inverted one. Doesn't give us the right result at higher frequencies where the panel would be beaming. Should give some insight into lower frequencies once we are below the point where the front and rear wave are canceling each other.

Firs the anechoic result of the woofers.
They mostly cancel each other out, increasingly so with lower frequencies.

Here is the simulated result in room. Notice there are of course peaks and dips, but we'll hear the peaks, and they are about 10 db higher than the anechoic result. So such a dipolar woofer would have louder, greater perceived bass than the anechoic curve suggests. Which isn't a surprise. It might not sound 10 db louder than anechoic predictions, but it will be discernible vs not by several db.