How to Measure Magnepan LRS

[amirm]

Maybe @NTK could comment on measuring a panel, as he seems to be quite familiar with the Klippel.

https://www.audiosciencereview.com/...-how-the-klippel-nfs-works.13139/#post-393407

We already had the theoretical discussion in that thread or another. Time for the real deal.

 

[abhijitnath]

I use absorption behind to reduce comb filtering -- that is one thing that always drove me nuts (admittedly a short drive) and I rarely had a big enough room for it not to be a problem.

The smaller models are more like line sources over a wider range than the larger models.

Placement of dipoles is a PITA but the rewards are great. For measured response, I must bow to the Scrutinizer ( @RayDunzl ), who has shown great in-room responses for any speaker let alone dipoles.

Any tips for placement? I'm still struggling with my spatials

 

[Newman]

Maybe @NTK could comment on measuring a panel, as he seems to be quite familiar with the Klippel.

https://www.audiosciencereview.com/...-how-the-klippel-nfs-works.13139/#post-393407

In that thread posts 31-38.
cheers

 

[RayDunzl]

In that thread posts 31-38.
cheers

Thank you.

I'm at the point here now where I can't even remember my own posts.

 

[Vladimir Filevski]

It illustrates a symptom of what I hear - more focused vs less focused "image" in-room.

1/48 View attachment 80962 1/24 View attachment 80963 1/12 View attachment 80966 1/6 View attachment 80967 1/3 View attachment 80968

Frequency response is not a noticeable problem, but sense of location is affected, when critically listening in the sweet spot.

Thanks! Could you post impulse response up to 35 ms (of both JBL and ML, on one diagram), with 1/24 octave averaging, please?

 

[Mittomen]

There is a letter to Stereophile from Siegfried Linkwitz on the topic, here:
https://www.stereophile.com/content/magnepan-magneplanar-mg36r-loudspeaker-more-comments

Comments on Nearfield Measurements of Panel Speakers: (The following letters were received but were not published.)
Siegfied Linkwitz comments

Editor: The review of the Magneplanar MG3.6/R in the August Stereophilecaught my attention. I am a proponent of open-baffle speakers because of their room acoustic advantages and the absence of sound coloring boxes. So I looked with great interest at figs.2 & 3 on page 89 showing individual driver frequency responses and their summation.

The nearfield measurements of woofer and midrange in fig.2, presumably taken only an inch or so from the driver surface, are a valid set of data. You also could have measured the tweeter at such close range and obtained useful information. Where things fall apart is in fig.3 when you form the complex sum of nearfield measurements and the 50" tweeter "farfield" measurement. This curve does not represent the frequency response a listener might experience at any distance and is therefore extremely misleading.

The nearfield frequency response of an acoustic source is only proportional to its farfield response if the source is small, ie, omnidirectional, and if it is in free-space. Summing a driver diameter corrected woofer nearfield response to a farfield midrange response works for a small monitor on a stand, but already has errors when the speaker is larger and the woofer is close to the floor—when the conditions move away from free-space or anechoic.

The Magneplanar is clearly not a point source and, being open-baffle, it has an acoustic short circuit between front and back. This causes a 6 dB/octave low-frequency roll-off in the farfield response. So from all open baffle nearfield measurements you have to subtract first a 6dB/octave (= 20dB/decade) slope before you can sum the data with other farfield measurements. When you apply this correction to the MG3.6 woofer response you see that it flattens from 400Hz to 60Hz and shows a peak at 47Hz. Similarly the midrange has to be corrected before you can use it for the composite response. The actual room response is still different from this composite, though, primarily due to the effect of the floor on woofer radiation.

You might consider to add a measurement taken with a 50ms time window at your listening position, spatially averaged and half-octave smoothed to include the room. I think as a measurement that allows true comparison between speakers, this would be more useful than the composite data that are correct only in a few special cases.

I hope this letter helps your readers to understand the difficulties in describing a loudspeaker by measurements.—Siegfried Linkwitz, Corte Madera, CA, www.linkwitzlab.com.

 

[raindance]

It appears that they are normally attenuated. You are only supposed to bypass it if the room is especially dead sounding.
/QUOTE]
No, the resistors are added by the user as needed. I don't recommend them as the tweeter crosses very low, and they attenuate mid band as well as treble.

 

[abhijitnath]

In that thread posts 31-38.
cheers

Haha thank you!

 

[Newman]

No, the resistors are added by the user as needed. I don't recommend them as the tweeter crosses very low, and they attenuate mid band as well as treble.

I was referring to what the manual says in post #3.

 

[raindance]

I was referring to what the manual says in post #3.

I read the manual and have owned many Maggies . Tweeter is "full volume" with the factory jumper in place and can be attenuated by replacing it with a resistor.

 

[Newman]

There is a letter to Stereophile from Siegfried Linkwitz on the topic, here:
https://www.stereophile.com/content/magnepan-magneplanar-mg36r-loudspeaker-more-comments

Have a read of the posts that I linked to the Klippel thread, and I believe it is clear that the kind of commentary that SL is making is not relevant to the Klippel system.

 

[Mittomen]

Have a read of the posts that I linked to the Klippel thread, and I believe it is clear that the kind of commentary that SL is making is not relevant to the Klippel system.

I did. But I'm no expert - so I thought it's better to post even it is irrelevant than to not post something what might be relevant.

 

[Ville]

It is less than 1/3 meter.

Hello! This is my first post. After reading the forum for quite some time and learning a lot --- thank you all for that --- this upcoming measurement using the Klippel finally caused me the urge to say something. This measurement and its results will be very interesting!

This might belong to the NFS math/physics thread...

The series expansion solution of homogeneous PDE for the sound field is valid it the source-free region (volume) between internal (loudspeaker) and external (walls/reflections) sources. From what I read Klippel probably uses spherical expansion/harmonics*). This means the source-free region is a spherical shell. Now, we measure some source that is about 1 meter tall (?) at the distance of 1/3 meter. Then part of the source will be inside that spherical shell because its inner surface is determined by the measurement point nearest to the expansion origin. In the series expansion, this part of the data fits both parts, the internal and external series. This would likely manifest itself as "leaking" of SFS or poor rejection of reflections. So excited to see the results! The physics work, you know.

I have some hands-on experience with partial differential equations (PDE) and spherical harmonics. The sound field separation (SFS) appears to be analogous to multipole expansion in electromagnetic field theory. I am not yet comfortable with sound field equations/physics but still, I think this particular measurement may reveal us some limitations of NFS/SFS or, hopefully, help us reverse-engineer it.

BR,
Ville

*)The math with spherical coordinates is relatively speaking "easy" with analytical equations. Although the measurement surface, or rather shell, is cylindrical, I really doubt they'd use cylindrical expansion. It is only required that the measurements are made in the source-free region and rest is fitting/regularization... Most importantly, math for finite, i.e., not infinitely long, cylinders is either tricky, requires numerical methods, or both.

 

[Deleted member 2944]

Where did you get that idea?

I already explained in my previous post. You seem flummoxed by the basic physics of this.
The microphone is way inside the nearfield of a longish transducer. It would be impossible for a single measurement 'not' to exhibit some form of comb-filtering. If you can't grasp that I'm not sure what to say.

Cheers.

Dave.

 

[richard12511]

MartinLogan (black, 30 degree curved panel) vs JBL LSR 308 (red), both with EQ to "flat".

The JBL are immediately to the outside of ML, so, almost the same location.

My interpretation of this unsmoothed frequency response taken at the listening position is less "interference" from room reflections upon the direct sound.

View attachment 80951

The hole around 50Hz is a room problem, the phase of the standing wave of the bass from the two channels goes into 180 degree opposition.

This is how it looks on an extended impulse response:

View attachment 80954

I follow what you're saying here. Makes sense.

 

[NTK]

Maybe @NTK could comment on measuring a panel, as he seems to be quite familiar with the Klippel.

https://www.audiosciencereview.com/...-how-the-klippel-nfs-works.13139/#post-393407

Thanks for over-estimating my knowledge of the Klippel NFS I am 100% sure by the person by far most knowledgeable about the NFS is Amir (unless some Klippel people are lurking here without announcing themselves). I may be able to throw out an equation or two occasionally, but it is totally incomparable to Amir's real experience with testing over 100 speakers.

I agree with the boss man here. This is rock 'n roll time, damn the torpedo, full speed ahead.

 

[DonH56]

Any tips for placement? I'm still struggling with my spatials

Not really, play around? Hand waving follows... They work best at least 3' and even better 5' or more from the rear wall. They are not terribly sensitive to side walls but are not completely insensitive, so if they are a foot or so away probably not a problem. I am not familiar with the LRS explicitly but generally Maggies sound (and measure) best to me with the tweeter inside and the panels toed to face the listener, maybe just a little out from that, but tweak them in small increments and listen for the image to "snap" into place. It may be easier to use a mono source (or set the preamp/processor to mono) and sometimes pink noise makes it much easier to hear things like comb filtering and such. Magnepan's big panels (and sometimes crossover) makes the tweeter's sound "lead" the bass by a bit so toe-in puts the tweeter more in line acoustically with the bass panel.

You can also move the listening position back and forth. Some people prefer them very nearfield, with the listener only a few feet in front of the speakers, much closer than the typical equilateral triangle positioning of speakers and listener. The speakers would not be toed in all the way, just leave them where they sounded best when sitting farther away. Try it and see -- I find it nice at first, and for some music, but sort of "overwhelming" in the long term.

As an aside, new owners often ask about the "dots" placed on each panel, and why they are not in the same place for left and right speakers. Those big panels have their own modes, frequencies at which they exhibit resonant peaks and dips, due to the acoustic signal across the panel. Similar to break-up in conventional cone drivers. The "dots" fix (tie down) the panels at spots to reduce those modes. They are different for each side so that the remaining modes are not reinforced by both speakers at the same frequency.

HTH, FWIWFM, etc. - Don

 

[amirm]

The series expansion solution of homogeneous PDE for the sound field is valid it the source-free region (volume) between internal (loudspeaker) and external (walls/reflections) sources. From what I read Klippel probably uses spherical expansion/harmonics*). This means the source-free region is a spherical shell. Now, we measure some source that is about 1 meter tall (?) at the distance of 1/3 meter. Then part of the source will be inside that spherical shell because its inner surface is determined by the measurement point nearest to the expansion origin. In the series expansion, this part of the data fits both parts, the internal and external series. This would likely manifest itself as "leaking" of SFS or poor rejection of reflections. So excited to see the results! The physics work, you know.

NFS doesn't use spherical scan but a modified one using a cylinder and flat top. A spherical scan would require much more horizontal space than this method.

As to distance, that is something the user sets. In my setup, I don't have the room to go much past 1/3 of a meter. I have the horizontal room but not the vertical height for a wide scan of that sort. The scanning distance impacts accuracy in very low frequencies. But then there are other issues there with noise, etc.

Note that I have measured pretty tall speakers already and those have excellent agreement with anechoic chamber results. Furthermore, the NFS system makes redundant measurements so that it can then provide a measure of error in its computation versus reality. The matching is usually excellent and if not, I can up the measurement points to get there.

All of this should work just as well with this speaker in theory but we will test to be sure.

 

[pozz]

I did. But I'm no expert - so I thought it's better to post even it is irrelevant than to not post something what might be relevant.

Thanks for posting it.

 

[Ville]

Thank you for the reply @amirm !

I am quite confident to claim that the spherical expansion coefficients can be fitted to measurements done in any shape. Sub-optimal measurement grid likely causes less than optimal numerical stability but it should still be manageable with almost full 4pi spatial angle coverage. Thanks for reminding me about all the physical constraints, they are what they are and the algorithm is probably doing something clever to work around those. One of the means being acquiring more sampling points. Or they might be using the scary-ugly cylinder-math...

As you say, you have measured tall speakers and the results look great. In (some of) those there is a rise or flat response near 20Hz region at very low level. Is this perhaps related to noise level or could the port be the cause if very low and possibly inside the measurement shell (I was talking about)? Not that there would be anything to fix with those measurement, a brain fart: a tall speaker could be measured lying on its side to increase the measurement distance.

I guess my curiosity would be satisfied by "breaking the rules of sound field separation" and forcing some part of the speaker in there (source-free region). Yes, not very efficient use of time to reverse-engineer but Klippel does not tell us the juicy details.

Can I read somewhere about the scanning distance and the accuracy in very low frequencies? The governing PDE seems to tie temporal and spatial sampling requirements together in non-obvious way, to me at least. Btw, reading the forum and also equations, physics and psychoacoustics here at ASR, also the subwoofers&room-EQ why's and how's are somehow making more and more sense.