KEF LS50 Meta Spinorama and Measurements

[hardisj]

Great review, as always. Thanks for the spins!

 

[jhaider]

"Triamp", right? They don't have 3 pairs of posts, sadly.

If you really want access to each driver, throw the plate off and replace with a new one. That way one can also expunge the dumb legacy connectors and use a modern 8-pole Neutrik Speakon.

 

[napilopez]

@Maiky76 if you want to try to EQ/score with my measurements, here's the MDAT with the spin files. Removed the 1/24 octave smoothing in my posted measurements since I'm not sure if Amir's measurements use any smoothing when you process them.

Curious to see how it differs vs KEF's measurements, though as noted earlier it's likely I was not perfectly aligned on-axis so the on-axis diffraction doesn't show up. (I think since I'm measuring outdoors now the ground was not perfectly level so I might've been off by a degree or two in the vertical plane).

 

[mieswall]

I've had KEF's new darling, the LS50 Meta, for a couple of months now, but have only just been able to complete measurements of them. I expected this one to be a fairly straightforward measurement, as luckily KEF provides its own spinorama for the LS50 Meta, complete with early reflections curves and all. It is available in the company's whitepaper(p10), but for ease of comparison and scale-matching, I've imported that spin to REW:

View attachment 98826

We see a pretty darn good speaker, whose deviations from neutral are just a dip in the upper mids, some on-axis diffraction, and a slightly warm tilt; we can see the on-axis and listening window curves have a slightly negative slope rather than being completely neutral.

Though my results are largely similar, there are some small but notable differences. These may be due to differences in measurement technique or my own error, but if anything, the speaker comes out looking better in my spin. Here's mine:

View attachment 98835
(Please note that some of the bumps from 400-800Hz are likely due to my new outdoor measurement rig, which I am still trying to refine.)

The differences are subtle, but worth noting:

Though the dip in the upper mids remains, my measurements show much better behavior on-axis. The most likely explanation for this is simply that I was not perfectly aligned with the center of the tweeter for the diffraction to take hold.

If so, it just shows how meaningless on-axis diffraction is so long as it gets smoothed out in the listening window. I repositioned my microphone about five times and was not able to match the on-axis diffraction shown by KEF. In any case, this is all but irrelevant considering KEF explicitly tells you to listen off-axis in the manual:

View attachment 98838

The more important difference is the tilt or slope of the curves. Both KEF and my measurements show smooth, mostly flat lines, however, KEFs are more tilted/darker/warmer than mine. For example, here are my Listening Window and Early Reflections curves vs KEF's (dotted):

View attachment 98857

If we line up the measurements in the low mids, KEF's measurements show roughly 2 dB less energy after 2kHz. This would definitely be an audible difference; while both sets of measurements indicate a good sounding speaker, KEF's suggest a slightly warm titlt, while mine suggest something more neutral.

I spent a while thinking about this difference in tilt. I thought it might be temperature, as it's getting colder here, and we've seen how temperature can affect woofer output in Amir's measurements. But the speakers were indoors right before I brought them out to measure, and 56 degrees isn't that cold.

So I decided to check out Soundstage Network's measurement of the LS50 Meta, as my measurements have always agreed closely with theirs. Indeed, digitizing their on-axis measurement, they seem to have the same achieved the same tilt as me:

View attachment 98861

The results are, for all intents and purposes, almost identical.

Moreover, if we compare my DI curves with KEFs, they are pracitcally identical too, which leads me to believe there's just a difference in our mic calibrations for some reason.

View attachment 98865

Either way, you're getting a very good speaker, but my results show a more neutral tonality than KEFs. Until someone else with reliable measurements gets their hand on the LS50 Meta, this might remain a bit of a mystery.

Still, I'm not complaining. I'll update with more detailed directivity stuffs and quick impressions a bit later. Directivity measurements and quick sound impressions in post #6.

Great work! Given how close is your curve with that of Soundstage (that I understand it is made by the Canadian Research Council, using anechoic chambers), how do you emulate the anechoic conditions for your measurements?

Also, a side note: It is commendable that KEF publishes true figures, no matter if it is a bad marketing strategy: although in-room response extends much lower, their specs say 79 hz @-3db, which is exactly what every measurement is showing. AFAIK, many other brands overstate the bass response of their loudspeakers closer to in-room figures instead of the true anechoic measurement.

 

[Kachda]

Great work! Given how close is your curve with that of Soundstage (that I understand it is made by the Canadian Research Council, using anechoic chambers), how do you emulate the anechoic conditions for your measurements?

Also, a side note: It is commendable that KEF publishes true figures, no matter if it is a bad marketing strategy: although in-room response extends much lower, their specs say 79 hz @-3db, which is exactly what every measurement is showing. AFAIK, many other brands overstate the bass response of their loudspeakers closer to in-room figures instead of the true anechoic measurement.

Yeah. And also brave of them to state the -3db figure and not -6db that many seem to use. Canton speakers on a4less have specs at -10db I believe .

though as a company trying to sell products maybe kef should also state the typical in room response.

 

[thewas]

They give also the -6 dB point and typical room bass response:

FREQUENCY RANGE (-6DB)
47 Hz - 45 kHz

TYPICAL IN-ROOM BASS RESPONSE (-6DB)
26 Hz

Source: https://uk.kef.com/products/ls50-meta

 

[napilopez]

Great work! Given how close is your curve with that of Soundstage (that I understand it is made by the Canadian Research Council, using anechoic chambers), how do you emulate the anechoic conditions for your measurements?

Thanks! It's not at alll the difficult to do if you only want to measure the axial response and or a few horizontal angles.

I'm not sure if you're familiar, but for the midbass to treble the technique is called gating or time-windowing the impulse response. MiniDSP has a rudimentary guide on how to do this with REW here. A more elaborate generalized guide by the late Jeff Baby here.

The basics are simple. Essentially, all you're doing is cutting off the impulse response data right before the first big reflection 'hits' the microphone. That data is then ignored, and you're left with a snippet of time that is essentially anechoic.

The only problem is that the short amount of time before the first reflection hits is rarely enough to get much resolution for the lowest frequencies, but it provides a useful approximation for the upper portion. The only thing you really have to do, then, is get the speaker as far as possible from major reflecting surfaces AKA walls.

In my case, measuring at 1m and keeping the speaker about 1.6m from the nearest surface I can achieve a roughly 7ms gate. That gives me data down to about 143 Hz, which is also the resolution of the gated measurement(It won't be able to properly represent deviations narrower than 143 Hz intervals). This makes the technique less useful the lower in frequency you go, but above 1-2kHZ the results should have no reason to deviate from anechoic results.

The lowest frequencies can be captured using the nearfield summation technique in the Baby white paper, or another method such as a ground plane method. This low end response is then visually aligned with the gated response, and the data is spliced or blended together.

If all you wanted to do is take a quick measurement of information above, say 300 Hz, all you really have to do is get the speaker away from walls, take a measurement, and set the window before the first reflection hits.

The only hard part is setting up the speaker away 5-6 feet from walls if you have a small home and nowhere you can measure outdoors. Previously, I lived in an apartment with high ceilings and did all my measurements by placing speakers on a stand, on top of a plastic turntable, on top of my kitchen island. Now I basically do the same, but in my back yard. Didn't have to deal with bad weather and noise when it was indoors though!

 

[Maiky76]

@Maiky76 if you want to try to EQ/score with my measurements, here's the MDAT with the spin files. Removed the 1/24 octave smoothing in my posted measurements since I'm not sure if Amir's measurements use any smoothing when you process them.

Curious to see how it differs vs KEF's measurements, though as noted earlier it's likely I was not perfectly aligned on-axis so the on-axis diffraction doesn't show up. (I think since I'm measuring outdoors now the ground was not perfectly level so I might've been off by a degree or two in the vertical plane).

Could you export as txt files? my REW (Mac) is does not recognize the file and gives me this message: "file is not .mdat format"

 

[napilopez]

Could you export as txt files? my REW (Mac) is does not recognize the file and gives me this message: "file is not .mdat format"

Weird! Here you go.

 

[kschmit2]

some more limited measurements: https://www.stereo.de/hifi-test/produkt/kef-ls50-meta-1904

 

[Matias]

Stereophile review is online:
https://www.stereophile.com/content/kef-ls50-meta-loudspeaker-measurements

 

[ernestcarl]

Interesting. One of the excesses of the another coaxial (Sceptre S8) which I have is there’s too much upper mid — or at least a section of it has a peak. Obviously a completely different design! But one of the things reviews have complained about. So my expectations and point of comparison is quite biased against any rise there.

I took another closer examination at the S8's curves and I will have to take back what I said about the upper mids in this speaker being too excessive. Sorry! Yes, there is a bit of a small hump there, but... It only superficially looks like that due to a shallow but wide response dip located between 1 - 2.5kHz.

no speaker EQ correction (other than 2 PEQs between 16-17kHz applied to pair-match right speaker to left)

So I took another stab at my own crude nearfield measurement technique in an attempt to create PEQ filters that could fit within the limitations of a miniDSP 2x4HD. I know I've been spoiled by the near limitless number of filter banks available to JRiver users, so I really had to curb my tendency for over-doing things here.

'Tssst' / 'tsssk' / 'ssshhh' from instruments and the female voice, as well as other HF plosives I think are more the real culprit here when it comes to this speaker's excesses, I believe. This is because the curves between 5 to 11- or 12kHz do not attenuate as much. For example in trace 1: 45 deg Vertical there's quite a prominent peak. And notice how the curves closely bunch up together between 7-9kHz. I mean, it's nice that we don't lose as much energy off-axis, but this can over-time prove itself to be too much.

My speaker EQ correction cuts off the some of the highs enough that someone else might prefer a shelving boost.

However, caveat emptor: one really needs to make sure their speakers are "pair-matched" and fixed for any channel-dependent deficiencies prior to speaker EQ.

I find that corrections made at a distance (1-2m) isn't nearly as effective as matching in the very nearfield. This is because nearfield measurements will magnify the differences between the speakers. You only need the on-axis curve and perform an A/B trace arithmetic to quickly auto generate PEQs to match one channel to the other. My left speaker has a larger dip between 1-2kHz, while my right speaker has a narrower dip between 16-17kHz. Once manually pair-matched, both frequency responses now closely track each other at any given distance.

Just to reiterate, my technique is quite crude and done indoors only with a UMIK-1, tripod, and a small midori multiruler:

nothing fancy here folks!


View attachment 99662

Curves separated and smoothed for visual clarity
View attachment 99663
I'm only trying to see the 'big picture' for my general speaker correction.

If you think the curves for this horn type coax are quite uneven, you are indeed correct.

Here's a comparison between the S8s horn and the LSR305's (superior) waveguide:

*drooping as surrounds' location is behind.

HOWEVER, what is surprising is how much I actually often prefer what I hear from my coaxes vs other speakers. The LSR305s HF sound unnatural in timbre. The KH120s are too dark at midfield distance. The Mackie MR524, and another 8-inch pair (can't remember the discontinued model) are too harsh... could be the dynamic-range, vertical directivity, power-response etc.

But I'm now just talking about a different speaker (while both are coaxes) from the OP spinorama review about the LS50 meta -- yeah, I'll have to shut-up and stop there. LOL

 

[ethanhallbeyer]

One subjective review I read said the Meta was an improvement, but not worth an upgrade from the originals. I bet with EQ they would be indistinguishable or nearly so. Too bad the piano black finish is gone. When I added subs my LS50's really went from good to great.

If you have LS50's or any of several other good small speakers, EQ and get subs.

which subs? rel t7i? the ls50 black edition finish was a matte textured finish. more rugged and not a fingerprint magnet.

 

[Maiky76]

Weird! Here you go.

Here are the scores and some EQ:

Score no EQ: 6.00
With Sub: 8.14

Spinorama with no EQ:

• Pretty good
• Just a trough and a mild peak

I have generated two EQs. The APO config files are attached.

• The first one, labelled, LW is targeted at making the LW flat
• The second, labelled Score, starts with the first one and adds the score as an optimization variable.
• The EQ score ends up with a high shelf type of response which makes the ON flatter? Needs listening test...

Score EQ LW: 6.30
with sub: 8.44

Score EQ Score: 6.58
with sub: 8.72

Kef LS50 Meta APO EQ LW 96000Hz
December172020-110015

Preamp: -2.2 dB

Filter 1: ON HPQ Fc 48.8 Hz Gain 0 dB Q 1.15
Filter 2: ON PK Fc 373.5 Hz Gain -0.61 dB Q 1.77
Filter 3: ON PK Fc 1865 Hz Gain 1.11 dB Q 3.23
Filter 4: ON PK Fc 2570 Hz Gain 1.9 dB Q 1.73
Filter 5: ON PK Fc 4350 Hz Gain -0.61 dB Q 5
Filter 6: ON PK Fc 16700 Hz Gain 1.11 dB Q 2.08

Kef LS50 Meta APO EQ Score 96000Hz
December172020-105832

Preamp: -2.2 dB

Filter 1: ON HPQ Fc 49 Hz Gain 0 dB Q 1.16
Filter 2: ON PK Fc 466.5 Hz Gain -0.88 dB Q 4.61
Filter 3: ON PK Fc 1828 Hz Gain 0.81 dB Q 4.49
Filter 4: ON PK Fc 2589 Hz Gain 1.51 dB Q 2.19
Filter 5: ON PK Fc 4193 Hz Gain -1.22 dB Q 7.62
Filter 6: ON PK Fc 7201 Hz Gain -1.11 dB Q 1.05

Spinorama EQ LW

Spinorama EQ Score

Zoom PIR-LW-ON

Regression - Tonal

Radar no EQ vs EQ score
Some minor improvements

 

[napilopez]

Here are the scores and some EQ:

Score no EQ: 6.00
With Sub: 8.14

Spinorama with no EQ:

• Pretty good
• Just a trough and a mild peak

View attachment 99684
I have generated two EQs. The APO config files are attached.

• The first one, labelled, LW is targeted at making the LW flat
• The second, labelled Score, starts with the first one and adds the score as an optimization variable.
• The EQ score ends up with a high shelf type of response which makes the ON flatter? Needs listening test...

View attachment 99682
Score EQ LW: 6.30
with sub: 8.44

Score EQ Score: 6.58
with sub: 8.72

Kef LS50 Meta APO EQ LW 96000Hz
December172020-110015

Preamp: -2.2 dB

Filter 1: ON HPQ Fc 48.8 Hz Gain 0 dB Q 1.15
Filter 2: ON PK Fc 373.5 Hz Gain -0.61 dB Q 1.77
Filter 3: ON PK Fc 1865 Hz Gain 1.11 dB Q 3.23
Filter 4: ON PK Fc 2570 Hz Gain 1.9 dB Q 1.73
Filter 5: ON PK Fc 4350 Hz Gain -0.61 dB Q 5
Filter 6: ON PK Fc 16700 Hz Gain 1.11 dB Q 2.08

Kef LS50 Meta APO EQ Score 96000Hz
December172020-105832

Preamp: -2.2 dB

Filter 1: ON HPQ Fc 49 Hz Gain 0 dB Q 1.16
Filter 2: ON PK Fc 466.5 Hz Gain -0.88 dB Q 4.61
Filter 3: ON PK Fc 1828 Hz Gain 0.81 dB Q 4.49
Filter 4: ON PK Fc 2589 Hz Gain 1.51 dB Q 2.19
Filter 5: ON PK Fc 4193 Hz Gain -1.22 dB Q 7.62
Filter 6: ON PK Fc 7201 Hz Gain -1.11 dB Q 1.05

Spinorama EQ LW
View attachment 99683

Spinorama EQ Score
View attachment 99687

Zoom PIR-LW-ON
View attachment 99688

Regression - Tonal
View attachment 99686

Radar no EQ vs EQ score
Some minor improvements
View attachment 99685

Thank you! I got a similar score with my modified version of MZKM's calc, but didn't want to post as it's been a while since I checked it's accuracy. Curious to see what happens if/when amir gets his hands on one of those, for consistency's sake. I know the messiness from 400-700 Hz is partly from my setup, but the dip at 850ish Hz appears to be from the speaker itself.

Not that I put that much weight on the scores, but I think it's safe to say these are a significant improvement over the original LS50s in some key ways, or at least the ones amir measured. Really, the listening window on mine is just as good as some of the studio monitors, and it appears to be free of major resonances. The curves not only trend towards flat, but are also quite smooth.

An 8.14 sub score basically puts it on par with the R3 (8.16), although it's worth noting that in my own measurements of the R3, the speaker appeared less smooth than the one amir measured and it "only" got a 7.8.

I liked the original LS50, but imo the Meta finally lives up to its reputation. Especially in a few years, when the price comes down.

 

[Juhazi]

https://www.stereophile.com/content/kef-ls50-meta-loudspeaker-measurements

(Atkinson's method is such that there should be elevated response below 200Hz)

 

[thewas]

Also the vertical directivity

and vibration spectral-decay plot

are worth showing imho.

 

[thewas]

Sweet, you're the one I was hoping would grab them lol. I'm really curious about 2 things when you have a chance to measure and listen some more. The first is there seems to be a difference in the bass response between the 2, I think some MMM measurements of both might clear that up. Also, once you have a chance to EQ both to the same target are they basically identical or are there any improvements in the meta?

For people without EQ, the meta does seem like a clear improvement for sure.

To answer the last question first, it definitely is, below are the measurements you and I wanted to see, all from my normal loudspeaker placement so either quite early gated or MMM.

First the gated on-axis which matches the one of @napilopez well:

A MMM based listening window approximation in 50 cm from the loudspeaker which seems to be a bit more decreasing at the highs which could be also due to too large angles at my measurement:

A MMM based sound power approximation (rotating the mic the loudspeaker at 2x 360° circles), again very good match, just a similar deviation in the highs:

As to be expected than my listeners MMM matches the PIR also well except the high frequencies:

A part of the difference can be that he used the 90° orientation and calibration of his mic while I used the 0° one, although my LP MMM usually matches the ASR based one in the highs quite well.

Now to the question of the frequency response difference in the bass of the Meta vs the old Anniversary one which I also own:

Low bass seems quite identical, differences are mainly above where the Meta is much smoother, to visualise them better I computed the differences of both L, R and L+R MMM which all match well:

From those I computed the average and created manually minimal EQ for people wanting to test Meta tonality on their Anniversary ones, that is minimising their LP MMM differences:

Filter  1: ON  PK       Fc   55.30 Hz  Gain  -1.00 dB  Q  3.000
Filter  2: ON  PK       Fc   140.0 Hz  Gain  -1.00 dB  Q  2.000
Filter  3: ON  PK       Fc   450.0 Hz  Gain   1.50 dB  Q  1.400
Filter  4: ON  PK       Fc    1500 Hz  Gain   2.00 dB  Q  3.000
Filter  5: ON  PK       Fc    2600 Hz  Gain  -3.00 dB  Q  1.700

Please keep in mind that this EQ will only approximate a Meta but not make it identical to it, as both generations have slightly different crossover frequencies, slopes and thus also directivity and also different distortion behaviour but like nowdays several posted EQ settings of the Anniversary, make it sound more neutral.

Talking about distortions, the KEF paper again doesn't lie being reduced at the Meta noticeably both in the bass, mids and treble, both measured at the same, high SPL level, in absolute THD:

In relative THD:

And their components, Anniversary:

Meta, above 100 Hz the dominating component is the 2nd (3rd and above are reduced compared to the Anniversary) which is less audible and less disturbing:

Detailed listening impressions and differences I would like to post in a later point as I want to listen more to them.

 

[mtmpenn]

Great review and interesting measurements!

Selfish question, since I am strongly considering purchasing a pair of concentric bookshelf speakers:

Does anybody have thoughts about how the new meta's might stack up against the Technics SB-C700 (which are discontinued but can still be found)?

Stereophile measurements available here.

The trade-offs that I can see are that the Meta's have a flatter on-axis response and a smoother off-axis response, at least in the high frequencies, but horizontal dispersion is wider with the Technics. Meta's are also less sensitive (84.5 vs 87.5) and have a smaller mid woofer, although maybe that is less of an issue with a subwoofer.

 

[Ron Texas]

Measurements from Stereophile. Meta in red, original in blue. I can't help but think with the 2.5 khz bump equalized out they would be hard to tell apart.