KEF Q350 Speaker Review

[MZKM]

Science doesn’t lie.
Observations, even subjective, warrant additional investigation.

The KEF Q350 is better than the XPL 90 In many regards but the XPL90 is smoother from 100 Hz to 1500Hz or so.

Might the preference score underestimate the value of smoothness in this region compared to somewhere else (say above 16 kHz)

Keep in mind that the formula does not weight frequencies differently, this speaker has a dip ~4kHz, which is in the presence region so it's more audible than the same dip but at 10kHz. This is likely one reason the accuracy is not much higher.

Also, Amir's listening room is open to one side and is a loft if I recall, that may be one reason he really likes wide dispersion (though preference for wide dispersion seems to have been documented in studies).

 

[tecnogadget]

Those 2 resonances given aside, the overall On-Axis still looks good and it's really not that different from the Elacs DBR-62 (which looks like a direct competitor in both price/performance) but the Directivity Index looks much better here...

While reading the review I though amirm would like them after watching the In-Room given his preference for bass boost but this has not been the case. It makes sense since the boost is centered at 200Hz. This review may strengthen the correlation between measurements and preference in the bass range, with the notion of a 200Hz boost messes things out giving a muddy/rude sound, but if the same boost where at 100Hz it would be perceived as richer/authoritative (aka Revel M16).

Has somebody noticed the excellent low distortion of the tweeter past 1kHz ?? Superbly clean!

 

[tecnogadget]

May not be the driver - look at the "glitches" in the impedance plot.

Sure, I've seen the graph but that still does not troubleshoot me enough to understand what causes them. It's out of the scope for me but maybe somebody has a hint of the answer.

 

[beefkabob]

Not Harman enough?

You're supposed to by the sycophant! Cut it out, man. Gonna ruin ASR's rep.

 

[Ron Texas]

FWIW in the US LS50's are $900 right now. Would you buy Q350's for $700 over the LS50's because they have a higher score and cost less?

 

[StevenEleven]

Those 2 resonances given aside, the overall On-Axis still looks good and it's really not that different from the Elacs DBR-62 (which looks like a direct competitor in both price/performance) but the Directivity Index looks much better here...

While reading the review I though amirm would like them after watching the In-Room given his preference for bass boost but this has not been the case. It makes sense since the boost is centered at 200Hz. This review may strengthen the correlation between measurements and preference in the bass range, with the notion of a 200Hz boost messes things out giving a muddy/rude sound, but if the same boost where at 100Hz it would be perceived as richer/authoritative (aka Revel M16).

This is what I think but you wrote it so now I don’t have worry about it. I would extend this to my own personal preferences, I would way way prefer a smooth bass boost centered around 100 hertz (or even a shade lower), rather than having it centered more on 200 hz and rolling off in that scooped way beginning just above 100 hertz. The LS50 has the same kind of 200 hz-ish bass peak as these 350s, it’s obviously an intentional choice, that‘s a very big deal to me, that’s the fundamental difference between the DBR62s and the two KEF speakers for my [imagined?] personal preferences. Maybe @Soniclife ’s point about room gain has some merit, for some rooms, I dunno. If someone else prefers the KEF smooth bass peak centered around 200 hz, I can see that, it’s okay, it’s just not me. Then maybe they should get these.

 

[GXAlan]

Keep in mind that the formula does not weight frequencies differently, as such an issue at 3kHz and an issue at 10kHz are treated roughly the same even though the former is more audible as it's in the presence region. This is likely one reason the accuracy is not much higher.

Hmm. we already have a subwoofer preference calculator. The Burchardt that Amir didn't like also had a glitch in this region. The KEF has a glitch in the region.

What about midrange purity score?

The question would involve figuring out what range seems to correlate with subjective experiences

250 to 4000Hz?
100 to 3000Hz? (Human voice, fundamentals)
100 to 1500Hz? (Where the XPL-90 seemed to shine and was Amir's "this measures poorly but it sounds good")

One fun fact that I encountered while looking for a good FR to measure is this paper:
https://asa.scitation.org/doi/10.1121/1.2761883
which reports that some people can hear to 28 kHz provided that it's >100 dB.

 

[dc655321]

I've seen the graph but that still does not troubleshoot me enough to understand what causes them.

Enclosure panel resonance at ~700Hz?
Standing wave seems unlikely at those frequencies...

The ~1.2kHz resonance does not appear in the impedance plot and does not disappear off-axis, so not a diffraction effect.
Dunno...

 

[amirm]

I think at this rate, in a few months, I will be the most studied audio listener in the world!

Seriously, there is not any listener out there become compared to such specific objective measurements. I think the guesses so far are pretty close to mine.

I am starting to think that the range up to 1 to 3 kHz is far more important than the rest. It is incredible how much difference it makes to pull a small peak down a few dB. And directivity definitely plays a part which until this review, I had not realized as much.

Of course, subjective listening tests are error prone but maybe with enough data points, some of that can be averaged out.

Longer term, I am keeping a collection of the budget speakers which I hope can be blind tested with our audiophile society here.

 

[Guermantes]

@amirm I'm wondering if your subjective impressions of these KEF coaxials would improve when listening to a phantom image from a stereo pair? Especially, it would be interesting to consider how the directivity of the two speakers interact in producing a "soundstage". I say this because this is the practical use-case for most of these designs. Just going down the subjective rabbit hole here . . .

 

[amirm]

@amirm I'm wondering if your subjective impressions of these KEF coaxials would improve when listening to a phantom image from a stereo pair?

Yeh, I had that thought this morning as I was listening. I shipped these back already so will have to test this in a future review.

 

[MZKM]

I am starting to think that the range up to 1 to 3 kHz is far more important than the rest. It is incredible how much difference it makes to pull a small peak down a few dB

It’s not called the presence region for nothing (though the exact bounds aren’t defined).

 

[js_k0914]

Here is our distortion and in-room response with or without plug:

This is a little weird. How close you measured this graph? using or not port plug will not be the shape of those graph. If the measuring point were far enough, bass FR grap shoud show very difference. And or if the point were very close to unit, the port resonance point should show and also it vary by port. The all three samples just look like response of sealed speaker. So I also have a little suspicion of distortion graph using that in room response.

I'm not a native, Even if it's awkward, understand it.

 

[aarons915]

These are pretty good measurements, I personally preferred the Q150 when comparing to these for similar reasons, the Q350 seemed to have too much bass and sounded too dark because of it. Having these measurements they would be very easy to EQ that out along with the resonances and probably have a very good speaker for the money.

This is the first time I've heard anyone complain about the point-source nature of a coaxial being a bad thing though, it's always sounded more natural to me than typical speakers with vertically aligned drivers.

 

[BYRTT]

Some overlays of technical curves for KEF Q350 based on Amir's shared spindata and curves:

Spin plus PIR and audioband related system impedance:

 

[tuga]

According to Stereophile, the croossover takes place at 2.5kHz with first-order slopes which may explain the peak at 1.2kHz.

https://www.stereophile.com/content/kef-q350-loudspeaker-specifications

The port measurement may point to the likely cause of the dip at 900Hz:

https://www.stereophile.com/content/kef-q350-loudspeaker-measurements

 

[Blumlein 88]

You're supposed to by the sycophant! Cut it out, man. Gonna ruin ASR's rep.

Are you claiming a lack of Harmany?

 

[bobbooo]

Of course, subjective listening tests are error prone but maybe with enough data points, some of that can be averaged out.

I would have thought a large proportion of the error in sighted subjective listening tests would be 'systematic', rather than random, so I'm not sure much of it could be averaged out. In the review you mentioned you listen to the speaker in mono "to the left". I don't believe this is how the listening tests were done by Olive and his team that were used in formulating the preference rating formula (I think each single speaker was placed directly in front of the listener). This could be one of those systematic errors. It might turn out that factors such as narrow directivity are more noticeable when a single speaker is placed to one side rather than directly in front of the listener.

Longer term, I am keeping a collection of the budget speakers which I hope can be blind tested with our audiophile society here.

Now those kind of blind tests with multiple listeners could be of real scientific value, and would be great to correlate with the calculated preference ratings.

 

[amirm]

In the review you mentioned you listen to the speaker in mono "to the left". I don't believe this is how the listening tests were done by Olive and his team that were used in formulating the preference rating formula (I think the speakers were placed directly in front of the listener).

They have different rooms. This is the smaller one that I have been to a couple of times:

I think as long as speakers are always tested in the same position, then the test is valid. Otherwise, how useful would their score be if it did not apply to a typical home use?

 

[amirm]

I would have thought a large proportion of the error in sighted subjective listening tests would be 'systematic', rather than random, so I'm not sure much of it could be averaged out.

Well, we are going to have the data and then we will know. For now, I feel some trends are reaching to the surface in a much more controlled way than any reviewer out there. Same single speaker, in the same location with objective test data.