Kef blade 2 meta frequency response

[617]

To think of equalizing as "messing" and/or "inferior" is an opinion I wish we left behind in the 90s.

I have a custom made equalizer made of capacitors and inductors haphazardly soldered together inside my speaker, works great.

 

[thewas]

On a speaker with good directivity, on applying an equalizer, we are essentially messing with the signal which goes to the speaker, meaning the direct sound has already become “inferior”(can’t we just call it altered?)

If the loudspeaker has smooth directivity but not linear direct sound, equalising it will make both the direct and reflected sound more neutral. A loudspeaker has its own transfer function so a linear input signal doesn't mean its output will be automatically linear. It's all also written in the link I recommended you above.

 

[Sancus]

Here's the Reference 5, similar slope there too:

The Blade Two is actually still darker than the Reference 5 Meta.

(spoilering images to reduce space)

Spoiler

Granted, I do think we're splitting hairs here a little. We are not talking about abandoning Harman/Toole's research entirely and making something completely random. This is a couple of dB of tilt one way or the other.

I'm not fully convinced the PIR encapsulates tonality in a good way either, so it's probably too inaccurate of a measure to use for this type of comparison with very different speakers.

This is based on the assumption that completely flat on-axis anechoic response is best in a domestic situation. For some reason everyone assumes this is so. But we don't really have a reliable reference or evidence to that fact.

Wait.... we don't? What lol. I think that's taking it a bit far.

Starting with my earliest listening tests in the late 1960s, through a prolific research period in the 1980s (see Figure 5.2), up to the present (see Chapter 12), it has been a monotonous truth that in double-blind listening tests, the highest rated loudspeakers had the flattest, smoothest on-axis and listening-window frequency responses. Listeners liked neutral, uncolored, direct sound. Beyond that, loudspeakers that exhibited similarly good behavior off-axis achieved even higher scores—reflected sounds would then have similar timbral signatures. These findings have remained valid in many different rooms over the years. These were small rooms: stereo listening rooms, home theaters and recording control rooms. As has been discussed earlier, listeners have a significant ability to separate the sound of the source from the sound of the room (Figure 5.16). The two sets of information appear to be perceptually streamed, with the result that loudspeakers retain their relative sound quality ratings in different rooms (Section 7.6.2).

That said, I'll grant that most of the research has been done with speakers that were considerably worse than today's speakers. So it's possible we'd get more interesting results in a double blind study of speakers that were all generally well designed, free of resonances, etc, but had different directivity and on-axis.

And of course, it's also established that *individuals* prefer different amounts of treble and bass.

I do find it questionable to buy a speaker based on the "voicing" and not EQ at all. Like, how would a normal consumer ever audition enough speakers to "find the speaker with the right tonality for their recordings". It seems practically impossible.

 

[sigbergaudio]

@Sancus

Blade 2 vs Reference 5 Meta: The main point is that they're clearly not perfectly neutral anechoically.

With regards to Toole. I suspect there's a strong correlation between speakers that are flat as well as being smooth. So which of those parameters are most important? Toole emphasized that what the listeners favored was the absence of resonances. We could then further speculate that it may have been the fact that the response was smooth and without major resonances and deviations that was the most important thing, potentially less so than the slope (or lack of slope).

If we look at our SBS.1 speaker (using them mainly because I know them intimately, not in an attempt to plug them) at the intended listening axis, and following the intentional slope, they're within +/-2dB, and that's with far less smoothing than is often used when coming up with the accuracy numbers. Let's say we had a speaker with more significant resonances, but that were completely flat - and let's say it used drivers with the same inherent quality. So the differences are the slope and the lack of smoothness. Which one would most people prefer? I don't really have the answer, but I suspect deviations (lack of smoothness) is more troubling than the slope. And I of course also (as I've argued) think that the slope is actually beneficial.

With regards to finding the "right" voicing: To avoid EQ isn't a goal in itself in my opinion. Especially in the bass, each room will be different - and to some extent your musical preferences will affect your tonality preference as well. So to have EQ available will be positive in any case.

Beyond that I think you will find relatively few speakers that are voiced with an upwards tilt towards the treble. And I think you will find even fewer people that find them to sound good. So then you're left with speakers that are either neutral, with a slope emphasizing the bass or finally (relatively common) with the smiley-graph (emphasis on both bass and treble). In the long term I would suspect most people will prefer speakers that are neutral, or with slight emphasis towards the bass.

 

[Sancus]

With regards to finding the "right" voicing: To avoid EQ isn't a goal in itself in my opinion. Especially in the bass, each room will be different - and to some extent your musical preferences will affect your tonality preference as well. So to have EQ available will be positive in any case.

Beyond that I think you will find relatively few speakers that are voiced with an upwards tilt towards the treble. And I think you will find even fewer people that find them to sound good. So then you're left with speakers that are either neutral, with a slope emphasizing the bass or (relatively common) you with the smiley-graph (emphasis on both bass and treble). In the long term I would suspect most people will prefer speakers that are neutral, or with slight emphasis towards the bass.

Yes, I think that's fair. For active speakers, honestly, you can make the tilt simple enough to change, and it would probably be a good form of simplified tone control. It would be an interesting way to gather data as well as long as you had enough products out in the wild. I know that Buchardt briefly talked about doing something like this.

I think it would be pretty interesting if someone gathered a few thousand data points this way.

 

[sigbergaudio]

Yes, I think that's fair. For active speakers, honestly, you can make the tilt simple enough to change, and it would probably be a good form of simplified tone control. It would be an interesting way to gather data as well as long as you had enough products out in the wild. I know that Buchardt briefly talked about doing something like this.

I think it would be pretty interesting if someone gathered a few thousand data points this way.

Yep, the plan for our next active is one "reference" setting along the lines of what we've been discussing here, one that is perfectly flat, and one with even more downwards slope for reflective rooms.

I'm afraid it will take a number of years to collect a few thousand data points though.

 

[617]

Yep, the plan for our next active is one "reference" setting along the lines of what we've been discussing here, one that is perfectly flat, and one with even more downwards slope for reflective rooms.

I'm afraid it will take a number of years to collect a few thousand data points though.

You gotta deploy a few thousand speakers, let the users select an EQ, and then collect that data somehow.

 

[Sancus]

I think a blind tuning algorithm would be pretty awesome, but would probably require a web UI that is capable of playback. Which is maybe too complicated for most people.

 

[sigbergaudio]

You gotta deploy a few thousand speakers, let the users select an EQ, and then collect that data somehow.

I'd rather sell than deploy if possible.

 

[jackocleebrown]

Hi, thanks for your interest in KEF loudspeakers. @sigbergaudio, thanks for the tag! I'm sure you all know our approach to loudspeaker design very well. In general you could say that we follow the same approach endorsed by Toole. We try and design products that have smooth and resonance-free on-axis and off-axis responses. Uni-Q was introduced in the late 80s precisely for this purpose. However, it's certainly not the case that you can design a loudspeaker based only on objective data or the Harman preference score. We find that the final balancing of the loudspeaker still needs to include listening to get the best possible result.

No loudspeaker or driver is perfect: look closely at any response at any price and you can find imperfections. With a given set of drivers and cabinet you can come up with many different crossovers that will perform similarly well on objective data. It's not always obvious from objective data which one will sound best. Listening is really critical in the final stage to find the best compromise. We have found great success during balancing when strongly focusing on smoothness, and particularly off-axis smoothness in all directions. It's also our experience that second order effects, not included in the Spinorama measurements, can have a small impact on the perceived response. For example, distortion or power compression. Finally, the overall perceived timbral balance (especially "brightness") can vary a lot depending on room size, position, acoustics and personal preference.

Ultimately our target in the balancing stage is to design loudspeakers that work well for real-world users in typical rooms. It's rather challenging to do this because there's such a wide range of use cases. We've designed our listening room to be a compromise between a controlled environment and a domestic room. Final balance is done in there. We reference all products back to our flagships during balancing. We do also take account of product size, price and likely room size and position they'll be used in. We also try and routinely get info on how our speaker sound the real world rooms and feed this back into the listening and tuning process.

The slightly downward sloped on-axis response is a result of this approach. On-axis data has a historical emphasis, and it's hard to escape from this, but it's our experience that other metrics are better at predicting the perceived in-room response. Keep in mind that in many typical domestic listening situations more reflected sound reaches the listener than direct.

 

[sigbergaudio]

@jackocleebrown Thank you for taking the time to respond! Your answer is very similar to to my own sentiments, so I don't have much to add - other than the fact that I'm happy to see you confirm them.

I also share the same challenges of trying to balance the sound, it will always be somewhat of a compromise as the sound differs so much depending on the room. And as you mention, to some degree personal preference vary as well - perhaps guided by the type of music the person mostly listens to. Also interesting to hear that your listening room is a compromise between a controlled environment and a domestic room, this is also similar to our approach. And of course try to get prototype units out into actual domestic rooms to verify.

 

[Snickers-is]

@jackocleebrown Thanks for your post regarding tuning of loudspeakers. I have discussed this widely with Mr Sigberg and others.

After studying third party measurements from several of your products, a slight pattern appears, which I wonder if you would like to confirm. The larger models, like R11, Reference 5 and Blade 2 meta, seems to have a bit more tilt than for example the LS50W. Is this simply a result of doing listening test for the larger models generally in larger rooms and on longer listening distances?

It is great to see that KEF has such a scientific approach. Many manufacturers claim to have that, but in your case, it seems to be a bit more than just a claim. I especially love the small tricks with the pole pieces and the bass ports used in the R-series.

 

[dogmamann]

@jackocleebrown Thanks for your post regarding tuning of loudspeakers. I have discussed this widely with Mr Sigberg and others.

After studying third party measurements from several of your products, a slight pattern appears, which I wonder if you would like to confirm. The larger models, like R11, Reference 5 and Blade 2 meta, seems to have a bit more tilt than for example the LS50W. Is this simply a result of doing listening test for the larger models generally in larger rooms and on longer listening distances?

It is great to see that KEF has such a scientific approach. Many manufacturers claim to have that, but in your case, it seems to be a bit more than just a claim. I especially love the small tricks with the pole pieces and the bass ports used in the R-series.

End of the day, they do tune the final part based on subjective listening , as opposed to the popular theory of looking at the graph and deciding. When other manufacturers say that, people here get attack them back , which isn’t fair in my opinion. No manufacturer produces their speakers without their listening tests, by real people, and tweaking based on what is listened. Simple.

 

[jackocleebrown]

End of the day, they do tune the final part based on subjective listening , as opposed to the popular theory of looking at the graph and deciding. When other manufacturers say that, people here get attack them back , which isn’t fair in my opinion. No manufacturer produces their speakers without their listening tests, by real people, and tweaking based on what is listened. Simple.

If I may interject here, that's not what I wrote. Objective data is not discarded at the listening room door. Nevertheless, feel free to judge our product however you see fit: objective or subjective, we aim to do a good job.

 

[Snickers-is]

It is interesting how some people seem to believe that objective data and listening somehow tells two entirely different stories. Not to point my finger at anyone, I just find it interesting.

 

[dogmamann]

If I may interject here, that's not what I wrote. Objective data is not discarded at the listening room door. Nevertheless, feel free to judge our product however you see fit: objective or subjective, we aim to do a good job.

What I understood is your company uses objective data to develop the speakers but your final product still is tuned by ears on the top of it.

 

[dogmamann]

It is interesting how some people seem to believe that objective data and listening somehow tells two entirely different stories. Not to point my finger at anyone, I just find it interesting.

I dont think that’s the case, it’s possible to understand the sound profile of any speakers by the data. But to do that, you need some real world experience of real speakers, some measuring bright, some dull, some with wide radiation, some with point sources and so on.

Owning just one type of speakers and assuming about everything else can be wildly wrong.

 

[Koeitje]

Kefs cannot produce the cymbals exactly like how they are in the recording unless you eq it up. Agreed it’s pleasant, but then again it’s not how it is in the recording.

Most folks who end up buying these won’t care about eqing. As per this logic, there are 1000s of “pleasant” sounding speakers in this world each of them thinking this is how I make my speakers fatigue free.

Good thing is Kef can be eqed, but most others cannot be.

How do you know this is the case? Do you record your own cymbals?

 

[fpitas]

I have a custom made equalizer made of capacitors and inductors haphazardly soldered together inside my speaker, works great.

The dirty little secret of audiophilia: your speaker's passive crossover includes..(drum roll)...EQ!!!!

 

[jackocleebrown]

What I understood is your company uses objective data to develop the speakers but your final product still is tuned by ears on the top of it.

Yes, you understood correctly. I took your previous post to mean that the final part is done by listening only, which is not the case at KEF. Listening is also used during development, we typically make quite a few early prototypes too. Personally I think it would be rather remiss of us not to listen whatsoever. Even if you take a view that objective data is enough to fully understand how a speaker sounds, this comes from experience and we have to build and maintain that across the team. Plus, our whole team are audio and music fans - listening products we've worked on is very motivating, and trying to understand the perspective of our customers is very important too.