KEF R3 OR BOWER & WILKINS S706 s2 ???

[lateralous]

Regarding the topic of engineering capabilities of large companies such as B&W, I think it would be fascinating to understand the actual performance metrics their engineers are trying to meet. On top of that, what are their constraints; are they given a pretty box marketing has ensured will sell through consumer trials? I would imagine they are sitting on a gold-mine of data on listener preference we all might learn from, even if it has been skewed towards generating a speaker that will sell, rather than perform at our best understanding of "state-of-the-art".

 

[Steve Dallas]

About the BBC dip...

SETUP

To investigate the BBC dip, I first searched for a definition. I found reliable sources which define it as a mild dip of about -3dB beginning at 1000Hz and ending between 3000 and 4000Hz. Using that information, I dialed in such a dip in Dirac Live 3, which I use daily in my home office, where my KEF R3s delight me daily. I normally correct to to 1000Hz, so I had to draw a custom full range curve, then modify that with an approximation of the BBC dip as previously defined. This is how that looks:

The Dirac Live Processor allows you to save up to 8 presets and switch between them instantly:

Then I measured the results, which came out pretty close. The dip extends a bit beyond 4000Hz, but the delta is < 1dB, so I decided to go with it.

LISTENING TESTS

[Signal chain: PC -> USB -> Schiit Modius -> Peachtree Nova 300 -> KEF R3]

In listening to unfiltered pink noise while taking those MMM measurements, I could very easily hear the dip and therefore though listening tests would be a cakewalk. I was wrong.

My initial test involved listening to an Amazon Music HD playlist I made years ago featuring music from the Guardians of the Galaxy movies. A lot of classic rock has what I would call edgy glare in portions of the program when heard on a capable system. You know, where it sounded fine to you in the car for 20 years, and then you discovered it was not as well-produced as you thought. Several songs in that playlist have that sort of glare, and I imagined it might fall in the frequency band of the BBC dip and be helped by the broad reduction there.

As I listened to the playlist, I swapped presets every time I heard that edginess only to find it was not improved. In fact, under almost all circumstances, I heard no difference at all between flat full correction and BBC full correction. Hmmm...

The only differences I could be sure I heard were limited to drums and percussion and some piano passages. Why? Because only the 2nd and 3rd octave overtones of rock instruments and voices fall into the range covered by the dip, and they are -10dB or more down in SPL. Soprano voices and 6 string guitars end around 1000Hz, for example. There simply is not much program material in that region in rock and roll. A spectrum analyzer clearly shows this to be true. (Jethro Tull excepted, of course.)

Knowing this, I switched to Santana's Abraxas album. The difference there is much more palpable due to the amount of latin percussion in the music. I had a definite preference for the flat curve, as the BBC dip sounded artificially dull with instruments having inaccurate tonality.

I then listened to The London Symphony Orchestra Performs the Music of Sting. Here, the difference was often obvious, depending upon the mix of instruments playing. Again, I very much preferred the flat curve.

However, I can see a case being made for a -1 or -1.5dB in that region to reduce some of the forwardness of woods, strings, brass instruments in the latter album. I can also see how the BBC dip may be preferred by someone with hearing sensitivities within its range.

The BBC could use this dip to reduce a speaker cone resonance for all those years, because there is not much signal occupying that space in a typical TV broadcast comprised of voices and background music. Certainly not such that anyone would notice nor care.

Finally, to throw @preload a bone, I dialed in +1.5dB peak at 10KHz. This added stick presence to cymbals. It sounded artificial to me, like the midrange and treble part of a smile EQ was in play, or someone mashed the loudness button on the boombox.

The last point I will make is I prefer Dirac with correction limited to 1KHz over both of these new curves. This is why - the normal response of these speakers in this room and what I listen to daily:

For reference:

 

[test1223]

About the BBC dip...

SETUP

To investigate the BBC dip, I first searched for a definition. I found reliable sources which define it as a mild dip of about -3dB beginning at 1000Hz and ending between 3000 and 4000Hz. Using that information, I dialed in such a dip in Dirac Live 3, which I use daily in my home office, where my KEF R3s delight me daily. I normally correct to to 1000Hz, so I had to draw a custom full range curve, then modify that with an approximation of the BBC dip as previously defined. This is how that looks:

View attachment 251609

The Dirac Live Processor allows you to save up to 8 presets and switch between them instantly:

View attachment 251619


Then I measured the results, which came out pretty close. The dip extends a bit beyond 4000Hz, but the delta is < 1dB, so I decided to go with it.

View attachment 251614








View attachment 251615

LISTENING TESTS

[Signal chain: PC -> USB -> Schiit Modius -> Peachtree Nova 300 -> KEF R3]

In listening to unfiltered pink noise while taking those MMM measurements, I could very easily hear the dip and therefore though listening tests would be a cakewalk. I was wrong.

My initial test involved listening to an Amazon Music HD playlist I made years ago featuring music from the Guardians of the Galaxy movies. A lot of classic rock has what I would call edgy glare in portions of the program when heard on a capable system. You know, where it sounded fine to you in the car for 20 years, and then you discovered it was not as well-produced as you thought. Several songs in that playlist have that sort of glare, and I imagined it might fall in the frequency band of the BBC dip and be helped by the broad reduction there.

As I listened to the playlist, I swapped presets every time I heard that edginess only to find it was not improved. In fact, under almost all circumstances, I heard no difference at all between flat full correction and BBC full correction. Hmmm...

The only differences I could be sure I heard were limited to drums and percussion and some piano passages. Why? Because only the 2nd and 3rd octave overtones of rock instruments and voices fall into the range covered by the dip, and they are -10dB or more down in SPL. Soprano voices and 6 string guitars end around 1000Hz, for example. There simply is not much program material in that region in rock and roll. A spectrum analyzer clearly shows this to be true. (Jethro Tull excepted, of course.)

Knowing this, I switched to Santana's Abraxas album. The difference there is much more palpable due to the amount of latin percussion in the music. I had a definite preference for the flat curve, as the BBC dip sounded artificially dull with instruments having inaccurate tonality.

I then listened to The London Symphony Orchestra Performs the Music of Sting. Here, the difference was often obvious, depending upon the mix of instruments playing. Again, I very much preferred the flat curve.

However, I can see a case being made for a -1 or -1.5dB in that region to reduce some of the forwardness of woods, strings, brass instruments in the latter album. I can also see how the BBC dip may be preferred by someone with hearing sensitivities within its range.

The BBC could use this dip to reduce a speaker cone resonance for all those years, because there is not much signal occupying that space in a typical TV broadcast comprised of voices and background music. Certainly not such that anyone would notice nor care.

Finally, to throw @preload a bone, I dialed in +1.5dB peak at 10KHz. This added stick presence to cymbals. It sounded artificial to me, like the midrange and treble part of a smile EQ was in play, or someone mashed the loudness button on the boombox.

The last point I will make is I prefer Dirac with correction limited to 1KHz over both of these new curves. This is why - the normal response of these speakers in this room and what I listen to daily:

View attachment 251620



View attachment 251621


For reference:

View attachment 251623

Nice experiment but the dip isn't right.The q-factor of the biquad filter is to low. At 1.9kHz and 4kHz there should be almost no attenuation. The max attenuation should be about 2.5dB at about 2.85kHz. You can try to shift it a bit up and down in frequency, since it is related to your head related transfer functions. Also the shape of the dip should be more symmetrical. If you try more attenuation you have to have a speaker with a wider high frequency beam or higher listening distance. In general the B&W like frequency response doesn't work that good in the near field, where the higher frequencies are to bright and the tonality isn't formed that much with later room reflections.

 

[TrevC]

Finally, to throw @preload a bone, I dialed in +1.5dB peak at 10KHz. This added stick presence to cymbals. It sounded artificial to me, like the midrange and treble part of a smile EQ was in play, or someone mashed the loudness button on the boombox.

A golden eared audiophile at work!

 

[Steve Dallas]

Nice experiment but the dip isn't right.The q-factor of the biquad filter is to low. At 1.9kHz and 4kHz there should be almost no attenuation. The max attenuation should be about 2.5dB at about 2.85kHz. You can try to shift it a bit up and down in frequency, since it is related to your head related transfer functions. Also the shape of the dip should be more symmetrical. If you try more attenuation you have to have a speaker with a wider high frequency beam or higher listening distance. In general the B&W like frequency response doesn't work that good in the near field, where the higher frequencies are to bright and the tonality isn't formed that much with later room reflections.

On what are you basing this? The commonly accepted definition of the BBC dip (please link) or some version of the B&W dip? You realize the horizontal scaling of the graphs is not symmetrical, yes? In any case, the changes you suggest would not change the outcome, as the differences are at 0.5dB or less.

Also, I can draw one target curve to be shared between two speakers. Precision is not guaranteed. I could get a closer match with more effort, but this experiment does not turn out to be worth that effort.

BTW, the speakers are 7.5' apart and 8' from the listening position.

 

[preload]

On what are you basing this? The commonly accepted definition of the BBC dip (please link) or some version of the B&W dip? You realize the horizontal scaling of the graphs is not symmetrical, yes? In any case, the changes you suggest would not change the outcome, as the differences are at 0.5dB or less.

Also, I can draw one target curve to be shared between two speakers. Precision is not guaranteed. I could get a closer match with more effort, but this experiment does not turn out to be worth that effort.

BTW, the speakers are 7.5' apart and 8' from the listening position.

Neat experiment, was very interesting to read, and thanks for taking the time to write it up. This is what science is all about!

Steve, you may have learned this during your research, but there are quite a few variations on the lower-treble dip that is often voiced into loudspeakers. The "BBC" dip is one of them. There's lots of different versions, none of which seem to have a strict definition - there's a gundry drip, a presence dip, the B&W dip/boost, etc.

I can share my observations as well. As you know, I have the Genelec 8351B's, which I can quickly calibrate to their "flat target" using the GLM software and the Genelec calibrated mic. From there, I'd been playing around with the PEQ in Roon in order to create my own target curves. What I noticed when trying to reproduce the B&W 802D room curve (captured in a different listening room) was that I had to match it pretty closely, and it required multiple stacked filters to get the shape right. When I just initially created "close enough" curves that were off by a dB here and there, the result just didn't sound right at all. It wasn't until I got a very close match (verified with an MMM measurement) that I got to something I genuinely preferred over the default calibration curve.

This is something to keep in mind. The curve you reproduced on DIRAC wasn't the B&W target curve, and so I'm not surprised that it didn't sound good, and I believe that it didn't. I doubt you'd like to go through the effort of reproducing the B&W curve I posted earlier, and even if you did, it still wouldn't be a perfect match because an MMM measurement with a UMIK-2 isn't going to exactly translate into a DIRAC calibration with your measurement mic.

Nevertheless, it's interesting that you tried it out some sort of dip, and thanks again for sharing.

 

[test1223]

On what are you basing this? The commonly accepted definition of the BBC dip (please link) or some version of the B&W dip? You realize the horizontal scaling of the graphs is not symmetrical, yes? In any case, the changes you suggest would not change the outcome, as the differences are at 0.5dB or less.

Also, I can draw one target curve to be shared between two speakers. Precision is not guaranteed. I could get a closer match with more effort, but this experiment does not turn out to be worth that effort.

BTW, the speakers are 7.5' apart and 8' from the listening position.

Yes there were many versions. The version I described is very common, which you will find in many engineer driven speakers. E g. the Kef Bade 1 meta has a smaller version of this dip in exactly this frequency range. https://spinorama.internet-box.ch/speakers/KEF Blade 1 Meta/KEF/index_vendor.html

This frequency range around 2.8kHz gets very easy offensive and is a key area in our head related transfer functions. If you tried to design a speaker this is one very important part which can mess up a good design.

I think a lot of people take linearity way to serious. Our whole hearing is none linear in many aspects. E.g. a loudspeaker which is used to play back records at 70dB spl needs to be fine tuned differently in comparison to a speaker which is used to play at 85dB. Deviations of +/-2dB aren't perceived as bigger deviations by a lot of people anyways.

In this thread and here at asr I have the feeling that a lot of people have the attitude that most engineers which design speaker are accused to be simply stupid if the direct frequency response isn't flat, which is arrogant and often times the other way around.

 

[Ageve]

Sorry, I'm not letting you off the hook this time. This needs to stop.

I'm sure you're not.

I have a suggestion for you. Before posting, read through your text and think about how it will be perceived by others. Your attitude is not constructive.

You totally missed that the treble level dropped:
15 degs: -1dB over a 4 octave BW
30 degs: -3dB over a 4 octave BW
45 degs: -5dB over a 4 octave BW

I guess you somehow missed this:

Based on the 805 D4 measurements (same story with 804 D4, but it's easier to illustrate with the Soundstage images):
A slight angle (15 deg) will reduce the tweeter level by a single dB.
A not so slight angle (30 deg) will reduce the tweeter level by just 2 dB.* It will also make the 3 kHz dip even worse.
...
Difference between 500 Hz and 8-9 kHz **:
0 deg: 8 dB
15 deg: 7 dB
30 deg: 6 dB

* It's close to 2 dB. Hard to tell from a low resolution image.
** 500 Hz - 8 kHz = 4 octave bandwidth.

The response at 45 deg is not relevant when it comes to direct sound, especially not with toe-in, as it would result in an awkward listening position* (the distance between the speakers would be greater than the distance from the speakers to the listening position). 45 deg is obviously interesting because of room interaction though.

* I don't know your definition of a slight toe-in, but with just 10 deg, it would look like this. Optimal listening position would be somewhere between 15 and 30 in this picture:

As to why I didn't reply to "why the slope of the FR regression line is so relevant", it's because I didn't even mention it. It was you, and I don't want you to change the subject when replying to me.

The reason for choosing the frequencies I did, was: The response is a roller coaster. I mean where do you even begin? It's one of the worst measurements I've seen on Soundstage (I haven't looked at all of them). 500 Hz is just below the first resonance (Batman ear). 1.5 kHz is just above the second resonance. 3kHz is the dip.

The whole point was that the treble level stays almost the same in relation to lower frequencies, not in absolute terms.

I'm sure you would agree that the tonality of a speaker wouldn't change if you just turned the volume down by 2 dB (I'm not talking about the Fletcher Munson curve).

Yet, that's basically what happens at 15 and 30 deg off-axis. Just look at the 804 D4 measurements above.

The same thing can be seen in the 805 D4 measurements. The treble level has decreased by 2 dB (or 2.2-2.3...) at 30 deg, but that decrease is present all the way down to 800 Hz. That was my point - The tweeter is still just as bright compared to the rest of the sound.

And as I said, a similar response on- and off-axis is not a bad thing in itself (obviously). It means that the dispersion is even.

It does however, also mean that it will not be possible to "tame" the tweeter by moving off-axis (i.e. reduce the treble to an acceptable level). And that's really all I said to begin with.

 

[preload]

The whole point was that the treble level stays almost the same in relation to lower frequencies, not in absolute terms.

No it doesn't. And I've already posted the 805D4 off-axis charts twice where the decreasing treble levels relative to midbass are plainly observable by anyone.

If you cannot see that, there's honestly nothing else to discuss.

 

[Ageve]

No it doesn't. And I've already posted the 805D4 off-axis charts twice where the decreasing treble levels relative to midbass are plainly observable by anyone.

If you cannot see that, there's honestly nothing else to discuss.

A step in the right direction, I guess.

 

[preload]

In this thread and here at asr I have the feeling that a lot of people have the attitude that most engineers which design speaker are accused to be simply stupid if the direct frequency response isn't flat, which is arrogant and often times the other way around.

I think this should win post of the year. Couldn't have said it better.

 

[TheBatsEar]

The Meta tweeter makes a big difference. The non-meta tweeters sound veiled. The bass is still not clean though.

Agreed about the bass, but it's a tiny speaker. Best to add a subwoofer and cut out the bass as high as you dare using DSP.

I couldn't hear much of a difference between Meta and not Meta and i don't see anything decisive in the measurements either. They should have named them LS50 MetaHype.

The ATC SCM19v2 clearly won for me, they play louder, they play deeper. Sadly, they look like a lump i found in my armpit once. The LS50 are so much cuter.

Can't say anything about B&W, last speakers i had from them where the DM601, first version. Not terrible, but not brilliant either.

 

[Kostas_96]

https://www.reddit.com/r/audiophile/comments/t2x0j3
This is why we always must listen everything. This guy was ready to "love" kef but in the end he was "unhappy" because b&w was better.

 

[Everett T]

https://www.reddit.com/r/audiophile/comments/t2x0j3
This is why we always must listen everything. This guy was ready to "love" kef but in the end he was "unhappy" because b&w was better.

Could have been they weren't level matched and since the BW are a couple of dbs more sensitive, created a larger imaging perception, that and the midrange hump.

 

[Kostas_96]

Could have been they weren't level matched and since the BW are a couple of dbs more sensitive, created a larger imaging perception, that and the midrange hump.

Of course i agree with you. But we shouldn't reject any speaker without listening it. Everyone has different ears. They are 2000$ speakers..they are both amazing, its just personal which we like more.

 

[Steve Dallas]

Of course i agree with you. But we shouldn't reject any speaker without listening it. Everyone has different ears. They are 2000$ speakers..they are both amazing, its just personal which we like more.

Listening was also conducted at Best Buy / Magnolia, which is one of the worst possible places to audition good speakers.

 

[Everett T]

Of course i agree with you. But we shouldn't reject any speaker without listening it. Everyone has different ears. They are 2000$ speakers..they are both amazing, its just personal which we like more.

I agree, evaluation is important and so is how the evaluation is conducted. I have a bias against B&W speakers below the 802D and the original Matrix 801 S2, but that's just me and my sound and design preference. I've only bought two speakers without listening to them, one I knew the designers and their goals (heard a variant of it but with a different midrange), and the other purely off measurements, still enjoying them both. In the end it's about enjoyment just that sometimes the road to that can be a little bumpy.

 

[Tonygeno]

https://www.reddit.com/r/audiophile/comments/t2x0j3
This is why we always must listen everything. This guy was ready to "love" kef but in the end he was "unhappy" because b&w was better.

Of course long term satisfaction cannot be predicted based on a demo in a store.

 

[DMill]

Flawed test. But more importantly, even if the test were valid, one person preferring a speaker over another means absolutely nothing. There is somebody out there that prefers a hot dog vs. a dry aged prime rib. Most people don’t though.

 

[TrevC]

https://www.reddit.com/r/audiophile/comments/t2x0j3
This is why we always must listen everything. This guy was ready to "love" kef but in the end he was "unhappy" because b&w was better.

Most of those expressing a preference for the KEFs have never heard the Bowers, nor have they tested them. The 706 is astonishingly good. I certainly have no regrets about buying them. Well apart from them being cheaper when the S3 came out that is!