Kef R3 META, digital crossover?

[LionIT]

Hi all,

I am an audio enthusiast and I really like to experiment with my system to seek an increasingly engaging but also technically more perfect experience. It's a bit like my hobby, let's say.
I currently own KEF R3 Meta and two amplifiers, a Hypex NC252MP and a Neurochrome Modulus 86 (old model with single-ended output).
I also have two 10" subs with NC502MP amp and multichannel audio interface with FIR filtering on PC (Acourate).
I had taken the Neurochrome to biamp the old DIY speakers, which I then sold. So now I have this amp and I'm thinking of using it to biamp the R3s, and use digital filters to better linearize phase and magnitude in the crossover region.
The R3 already have double binding posts, one for the LF driver and the other for the Uni-Q (MF+HF), therefore the most practical option would be to dedicate the Neurochrome to the Uni-Q driver. but it has "only" 65W on 4 ohms and from the measurements of the R3 it seems that the passive crossover between LF and MF already works perfectly, both for FR, PR and DI.
What seems to be sub optimal is the transition between MF and HF, that involves a good PR, but to the detriment of FR and DI (this is also confirmed as a trade off by KEF in its white paper).
Now, in practical terms, it would not be a problem to disconnect the HF driver from the passive crossover to connect it directly to the Neurochrome (I could temporarily pass the cable in the reflex port). Without changing the passive crossover, however, the LPF filter of the MF driver would remain, which would limit the possibility of playing with digital crossover a little.
Surely I could better linearize phase and FR in the region, and I could also use much steeper filters to better separate the bands (considering also the drivers of the Uni-Q are at opposite polarity, therefore integration worsens by moving away from the crossover point).
The main doubt is in terms of PR and DI, which I have no way of measuring too easily / precisely.
On Erin's website there are the Uni-Q measurements of the R300s including off-axis at 30 and 60 deg, which show a rather uniform directivity. That unit should not be very different from that of the R3 Meta, so this makes me think that playing with a digital crossover I could improve the performance a bit.
I'm not sure which approach to use though.

What would you do?
Can anyone give any suggestions or ideas about this?

Keep in mind, as I said at the beginning, this is a bit of a hobby for me. I'm not thinking that R3s have a problem to solve, nor that any audible improvements can come out. It is only to pursue technical improvements, fun and even peace of mind!

 

[Miley_B]

What seems to be sub optimal is the transition between MF and HF, that involves a good PR, but to the detriment of FR and DI (this is also confirmed as a trade off by KEF in its white paper).

From the linked Kef R series with MAT white paper:

"The R3 Meta is a special case within the range as the LF section does not form a symmetrical arrangement around the Uni-Q, since it has only one LF driver. For this reason, the Uni-Q is mounted close to the top edge of the enclosure, and this results in a dip in its acoustic power directivity index just above 2kHz. Designing a crossover considering only the axial response smoothness would result in a corresponding bump in the acoustic power average. The studied approach, however, allows a dip in the axial response, prioritising a smooth power average. This represents the approach refined during the development of Reference Meta [8] as well as extensive listening tests in different rooms."

The compromise appears to be with the cabinet design not the crossover so if the reason for using a DSP crossover is to fix the directivity index dip I don't see that this could be achieved with DSP whilst keeping the drivers within their operating window. I'm no speaker designer though.

I think the test of the R300 driver was done in free space not on a baffle.
Erins R300 driver measurements pre date his Klippel NFS so are unlikely to be as accurate as his current work.
That said, if you want to see what you can achieve and you're not worried about the possibility of damage when you're experimenting then I say go for it!

 

[Oddball]

Honestly does not seem like a good reward benefit case. But people do all kinds of things for hobby. There are modified cars that cost quite a bit more than these speakers, and people have done it knowing they would last for only 5k or so km.

I would personally get better speakers for the risk premium and enjoy the content more with those.

Or stay with what you have and if you need to tweak, probably better to do it in software space. Many other EQ solutions available albeit might require some hardware.

 

[LionIT]

Good observations.
Kef explains that due to the vertically asymmetric design they have developed the passive crossover of the R3 with the aim of having a linear PR, rather than a linear FR on the axis.
From a purely mathematical point of view, however, I suppose there could be n solutions with linear PR and different FR and DI. And among these solutions there could be some that are better than the one implemented with the passive crossover, which by nature does not have the versatility and potential of a digital filter.
In practical terms, I imagine as an improvement a reduction of that horizontal dispersion peak at 2.5kHz, maintaining a sufficiently linear PR. But it is only my hypothesis, and it is not certain that it is achievable.

A Klippel NFS would help... but in the absence of it, is there any valid approach in your opinion?

From a theoretical point of view, in your opinion, could we already identify a filter that is definitely better than the passive one?
For example, steeper filters plus phase linearization shouldn't sum better off axis, being a coax driver?

 

[Ellebob]

Another thing to remember is you are not bypassing the crossover circuitry in the Kef when you bi-amp. There are still filters on each connection to filter out the highs from the woofer and filter out the lows from the coaxial. To get the true benefit of a digital crossover you would want to directly power the woofer and coax separately without the internal crossover in place.

 

[LionIT]

Another thing to remember is you are not bypassing the crossover circuitry in the Kef when you bi-amp. There are still filters on each connection to filter out the highs from the woofer and filter out the lows from the coaxial. To get the true benefit of a digital crossover you would want to directly power the woofer and coax separately without the internal crossover in place.

Yes, of course.
But for me, opening the speakers is not a problem. In fact for the tweeter I can connect directly to its internal terminals, while for the mid I do not have a dedicated amp to drive it, so I have to keep the passive crossover to supply the power and maintain the integration with the woofer.
To fully benefit from the digital crossover the LPF part of the crossover for the mid should be removed, but I don't want to make irreversible changes.
With a digital filter, however, I can still linearize the phase of the mid LPF, and adjust the magnitude a little bit, depending on how the passive crossover is rolled off.
Clearly this discourse derives from the consideration that the crossover between woofer and mid is almost perfect so it would make no sense to intervene on it.

 

[charlielaub]

@LionIT Have you designed crossovers before? Can you make accurate quasi-anechoic response measurements on the drivers using a calibrated microhpone? If yes to both, then sure you can redo/remake the crossover in any way you see fit. It can be fun and fulfill your hobby desires, but whether it will improve on the loudspeaker is unclear.

I have KEF R3 Meta speakers set up in one of my listening rooms. I use DSP to change the overall tonal balance, which is too bright for my tastes, and to crossover to a subwoofer. I use the R3s with the ports plugged. You could try to see if similar gentle EQ of the response above 1kHz helps. Since the speaker has close to constant directivity, EQ is an effective and practical way to change how the speaker sounds overall without having to get into the mud with a crossover (DSP or hybrid) redesign.

 

[LionIT]

@LionIT Have you designed crossovers before? Can you make accurate quasi-anechoic response measurements on the drivers using a calibrated microhpone? If yes to both, then sure you can redo/remake the crossover in any way you see fit. It can be fun and fulfill your hobby desires, but whether it will improve on the loudspeaker is unclear.

I have KEF R3 Meta speakers set up in one of my listening rooms. I use DSP to change the overall tonal balance, which is too bright for my tastes, and to crossover to a subwoofer. I use the R3s with the ports plugged. You could try to see if similar gentle EQ of the response above 1kHz helps. Since the speaker has close to constant directivity, EQ is an effective and practical way to change how the speaker sounds overall without having to get into the mud with a crossover (DSP or hybrid) redesign.

I have already created some digital crossovers with my old DIY speakers using Acourate.
On quasi anechoic measurements I never tried to do them because in the end to create digital crossover filters with Acourate I always had to use the measurements in the listening point.
According to someone, with an appropriate FDW the FR obtained with such measurements does not differ much from the anechoic one.
Of course here it would be necessary to see PR and DI to know with certainty the result obtained compared to the original passive crossover.
The discussion is precisely to understand if there may be some approach, even approximate, sufficiently valid.
I was actually thinking of trying quasi anechoic measurements.

 

[Ellebob]

If you are willing to bypass the built in crossover between the woofer and coax unit, I say go for it!

 

[LionIT]

If you are willing to bypass the built in crossover between the woofer and coax unit, I say go for it!

Between woofer and Uni-Q it doesn't seem worth it to me, because in that region FR, PR and DI are almost perfect with the passive XO (but if I miss something correct me!).
Having to choose I would opt to bypass the XO between mid and tweeter. Beyond FR / PR / DI, at shorter wavelengths could be more beneficial to have a perfectly linearized phase between drivers, especially with coax.
But I repeat, if I miss any advantages to digitize the XO between woofer and Uni-Q explain to me as well!

 

[Ellebob]

There are a lot of advantages to a digital crossover over passive for more precision. Powering the drivers directly will also decrease distortion. How much of a notable difference is the question.

 

[LionIT]

How much of a notable difference is the question.

Honestly, I think the only thing I would notice is a nice step response chart. Perhaps a little better definition in the soundstage, but the xo band is very narrow so I'm not sure.
However, looking at that chart and thinking that at least from the speaker all the frequencies come out exactly at the same time is already a satisfaction

 

[Digital_Thor]

I tried to listen to so many speaker systems, and found out - finally - that I prefer a smooth power response the best. This made me fall for the KEF coax, which I first only thought could be obtained, by buying the finished speaker, which would also relieve me from the burden of making a good-looking speaker cabinet
For me, the "standard" KEF sound and approach, was not fully to my liking. The modern trend with small drivers, do make sense in several aspects, but I do also like the more "full" and authoritarian sound that bigger woofers can bring to the sound, from below around 400Hz. Further, the slight rise at the top ( yes I know about the toe-in) was not my taste either.
Since I was never blessed with the patience to learn how to build a passive filter, I jumped straight to active filtering with DSP and an amplifier for each driver unit.
Above 400Hz I only use the DSP to make the speakers measured performance look like what you often see in NFS scans. Below, when you enter the long-wavelength territory, I start fiddling with the DSP to sum the mains with my subwoofers and the room.

Finding a set of "loose" KEF unit, was my happy trip down DIY-lane. And now, I think I would just buy a set of R3's and "steal" the Coax to build what I want.
And I want relative narrow towers, with larger woofers and closed boxes, point source sound, that still reach subwoofers with a smooth downwards roll-off.
And I think the result sounds awesome.... looks are getting there.. but I'm no champ in wood working - so I had to live with some compromises
You could just pull the passive filter and do the active swap directly in the original cabinet... if you wish. Cause if you are already going that far... what else could be fun to do too?

So I had to decide to take that chance. Maybe I could live with a set of R7, with plugged ports and some EQ to "clean" up the linear problems in the upper frequencies.
But ask yourself - are your DIY-skills ready to rumble? Because it can be done... you just gotta be willing to fight a bit for it. Though.... I would also never shame anyone for not doing it. I'm just me, with my little personal HIFI adventure

 

[charlielaub]

I have already created some digital crossovers with my old DIY speakers using Acourate.
On quasi anechoic measurements I never tried to do them because in the end to create digital crossover filters with Acourate I always had to use the measurements in the listening point.
According to someone, with an appropriate FDW the FR obtained with such measurements does not differ much from the anechoic one.
Of course here it would be necessary to see PR and DI to know with certainty the result obtained compared to the original passive crossover.
The discussion is precisely to understand if there may be some approach, even approximate, sufficiently valid.
I was actually thinking of trying quasi anechoic measurements.

Well IMO one problem with this approach is using the in-room measurement for the basis of FIR global correction. I don't mean to try to offend, but I see this as a typical DIYer mis-use or over-use of the amazing capabilities of FIR filtering and Acourate.

What you want from a loudspeaker is that its acoustic output is well integrated between all the drivers and on all axes. The R3 Meta already has this sort of characteristic, right out of the box. In order to know what the loudspeaker itself is doing you need to remove the room from the equation. Making a quasi anechoic measurement is not always simple if you want it to be valid at frequencies below about 300Hz, and with sufficient resolution, so people resort to doing what you do (include the room). This can cause problems even up around 1kHz or so. Maybe this explains your concerns with the crossover?

The room interaction and treatment is a separate issue and I would not use FIR/DIRAC for the most part. I really do not have a positive impression of room correction except when used very mildly and with gentle corrections, and for the lowest bass frequencies e.g. room resonances. I have the same sentiment towards FIR linear phase filtering. These techniques are only valid for one location anyway. But you hear sound from the loudspeaker and reflections from all the room surfaces, and these are all arriving at different times, so you cannot truly fix these sorts of problems via DSP or the crossover. Knowing which to fix and which to leave alone is important. Hint: fix the speaker itself first, then look for problems with the room that you can address by changing the characteristics of the room. You might need to make a wall MORE reflective instead of adding diffusion or adsorption. Reflections are not all bad, and it turns out that they can be useful/used to improve the sound quality. So don't always try to bash them over the head with Acourate...

 

[LionIT]

Thanks everyone for your comments, very interesting!

I am about to verify the practicability of the modification.
There is already a first thing that seems less easy than expected.
How do I remove the plastic ring around the drivers (shadow flare)?
From the videos and online info it just seems to be pushed in by pressure and kept there for interference (in fact there is variability of FR due to the misplacement of it).
But how to extract it, without ruining it, is less intuitive than expected.
That of the woofer has enough space to put on something that can pull, but that of the Uni-Q does not.
Is there any particular trick?
If I use double-sided gel tape (the one normally removable and reusable)? Is there a risk of remove the paint from the ring or is it just plastic?
Among other things, I would like to take advantage of this operation to better align the driver to the center of the ring and look for a ring positioning system that prevents it from coming out due to vibrations.

 

[ROOSKIE]

Hi all,

I am an audio enthusiast and I really like to experiment with my system to seek an increasingly engaging but also technically more perfect experience. It's a bit like my hobby, let's say.
I currently own KEF R3 Meta and two amplifiers, a Hypex NC252MP and a Neurochrome Modulus 86 (old model with single-ended output).
I also have two 10" subs with NC502MP amp and multichannel audio interface with FIR filtering on PC (Acourate).
I had taken the Neurochrome to biamp the old DIY speakers, which I then sold. So now I have this amp and I'm thinking of using it to biamp the R3s, and use digital filters to better linearize phase and magnitude in the crossover region.
The R3 already have double binding posts, one for the LF driver and the other for the Uni-Q (MF+HF), therefore the most practical option would be to dedicate the Neurochrome to the Uni-Q driver. but it has "only" 65W on 4 ohms and from the measurements of the R3 it seems that the passive crossover between LF and MF already works perfectly, both for FR, PR and DI.
What seems to be sub optimal is the transition between MF and HF, that involves a good PR, but to the detriment of FR and DI (this is also confirmed as a trade off by KEF in its white paper).
Now, in practical terms, it would not be a problem to disconnect the HF driver from the passive crossover to connect it directly to the Neurochrome (I could temporarily pass the cable in the reflex port). Without changing the passive crossover, however, the LPF filter of the MF driver would remain, which would limit the possibility of playing with digital crossover a little.
Surely I could better linearize phase and FR in the region, and I could also use much steeper filters to better separate the bands (considering also the drivers of the Uni-Q are at opposite polarity, therefore integration worsens by moving away from the crossover point).
The main doubt is in terms of PR and DI, which I have no way of measuring too easily / precisely.
On Erin's website there are the Uni-Q measurements of the R300s including off-axis at 30 and 60 deg, which show a rather uniform directivity. That unit should not be very different from that of the R3 Meta, so this makes me think that playing with a digital crossover I could improve the performance a bit.
I'm not sure which approach to use though.

What would you do?
Can anyone give any suggestions or ideas about this?

Keep in mind, as I said at the beginning, this is a bit of a hobby for me. I'm not thinking that R3s have a problem to solve, nor that any audible improvements can come out. It is only to pursue technical improvements, fun and even peace of mind!

Huge DIY proponent.
Therefore I think you should play around a little with this if this is a current hobby because why not?
But really only enough so that you realize you enjoy playing and then turn to toward something more worthwhile, as I think this is a lot of work that could be applied more smartly to a different DIY project.*

*In fact why not play with just applying some PEQ to get a feel for small changes and decide if such things are audible to you and if the improvement level brings benefit. The example PEQ @Spinorma.org brings the 'score' up about 1/2 of a point. You could implement and AB or ABX and see what that kind of change brings. https://www.spinorama.org/speakers/KEF R3 Meta/ErinsAudioCorner/index_eac-v1-ported.html

That said going back to any unpublished DIY, you would have to be able measure quite accurately. Lucky for you gated filtering in the upper mids and high frequency range leaves a measurement that is pretty accurate. Especially outdoors, but even so in a large room.

You are dealing with a KEF marketing department that wanted industrial design. The sharp edges are a factor. KEF has mitigated them very well but adding side panels with big rounded over edges and a top panel with a smoother transition will reduce or remove the left over edge diffraction artifacts. You can not remove acoustic interferences with DSP or active crossovers.
The tweeter is in a waveguide so it doesn't see the edges much(the waveguide is the tweeters edge) but the woofer does see them and so you simply will have interference issues. If the tweeter was crossed very low then it would not be as much of an issue. That tweeter can't handle that though. Steeper slopes will not help much, they will likely just shift the issue around in frequency a bit. If you could cross down at like 1000-1200hrz maybe even 1500 you'd be good, but anyway that required a beefy tweeter in a huge waveguide ala JBL.
Get the Edge diffraction simulator and play around with examples. That might give you some ideas.

So yah, the major part of the solution for this left over acoustic interference would be a very curved cabinet like Genelec but that is not what KEF is selling. I'm fine with that as the whole industrial design of KEF products looks cool to me and they did really well reducing those issue to very minor things. Look at the issues other companies are selling in sharp edged boxes.

My ring on my regular R3 set only came forward in shipping. It doesn't seem to move from normal use.
Where do you have your sub(s) crossed in? The R3 sounds bad to me below 60ish hrz at higher SPL, that would be a very nice place to work on improvements. I haven't used mine in awhile but I think I ended up around 90-100hrz to the subs, major subjective and objective benefits. Though you may have already realized any benefits with your current tune.

PS. Hey!!, here is what I think you 'should'/could do. Drop $300 on a pair of Q150's on sale and see how close you get them to the R3 above 150hrz. Pull the crossover, brace the cabinets a bit, play with stuffing, make a full active crossover with DSP PEQ and really eek it out. Then blend with the subs as '3 ways' around 150-180hrz. While you are at it put some crazy paper backed veneer https://www.veneersupplies.com/ on the cabinets with iron on glue(not a lot of tools required for that) and just enjoy some weird DIY time. https://www.joewoodworker.com/veneering/iron-on-veneering.htm and a lot is still on the table in the Q150 (or Q350). KEF pretty much crushed the R3meta with epic super engineering skills given that desired aesthetic. Some PEQ applied to the Klippel measurements data will help a bit but I bet no amount of DIY x-over work beyond that (unless modding the cabinets) will make them better, only different.

 

[davidbosch]

Hi all,

I am an audio enthusiast and I really like to experiment with my system to seek an increasingly engaging but also technically more perfect experience. It's a bit like my hobby, let's say.
I currently own KEF R3 Meta and two amplifiers, a Hypex NC252MP and a Neurochrome Modulus 86 (old model with single-ended output).
I also have two 10" subs with NC502MP amp and multichannel audio interface with FIR filtering on PC (Acourate).
I had taken the Neurochrome to biamp the old DIY speakers, which I then sold. So now I have this amp and I'm thinking of using it to biamp the R3s, and use digital filters to better linearize phase and magnitude in the crossover region.
The R3 already have double binding posts, one for the LF driver and the other for the Uni-Q (MF+HF), therefore the most practical option would be to dedicate the Neurochrome to the Uni-Q driver. but it has "only" 65W on 4 ohms and from the measurements of the R3 it seems that the passive crossover between LF and MF already works perfectly, both for FR, PR and DI.
What seems to be sub optimal is the transition between MF and HF, that involves a good PR, but to the detriment of FR and DI (this is also confirmed as a trade off by KEF in its white paper).
Now, in practical terms, it would not be a problem to disconnect the HF driver from the passive crossover to connect it directly to the Neurochrome (I could temporarily pass the cable in the reflex port). Without changing the passive crossover, however, the LPF filter of the MF driver would remain, which would limit the possibility of playing with digital crossover a little.
Surely I could better linearize phase and FR in the region, and I could also use much steeper filters to better separate the bands (considering also the drivers of the Uni-Q are at opposite polarity, therefore integration worsens by moving away from the crossover point).
The main doubt is in terms of PR and DI, which I have no way of measuring too easily / precisely.
On Erin's website there are the Uni-Q measurements of the R300s including off-axis at 30 and 60 deg, which show a rather uniform directivity. That unit should not be very different from that of the R3 Meta, so this makes me think that playing with a digital crossover I could improve the performance a bit.
I'm not sure which approach to use though.

What would you do?
Can anyone give any suggestions or ideas about this?

Keep in mind, as I said at the beginning, this is a bit of a hobby for me. I'm not thinking that R3s have a problem to solve, nor that any audible improvements can come out. It is only to pursue technical improvements, fun and even peace of mind!

Hi @LionIT Some of the comments above are quite right, you won't be modifying the inherent acoustics of the system, and the directivity of each way is the result of the driver and system acoustics. You can certainly play around with either smoothing the on-axis or the power average if you have a directivity discontinuity, particularly around the crossover region, where the output from the two drivers combine. We've found that a smooth power average is generally more desirable, but we do have a compromise (and it depends on how discontinuous the directivity is for a particular speaker). You could certainly make some improvements in the crossover with steeper roll-offs, not only in the MF-HF but definitely also in the LF-MF, where you want to avoid driving the MF to large displacement (which leads to distortion) and where you want to minimise the LF reproducing midrange, as there are various resonances that can be excited in that region (large acoustic cavity, suspension, large enclosure panels, etc.), but to do this there is no easy way without proper measurements. If you are to redesign the crossover you need good quality anechoic measurements or you'll be shooting in the dark. Designing a crossover by only listening to it is very difficult if impossible to do correctly even with a trained ear so you run the risk of not only not improving on the already carefully designed passive crossover but ending up with something worse. You also need to know what's the lower limit of the MF and HF so that you don't drive them into distortion. Having said that, you can try measuring the transfer functions of the LF, MF and HF passive crossovers and take those as a starting point for your DSP crossovers and make an on axis and an off-axis (30 or 40 deg horizontal?) measurement in a semi-anechoic environment close to the speaker but not in the nearfield and use those as a broad reference. Then potentially you can do adjustments to the LP on the MF and the HP on the HF that yield the same combined response on axis and - hopefully - a better response off axis. Finally, if you're going digital, go all the way and completely delete the passive crossovers. Passive crossovers both add distortion and dissipate power, so you can have a better speaker and happier amps with a digital crossover. You'll need more amps but it's probably the only way in which this project is worth it in my opinion. Hope this helps.

PS. I would recommend not changing the MF-HF crossover frequency. That frequency is generally set by the Uni-Q geometry (and the filters' shapes, so you have some wiggle room). Uni-Q, as the name suggests, allows to match the MF and the HF directivity at the crossover frequency, by narrowing the HF directivity at its lower end through the use of a waveguide. If you decrease the crossover frequency, you'll cross-over to the HF when the MF's directivity is wider than the HF's, and you'll cause a discontinuity in the system's directivity. Also, moving the HP on one of these tweeters to anything below 1.8 kHz will surely result in early distortion. I wrote a little section on the Blade/Reference Meta white paper that exemplifies this (you can find it here: https://assets.kef.com/documents/reference/KEF_Blade_Ref_Meta_Tech_Paper.pdf)

 

[LionIT]

Wow. I didn't expect these contributions. Thanks to all really.
You are giving me very precious and meaningful notions, and I will take note of it.

@davidbosch
An element that intrigues me about the Uni-Q is the relative phase between MF and HF in the crossover region.
Based on your experience / knowledge, is it possible to define a relationship between relative phase and directivity?
Or in simpler terms, how could the directivity of a Uni-Q with perfectly phased drivers (via DSP) could be compared to the current passive XO?

 

[Winemaker]

Good morning everyone, but comparing the "normal" R3s to the R3 Metas, is there a big difference in terms of sound or are they more or less the same thing? Are they a real upgrade or is it just a marketing gimmick?

 

[thewas]

Hi @LionIT Some of the comments above are quite right, you won't be modifying the inherent acoustics of the system, and the directivity of each way is the result of the driver and system acoustics. You can certainly play around with either smoothing the on-axis or the power average if you have a directivity discontinuity, particularly around the crossover region, where the output from the two drivers combine. We've found that a smooth power average is generally more desirable, but we do have a compromise (and it depends on how discontinuous the directivity is for a particular speaker). You could certainly make some improvements in the crossover with steeper roll-offs, not only in the MF-HF but definitely also in the LF-MF, where you want to avoid driving the MF to large displacement (which leads to distortion) and where you want to minimise the LF reproducing midrange, as there are various resonances that can be excited in that region (large acoustic cavity, suspension, large enclosure panels, etc.), but to do this there is no easy way without proper measurements. If you are to redesign the crossover you need good quality anechoic measurements or you'll be shooting in the dark. Designing a crossover by only listening to it is very difficult if impossible to do correctly even with a trained ear so you run the risk of not only not improving on the already carefully designed passive crossover but ending up with something worse. You also need to know what's the lower limit of the MF and HF so that you don't drive them into distortion. Having said that, you can try measuring the transfer functions of the LF, MF and HF passive crossovers and take those as a starting point for your DSP crossovers and make an on axis and an off-axis (30 or 40 deg horizontal?) measurement in a semi-anechoic environment close to the speaker but not in the nearfield and use those as a broad reference. Then potentially you can do adjustments to the LP on the MF and the HP on the HF that yield the same combined response on axis and - hopefully - a better response off axis. Finally, if you're going digital, go all the way and completely delete the passive crossovers. Passive crossovers both add distortion and dissipate power, so you can have a better speaker and happier amps with a digital crossover. You'll need more amps but it's probably the only way in which this project is worth it in my opinion. Hope this helps.

PS. I would recommend not changing the MF-HF crossover frequency. That frequency is generally set by the Uni-Q geometry (and the filters' shapes, so you have some wiggle room). Uni-Q, as the name suggests, allows to match the MF and the HF directivity at the crossover frequency, by narrowing the HF directivity at its lower end through the use of a waveguide. If you decrease the crossover frequency, you'll cross-over to the HF when the MF's directivity is wider than the HF's, and you'll cause a discontinuity in the system's directivity. Also, moving the HP on one of these tweeters to anything below 1.8 kHz will surely result in early distortion. I wrote a little section on the Blade/Reference Meta white paper that exemplifies this (you can find it here: https://assets.kef.com/documents/reference/KEF_Blade_Ref_Meta_Tech_Paper.pdf)

View attachment 400389

What an unusually helpful and detailed answer from a manufacturer where most others would give no advice and say leave it as it is, one of the reasons I really like KEF.