KEF R900 Improvements with DSP

[Waveform Fidelity]

Hello Forum members,

I have just joined ASR to share my experiences in upgrading my KEF R900 with a DSP crossover.

My first question is the SP1632 driver has a very high output at 5kHz, which I understand is diaphragm breakup. The directivity goes bad at this frequency and the out of band output is higher than in band. I read in KEF AES paper that they using a viscous rubber damper beween voice coil and diaphragm to decouple the driving signal from the cone and the claim is that that fixes this breakup response peaks occuring at 5kHz and after from being audible.

Anyone have any thoughts on this?

 

[DVDdoug]

Since that's not a "cheap speaker" and KEF is a reputable company with skilled engineers, personally I wouldn't try to re-design and re-build it. Although if your modifications are reversable you wouldn't hurt the value.

EQ is worth trying and it can be free on a PC so you can try it even if you want to eventually use "hardware" EQ/DSP as your solution.

Another more practical solution is replace the speakers with something you actually like, and you can optionally sell the R900's.


I do have a pair of "cheap generic speakers" (not my main speakers) that had rotted woofer surrounds so I replaced the woofers with something "random" without running it through speaker design software or anything. And the highs seem a bit weak so I might replace the tweeters... someday...

 

[Waveform Fidelity]

I have been modifying my KEF R900 speaker from it's passive crossover to a DSP one and wanted to share with members my findings if you are trying a similar approach.

I am using DSP filter/EQ with a maximum of 512 Taps and equalizing both the amplitude and phase

First the results. On-axis and 50 degrees off-axis. Improvements to flatness and ripple are evident from the plot:

Some comments and conclusions

1. The curves are just covering the SP1632 UniQ driver array in the R900 cabinet
2. I used a crossover frequency of 2.8kHz (OEM is 2.9kHz) as the polar response goes really bad at around 3kHz on the MF unit. There is also terrible cone breakup and peaks out of band in the MF unit which have been fully suppressed.
3. I tried a number of crossover slopes to get a seamless transition and found the optimum was LR 48dB/OCT acoustic
4. One of the limitations with FIR filter is the dynamic range when the you need low frequency suppression - you can get very sharp and deep filters but they tend to still have significant response at low frequencies - this is especially important for the tweeter and mid-range that must be protected from low frequency signal. I ended up using capacitors to obtain the additional roll-off ( provided DC blocking too)
5. You will see the DSP responses are smoother, but the ripple actually comes from FIR filter ripple due to large cut-off slopes.
6. Getting the response flat on -axis doesn't help with the fundamental limitations of the speaker off-axis - due to the tangerine wave-guide and cabinet effects.
7. The factor EQ makes the speaker "bright" in the 3-5kHz range as you can see from the slopes.

Hope this is helpful to members, that might be considering an upgrade.

More to come on the dynamics of active vs the passive crossover.

 

[Digital_Thor]

Lovely speaker. I just copied it with coax from a R3 and two Satori WO24P, fully active, LR24dB at 400 and 2200Hz

 

[Waveform Fidelity]

Nice one! The later, improved UniQ drivers can be crossed over as low as 2.2kHz, but the earlier version in the R900 needs to stay above 2.7kHz to avoid distortion. Is your speaker like the R900 which has seperate enclosures for each woofer? And is it ported? I block the ports on mine (-6dB around 45Hz). I have just finished doing the crossover LR 48 at 400MHz and equalizing the phase that stays at 120 +-/30 degrees over 45-400Hz. I am curious how you align the 400Hz crossover points? The SPL levels - with LR there is a power dip as the woofers pair contribute “equally” to the UniQ. I use a MLS signal which has to be windowed, which then looses validity of the low frequency measurement - so I go outside and with the lower woofer is 1.6m from the ground, but the window is still not wide enough to get good low frequency response. If I use a swept sinewave I get reflections from the ground, so I am thinking of putting the speaker on its back so the drivers as facing skywards to avoid the floor bounce.

Are you doing the crossovers in DSP or analog active filters (opamps)

How does the speaker sound? Happy with it?

 

[Digital_Thor]

Exactly. I chose this driver because it was the first to include the rubber-coupling between the voice coil and the cone, namely to counter the break-up mode in the midrange cone.
I did compare it to a set of LS50 Meta, but found the dedicated midrange and active solution + EQ ( before the room) to smooth out any noticeable difference between the two.
Furthermore, I actually wanted to copy the R900 more directly, but decided that maybe it was not that important, since I like closed cabinets and the Satori plays nicely in one. Also, I want a smooth downward tilt towards the deeper frequencies, with minimum phase issues from port, passive driver or the like, because that made it easier to blend them with my 4 subwoofers.
I chose the WO24P because they are quite broadband, and smooth in their roll-off characteristic, which gives the filters a better chance to work as the math behind tries to dictate.
Maybe, because I see above 500Hz as a speaker without room, and below around 500Hz as an increasingly room dependent collaboration between less and less direct sound and more and more reflective sound. By this theory, I focus on only correcting the speaker above around 500Hz - because this is the lowest I can measure reliably in-room - with a pillow at the floor (first reflection) and gated.
Then I both listen and measure at the listening position with all 4 woofers, and then add the 4 subwoofers, as I find a coherent sum, so that I correct for both the room, my preference and listening position with now 6 bass-sources in total.
My pre is 8 channels, but I direct 2 to a 4 channel DSP, so that I can have fully active 3-way mains and 4 individually controlled subwoofers.
3x200W@8ohms for mains and 4x700@8ohms for subwoofers - meaning absolutely no limits to power or SPL, unless I want hearing damage or fried drivers.
The result is totally effortless dynamic sound at any preferred listening level and anything from thundering deeps to airy highs, in one unison - which of course is my interpretation - but some that heard the system - do say that I'm not totally off the rails
I'm a bit bonkers, but it is great fun to make your own stuff

 

[Digital_Thor]

Hello Forum members,

I have just joined ASR to share my experiences in upgrading my KEF R900 with a DSP crossover.

My first question is the SP1632 driver has a very high output at 5kHz, which I understand is diaphragm breakup. The directivity goes bad at this frequency and the out of band output is higher than in band. I read in KEF AES paper that they using a viscous rubber damper beween voice coil and diaphragm to decouple the driving signal from the cone and the claim is that that fixes this breakup response peaks occuring at 5kHz and after from being audible.

Anyone have any thoughts on this?

Mine are the SP1753AA, which should be the 2018 version, the dedicated midrange version from the R3 - just before Meta.
It is right here:

https://assets.kef.com/documents/rseries/rseries2018-white-paper.pdf

You should be able to EQ yours to be better, even though taming the break-up, is always tricky.
If it was me, I would just buy a set of R3 or R3 Meta and pull the driver.... but that's just me

 

[terryforsythe]

Hello Forum members,

I have just joined ASR to share my experiences in upgrading my KEF R900 with a DSP crossover.

My first question is the SP1632 driver has a very high output at 5kHz, which I understand is diaphragm breakup. The directivity goes bad at this frequency and the out of band output is higher than in band. I read in KEF AES paper that they using a viscous rubber damper beween voice coil and diaphragm to decouple the driving signal from the cone and the claim is that that fixes this breakup response peaks occuring at 5kHz and after from being audible.

Anyone have any thoughts on this?

I have removed the passive crossovers from my Elac Uni-Fi Reference bookshelf speakers and gone all active. The way I approached it was to take frequency response measurements of the drivers on and off axis, then model it in VituixCAD. Not only did I pay attention to on-axis frequency response, but also to the directivity index and estimated in-room response. The directivity index and estimated in-room response will guide you on how high in frequency you can push the midrange.

The benefit with DSP is that it opens up a lot of flexibility with respect to crossover topology, e.g., crossover slopes. Also, removing inductors from the woofer circuits improves damping factor significantly.

Initially I only meaured the drivers at 0, 30 and 60 degrees horizontally. That is not sufficient. I am preparing to do a new set of much more thorough measurements. Here is a link to my thread on the project:

Modifications to Elac Uni-Fi Reference UBR62 Bookshelf Speakers

I have removed the passive crossovers from my Elac Uni-Fi Reference UBR62 Bookshelf Speakers and gone all active. In doing so, I moved the woofer/midrange crossover frequency from 200 Hz to 450 Hz. Side Note: Elac specifies the woofer/midrange crossover point as being 260 Hz, but my pair...

www.audiosciencereview.com

There will be more to come after I get a more complete set of measurements and update my VituixCAD model. Nonetheless, as it is, with the modifications I have made and implementing DSP I have improved the sound of the speakers both objectively and subjectively.

 

[Waveform Fidelity]

Yep, I’ve been building speakers for a while in 1977 I read Siegfried Linkwitz paper and design in Wireless World ( a UK magazine) and built it - designed around the KEF T27 and B110… as a teenager. Now I am retired as a wireless/Satellite Comm Engineer and am discovering the fun of loudspeaker projects once again.

 

[Waveform Fidelity]

Mine are the SP1753AA, which should be the 2018 version, the dedicated midrange version from the R3 - just before Meta.
It is right here:

https://assets.kef.com/documents/rseries/rseries2018-white-paper.pdf

You should be able to EQ yours to be better, even though taming the break-up, is always tricky.
If it was me, I would just buy a set of R3 or R3 Meta and pull the driver.... but that's just me

I think the bolt pattern maybe different and/or depth of the basket. Also the flare ring changed. I have LR 48dB crossover set at 2.8kHz, so no chance of exciting the breakup resonance, but I did think about upgrading to the later UNIQ or UniQ meta. I have not seen anything from either KEF nor the Audio industry showing the reflections with and without absorber on the tweeter - which is strange as this is the whole purpose of the meta material absorber. The impulse response on the SP1632 is highly oscillatory, mainly (I think) because using a Wideband MLS sequence excites a strong resonance at 24kHz, but it could be reflections to/from the dome and back of tweeter enclosure. I guess I need to do some detective work to see if it is reflections. If it is, and they are audiable then, if the R3 midrange fits, I might do the retrofit. The LS50 apparently is a replacement, but I am sure that optimized also as a woofer, whereas the R series midrange are optimized for the midrange.

Thoughts?

 

[Waveform Fidelity]

Woofer Equalization using FIR filter.

Just wanted to share what I have found possible regarding KEF R900 woofer equalization - constrained to 512 Taps.


A few observations:

1. Below ~100Hz FIR equalization is not useful (48kHz sampling rate) - so I only applied curve EQ starting around 120Hz and then LR -48dB/Oct acoustic roll-off (which is performing well)
2. I am attempting to fully compensate the phase shift to a constant value of +120 degrees for all drivers - you can see the limitation of 512 taps - still, the phase shift is held within +/-30 degree variation within the passband which is better than I anticpated. The OEM response phase wraps +/-180 4+ times over the same bandwidth ~ 1440 degrees, so there is a lot of phase compensation.
3. The amplitude response is held within +/-1dB from 70-300Hz with intended crossover landing exactly on the 400Hz cross-over point at -6dB point. This chops off signifcant response of the woofer that has strong resonant peaks in the kHz range.
4. The sound?

With the full range speaker (all flat 120 degree phase shift up to 15kHz) - the speaker sounds more neutral that the OEM crossover (which has a brightness hump ~ 1-3 kHz), but the bass has a noticibly stronger impulsive sound - drums especially sound more dynamic with sharper transients. This is a common observation with linear phase bass speakers and was discussed by Laurie Fincham of KEF in the early 80's in an AES paper. I am very pleased with this result. I keep the ports plugged to maintain good transient response and the expense of a higher cut-off frequency of 48.5Hz.

Hope this information is useful to others contemplating DSP equalization using FIR filters, with limited number of taps. The latency is 7.5msec.

 

[Digital_Thor]

Yes the LS50 is a bass/midrange.
I found that it didn't have bass to my liking, simply by being way to tiny for my room. The R3 surprised me, but still no near the dynamics of bigger woofers - IMO.