Go Back   Home > Forums > >
Home Forums Rules Articles diyAudio Store Blogs Gallery Wiki Register Donations FAQ Calendar Search Today's Posts Mark Forums Read

Multi-Way Conventional loudspeakers with crossovers

Crossover instead of Kube on KEF?
Crossover instead of Kube on KEF?
Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Thread Tools Search this Thread
Old 3rd March 2009, 11:56 PM   #11
mwaters10 is offline mwaters10  United Kingdom
diyAudio Member
Join Date: Nov 2004
Location: UK
Attached is the LF schematic for the Kef 107
Now I have the inductor values I can work out the notch filters.
It seems odd to have 2 notch filters where the LF roll off is 150Hz.
My colleague told me that the lower frequency notch may be to filter rumble.
You'd expect this in a phono stage, but not in the crossover.
Perhaps I'm missing the purpose of C4/L1/R2 and C5/L2/R2 ( as in my drawn out circuit, not the Kef original )

I think the bass from the 107s are spot on with the Kube in circuit.
However, the Kube makes the sound less refined, but you definately notice a loss in bass balance without it.
Attached Images
File Type: jpg 107-lf.jpg (77.8 KB, 644 views)
  Reply With Quote
Old 4th March 2009, 12:04 AM   #12
mwaters10 is offline mwaters10  United Kingdom
diyAudio Member
Join Date: Nov 2004
Location: UK
Sorry, here is the original Kef schematic
Attached Files
File Type: pdf 107-crossover-lf.pdf (97.5 KB, 466 views)
  Reply With Quote
Old 4th March 2009, 10:21 AM   #13
sreten is offline sreten  United Kingdom
Join Date: Nov 2003
Location: Brighton UK

Most of the complexity of that crossover is due to impedance
compensation, the 107 appears nearly a constant resistive
load to the amplifier. Arguably it is total overkill and not a
good idea, IMO maximising the impedance whilst minimising
to a degree the phase angles is the most amplifier friendly.

FWIW (seeing this is diy forum) it is quite possible to build
a 2nd order high pass filter with a Q ~ = 2 (max gain 6dB)
around the feedback loop of the power amplifier, assuming
the amplifier has the standard "op-amp" topology.

  Reply With Quote
Old 4th March 2009, 12:39 PM   #14
mwaters10 is offline mwaters10  United Kingdom
diyAudio Member
Join Date: Nov 2004
Location: UK
Thanks for responding
I'm using 805 valves from Meixing, so no op-amps. On first examination it appears to be a 2nd order ( 12db ) crossver, with impedance stabilisation circuit, and notch filters.
Isn't the LF rolloff set by L3 ( - L5 ) and C2 ?
L5 in parallel is a small inductor coil ( subtracting from the series inductor ). I maybe wrong about this. I was wondering if Kef used the small inductor L5, to reduce the inductance on L3, just to use a stock value that they had an abundance of. Who knows their thinking ?

I know that the Kef 107 crossover point is 150Hz.

I think that R1/C3/R2 acts as impedance stabilsation circuit,
and C4/L1/R2 and C5/L2/R2 are notch filters.

I'm thinking of ditching the impedance stabilisation circuit all together, and seeing how the amp reponds to the Kef B250 behaving in a non-linear response. I guess Kef don't have this luxury, they won't come to my house and see how my amp behaves when they're designing their impedance stabilisation circuit.
  Reply With Quote
Old 5th March 2009, 04:45 PM   #15
mwaters10 is offline mwaters10  United Kingdom
diyAudio Member
Join Date: Nov 2004
Location: UK
I think I understand what's going on in this crossover now.
Here is what Stereophile wrote about the 107
That is the question, and the 107 answers in the affirmative. The KUBE cuts through several vexing problems that have traditionally compromised subwoofer designs. Even my six-year-old, Dahlia, associates bass sensitivity and extension with large enclosures. And, indeed, the laws of physics conspire to make it so, as Richard Small (now working as Head of Research at KEF in the UK) and A.N. Thiele so aptly demonstrated almost 20 years ago. I'll get a bit didactic and elaborate:

In order to achieve flat LF response to, say, 20Hz, you have to start with a woofer whose free-air resonance is no higher than 20Hz, and preferably even a bit lower. Remember that below the driver's resonant frequency, the output will droop as the driver is no longer mass-controlled and cannot maintain a constant acceleration as frequency decreases. Next, it turns out that a large woofer is much more adept at pressurizing a given-size box than a small woofer, and a very large air volume is needed to prevent the air stiffness of the enclosure from significantly raising the system's resonant frequency. Reduce the volume and low bass can still be squeezed out of the design, but only at the expense of severely curtailing sensitivity.

If you want low bass and a usable sensitivity, therefore, what you end up with is a very large box that is not only expensive to construct and ship, but also, most importantly, almost impossible to make rigid enough to minimize colorful cabinet radiations. Have you ever seen the WAMM subwoofer? Do you know how much carpenters get paid these days? Now you know why that system costs over $40k. So it appears that, at least for realistically sized production-line speakers, the hopes of realizing a 20Hz response using a natural alignment are doomed.
So I would expect to see a resonance notch filter somehere around
20Hz, right ?

Not all all. Either Stereophile didn't get the chance to look at this crossover, or they were fed some BS by the Kef Marketing department. This reminds me of their so called speaker/crossover matching and having your serial numbers on file .. ;-)
The service people at Kef, know this is BS too. They never asked me for the serial numbers when one of my drivers blew, and all the crossover values are the same.
Hand picked capacitor values, possibly, but I think not ..

L3 / L5 and C2 set the LF-roll off at 122Hz ( 6mH + 1mH and 240uF )
I worked out that the LF crossover with 7mH and impedance of 4 ohm for 2nd order Butterworth would make C2 at 220uF.
This would have made the crossover 128Hz
Maybe Kef had a surplus of 240uF values in stock, and I suspect that the use of the 1mH L5, is to get the desired crossover cut-off Fc.
Actually it makes little sense, why not just use a 7mH coil, as they have for L2.

The MF filter of 200uF and 6.4mH sets the high pass at 150Hz which goes off to the MF/HF unit.

L1/C4/R2 ( 33mH / 480uF / 4 ohm ) is a notch with resonance peak
at 38Hz.
I'm confident this is the free air resonance of the Kef B250. I can't find the specification for this unit anywhere .. HINT .. HINT

Anyway, it's not 20Hz ;-)

L2/C5/R2 ( 7mH / 200uF / 4 ohm ) is a notch at 128Hz

R1/C2/R2 is the conjugate loading or impedance equilisation as commonly known.

Having dug out one of my old audio design books and read up on amplifier / speaker stability, it started to makes sense. The second notch is at the crossover point.
I agree sretan, this is overkill.
I like that they are going for a flat response, but with a roll-off at 128Hz, you have to wonder if the second notch is a step too far
I think that Kef have tried to make the speaker load seen by the amplifier as purely resistive. As you say, this has it's own problems with phase shift etc..
I wonder how many other crossover designers use a second notch at the roll-off point
Here is an updated drawing I made. The Kef original is not the best quality
Attached Images
File Type: jpg kef107-lf2.jpg (90.3 KB, 599 views)
  Reply With Quote
Old 6th March 2009, 10:26 PM   #16
AndrewJ is offline AndrewJ
diyAudio Member
Join Date: Sep 2001
Rather than the guessing that is going on here, perhaps I can fill you in on exactly how this network was designed.
In refering to component functions I will use the component nomenclature used in mwaters10 latest diagram, as this is easier to read than the supplied KEF drawing.

First, the impedance compensation (conjugation) is for the purpose of flattening the speakers complex impedance curve before adding the low pass filter network. The bass system utilizes a coupled cavity bandpass configuration, the by-product of which is an impedance curve that reacts very badly when trying to drive from a passive low pass filter. The upper impedance peak of the system is very close to the desired xover frequency. Without impedance compensation it is next to impossible to design a low pass filter of the required caracteristics without suffering big response irregularities and/or severe dips in the input impedance.
Therefore full conjugation of the equivalent circuit of the loudspeaker is employed. Since a coupled cavity produces a double tuned impedance characteristic, similar to a reflex system, then two branches are used to compensate this, components L1 - C4 and components L2 - C5. Additionally C3 - R1 compensate the voice coil inductance, and R2 the voice coil resistance. This configuration is an exact dual, placed in parallel, to the equivalent circuit.
So, it's main function is not to make an easier load to the amplifier, but to make an easier load for the network.
Under no circumstances would I advise that you ditch the conjugation components.
Now, L3, L5, C2. These form a low pass filter, but a notching low pass filter (C2 and L5 producing the notch at around 330Hz). Because of the long length of the port in the coupled cavity, there was peak in the response of the vent (the vent being the sole output from the system) just as the acoustic response was rolling off. The notching low pass filter compensated for this peak, so that overall the filter plus speaker, rolled off as a fourth order acoustic response.

As for driver and network component matching, I cannot attest to what is done these days, but in my time we measured every driver that was built and sorted them into tight tolerenced sensitivity bands.Then we would draw up a matching table for how to match bass, midrange and tweeters together from different sensitivity bands to insure the best overall system responses.
The computers that were used for design were the same as for production measurement, and were located in the same room. At any time I could glance over and see how close the daily production was to the reference driver, and I could exctract that data and input it to my design programs if needed.
Records were kept of all drivers and their sensitivity tolerence and the serial numbers of which systems they went into. Whether service continued to use this info for subsequent replacement I cannot say, but the information was certainly available.
For the xovers, all inductors were wound in house. Design engineering would specify how many turns were needed to wing the inductors into specific 1% tolerence bands around the nominal value, so we made inductors over a +/-10% tolerence band.
We then measured and sorted the capacitors and resistors into 1% bands from initial incoming of 10%.
This then gave us a stock of components all sorted into 1% tolerence bands. We would then look at the network design, all of which was done with computer optimisation algorithms written in-house, and perform a component sensitivity analysis. We would then deliberately perturb, one by one, each capacitor by 1% at a time, re-optimise all the network components and draw up a compensation table that told us how to pick the components values to match together to get back to nominal response for the network.
This allowed us to use components from all the tolerence bins, avoiding the very high cost of buying tight tolerence components.
These are the tables you see all around the network on the KEF

Next, the KUBE itself.
Yes, the system was designed with the KUBE as an integral part of the overall bass alignement. You will definitely get an odd bass response if you don't use it.
Yes it is full of opamps, and I know that these don't measure up to modern types.
If you are into doing so much modifications, the best route would be to convert the system to bi-amplification and then insert the KUBE only in the bass amplifier chain. This would avoid any of the concerns of the sound of the KUBE throughout the mid and treble at least.
Unfortunately I don't have samples of these KUBES any longer, otherwise I would measure them and offer a suggestion of how to build a better version.

Hope this gives you all a better idea of what went into the design of these systems.


  Reply With Quote
Old 7th March 2009, 02:29 AM   #17
mwaters10 is offline mwaters10  United Kingdom
diyAudio Member
Join Date: Nov 2004
Location: UK
Fair play to you AndrewJ,
It was the first time I looked at this crossover , I think my guessing wasn't miles away , but you are absolutely right, the reasons for the upper impedance peak of the coupled cavity bandapss design was not fully understood, nor the notch at 330Hz.
I had already realised by the time of my last mail that the impedance stabilisation circuit was necessary and would not work well without it ( forgot to mention) .
Thanks for taking the time to look into this.
Your insights into the Kef design and their processes is very interesting and your explanation of the crossover is very helpful
In many ways the 107 design works well, but and as you rightly say, the bi-amp option and upgrade of the Kube makes perfect sense.
As for Kefs' matching, I'm being cynical, but most people would want their speakers matched within a certain tolerance. I'd expect it to be part of normal process, who would think an unmatched pair of speakers was a good idea ?
I have to admit that Kef have always been a good company to deal with, and their service department are still pretty good, in comparison with most modern companies and their annoying automated voice mail systems.
One thing that does surprise me is the absence of resistors ( L-pad ) in the MF crossover.
I expected the B110 MF unit to be a higher sensitivity than the B250.
Do you think is it possible that the crossover for the B110 and T33 is
matched for sensitivity only when the Kube is in circuit ?
Do you think the MF/HF units sensitivity could benefit from being attenuated when the Kube is not in circuit ?
The MF/HF drawing is coming next ..
  Reply With Quote
Old 7th May 2009, 11:41 PM   #18
NATDBERG is offline NATDBERG  United Kingdom
diyAudio Member
Join Date: Nov 2006
I've a pair too, Mk 1s, and wanted to convert to bi-wire for the same reason. Going by your drawing, would it be a case of separating the MF/HF connections and placing a copy of the L3, L5 and C2 filter between MF/HF crossover and inputs?

An alternative perhaps could be (if you're not adverse to digital EQ in your system) to derive an EQ curve from the Kube, with combinations of settings, with a frequency sweep and program a good quality digital EQ. Meridian created a digital EQ for the Kef 107 in their 601 digital pre-amp and so must have the original Kube response curve somewhere even if Kef don't (Meridian seem to have circuit diagrams and manuals at hand for all their old products).

I think Meridian were bought by Kef around the early 90s, (is that so AndrewJ?) so I'm sure there would have been free flowing info between them.

I do have a copy of the manual for the 601 which provides an EQ curve for the emulation. I can post it up if it's helpful..

Alas, my Kefs are/were a free repair project that I've not got around to starting in 2 years. The Kube took a shock when the previous owner dropped a screwdriver into his power amp that was left open to easily replace blowing fuses.. I think the Kube saved the drivers and crossovers! The opamps have had their tops blown off..
  Reply With Quote
Old 8th May 2009, 12:12 AM   #19
NATDBERG is offline NATDBERG  United Kingdom
diyAudio Member
Join Date: Nov 2006
Originally posted by mwaters10

One thing that does surprise me is the absence of resistors ( L-pad ) in the MF crossover.
I expected the B110 MF unit to be a higher sensitivity than the B250.
Do you think is it possible that the crossover for the B110 and T33 is
matched for sensitivity only when the Kube is in circuit ?
Do you think the MF/HF units sensitivity could benefit from being attenuated when the Kube is not in circuit ?
The MF/HF drawing is coming next ..
The diagram in the 601 manual shows a mid-band shelf drop that isn't there in the Kube 200 emulation. I'll post it up..

Obviously it is based on the 107/2 Kube with LF HF Contours rather than the Extension, Q-Factor and Contour controls of the previous version. Does anyone know if the resulting curves are much different?
Attached Images
File Type: jpg 601-kube-107-emulation.jpg (29.2 KB, 501 views)
  Reply With Quote
Old 8th May 2009, 01:57 AM   #20
NATDBERG is offline NATDBERG  United Kingdom
diyAudio Member
Join Date: Nov 2006
To correct myself, looking at the Kef diagram this time, all I need to do is cut into the tracks before C7, C8 and L5 (using the Kef nomenclature - and not named on mwaters10's diagram) and have the MF/HF input enter there?

Or.. some free thinking here.. could a line level filter be applied to the signal to direct low fequencies through the kube and the rest via delay/phase matching circuit to be re-combined before the power amp? This would save the expense of bi-amping.
  Reply With Quote


Crossover instead of Kube on KEF?Hide this!Advertise here!
Thread Tools Search this Thread
Search this Thread:

Advanced Search

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
Kef R105/3 Kube equaliser schematic jez Multi-Way 89 30th August 2018 09:56 PM
Kef 105.3 / Kube 200 ANTHONY2181 Multi-Way 7 21st January 2013 12:29 PM
KEF KUBE with 107/2 wiring setup terryd2053 Multi-Way 2 10th June 2008 11:25 AM
KUBE for KEF 107/2 buril2000 Everything Else 0 4th December 2005 12:00 PM

New To Site? Need Help?

All times are GMT. The time now is 05:14 PM.

Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 15.79%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2018 DragonByte Technologies Ltd.
Copyright ©1999-2018 diyAudio