KEF Q200c

Years ago I built my own centre speaker using two 5″ full range drivers inversely mounted together in an isobaric chamber, and although it sounded good, it was too directional. So after some research, I chose a Kef Q200c to replace it because apparently the sound stage is excellent, although how that works with one speaker, I’m not sure 😉 I already have a pair of the original Kef Eggs (HTS2001) and they really do have a wide, even dispersion. However, having investigated the speaker thoroughly, and spent hundreds of hours listening to it, I do not recommend it because it resonates and either needs a better enclosure, major modifications, or parametric equalisation.

kef-q200c

I am a bi-amping fan, and I’m not the only one. The Q200c is a three-way but I only have two amp channels available (a modified Tripath Ta2020 powered by a Paul Hynes PR3 regulator) so I’m planning to use a conventional first order speaker crossover for the mid and tweeter, and then bi-amp that and the woofer by trying various crossovers to compare them (in the end I chose an electronic 4th order Linkwitz-Riley but that needed modifying too). One of the reasons for choosing the speaker was that apparently KEF designed these Uni-Q drivers for first order crossovers so this should be possible. This isn’t true for all of their co-axial drivers – the Eggs have a 2nd order with a zobel, for example.

What’s in the box?

The cabinet seems well made and braced, although the interior fibre board is not the best. After removing the terminal back plate, the crossover is visible behind the passive radiator, which is the cone nearest to the KEF logo. It can’t be removed from the rear – it’s too big and hard to get to. It’s even hard to photograph through the opening.

1070540

KEF have a unique way of (dis)connecting the drivers’ crossovers to allow bi-amping – a hollow cylinder of metal can be screwed in/out onto wide copper traces on a PCB to make a connection across the traces. Unconnected, by screwing both blocks out, it’s fine to use a bridge tied load (BTL) amp for bi-amping because the circuits are independent. This bi-amping assembly is a bolt, and it has a circlip on the end to stop people from screwing it all the way out. The bolt pushes a top-hat-shaped metal cylinder onto the copper traces. The top hat is spring loaded to ensure that when screwed out, the top hat moves and stays out.

Accessing/removing the xo requires a fingernail (or better tool) under the plastic ring surrounding the passive radiator. This lifts out, rather awkwardly and not that easily, but like a speaker grille. Six screws later, the radiator is out and after removing wadding, there is a large crossover with 3 inductors, including 2 iron core, 6 resistors, and 4 capacitors, including 3 non-polar electrolytic. No wonder it sounded rough when I first hooked it up – it took several hours for the caps to break in. Actually, the drivers themselves are still loosening up even after 100+ hours, producing more bass.

The co-axial tweeter-mid is sealed in its own “cube” enclosure, which is roughly 2.5 litres, less the driver’s volume, about 0.5l, so about 2 litres. The enclosure dimensions are similar in each direction, hence, a cube. This is bad for resonance – quarter wave resonance is easily heard using frequency generation in WinISD. It tests with output peaks at 350Hz and 430Hz – pretty much what the dimensions suggest [f = 340/(4x)] – and only slightly lower than KEF’s stated 3rd order crossover point of 500Hz. It is damped only with polyester wadding so after checking out the cube, I added some “egg box” foam to improve damping, but I’m still very disappointed by this cube. The volume is fine according to Win ISD software, but the shape is an issue because I can occasionally hear those resonant notes in music and it’s annoying when the hi-fi distracts from enjoying the music. So another aim is to remove them by using a higher order crossover.

p1070571

The woofer, centre-vented and marked SP1589, 4 ohm, is on the right, and pushes air around behind the cube to the passive radiator/ABR on the left, marked SP1821, 41g. The woofer driver has a cast basket, 2-inch voice coil and paper cone that has the outwardly visible aluminium cone adhered to it at its outer edge. It looks like the drivers share some common components.

p1070569

The mid-tweeter is marked SP1587.2 – perhaps the same as the SP1587 driver in the Q100, for which there is a lot of info:

http://medleysmusings.com/kef-q100-drive-unit-testing/

http://www.zaphaudio.com/blog.html

Here are the frequency response charts for the Q100 showing a 1st order crossover around 3KHz should work very well, with the tweeter about 3dB brighter. Kef designed the Q100 with a 2.5khz crossover and 1.5dB attenuation.

kef-q100hf-fr

kef-q100lf-fr

And here is the Win ISD simulated plot for the sealed 2 litre enclosure. Since I’m going to be tweaking it, this is the Q200T. 🙂 You can download Win ISD from Facebook.

Q is estimated at 0.63 and Fsc at 110Hz. Q of 0.7 is 1.4 litres and Fsc at 123Hz.

q200t-winisd

The crossover

The woofer has a separate circuit and is a non-standard 3rd order low pass: 1.8mH + 0.5R series, 1.8R + 100uF to negative, and 0.5mH series: both are ferrite core inductors with an electrolytic capacitor so those components have to go. According to simulators, the values are somewhere between a Bessel filter Q = 1√3 and a Butterworth filter Q = 1√2.

p1070549

The mid and tweeter share another circuit with a first order high pass on the tweeter. The mid positive is tied to the tweeter negative, so it is phase inverted. The Q100 is in phase.

Tweeter => 0.5R + 4.3uF in series.

Mid => 2.2R + 100uF in series, 6R + 0.95mH to negative, and 10uF + 8.2R to negative.

Putting the second order high pass into a calculator shows it has a high Q and may ring at 482Hz. However, I’m sure the resistors and following leg prevent this.

Note the low-pass/zobel leg doesn’t match the reported Re and Le results for a pure zobel. Re is 3.09 and Le is 0.256mH resulting in a “standard” zobel of 17uF and 3.86R. By calculating power dissipation, the low-pass/zobel attenuates the mid by 1.6dB @2.5Khz, 2.5dB @4Khz, and 4.4dB @11Khz. Here’s the simulated output of the mid-range with different filters:

Q200 filters.jpg

The series resistors are sufficient to balance the drivers’ output volume: 0.5R on a 3R tweeter is -1.3dB overall.

t-pad-attenuator

As an example, this t-pad drops the output by 2dB and creates 3.5R as the impedance for crossover calculations. Here are the Q100 impedance plots from Zaph’s site. A kind fellow called Davey also shared exact numbers in these two txt files – the tweeter and the mid-woofer

kef-q100lf-imp-2kef-q100hf-imp-2

First, I’ll try a 3rd order low pass and high pass passive line level filters on the output of the pre-amp channels (a PGA2311PA that can drive 600 ohm loads) before the power amp by using 1/2 values of capacitors, e.g. 0.47uF, 0.22uF and 0.1uF, plus doubled resistors: 680R, 1k5R and 3k3R respectively for a Q of 0.40, although 0.5 is the target to avoid a dip in output. PLLXO’s are effective, sound better than a speaker crossover, and are much cheaper too. One disadvantage is that it will lower the mid’s output, but this can be resolved with tweaking the other drivers. Another is over-damping, creating a dip in output although whether this is an issue depends on the drivers. In addition, the cube and the original design mean some trial and error will be needed to get this sounding “right”. Anyway, worth a try, and there’s an excellent calculator here for 2nd order.

Another option is to modify the main amplifier, a TA2020, because that has an inverting op amp input which could be hacked to create a Sallen-Key filter. Possible but perhaps unpredictable because of the dc bias voltage present at the op amp inputs. Yet another possibility is to add a first order filter on the input to the PGA2311. And of course, I can try a mix of these – for example, 680R + 0.47uF on the output of the PGA2311, plus 20KR + 1.6nF on the input of the Ta2020 amp gives a second order with Q = 0.492.

p1070618

Finally, for sure I plan to try an active 4th order active crossover, as above, using a couple of op amps after the PGA2311. The disadvantage is one more component to add THD+N but if the op amps are good and parts are matched to avoid errors, then this is probably the best choice because it can give very predictable, and excellent results. Rod Elliott has made an excellent calculator to work out values.

Modifications

First thing was adding egg box foam inside the cube; then also on the rear side panels of the woofer/ABR cabinet to avoid a standing wave along that back section and on the back panel to limit reflections back out through the woofer/ABR. I still get resonances so damping is a work in progress.

p1070570

Then the alchemy of making a sweet sounding crossover….

Since the drivers in Q100 and Q200 might be the same, the initial plan was to copy the simple Q100 crossover but shift the point from the Q100 2500Hz to the Q200c 2800Hz. The Q100 seems to use a 0.6mH coil on the mid and a 0.68R resistor and a 3.9uF cap on the tweeter, and no zobel. Dropping those values to around 0.51mH and 3.3uF should have done it. However, this sounded awful. So I started to experiment, do some maths based on the measurements of the SP1587, experiment some more.

In case you don’t know, the maths is based on 2 equations and log scales for dB changes due to voltage. R=2πfL for the low pass inductor and R=1/(2πfC) for the high pass capacitor, and -6dB = 20log(0.5), with 0.5 being Vout/Vin. 10Log(0.5) is for power not voltage. 20log(0.7) gives -3dB. I still get confused because most online Butterworth calculators give -3dB values which will produce a peak at the crossover frequency (Q=0.7). To get Q=0.5, the factor is √2 or 1.4, however, this calculator will do -6dB.

p1070615

I made a temporary 3rd order pllxo, as above, in the amp next and hooked that up. It is as expected – not great in many ways but there is more definition. I don’t notice a dip in response at 500Hz, but that would be quite hard to hear in music – need to test that with a frequency generator. After some maths and tweaking, this sounded okay with the pllxo:

5.5R + 3.8uF in series with the tweeter.

hand-wound ~0.49mH 19awg inductor in series with the woofer.

However, I think I can still hear the resonant notes so I want to try a 4th order, and that means going active. So the next thing was to replace it with a Linkwitz-Riley 4th order. I ordered parts, plus some better caps & coils for the mid-tweeter that match the signature, especially the capacitors. I have been using Mundorf silver/oil but these seem a little aggressive on the tweeter and actually, I find I’m not that keen on the tweeter’s signature – I prefer Klipsch RF52 horns. Anyway, I also want to upgrade the inductor to a foil coil. I decided to go for Jantzen and ordered the Superes 5W resistors, Superior Z-Cap, and Cross Coil from HiFiCollective. The prices are good – similar to Taobao (mostly fakes), cheaper than Ebay (some fakes and high fees).

Next I built the 4th order LR active crossover. The caps are polypropylene “HiFi box” 220nF from my local shop, matched to 0.9%. Where I needed 440nF, I used 2 in parallel, one mounted on the underside. Matching took 15 minutes of testing about 50 caps to get 12 that are similar, which I tested again at home to pick 10. Resistors are 1k & 2k 1/4 watt 1% metal film. Not fancy, and when measured they are a 0.5% range. The op amps are AD8620 – great sounding op amp that doesn’t need high supply voltages to perform, and can easily drive the circuit loads even at low supply voltages. This is very important because they will use the same +/-5.25V power as the PGA2311 (from a Kubota regulator modified to lower noise but maintain the low output impedance). The routing isn’t that good and to get a 4th order mono I had to add two routing wires, trying to avoid paralleling signal/power lines, and crossing perpendicular instead, and over the ground plane for return signals. I added trim pots as volume controls.

p1070613

I hooked it up to test – it works, but not as well as expected: the mid bass is a little too warm so it needs modifying. Even though the 4th order is supposed to be phase coherent, I found inverting the mid-tweeter by switching positive with negative made a significant difference. I assume that this is because the first order on the mid has shifted the phase of that driver relative to the woofer. I did some testing of the mid and woofer separately using frequency generators and found there is too much output at 500Hz and this is mostly from the mid. The cursed cube resonance? Anyway, it only took -3dB in a broad band equaliser at 500hz to fix. So I need to modify the LR crossover.

On the plus side, the deep bass sounds tighter than the pllxo and the op amps stay cool and the sound is clear so I don’t feel the need to put a scope on it to check for oscillation and distortion. The woofer output is higher so I turned that down using the trim pots. However, this was a pain for fine adjustment, so these were replaced – I fitted a 20k pot as a volume control to the panel at the back of the amp case so I could adjust the bass volume without taking the case cover off. This helps with adjusting for placement and it’s set around -4dB to -5dB depending on speaker placement. The volume control mod added a hum whereas before there was no audible noise of any kind. All the circuits are floating so I tied the ground of the volume control to a star earth point on the chassis and the noiseless background returned.

So, to modify the LR active crossover, I used two calculators, the ESP one by Rod Elliot and the  Okawa-Denshi. The easiest option was reducing the 2k resistors in the high pass to 1k – just parallel another 2k. This will push that filter up from 510Hz to 720Hz but lower the Q: see below. However, if I halve C2 from 220nF to 100nF, and keep the resistance the same, the frequency will still be higher but it will not reduce Q significantly. Plus I can reverse/adjust this easily by paralleling caps. Output should drop by 3dB, and it’ll help to mask the cube resonances further. I tried and it worked exactly as calculated – so no equalisation needed and no major resonances in the mid-bass either. I didn’t fix the cube but it’s much better. There are also resonances in the lower bass but these are room interaction and not something I can correct with the crossover.

modify-lr-4th-order

When the Jantzen parts arrived I was initially very pleased to hear the effect – the tweeter was smoother and nicely balanced. I left the cap to burn in overnight and by the next evening, the sound had changed considerably and it was over-detailed, and seemingly louder. So, this was the set up I left to burn in even though upper female vocals were shrill:

Tweeter: 3.9uF Jantzen Superior Z capacitor, 1R series, 5R1 shunt 

Mid-range: 0.56mH Jantzen Cross Coil foil inductor.

However, the parts are just the right size to fit on the plastic panel after removing any unnecessary plastic risers. A bit of epoxy putty holds them securely.

P1070634.JPG

After playing around with equalisation, and MUCH trial and error, I got a sound that was very similar to the rear KEF Eggs but a little less warm and more revealing. However, the price of a better sound has been much lower efficiency, although that isn’t an issue since the maximum output is still more than ample. There is a modified 4th order electronic crossover at about 600Hz between woofer and mid-tweeter, with a volume control on the woofer: as above. There is a first order crossover between mid and tweeter:

Tweeter: 4.3uF (3.9uF Jantzen Superior Z capacitor + 4x 100nF Vishay MKP1840) + 7R in series. 

Mid-range: 0.56mH Jantzen Cross Coil foil inductor, with a zobel using 5.1uF Sonicap and 10R .

So eventually, it sounds good and I can enjoy listening! There are still issues because the frequency response varies significantly around the room – that is to be expected, and is something for another project (I’ve got my eye on a miniSharc from MiniDSP…). So the speaker is very nearly finished with only minor tweaking left, and the equalisation will be part of another project.

Finally, some more pics, click for bigger: