Humble Homemade Hifi - Soup Ceramique

Soup Ceramique

Soup Ceramique – The proof is in the eating

Loudspeaker designing is like cooking. From several different ingredients you hope to create a culinary masterpiece. The art is in choosing the correct ingredients and blending them together in a way that you find pleasing. Give ten different people the same ingredients and you will get ten different recipes. Here is my attempt at creating a musical masterpiece.

The tweeter

The Thiel & Partner C2-12/6 is a Ferro fluid filled tweeter with a 25 mm lightweight concave ceramic dome. Designed for 3-way systems or small, low-output 2-way designs. Very high resolution and very good dispersion up to 42 kHz. Low resonance frequency allows first order filtering and a crossover frequency as low as 2 kHz. The ultra hard ceramic dome material moves like a piston well above the audible frequency band and the high internal sound velocity features very low distortion and virtually no coloration.

The mid-woofer

The Thiel & Partner C2-89/T6 is a bass-midrange driver with a 90 mm light weight concave ceramic dome, high loss rubber surround and titanium voice coil former. Made without ears for high excursion. Designed for small 2-way systems as bass-midrange driver or high quality 3-way systems as low midrange. Very high resolution and very good dispersion up to 5 kHz. Low resonance frequency allows first order filtering and a crossover frequency as high as 4000 Hz. The ultra hard ceramic dome material moves like a piston well above the audible frequency band and the high internal sound velocity features very low distortion and virtually no coloration.

The woofer

The Thiel & Partner C2-220/T6 is a bass-midrange driver with 172 mm light weight concave ceramic dome, high loss rubber surround and titanium voice coil former. Made without ears for high excursion. Designed for large 2-way systems as bass-midrange driver or high quality 3-way systems as bass unit. Very well suited for subwoofer designs due to low resonance frequency. The ultra hard ceramic dome material moves without flexing losses like a piston from 20 Hz to 2.5 kHz. It features very low distortion and virtually no coloration.

The cabinet inside and out

The raw cabinets were constructed by Gydotron and made of thick 30mm mdf; the front baffle is double layer with 60mm total thickness. To create the angled baffle the four corners were machined to fit separate panels, if you don’t have the possibility to do this, the same visual effect can be done by making the four corners even thicker on the inside and then cut the angles away as can be seen during my factory tour. The outside of the cabinet can be finished to your fancy – I veneered my cabinets myself with a nice light coloured maple and 6 layers of clear water based, polyurethane reinforced varnish. Then a final layer of bee’s wax polished to a nice deep but not too glossy shine. Internally there are several bracings to give extra stability to the cabinet, the separate midrange enclosure also adds to this extra strength. All internal walls (including the midrange enclosure) are covered with a double layer of bitumen (2x4mm) to add extra mass and control panel vibrations. A layer of wedge-moulded foam covers the inside of the woofer compartment to minimize internal reflections. The midrange enclosure is filled completely with sheep’s wool. The bass volume is tuned to taste with bonded acetate fibre (BAF) filling the top section behind the midrange enclosure and a little directly behind the woofer. The area around the reflex port is left clear.

The 30mm thick black painted base-plate acts as a visual plinth to give the speaker a more graceful appearance and also works as a rear vent for the down-firing reflex port. The Soup is a ported design rather than a sealed enclosure. It is designed to have the transient response of a sealed box with the deeper bass extension of a ported design. This is done by tuning the port near driver resonance and filling the enclosure with damping material. The down-firing port is located on the cabinet bottom next to the input terminals. Single pair of input terminals is provided on the enclosure bottom. This means that the Soup cannot be bi-wired, and that the loudspeaker must be tipped on its side to connect loudspeaker cables. Three sharply pointed cones are provided - two for the rear and one for the front of each loudspeaker. The loudspeakers sound their best with their contoured grilles off; the Soup was designed to perform optimally with these black grilles off. The cabinet features a faceted front baffle; a shape that further reduces diffraction, allows for a thicker baffle, and contributes to wider dispersion. The entire faceted structure is a solid block made from two layers of mdf. The resulting structure keeps the cabinet weight reasonable. The cabinet's swept-back profile better time-aligns the drivers, and makes the speaker more stable by shifting the massive front baffle's centre of gravity rearward.

The cabinet measures 1025x260x405mm with an internal volume of 50 litres for the woofer tuned to about 27Hz it gives a –3dB point around 30Hz. The closed mid-woofer compartment is about 4 litres in volume and tunes the driver to a nice Q of 0,5. This compartment is filled with natural sheep’s wool. All walls are made of 30mm MDF except for the double layer baffle of 60mm. The inside of the cabinet walls is lined with 2x4mm thick bitumen and Pritex wedge moulded foam to minimise cabinet vibrations and standing waves. The reflex port is kept clear. Depending on the positioning of the speakers in the room and personal taste this fibre can be rolled up densely or loosely to give a tighter and dryer or a fuller and warmer sounding bass. All internal wiring is Van Den Hul The Breeze Hybrid.

Crossover and listening

This masterpiece took a lot of preparing. Starting out as a 2.5-way concept I went for a true 3-way design after finding the three-way to seem more forward and dynamic in the oh so critical midrange. The 3-way crossover saw many variations on a theme: shifting crossover points up and down, playing around with various combinations of slope steepness, balancing bass/mid/tweeter levels, etc. But after months of tuning the crossover I still wasn’t satisfied, the sound didn’t “click” into place. I could describe it with words like “you can hear right into the recording” or “amazing bass from only an eight inch woofer” or “ listen to the direction in which the brushes are stroking the snare-drum”. But I never once had the feeling I was listening to music – it was too perfect. Like Paris Hilton: technically perfect and great to look at but misses the texture, warmth and depth of the real thing.

So I decided to scrap the 3-way crossover and start all over again with the most minimal 2.5-way I could get away with. The woofer got the most “cleaning” done: I ditched the Zobel network, I ditched the damped 2nd order leg, I ditched the LCR-network that flattened the impedance peak in the bass, and ditched the parallel notch filter to take out the cone break-up at 3400Hz. The mid-bass sounded better without the parallel notch filter, but the treble got messed up if I left it out. So I took a different approach to cancel out the cone break-up peak: parallel to the main woofer inductor I placed a RC-network that, together with the inductor, forms a resonance circuit centred at the cone break-up frequency. Above this frequency the network resonates so to cutout the high frequency mess a relatively small capacitor is placed parallel to the woofer. The overall result is a more open sounding and less “twisted” midrange with more “one-ness” with the other drivers compared to the classic parallel notch filter approach. All that was left besides the resonance filter was a single high-quality air-core inductor that in combination with the woofer forms the baffle-step compensation for the midwoofer (opposite to the standard approach in which the upper midrange and treble are squashed to match the deepest possible bass; extra bass is added by means of a second woofer to meet the efficiency of the upper midrange and treble – like I did with the Proteus speaker). The midwoofer / tweeter network is a classic a-symmetric order series crossover that I have found to be successful with most 2-way driver combinations.

This “crossover-on-a-diet” led to a far more open, forward and articulate sounding speaker with better efficiency and more coherency than the 3-way approach. The end result can be described as an image with a complete “one-ness” standing on a firm foundation. The soundstage is large in all directions, especially well layered from front to back and left to right. All details are nicely highlighted without being over etched, the top-end is slightly rolled-off (a characteristic of the tweeter) but at no time did I have the feeling of missing clarity. This loudspeaker has a unique ability to reach deep down to the core of musical expression in a way that makes music listening a deeper and more gratifying experience. The sound has a coherence, clarity and resolution that are exceptional! Happy eating!

Crossover components 2004-version

L1 = 0,68 mH Intertechnik Tritec inductor 3,5 mm wire, R = 0,10 ohms (tolerance max 2%)

L2 = 0,56 mH Intertechnik Tritec inductor 3,5 mm wire, R = 0,11 ohms (tolerance max 2%)

L3 = 3,3 mH Intertechnik Tritec inductor 3,5 mm wire, R = 0,33 ohms (tolerance max 2%)

C1 = 10uF Mundorf M-Cap Supreme Silver/Oil capacitor (tolerance max 2%)

C2 = 0,01uF Vishay MKP1837 polypropylene foil capacitor (tolerance max 1%)

C3 = 4,7uF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance max 2%)

C4 = 0,01uF Vishay MKP1837 polypropylene foil capacitor (tolerance max 1%)

C5 = 3,3uF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance max 2%)

C6 = 0,68uF MKP polypropylene foil capacitor (tolerance max 5%)

C7 = 5,6uF MKP polypropylene foil capacitor (tolerance max 5%)

R1 = 15 ohms, 10 watts metal film resistor (tolerance 2%)

R2 = 1,8 ohms, 10 watts carbon film resistor (tolerance 2%)

R3 = 10 ohms, 10 watts carbon film resistor (tolerance 2%)

R4 = 10 ohms, 10 watts metal film resistor (tolerance 2%)

Update June 2007

After a few years “in the field” and many sets built, it was time for a revision of the crossover. The new crossover design leads to an even clearer image with a slightly less rounded-off top end and an even smoother approach to the music.

Crossover components 2007 version

L1 = 0,68 mH Copper-foil 12AWG or Tritec inductor R = 0,15 ohms (tolerance max 2%)

L2 = 0,56 mH Copper-foil 12AWG or Tritec inductor R = 0,13 ohms (tolerance max 2%)

L3 = 3,9 mH Copper-foil 10AWG or Tritec inductor R = 0,30 ohms (tolerance max 2%)

L4 = 0,56 mH air-core inductor R = 0,61 ohms (tolerance max 5%)

C1 = 10uF Mundorf M-Cap Supreme Silver/Oil capacitor (tolerance max 2%)

C2 = 0,01uF Vishay MKP1837 polypropylene foil capacitor (tolerance max 1%)

C3 = 6,8uF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance max 2%)

C4 = 0,01uF Vishay MKP1837 polypropylene foil capacitor (tolerance max 1%)

C5 = 5,6uF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance max 2%)

C6 = 0,56uF MKP polypropylene foil capacitor (tolerance max 5%)

C7 = 8,0uF (4,7+3,3uF) MKP polypropylene foil capacitor (tolerance max 5%)

C8 = 0,82uF MKP polypropylene foil capacitor (tolerance max 5%)

C9 = 15uF MKP polypropylene foil capacitor (tolerance max 5%)

C10 = 18uF (10+4,7+3,3uF) MKP polypropylene foil capacitor (tolerance max 5%)

C11 = 1,2uF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance max 5%)

R1 = 1,5 ohms, 10 watts carbon film resistor (tolerance 1%)

R2 = 15 ohms, 10 watts carbon film resistor (tolerance 1%)

R3 = 10 ohms, 10 watts metal film resistor (tolerance 2%)

R4 = 27 ohms, 10 watts metal film resistor (tolerance 2%)

R5 = 12 ohms, 10 watts metal film resistor (tolerance 2%)

R6 = 6,8 ohms, 10 watts metal film resistor (tolerance 2%)

R7 = 15 ohms, 10 watts carbon film resistor (tolerance 1%)

Measurements

These measurements were taken at 2 metres microphone distance with both speakers playing. The speakers were also 2 metres apart therefore forming an equal sided triangle between the microphone and the two speakers. I believe this to give a more realistic presentation of the “character” of the speaker. Such an approach can never provide "absolute" data about a speaker's performance, because it also incorporates a few minor reflections – even with the gated MLS measuring technique I use you can’t remove the room completely. However, it can be argued that this technique is actually more meaningful in terms of the actual listening experience. I've measured several speakers this way, and the results invariably show very good correlation between the measured and the subjectively perceived frequency balance of a specific loudspeaker. While the balance is by no means the only important factor in defining a speaker's performance, it is a very significant one. The in-room balance of a given speaker is like a fingerprint, distinct to that loudspeaker, that room, and the siting therein, and always says much about the overall character of the sonic presentation. Without defining the in-room balance, it's very easy to be "fooled" by minor anomalies that can make the speaker sound artificially attractive (or unattractive). Once the room response is properly defined, it becomes much easier to dig below the surface to discover the underlying qualities – it takes a bit of effort to determine the best measuring position that lets you “see” the speaker and not the room, but its worth it!

Overall output level 200-20.000Hz and individual driver responses; horizontal division 5dB. Waterfall spectrum horizontal division 2dB. The drop above 15kHz is due to the limitations of the soundcard.

Overall an extremely smooth response with an energy balance focussed toward the midrange. Each driver roll-off is very well controlled, the woofer showing no resonance peak what so ever and the midwoofer showing a minor “ripple” in the output level and a little “mess” in the waterfall – I could get smooth it out with a LCR-network so the measurements looked more pretty, but it sounded better without. There is a large overlap between the woofer and midwoofer output with its “blend-out frequency” at 600Hz. The midwoofer has no high-pass and the crossover point to the tweeter is at about 2400Hz. The top end roll-off is due to measuring at this distance, when the microphone is placed at the standard 1 metre (or less) distance it is flat to 15kHz, the drop above 15kHz is due to the limitations of my soundcard. Sensitivity is specified at about 88 dB at 2.83 V and 1 meter. The waterfall plot is very “quick” considering the fact that the measuring distance is 2 metres (compare it to The Monitor which is measured under exactly the same conditions).

Tony Gee, The Netherlands, October 2004, January 2005, updated May 2005, updated June 2007