Progress – crossover design
30-10-2003 Initial crossover design
Tweaking the crossover.
The speakers are playing! I put together a crossover based purely on simulations done with frequency curves traced from the official datasheets. These will be different to the ones of the drivers in the actual enclosure but I have found in the past that it can be a good starting point to work from. At least this way it excludes any abnormalities that could occur from measurements later on. Also it forms a reliable data set to compare the real-life measurements with. All drivers are connected in phase except for the midrange unit.
Lets start with the Excel woofer: I decided it would need an LCR to flatten the impedance peak at fc so that the filter function would work properly, an LCR to cut out the cone break-up just above 4kHz and then the actual low pass network being a damped 2nd-order network consisting of 8,2mH and 47mF (the capacitor is damped with a 1 Wresistor). I wanted the subwoofer section of the crossover to be a separate parallel stage so that it would give the possibility to use an active bass module later on without having to redesign the whole network.
Then we move up to the 3-way series network: A linked series network (the difference between a linked and a cascaded series network can be found on Andy Gradd’s site). It is a mix of 1st and 2nd order networks with a Zobel only on the mid-woofer. The tweeter has an L-pad to lower its output level. I added a small inductor in series with the mid-range unit to tailor its top-end response a little.
Sound: not bad but also not good. The overall response is reasonably smooth but it doesn’t grip me, nice, but one would expect more from such expensive drivers. Also the balance between everything above 400Hz and everything below 400Hz isn’t good. The sims showed a nice well-balanced output level but measurements showed a shelf in the response below 400Hz. Moving the microphone and the speaker around didn’t help so it wasn’t due to the measurement set-up. Also listening sessions proved an over detailed, over clear and lean sounding speaker. This meant the whole range above 400Hz must be lowered. Spatiality was very good, maybe to do with the open-baffle mid-range.
Crossover 01 based purely on simulations.
‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗
So I lowered the whole range above 400Hz. The resistor parallel to the tweeter is about half going from 15 ohms to 6,8 W. An L-pad was added to the midrange drive also consisting of 1,8 and 6,8 W. Besides lowering the output level of both drivers the L-pads also form a type of impedance correction networks lowering the peaks at fs and the induction rise due to the voice-coils. I played around a bit with the crossover point between tweeter and midrange changing them both from 2nd order to 1st order. This greatly improved imaging and spatiality. To create a flat response again the mid-woofer needed to run a little higher so its series inductor was lowered from 1,5mH to 1,0mH. The crossover point of the sub-woofer needed to be lowered to stop a bump occurring around the 100-200Hz region, it now has a 2nd order low pass of 15mH and 100mF. Furthermore the LCR to cut out the cone break-up just above 4kHz has been swapped by a CR-network (33 W/ 0,18mF) parallel to the large inductor. This also cutouts the peak at 4kHz without altering the response above f-peak. A standard LCR-network will just cut-off everything above f-peak. Another small advantage is that is uses one component less and the capacitor is a lot smaller (cheaper). I know this is a cost-no-object project but the principle works for any speaker.
Sound: Overall balance is a lot better, the music now has “body”. As stated imaging and spatiality are greatly improved and the midrange is “closer” and “softer”, you can hear more into the recording. Bass is deeper and stronger.
Crossover 02 optimised after measurements,
simulations and listening.
‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗
Interested in tuning the bass I looked what effect the sub-woofer network had on the Excel woofers response. All diagrams’ horizontal range 20-20.000Hz, vertical range 10-90dB with 5dB divisions.
Subwoofer response for crossover version 01. The same but with 15mH / 100mF.
Swapped the LCR for a CR-network. 15mH / 100mF without any network to cut out the cone break-up just above 4kHz.
Seeing as the low-pass function was set at a low 120Hz now the CR-network or LCR network didn’t seem much use. So I decided to leave it out. Next step was to give the lowest frequencies a little boost; this can be done by leaving out the LCR (15 W / 470mF / 22mH), which flattens the impedance peak at fc. Both diagram’s horizontal range 20-200Hz, vertical range 50-100dB with 5dB divisions. Result +2dB’s in the 50-70Hz range, this should give a fraction deeper sounding bass, but also there didn’t seem much reason to keep the network, so I left it out. The resulting low pass network for the sub-woofer is now just a simple 2nd order network without any corrections.
Now I was starting to get somewhere I did the next stage mainly by ear, remember I have ears and a computer doesn’t. I felt the midrange could be a little more forward and articulate so I let the 18W8531G-00 run even further by lowering its series inductor to 0,56mH and changing the 12M4631G-00’s series resistor to 1,5 W. This inductor needs big steps when changing otherwise they are not audible. Stepping up from a 2,5mm wire air-core inductor to a copper-foil version also improved depth and openness a lot; it lowered the noise-floor level even further. I also switched the 33mF M-cap to 20mF so that the tweeter would start a little higher (some recordings were a little to bright and “edgy” otherwise). Interesting also was the Zobel network parallel to the 18W8531G-00, it didn’t have any effect on the sound, and so I left it out. I will play around with this crossover concept nr. 03 for a while now, take lots of measurements, do lots of listening and see where it will lead.
Crossover 03 optimised after intensive
listening.
‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗
18-11-2003 After more listening and measuring
Crossover 04 further optimised after
intensive listening and measuring.
A few more tweaks since my last post. The crossover point between the tweeter and the midrange unit has been moved up slightly adding a little more openness to the midrange and giving the tweeter some extra protection. This meant that the midrange driver needed a little extra taming in its top region so L5 of 0,22mH was added. At the other end the low bass needed more energy so I made L1 a little smaller. The crossover point between the mid-woofer and the midrange unit was altered by changing L3 and L4 and putting back the Zobel for the mid-woofer. The “problem” with a 4 driver 3.5-way system is the large amount of variables. This makes fine tuning the crossover nice and difficult. For example to make the midrange a little more forward I could make L4 smaller or lower the value of R3 or leave out C7/R5 or change L3 to something smaller. All have a more or less same effect on the tonal balance of the whole system. Which one (or combination of) component(s) is the “best” choice? Now just that is the fun of designing speakers for me! This crossover version nr. 04 is already very pleasant to listen to: coherent, open, spacious and smooth. Still I feel there is room for even more improvement – to be continued.
‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗
09-01-2004 After even more listening and measuring – the final crossover!
Still tweaking the crossover.
Well it looks that I have finalised the crossover design! In the past few weeks the sub-woofer section has gone from 2nd order to 3rd order and back again. But the real problem was getting the lower midrange balance correct. This speaker is very misleading; the open baffle midrange creates a sound that is totally free of boxiness, which at first gave the impression that I was missing something in the lower midrange. This lead to increasing the overlap between the 18W and the 12M drivers, at one point I was running the 18W nearly full range! This gave more energy in the midrange but the tonal balance was lost; Diana Krall started shouting and Jos van Immerseel’s forte-piano sounded like a Casio keyboard. Then I realised what was missing: the overhang created by a cabinet. Compared to the Progress speaker normal speakers sound terribly “boxy” or “cloudy”. But even though I didn’t need to let the 18W run further (I actually enlarged its series inductor in the end) the speaker still lacked some weight, especially noticeable with the left hand of the piano. Just letting the subwoofer run a little further into the lower midrange wasn’t the right way to go, it gave the extra weight but also made the lower midrange less detailed. So I changed the connection of L3 directly to the input of the crossover instead of after the first inductor L2, thus going from a series-linked to a series-cascade model. All things come with a price: the gain in weight came with a compromise in sound stage depth; better tonal balance at the cost of flatter image. Seeing as imaging was still very good I decided that the better tonal balance was more important. Also it gave a new possibility to the crossover: the whole midrange/treble section can now be “dialled-in” with R5 (I ended up with 0,75 W made from 2x 1,5 W in parallel). This gives extra flexibility when matching this speaker with your equipment and personal taste. The very tricky thing about this speaker is the tweeter and midrange resistors, they have such a great effect on the overall balance of the system and there are 4 of them making the possibilities nearly endless! It’s very easy to go wrong here and start changing inductors and capacitors when all you need to do is tweak a resistor or two. R3 by the way is 1,1 W made from 2x 2,2 W in parallel – this explains how critical the resistors are; the standard 1,0 W and 1,2 W weren’t “it”. All the resistor values only work properly in combination with the Rdc of the inductors I used. All resistors are 10 watt MOX.
Crossover 05 further optimised after even
more intensive listening and measuring.
16-01-2004 No more tweaking!
Everything is soldered into place! I made one small but important alteration to the crossover: capacitor C4 (the 6,0mF Hovland Musicap) has been replaced by a 6,8mF Mundorf Supreme Silver in Oil cap. I originally used the Hovland Musicap to “brighten-up” the Mundorf Mcap’s a bit. It succeeded in doing this but the “s” sounds in vocals got a nasty edge. The Silver in Oil took away the rough edge and gave overall more depth to the upper range. It would be nice to replace all the capacitors in the series section of the crossover but then I would probably have to sell the house, I presume my wife would object. The photo shows the bottom of the bass cabinet where I made a separate compartment for the crossover. The bass section consisting of the large N-130 transformer and a few cap’s are situated inside the bass compartment and therefore not visible on the photo. All components are hard-wired directly to each other. I needed to add only two extra jumper leads and they are made of 2mm solid core copper wire. The cables leading to the drivers are connected directed to the filter components: Green Van Den Hul CS-122 Hybrid for the Scanspeak 18W8531G-00; crčme coloured Van Den Hul CS-122 Hybrid for the Scanspeak 12M4631G-00 and the Scanspeak R2904-7000; 4mm2 OFC copper wire for the Seas L26RFX/P. Note the parallel resistors to get the correct value.
Quite simple for a 3-way crossover!
The final crossover.
L1 = 12 mH Mundorf N130 Zero-Ohm transformer inductor, R = 0,10 W (tolerance 5%)
L2 = 0,33 mH copper foil inductor CFC-12, R = 0,09 W
L3 = 2,00 mH Tritec inductor 3,50 mm2 wire, R = 0,22 W (tolerance 2% - matched pair less than 1%)
L4 = 0,15 mH copper foil inductor CFC-16, R = 0,15 W
C1 = 33mF MKT polyester foil capacitor (tolerance 5%)
C2 = 6,8 mF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance 2%)
C3 = 4,7 mF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance 2%)
C4 = 6,8 mF Mundorf M-Cap Supreme Silver-in-Oil capacitor (tolerance 2%)
C5 = 47 mF Mundorf M-Cap polypropylene foil capacitor (tolerance 3%)
C6 = 47 mF Mundorf M-Cap polypropylene foil capacitor (tolerance 3%)
C7 = 0,1 mF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance 2%)
C8 = 1,0 mF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance 2%)
C9 = 0,1 mF Mundorf M-Cap Supreme polypropylene foil capacitor (tolerance 2%)
C10 = 33mF MKT polyester foil capacitor (tolerance 5%)
C11 = 2,2mF MKT polyester foil capacitor (tolerance 5%)
C12 = 0,01 mF ERO MKP1837 polypropylene foil capacitor (tolerance 1%)
R1 = 1,5 W, 10 watts metal film resistor (tolerance 2% - matched pair less than 1%)
R2 = 1,8 W, 10 watts metal film resistor (tolerance 2% - matched pair less than 1%)
R3 = 5,6 W, 10 watts metal film resistor (tolerance 2% - matched pair less than 1%)
R4 = 2,2 W, 10 watts metal film resistor (tolerance 2% - matched pair less than 1%)
R5 = 6,8 W, 10 watts metal film resistor (tolerance 2% - matched pair less than 1%)
R6 = 1,5 W, 10 watts metal film resistor (tolerance 2% - matched pair less than 1%)
R7 = 2,2 W, 10 watts metal film resistor (tolerance 2% - matched pair less than 1%)
Ls1 = Seas L26RFX/P – aluminium woofer
Ls2 = Scan speak R2904-7000 – Revelator ring radiator tweeter
Ls3 = Scan speak 12M4631G-00 – Revelator sliced paper mid-range driver
Ls4 = Scan speak 18W8531G-00 – Revelator sliced paper mid-woofer
‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗‗
13-02-2004 Never say no more tweaking!
I tried two more things to bet the maximum possible from these speakers. First of all I wanted to see (hear) what would happen if I added some absorption to the rear waves coming from the open baffle mid-range unit. I used a 10mm thick felt ring, which was placed over the shielded magnet leaving a gap of about 15mm between the felt and the rear of the baffle. The result in sound with the absorber was a little gain in clarity at the cost of dynamics and spatiality, so I decided to leave them out. The second tweak was the addition of a small low voltage capacitor parallel to the tweeter series caps – check out my cap test page for more details.
Without and with the absorber.
NOTE: This design is strictly for the home DIY enthusiast and not to be used professionally without my permission!
Tony Gee
The Netherlands