This may be an odd place to start, but why another speaker? The world is full of speakers and don't we have enough? From mails coming in, the appetite for new speakers in the diy community is endless. Curiosity and the passion for making something better than last time is the main driver. This must have been going on ever since we sat around the fire carving bows and arrows. How do I make this arrow fly faster, with more precision, with better penetration? If only.... then I would only have to go hunting half day and the afternoon could be spent sitting in the shadow of the trees playing my bone flute - until I get bored and start thinking: "What if..?" Anyway, this isn't about arrows or bone flutes, rather sophisticated loudspeaker drivers that may enable us to get a little closer to the music, reveal more detail and provide an overall more truthful sonic image of what was going on behind the microphones capturing an acoustic event.

The construction shown here will offer you an opportunity to reach into serious high-end, that is: If you make rock solid cabinets, if you damp them well, if you offer serious money on drivers and crossover components - and if you have a good amps and source material, the latter not least important. All of this doesn't come cheap and we're into the area of diminishing return from our investment. What we should get from expensive drivers is the ability to play louder with less distortion compared to el-cheapo drivers and here you get it. Some people question the value of expensive crossover components like super caps, etc. My response to this is that if you can't hear the difference between a standard cap and a super cap, your amps, your CD player or turntable cartridge simply isn't good enough.

During construction and crossover optimisation, the Morel CAT-308 was the tweeter in place and did a damned good job. However, modelling a large number of tweeters, it was found that the venerable ScanSpeak D2905/9900 not only allowed better phase integration, simpler crossover, better sound, thus the 9900 it's going to be. The slightly concave face plate adds distance to the dome and impulse response could be further improved. The low Fs allowed a simple 2nd order LR to be established with as few components as possible. I like that! Ideally the CAT-308 would require a single cap and two notch filters to shape the 2nd LR roll-off although a single LCR circuit did well for the listening test. With SS 9900 in place a perfect 2nd order LR roll-off could be had from a RL circuit to ground. The 9900's impedance peak at 500 Hz needs flattening from a LCR circuit. Serious listening test were performed w/wo this notch filter and it was impossible to hear any difference, but for the benefit of the doubt it stayed in place. View schematics below and read comments on crossover development.

The PRELUDE will offer you a rock solid bass playing high or low, a fully integrated midrange and treble with some of the best treble I've had due to driver quality and the low-order filters employed. The PRELUDE will handle anything from classical guitar, vocals and trio-jazz to heavy metal. Give it some of the best 100 watts you can afford and you may even throw a Saturday evening party.

The aim of this cabinet construction is to make a rock solid enclosure for the AT drivers, so be generous with bracing. Next I wanted the midrange at ear height, thus a small stool to lift the speaker to an overall height of 1 meter. The stool is 16 cm high and some spikes or whatever adds to the final 20 cm base height. Finally I wanted a cabinet where most diy'ers can follow in terms of overall complexity. No curved panels or fancy forms that only CNC machines - or talented diy'ers - can cope with. No, this is not a skinny runt of 18 cm width, rather a 36 cm (14") wide baffle providing acoustic support for the drivers and enabling simplicity of crossover.

Regarding box size and tuning: Please read here: http://www.troelsgravesen.dk/10C77.htm

Final box shown here is around 55 liters net volume (for bass driver alone) depending on use of bitumen pads, amount of bracing, volume taken by crossover, etc. I recommend a port tuning of 30-33 Hz. Use 2 x vents @ 68 mm diameter and 220 mm length. Download excel file to calc. vent length here. Note excel file is set for one port. When you use two ports, double vent length.

Work on this crossover has been ongoing for 4-5 months, much longer than any other speaker I've made and eventually produced "Siri's Killer Note". The reason for this is not that the drivers are particularly difficult, and going the standard route of 2nd/3rd order filters, the crossover was in place in a few hours. But these drivers are not any drivers and upper mid and treble presentation suggested more work to be done, thus a whole range of crossovers were simulated and tested for the mid-tweeter section going from 3rd LR to 2nd LR to 1st order filters and eventually a combination of LR2 and 1st order Butterworth did what I was after. In reality true 1st order filters for all drivers cannot be realised and the acoustic roll-off is indeed often something between a true 1st order Butterworth and 2nd order LR and depends on how far above or below point of crossover we want the slope to follow target response.

What often happens during crossover development is that you reach a certain point where the sound is very good, but measurements suggest more to be done and slowly you improve measured performance - and decrease sonics. Reaching crossover version 7 mkII, everything looked great - but magic had gone. Now bass and mid integration was pretty much in place but what happens here will also have an impact on upper registers and in reality I had reached a rather perfect LR2 topology for both mid and tweeter. The close-to-first-order mid-tweeter filter of previous versions was gone - and so was the magic. Hmm... Back to LspCAD and some more late night hours' simulation and crossover soldering. When the final middrivers were at hand, another round of crossover fine-tuning had too be done, but nothing serious compared to the temporary 15H52-15-06 SD I had used for initial experiments.

LR2 between bass and mid was run for a long time, but having the mid on top, driver integration wasn't perfect, thus the 1st order high-pass to the mid was tried and an LCR circuit takes care of the mid's impedance bump at 80 Hz. The mid cabinet has a 40 mm vent and Fb = 47 Hz, but eventually the vent was stuffed lightly with damping material and the impedance flattening circuit introduced. This works. Actually the LCR circuit also works very well with the vent not stuffed, so take your pick. The sound is slightly different and I strongly recommend adding the vent for experimentation with closed, aperiodic and vented mid cab.
With a second order low-pass section to the bass driver and true 1st order high-pass to the mid, good driver integration could be achieved with regard to amplitude and phase. This step also included lowering point of crossover between bass and mid some 100 Hz, now close to 320 Hz. Not only did it measure well, it also sounded great.

The final crossover is asymmetric, providing LR2 slopes to bass and tweeter and 1st order to the middriver. Thus proper phase integration is achieved with enhanced impulse response. In a commercial add this would account for perfect transient response, time-aligned, zero time and phase distortion and bla bla bla.... But let's be honest and agree only e.g. QUAD ESL speakers come close to these virtues. Dynamic drivers can do a lot of things electrostatics can't, and as always there is room for improvements.

Looking at the step response below, we're close to a perfect time alignment between mid and tweeter and this truly pays off in terms of transient response and enhanced overall transparency. Adding an all-pass section to the tweeter to render same polarity for mid and tweeter was tried and didn't pay off. Serious listening tests were performed but I'm really not sure that same polarity pays off as long as proper phase integration is accomplished. It may be a give-and-take situation where the added series components detract from any possible benefits of same driver polarity.

The crossover seen below may seem a bit complex to some people and here are my comments, click here.

I've already had the question on the differences between the middrivers tested and I had three identical pairs except for voice coil former material, alu, kapton and kapton/alu. The voice coil former material has a significant impact on performance, particularly on upper mid and treble range and what works the best may depend on choice of tweeter and crossover topology. The impact on sound is very subtle and the choice was hard. The mids with alu formers had the best (smoothest) frequency response but eventually the kapton/alu was chosen based on sonics and please do not ask how they differ, because if I put in onto paper it would be black and white - and it isn't. The choice of voice coil former material will depend on the actual application, and we can't simply conclude one material is better than the other. Read here at Audio Technology website about voice coil former materials: What is the difference between Kapton and aluminum, question 3.

Left: Impact of mid high-pass section LCR circuit (10 mH + 270 uF + 6R8). The reason for showing this is that in a commercial speaker this circuit would definitely be
left out due to added cost, but as diy'ers, we don't necessarily have to be that fixed on cost although 270 uF PP doesn't exactly come cheap. You may use an electrolytic...
Although deep in frequency, the LCR circuit does add to proper phase integration between bass and mid. Measurements not shown.
Right: Distortion measurement done at 1 meter distance and ~95 dB sound level (high!). I was pleased to see this; no spooky dips or peaks in upper mid and treble range.

I think this is enough. Basically measurements tell very little - if anything - about how a speaker actually sounds.

Routing for the 9900 tweeter may discomfort some people, but it's easy. Your router bit (use an old one) will easily cut through the aluminium face plate, but it has to be done slowly. Lower router bit some 0.2-0.3 mm at a time and move the router slowly over the 9900 face plate. The 9900 tweeter comes with a fixed gasket and from my experience, some 5.3-5.4 mm depth is required to make the face plate flush mounted. Tightening the screws a bit further will take the face plate ~0.1 mm below surface plane and you won't scratch the face plate's anodisation when moving the router.
Cover the dome with a piece of gaffa tape (70 x 70 mm) and don't forget to add a piece of paper (50 x 50 mm) where it covers the dome!! After routing, clean the 9900 cut-out edge with fine grade sanding paper and run a black marker over the edge.
I'm sure I'll have the question if this is really necessary, and why can't I just mount the 9900 tweeter below the middriver? Well, vertical dispersion is not a nice view when dealing with low order filters and having the drivers as close as possible helps, so... just do it!

Front panels modified for 9900 tweeters.

I use two materials: 10 mm (green) felt and MDM3 (Monacor). Check materials here: http://www.troelsgravesen.dk/cabinet-damping.htm

1. Before adding damping material, 4 mm bitumen pads were added to all internal panels except front panel.
2. 10 mm felt was added - again to all panels except bottom panel (where crossover is placed) and front panel.
3. ½ roll MDM3 is folded and placed right behind bass driver on rear panel.
4. ½ roll MDM3 is placed on top of crossover at the bottom.
5. ½ roll MDM3 is used in mid cab placed on side panels and double layer on rear panel. Nothing (except felt material) on top and bottom.
That's all. 3 rolls of MDM3 - and please calculate yourself how much felt material is needed :-)

Upper: As can be seen we have some 2-3 dB difference at 100 Hz and the notch filter produces and overall smoother response.
Lower: What happens to the system impedance looks like this: Have added red to the +LCR lines (thick = impedance, thin = phase. Disregard > 1 kHz red.).

All in all this makes better impedance in the 100-200 Hz area but ~2 ohms lower around 30 Hz. The huge phase shift at 50 Hz is gone, but still remains around 25-30 Hz. Jantzen Audio does have a 18 mH coil, DCR = 5.7 ohm (coil # 1699), which seems perfect for the application. Only thing is that we may have to use a 330 uF electrolytic capacitor. The CrossCap range does include a 330 uF value, but it is huge: 68 x 110 mm and I hardly have room for more. And it's around 70 USD each.
Without having tested this addition, I render it an interesting option for adjusting the speaker to various room positions. A lot of people have major trouble in placing their speakers optimally with regard to room acoustics (WAF!) and having a elevated response around 100 Hz may render your speakers boomy.

The final LCR circuit, here temporarily set-up for test. The board will be placed on rear panel behind bass driver.
If you want to be able to adjust resistor value frequently, place an additional set of binding post on rear panel and use these for holding the resistor.

The sonic impact from adding the bass LCR circuit is a more balanced and subjectively pricise bass. Strangely enough, bass extension appear enhanced too. Conclusion: The LCR circuit has come to stay.

This is one thing you must do. Tune tweeter level to suit your taste, your room and not least speaker position.
I find the PRELUDE to sound the best only slightly toed in, that is not listening directly on-axis. Due to the constant
directivity face plate of the 9900 tweeter dispersion is somewhat different from conventional domes and needs attention.
Above can be seen tweeter levels from R2021 = 2.2 ohm, 3.3 ohm and 4.7 ohm. Try it out on a range of different music before deciding.

My final value for R2021 has changed from 3.3 ohm to 4.7 ohm. I guess most will settle on 3R3, possibly 3R9. 4R7 may be too laid back for some people, but to ears makes the best balance of basic notes and uppertones.
Extra 3R3 resistors have been added to parts list. If you want additional 3R9 and 4R7, specify on order.
BTW: Tweaking tweeter balance I used graphite resitors for R2021 and R2051. Better sound? Maybe :-)