Design notes

I wanted to build something better than the Asawari for my own use. I wanted something which would allow me to stretch the performance envelope and force me to learn new insights into speaker design. And I will call it the Darbari in memory of the original Dagar Brothers, whose 1973 performance of the Darbari Kanhra for HMV is still a legend (never published on CD).

Three's company

I have decided that one of the first steps forward is to move from the undeniable compromises of the two-way to the three-way speaker. The two-way speaker has the difficult task of delivering punchy bass and delicate, nuanced voices through the same driver. Imagine one cone moving in and out, in and out spasmodically to the beat of a big bass drum, and at the same time trying to vibrate in delicate ways to generate frequencies of the human voice. It's a crazy demand, and drivers usually do not do a very good job with it.

Therefore, one challenge the speaker designer has to take is to keep aside a separate driver for the two bottom octaves, up to about 80Hz.

Regarding this issue, there is one audiophile and DIY myth that comes to mind. There is this strange belief that the human voice stretches from about 300Hz to about 3KHz. Many a speaker designer has tried hard to keep this range of frequencies within one driver, so that the polluting effects of a crossover do not hit this most special of frequency ranges. The truth is that a male voice often has a fundamental frequency at 150Hz, and female voices often start from 250Hz or so. Therefore, if one has to reserve a single driver to handle the voice frequencies, this driver must stretch from 100Hz to about 3KHz.

The right way to do this would be, of course, to use a wide-range or so-called "full range" driver. and add assistance to the top and bottom ends. You need an extended-range driver which can handle the range of 100Hz to at least 4KHz smoothly. There is probably no better driver for this role than the Jordan JX92S or its successor the the Jordan Eikona, if you can afford it. I have lived with the JX92S for a few years in a single-driver enclosure, and I've seen very knowledgeable designers compare this driver with other similar-sized drivers. There is really nothing from Scan-Speak or other major brands which can beat the JX92S as an extended-range midrange driver. Those who have tried to describe the sound of the JX92S have used the phrase "electrostatic-like". You have to hear it to believe it.

I now no longer have access to the JX92S. Nor can I afford one now. So, for the Darbari I will use a more ordinary midbass driver.

Get active

Another thing I have decided to do is build an active crossover system. I want to see what I can do with electronics. I have been getting more and more interested in amp design and electronic circuit design, and it would be fun to channelise this curiosity into giving me better sound than a passive speaker. Secondly, I have always read about the benefits of active crossovers. It would be nice to see what a good active system sounds like.

When the bell tolls for me

A third aim is to work with low distortion metal cone drivers. These drivers deliver lower distortion in the bass and midrange regions than paper cone drivers, but have serious stored energy problems in the region of their cone break-up. At the cone break-up point, they often have peaks which rise 15dB above their base SPL, and also resonate, i.e. keep ringing after input signal stops. The only way to get good performance from such drivers is to cut the cone break-up region so aggressively that they do not have any audible effect on the sound. This is very difficult to do.

Conventional crossovers with fourth-order slopes just do not deliver the goodies. They will do a fine job with a metal-cone woofer if you want a 100Hz low-pass and the cone breakup is at 1KHz. But if you want to use a metal-cone midbass driver, then you will want to take the midbass up to about 2KHz, and the cone breakup will rear its head quite close by, at perhaps 4.5KHz. You will then have little over one octave to bring down the SPL by at least 80dB.

Many conventional speaker designs using the excellent Seas Excel range of midbass units have tried mating a notch filter with an LR4, in a passive crossover. This is, in my humble opinion, very difficult to do. You need a wide, i.e. low-Q, notch, since the cone breakup gives rise to a cluster of multiple peaks, and you have to cut the whole bunch. Getting a deep, low-Q notch to work well with a passive crossover is not very easy.

Some designers, including the very respected John "Zaph" Krutke and Joe D'Appolito, have been happy to cut the resonant peaks by 30-40dB for some of their speaker designs. The Seas Thor kit is an example of this sort of crossover. In my humble opinion, a cut of 40dB is simply not enough for this malaise. This is no benign second harmonic distortion, where some audible distortion can make the music a bit more euphonic. This is true ring-like-a-bell resonant ringing, of the kind that can slice your ears off with listening fatigue. Cutting these down by merely 40dB will not make good quality speakers.

The only good treatment of this disease that I have read about is the Modula MTM built by Jon Marsh in the "Mission Accomplished" DIY forums of www.htguide.com. There, he designs an MTM using metal-cone midbass drivers and crosses them over to the tweeter at a low 1,400Hz. And above 1,400Hz, he cuts the SPL with a Cauer-Elliptic passive crossover which starts out looking like a fourth-order and then steepens rapidly to an eighth-order slope. Only with this sort of all-or-nothing approach will a designer be able to make the ringing truly go away.

That's my aim too. I will be doing the same thing, but with active filters. I intend to build an LR4 low-pass, followed by a notch filter. Simulations of this topology with a metal-cone midbass driver's FRD and ZMA files has yielded SPL and phase results very close to the Cauer-Elliptic, and Jon Marsh had approved. I had built nothing physically at that time, but this time I'll bite the bullet.

My aim is very clear: the highest peak of a cone-breakup resonant spike in the SPL graph must be brought down 70dB below the flat base level of the pass-through portion of the SPL curve. Basically, there should be no portion in the breakup region higher than -70dB below the base SPL level.

Why struggle with metal-cone drivers at all? Simple: they have rigid cones, and these generate less distortion than any floppy-cone drivers, e.g. paper or poly cone drivers. Those designers who wish to keep simple crossovers use these cone materials because these drivers have smooth and benign SPL curves and can be mated easily with other drivers in a crossover. No cone breakups and nasty jagged edges to tame. But what you gain is a lot more low-order harmonics. Many listeners actually don't mind those harmonics much because they are often second-order and add a certain warmth and richness to the music. If you want to see one veteran designer's honest assessment of his recent 3-way design with a paper-cone midrange, check out Roman Bednarek's JDB3 and scroll down to the section titled "Listening Impressions." That analysis sort of captures it all.

But I want low distortion sound, and I want to learn to tame the beast of the tolling bell.

Configuration

I have therefore decided to build a three-way, but I do not want to go the sat-sub route. I would prefer to build a floorstander. Ideally, I will aim for a sealed bass enclosure with a woofer handling the sub-100Hz range. On top of this will be a separate mid-and-tweeter enclosure, physically separate, so that baffle vibrations from the woofer enclosure do not communicate to the mid-and-tweeter unit. I am thinking of placing a soft mat, like a folded towel or a felt mat, between the two enclosures, and placing, not fixing, the mid-tweeter enclosure on this mat. Should work just fine, and should cut out all vibrations above a few Hz.

I want to model the enclosures in shapes inspired by Neil Patel's Avalon speakers. Flat surfaces, bevelled corners, front baffles three inches of solid MDF, and of course, the extensive internal bracing as per the tradition first used with the Asawari.

And I will use separate amplifiers to power this floorstander. I want to build stereo amps in separate chassis, so that my curiosity about amplifiers is catered to as well. I would like to build amps in re-usable blocks, one stereo power amp per chassis, and swap amps in and out.

Therefore, I expect that there will be a large number of rack-mount boxes in the space between the speakers. There will be one box holding the active crossovers, one for the preamp, and three for the three stereo power amps. Wish me luck.

Drivers, drivers

I have decided to use a trio of drivers which are well-known as being low distortion and excellent value-for-money.

The tweeter: It was originally planned that this will be the Seas TDFC fabric dome tweeter. John "Zaph" Krutke confirms, in his tweeter tests, that this is one of the lowest distortion tweeters around. A lot of boutique tweeters costing twice to four times this little thing actually have higher distortion. A second lovely feature: this tweeter can be crossed over quite low. That, as you have read, is a necessity to let me tame the midbass beast.

I later read that the Peerless 810921 is an even lower-distortion tweeter. It's about 50% more expensive than the Seas TDFC. I decided to switch to this tweeter, increasing my total costs for the Darbari by about $50 for the pair. The Peerless HDS is now available not under the name Peerless, but as the Scanspeak Discovery D2608/9130 HDS dome tweeter.

The midbass unit: This will be the Dayton RS150 6" midbass driver. I tend to cross this over to the tweeter at about 1,500Hz or so, so that I get the maximum opportunity to kill its cone-breakup peaks. On the bottom end, this driver will happily deliver 80Hz and below, hence I have no apprehensions that I will be able to get it to deliver a Qtc of 0.7 with a Linkwitz Transform, in a sealed enclosure.

The woofer: This will be the Dayton RS270 10" woofer. I intend to use this from some point below 100Hz, down to about 25Hz, in a sealed enclosure.