Diatone Big Blue by Markus Karner This is the write-up of a roughly 3-month DIY project using the Diatone 610-MB driver in an MTM configuration with two 10"-Vifa woofers for the sub-310 Hz range, in an actively equalized open baffle dipole configuration, actively bi-amplified (two Burr-Brown OPA-548 chip amplifiers per channel, mounted on back of the baffle) and with active line-level crossover at 310 Hz (L-R 4th order). To give my credits at the start, the single driver idea and impressions came from this website (Single Driver Website), the dipole idea, theory, concepts and much of the circuit layout came from Siegfried Linkwitz's website, heavily modified and adapted to the wide-range driver used, and the crossover circuit came from Rod Elliott's Sound pages. Materials were purchased from Madisound (Vifa), Radio Shack, Digi-Key and Mouser (electronic parts) and from private party advertising on this website (Diatones, purchased 2nd hand but "new in box"). Thanks to Craig Schmidt for discussions, I almost used his JBL drivers but then decided on the Diatones. The concept I dropped on Siegfried Linkwit's dipole pages by chance, and got immediately attracted to the concept. In essence, it is about using conventional (as opposed to electrostatic or similar) drivers as dipoles, on an open baffle. As Linkwitz explains - and I cannot give him enough credit for his exquisite website with a massive amount of bullet proof engineering information - the advantages of this concept are: absence of box resonances and coloration; directivity/dispersion control, specifically of the bass and thus, less excitation of room modes and therefore, cleaner bass. The specific challenges to be dealt with are: dipole fall-off of 6dB/octave below a certain threshold frequency determined by baffle size. This is caused by rear-wave cancellation by the out-of-phase backside, occurring where front and rear waves meet. One part of this effect creates what you want, i.e., less sound going to the side walls etc. and more to the listener. But the overall rolloff means, it is necessary to use active equalization of the dipole rolloff, which is best done with line-level filters and thus requires bi-amplification. The active equalization approach means, demanding a very high volume displacement of the woofers in the equalization range. Thus, only high Q drivers of reasonable size and high x-max need apply :-) For more, visit Linkwitz Lab, and prepare to spend a few weeks to work yourself through the concepts :-). The design phase According to the constraints above and with the single driver concept in mind, helper woofers for the bass were a must according to calculations of necessary driver volume displacement. Since the Diatones are somewhat challenged in the x-max area (2 mm) a crossover at about 300 Hz seemed a good idea; that means separate boxes for the woofers were impractical: the woofers would have to be no farther than about 40" from the Diatones, ideally symmetrically and on the same baffle. So, using large, say, 12" drivers or larger, was made impractical due to resulting baffle size, and I chose to use two 10" woofers per side, with Q=0.5 and x-max=6.5 mm in MTM configuration. This puts me still on the lean side according to calculations! The limitations in useable SPL for the Diatones and the woofers are in rough agreement though and so the concept seemed coherent. The crossover is 2 octaves from Fs for the Diatones and the woofers have a useable frequency range exceeding 1000 Hz, besides I chose a steep 24 dB/octave crossover to limit stress on the Diatones - so all in all quite conservative choices. Baffle was chosen as a folded-back design of about 44" height, 14" width and 4" side "wings" (foldback). The baffle size was a mix of educated guesses, available materials, and desired size and looks. The size of the baffle puts the beginning of the dipole roll-off at about 300-400 Hz. The Fs of the Vifas is around 23 Hz, so I chose the equalization corner frequencies around 25 Hz on the low side, ending equalization at about 360 Hz. From 360 Hz up, in other words, the baffle size itself is enough to prevent rear wave cancellation. With 6dB/octave dipole rolloff below 360 Hz and the corresponding equalization boost, that gives a maximum boost of almost 24 dB of amplification in the bass region (reached at 25 Hz) - hence the very high displacement requirements for the woofers, because they have to play that much louder in order to achieve same SPL as they would have to in a closed box, due to the (now equalized) dipole roll-off. Since after investigation of their website I found out that Burr-Brown kindly offers free samples of some of their chips with no questions asked, I took the offer and so I had chips available for the power amplification (I do not own a 4-channel power amp nor did I want to buy one). It seemed practical to place each channel's equalization, crossover and the two power amp channels per side within the baffle because there is space to spare, it's well ventilated :-) and that saves an additional box standing around. So now the design becomes an active speaker, requiring an external power supply (I chose a 375 VA Amveco toroidal, producing +-26V DC after large capacitor regulation). The same power supply is used for further full regulation (with regulator IC's) to power the line level electronics. Purists will scoff but it doesn't seem to produce ill effects; in particular the high primary voltage (26 V regulated down to 15V) gives a lot of headroom for voltage variations. I must admit though that the slightest departure from star grounding took its revenge and had to be corrected! In the final incarnation I now don't have much noise or hum. The baffles are free standing. I chose MDF for the front panel, mahagony for the sides. The drivers are front plate mounted (with bolts through the entire panel) - albeit with the recommended rear brace - because I did not want to invest in a router-cum-appendices and equipment. So a $17.99 jigsaw had to do to produce the cutouts freehand (the circular attachment was too cheap to work properly). For the line level electronics I incrementally implemented several of Siegfried Linkwitz's models, recalculated for my design. Unfortunately I do not own the appropriate measurement equipment, so my choices for recalculations had to be made according to theory and educated guesses. First results made me correct some details after careful listening to test discs, music, and some primitive RS-SPL-meter pink noise measurements (does not produce true frequency response but gives hints at what's and why's of suspected problems). I ended up with the following configuration, and I repeat, the designs are by Siegfried Linkwitz for the most part: Common 140 kHz low pass (filters out radio interference) 100-200Hz 6dB shelving high pass (according to S.L., a 6 dB floor reinforcement in-room starts between 100 and 200 Hz and reinforces the lower bass disproportionately in all speakers. I thought I could get away w/o it because my woofers are high on the panel - but listening and pink noise tests convinced me otherwise and I built the filter anyway. The result: works marvelously well, can't do without it. Without this filter I could either set the woofer correctly for the300 Hz region and get boomy bass, or set the bass correctly using, say, 100 Hz pink noise as reference, and get a very audible through in the 250-400 Hz region) buffer Diatone channel 310 Hz 24 dB/octave crossover power amp BB OPA548 Woofer channel 310 Hz 24 dB/octave crossover level setting +- 6 dB (to correct differences in driver sensitivity between Diatones and woofers) 25-360 Hz dipole equalization (i.e., shelving low pass 6 dB/octave) power amp BB OPA548 Op-amps were BB 2134 (dual), extensively bypassed as of Burr-Brown's recommendations, as were the OPA548 power amps chips. Since I added bypassing gradually I can attest that it makes a huge difference in perceived distortion. Power supply: 375 VA, 30000 µF regulator capacitors, +- 26 V final output for power amps; +-15 V 7815/7915 regulated for line level electronics No group delay correction was attempted because the drivers are on the same panel, physically not too different, flat in fequency response with wide overlap in the crossover region, and the wavelength at 310 Hz is much larger than possible offsets. The 24dB/oct. L-R circuit itself will produce a group delay though, but frankly, I see/hear no problem in the crossover region, and on my Sheffield test CD I could not distinguish tracks with an artificially introduced group delay corresponding to a 7 ft. driver offset in the bass. So I let that one go for the moment... Construction My previous speaker DIY experience is a ported box using the "L'Audiphile Le Petit" design and RS40-1197 drivers, that I built for my girlfriend. They work very well, although with the known limitations in bass - and that's what encouraged me to go further. On the electronics side I had no experience whatsoever, neither in circuit design, nor did I even know how a capacitor looks like or how to choose between different types. I have a science background (biology) so I was able to read myself into the theory, but the practice was sorely missing. So I built a power amp to go with the 40-1197 speakers to warm up, and read "Practical electronics for inventors" with heightened interest. I can only highly recommend this book, the title is silly but it is extremely useful. Many details of the power supplies and regulation directly came from that book. Other circuit elements on the amp side came from data and application sheets, as .pdf's from Burr-Brown and National Semiconductor, which I adapted to my needs once I understood what does what and why. So once I knew how to wield the soldering iron and why it's not a good idea to mount tantalum capacitors the wrong way around, I built the line level circuits on Radio Shack standard circuit boards. The power amp chips were point-to-point wired to terminal strips with aluminum door railing as a heat sink (Home Depot). In my ignorance I had ordered surface mount chips which led to 2 consequences - I rapidly found out the difference between DIP and SMP, and my soldering skills quickly evolved in order to accomodate the necessary precision for 0.5 mm distanced pins :-). Since it's all mounted open but out of sight I did not bother much with beautiful cases or such but screwed it all to the back of the box with wood screws. I was afraid that either vibrations or magnetic fields from the drivers would cause problems, but that does not seem to be the case. I think my solder joints are of good quality; where my skills are lacking is in the terminals and connections between boards, here the space limitations and the star grounding mean that some connections are a wild bunch of wire that ends up in a ball. This may be corrected in the future as I learn to do it right. The woodwork was straightforward, although since I only have a freehand power drill and jigsaw, and all else is by hand tools (and I have no workbench either) I used simple braces/screws and some glue, no fancy dados/rabbets/dovetails. The mahagony of the side wings was not a good idea, it clearly vibrates more than the front MDF. With bracing and carpet backing as damping I got this problem down to a reasonable level. The front panel has additional damping through a fabric covered foam layer on the front (inverting a foam backed PVC tablecloth material), and very little vibrations actually make it through the front side. Since this is an evolving prototype, all connections are through the cheapest tin spades courtesy Radio Shack, or hardwired with 16 ga. speaker wire of the cheap kind (Home Depot). My project was built in many phases and I always listened extensively to the differences the latest addition would make. This was very instructional, because (in the light of the accompanying theory) it taught me a lot about how certain expected deficiencies sound. In fact often it sounded good already, with just some small perceived remaining issues - but after correction of these "minor" problems, it sounded always so much better that I thought, it's really easy to fool oneself. For instance the Diatones without any electronics or helper woofers on such a large baffle sound quite pleasing - certainly listenable. Once the woofers came in I suddenly knew what bass was. The complete setup then, but without 100-200 Hz 6 dB shelving high pass sounded almost complete, almost perfect - but with the filter, a certain lean-ness in the mid bass (because I had to set the woofers too low for the 300 Hz region in order to set the deep bass correctly) just disappeared and the lower voice region now sounded as natural as the rest. And so it was with a lot of little details I added gradually, too many to list. In essence, everything on that speaker has a purpose. Well, except for the mahagony maybe... :-) Of course this would all have been a lot easier with proper measurement tools. But I didn't want to spend $1000++ on oscilloscope, signal generator, mike, mike preamp, software, soundcard, and $3000 on a light truck to schlepp it all to the mythical open quiet parking lot for true free space measurements... so I accept the compromises and my inability to pinpoint remaining problems accurately. Sound It sounds great. The great premises, and promises, of dipole theory all came through. In the voice range, the Diatones are superb. They may not go to 20 kHz (fairly flat to 15kHz though) but I miss nothing. On the bass side, crossover integration is not detectable to me, that is, at least 2 m away from the speaker. The bass is tight and powerful, very detailed, I can hear a world of previously hidden marvelous bass instrumental details. The naturalness of drums and any percussion is phenomenal. So it all works out - the Diatones give natural, detailed voice and highs, the helper dipole woofers give detailed and natural bass. Room effects are truly much less severe, boomy resonance points are much less noticeable. The thing about dispersion is also really true: with dipole woofers you should get a similar dispersion in the bass as in the rest of the frequency range. This means, you get a more directional bass that allows actual bass imaging, but most importantly, the frequency response off-axis is similar in balance to the on-axis response. So, you don't have a narrow sweet spot, but a wide listenable area. If you walk around the room or, heavens forbid, listen from outdoors to speakers which are inside the room, you hear a fairly good replica of the frequency distribution. With regular speakers you mostly lose the highs and only hear some distant thumping from the bass. With these speakers, you hear the correct balance - only less loud. I did measurements with the Sheffield test CD, 1/3 octave pink noise, and the RS SPL meter (C-weighted). I do not trust the meter much, but considering Radio Shak's included correction curve, subjective impressions, and possible problems I knew about, I do indeed get a flat frequency response with little room effects from ca. 30-40 Hz up, at worst +- 3 dB and often better, with one exception: I still get a through in the 400-600 Hz region, which is possibly due to my design choices (discussed below). Things to improve or think about The frequency response according to the RS SPL meter still has a through in the 400-600 Hz region. I chose to place the dipole EQ only on the woofer channel and to start equalization at 360 Hz, so this may explain part of it. Baffle size, however, should be large enough to cover 400 Hz. In fact, considering theory, I expected a peak around 360-400 Hz if anything. But without real, accurate free space frequency response measurements I am somehow stuck here. I could tinker around with educated guesses but I can't hear the problem (unlike previous imbalances). Maybe it's the RS SPL meter :-) If the through is real: One possibility is that I may be getting a first null created by, again, rear wave cancellation. The rear wave, with increasing frequency, should alternately amplify and cancel the front wave according to perfect point source based theory. In a real speaker in the upper frequency regions this should not actually occur too much because of increased driver directivity at higher frequencies, and the first null should be alleviated by asymetrical driver placement on the baffle. My driver placement may not be asymetrical enough for this. Anyhow, this will rest for a while as is. Displacement limitataions are a reality. Even these monster woofers will somehow bottom out around 92-94 dB in-room (as by RS sound meter at about 2 m), and around this level the Diatones lose their naturalness too. It then starts to sound "loud". Below that level it is hard to really say how loud you are playing because it doesn't sound unpleasant (until the roommate complains about how loud you are playing). So, that means there is room for additional, bigger woofers :-). Unfortunately, you can't just double up the fullrange driver because you couldn't get them physically close enough to eachother to avoid lobing problems at the high frequencies. So I guess this is a design limitation. Maybe a driver such as the Fostex FX200 with better power handling could fix that. Due to the floor standing MTM design with big woofers, and too much space left between the woofers (for experimentations I never did) the Diatones ended up placed about 1 ft short of their ideal height (about 28" now, I wish it was at least 35"). I would do it differently next time around. Sometimes I think I can hear a limitation in the Diatone's transient renditions. This though only occurs on things such as high piano notes with sharp attack. Drums and cymbals, e.g., come extremely naturally, and bells or triangle comes so close to reality that you unwillingly turn your head to look for the bell. I'm not kidding you. So it may just be some of the recordings after all. One final detail: right now everything is hooked up to my Rotel integrated amp over the headphone output - obviously, not the finest of setups. Some of the consequences are, much too much gain, and somewhat more noise from the Rotel than I would like. I plan to build a pre-amp with appropriate characteristics eventually. Sonically though there are no really audible misgivings beyond the noise. Pictures are below. Enjoy! Markus Karner The whole back of the open baffle, with the back braces for driver baskets and damping foam. Note the heat sinks for the two power amps. I have added some light backing to the rear of the Diatone since the pictures were taken. Line level electronics. Note that the power amp chips (50W each) are so small that they disappear in the heat sink assmbly, you can only see the connecting wires. Top