SME

I did it! I am running my first live test with my custom DSP. I instituted a work-around for the Jack library problem. Although it's only really a partial work-around because I still intermittently fail. My problem may simply be that the Python code isn't realtime enough for Jack to not be fussy about it. The actual DSP code is written to be completely realtime safe. I did find and fix 3 bugs in my delay code, and I *think* it's flawless now. I'm listening live now with my own delay compensation active, and I confirmed its correct action by measuring a few impulse responses with REW. Woohoo! The imaging is definitely a lot better now. The upper mid-range got a bit stronger, as I expected. It measures 6 dB hot, which is exactly what I expected. And in fact, my level matching by ear of the woof and horn got them within 1 dB or so of flat. Tomorrow it will be time to migrate my existing bi-quad code over to my new stuff. Most of the work is already done for implementing specific filter types like HPF/LPF/peak/dip/shelf once I've migrated the generic bi-quad stuff to the realtime engine.

This is awesome stuff! Every one of your plots show superior in-room results, compared to the XXX woofers. Congrats!

Tonight, I installed my fourth Motu A16 unit. Thus far, it's working flawlessly. That's a big relief! It looks like this one was a brand new one that they opened up and tested by hand. As frustrating as this experience was, I'm glad they followed through on getting this solved for me. They also sent me another return shipping label. So far, I've been very happy with unit #2 aside from the bad display. Other than that, work has been a real pain lately, and I've managed to lose some of my time to other discussions on this forum. Nonetheless, I am very close to testing a live version of my custom code. I think the code that does the actual processing is working fine, but I want to see some good automated test results before I go live. I don't want to know what 50V to my horns would do to them or my ears in the events I somehow manage to send full-scale white noise to them. Right now, I'm struggling with the Jack audio library. It mostly works well, but does have bugs as well as confusing behavior that's not well documented at all. In any case, it's just a matter of time before I find a work-around to the issue and get the tests fully working. Hopefully I'll be doing measurements and/or crossovers by the weekend.

More thoughts. If the meter itself has a flat response down to DC and the loss of accuracy below 40 Hz is due to the "true RMS" integration time not being long enough, I think this would explain the oscillation. Furthermore, one would expect the spread of measurements to widen with decreasing frequency until the DC limit (vanishingly small frequencies) where it would oscillate between 0 and the the peak voltage. So if you always write down is the highest voltage you see, you'll be off by between 0 and +3 dB. Or you can subtract 1.5 dB, and now you're off by +/- 1.5 dB. It'd be nice to be able to do a better correction, but I think we'd need to know what kind of averaging filter was used in the meter. In time, I might think of a way to fudge things "a little better" though. There may be an easier way. Does the meter have a "peak hold" function? Assuming it samples quickly enough, you might have an easier time just measuring voltage peaks and subtracting 3 dB. For low frequency signals especially, it shouldn't be hard at all for the meter to get accurate voltage peak measurements. You do have to test with sine signals for this method to work, of course.

From casual observation, my meter seems to have a pretty wide range of validity. It only started oscillating badly below 20 Hz or so, IIRC. It also seems to exhibit some mild roll-off > 1 kHz, which may start at lower frequency. I'm not sure. I'll have to try playing with it some more. I'm thinking of getting a calibrated SPL meter to get my absolute levels a bit closer as I suspect my rat-shack meter is reading 3 dB or so too low, in which case, I'm calibrating everything 3 dB too hot. Edit: Don't know if this is the right meter, but here is an extended specs page that gives a range of 45-500 Hz for both volt and amp readings (page 3): http://media.fluke.com/documents/2155a.pdf

Interesting. I wonder if there is some measurement error somewhere? Are you using two Fluke meters for measurement? I don't know if yours has different capabilities than mine, but mine does supposedly offer True RMS. However, I notice that the readings oscillate more the lower the frequency I measure. IIRC, I went with the middle of the range for my number, but now I wonder if that might have been a high estimate. My signal chain roll-off measurements did seem a bit too good to be true. Maybe the 40 Hz measurement is correct and the 5 Hz and 10 Hz measurements are inflated. Just a thought. Edit: So to test this, maybe just run different tones at different frequencies but the same level to the amp. If you get a higher voltage for 20 Hz than 40 Hz, then the issue probably is measurement.

That's awesome! But I'm still curious why it can't be pushed farther at 5 Hz. Are you out of displacement there? Or does it throttle down too quickly? Am just curious what your thoughts are.

It looks to me like you're only -2 dB from rated power at 5 Hz. That's a pretty solid result. Although I suppose it might not do as well with both channels playing. Out of curiosity, why do you say that pushing it harder at 5 Hz wouldn't have made any difference?

Thanks for the offer, but I'm going fully active. That's partly what the Motu's for ... and the two Emotiva XPA-5s. In the long run, the crossover will be done as part of the room EQ. I can take measurements from each driver (or sub channel) and then run simulations that accurately predict the responses for whatever arbitrary filters I choose. Then I can iterate on the filters until I'm happy with the results. In time, I'll probably experiment with some automation of this process. I've already done this with remarkably positive results for my subs and MBMs with a MiniDSP 2x4 on the sub channel. I also use filters via OpenDRC-ANs on each of my mains to improve their integration with the subs and MBMs. With the Motu up and running and the software improvements I'm working on, I will be able to customize filters on each bass driver in my room for each input channel (i.e., FL, FR, FC, SL, SR, and LFE) independently. for example, say you're trying to optimize your sub delay for the best crossover response with the mains and you notice that 3 ms looks better for FL but 7 ms looks better for FR. Well with full-matrix DSP, I can have it both ways. I have 9 different bass sources in my room right now, and depending on what subs and amps I buy, I may end up with 11. Each of these 11 sources may use a different set of filters for each of the 5.1 or even 7.1 channels. (Even though my system will only be "5.1", there's no reason I can't customize the mix-down of the 7.1 channels to the 5 treble and 11 bass sources. This whole concept really blurs the notion of what a speaker is and how channels should be counted.)

Yes, absolutely! I have 40 uF in series with each CD, but that's all.

Just a quick update. I'm still using the Motu with the broken display and haven't heard back from Motu for a couple days now. I hope they are working finding me a fully working unit, but I will ping them tomorrow if I haven't heard anything. I feel slightly silly for listening to these speakers without a proper crossover, but they do sound quite good even without. I've been working on the PC that will run the DSP code. It's set up with the Motu on USB and Jack audio server on a PC running the Linux Mint 17.3. It took a bit of tweaking, but I got things up and running without too much hassle. This stuff is way easy compared to 15 years ago. My informal tests suggest I can reliably get latency down to 2 ms with the kernel I'm running. I bet I could go lower with a different kernel, but I'll probably opt for a built-in latency of at least 10 ms, probably more to improve computational efficiency and make room for some non-minimum phase correction. I can also confirm that the system (without my software added) successfully passes audio. So it looks like the Motu A16 and Linux work fine together via USB. That simply rocks. So now I'm back to just code. We'll see how I progress, but I hope to be working on a crossover by the end of the weekend.

I read a case report about a workplace environment complaint in which the employee complained about there being "bad air". Apparently many different specialists were called to address this problem including mold remediation specialists and so on. Nothing alleviated the complaint, until a clever engineer noted that part of the HVAC system was resonating and producing a strong infrasonic tone. Rebuilding the affected ducts mitigated the resonance, and solved the problem once and for all. The morale of this story is that people do not necessarily experience infrasound as either auditory or tactile sensation, at least until they actually understand what it is that they are sensing.

Bummer! At least movies like this are less common than they used to be. I agree that limits in the capability of monitoring equipment is a big factor in mix quality, and I'm sure this is especially relevant for ULF and bass filtering. When it comes to straight-up loudness, however, I don't know that monitoring equipment is really the issue. My experience has been that clipping is no less audible on low end systems, and in fact, it often sounds worse on low end systems. The effect is to make home viewers turn it down a lot more for the action scenes versus what they would have done for just the dialog, which means they have to ride the volume even more. I'm not sure what effect compression has on listening with lower end systems, but I know my own experience comparing the more compressed "Spirited Away" BD with the DVD was that the sounds in the DVD were much more distinct. In the more compressed BD, the dialog often got buried by the music, but in the DVD, the dialog punched through very well. I wonder if the dialog would have punched through better on the DVD versus the BD on a HTIB system too? Given that dialog intelligibility is a very common consumer complaint ("the music and effects drown out their voices"), could it be that efforts made to improve dialog intelligibility using compression often have the opposite effect? I fear a lot of the problem is that, even in this Hollywood, a lot of audio people just don't know better.

I got busy today. With help from the wife, I got my second XPA-5 installed into the rack above the first one. It involved plenty of swearing, but we managed to get it in a lot quicker than we expected. Then I made some new speaker cables and got the new amp connected to the Motu A16. When I was done, I dragged out the new speakers, set them up, and routed the left in to the left woofer and horn and routed the right in to the right woofer and horn. I don't have any kind of crossover yet, except for the protective 40 uF in series with each horn. I do have the speakers crossed to the subs at 110 Hz (same XO as used currently by the old speakers). I used the A16 web app to set the gains of the four channels by ear. I'm not using any room EQ. I am absolutely amazed at the sound quality despite the lack of crossovers and the fact that the horn CDs are mismatched. Yes, the sound is heavy on the upper mids due to the extended bandwidth of the AE woofers, and the left horn has quite a bit more top end than the right horn. I also don't have any delays set between the drivers much less room EQ. Nevertheless, the imaging is remarkably good and in many respects, it exceeds the performance of my current speakers in their fully optimized/calibrated configuration. I'm sure the imaging will improve a ton once I start getting some filters on these guys. Oh yeah, on another positive note, the noise floor with the MiniDSP OpenDRC-AN units removed from my chain is much lower. The speakers are hooked up to the amp, which is hooked up to the Motu, which is hooked up to the Denon. All connections are balanced except between the Denon and Motu. The noise floor is still audible, particularly in the horns, but it's not very annoying at all. With the MiniDSP units in the chain, we could hear the horns hissing from the bedroom of the house. So it does appear that the noise performance of the Motu is quite good. If need be, I can still pad down the sensitivity of the horns a bit, but I may be able to reduce it a lot by improving my cable hygiene. This second Motu unit seems to be doing fine with audio. The display actually worked really well for about 24 hours in a row, but it's back to being blank again. There's hope that I can make this thing work.

That's too bad. I forget where I read about it (and can't seem to find it on Google), but anyway. There's supposedly a camp in the Himalayas where climbers have been known to die mysteriously when storms passed through. Storms that should have been survivable for experience crews wiped them out. Investigation and interviews with survivors led to the understanding that something about the conditions in the environment caused many among the crews to literally panic and disregard precautions (like staying sheltered during the worst of weather). In the end, the best explanation offered was that nearby rock formations formed a resonant cavity through which the wind flowed through to produce intense infrasonic sounds, and that the mountaineers, lacking familiarity with high intensity ULF suffered from immense confusion, became consumed with panic, and then froze (or fell) trying to escape their predicament. I guess it sounds like a bit of an urban legend, but I think it's noteworthy nonetheless.

The capabilities I need from the Motu include the analog inputs and outputs, the USB digital audio interface, and the AVB audio support. The front panel display is not essential, but it is very nice to have because it displays both continuous average and peak hold levels for every analog input and output as well as the current sampling rate. Motu asked me to verify that the second unit is not exhibiting any of the symptoms seen in the third unit. I will probably test it more thoroughly this weekend, ensuring I've listened to sound from every output as well as operating with the AVR outputting a channel through each input. I'll test the cables with my multimeter to ensure the caps don't get energized again. I may also try to verify that a DC voltage presents itself on the third unit with it powered on. And hopefully I'll get time to do other stuff this weekend. I've got some design issues to think through before I can write more code, but after that, I hope to be able to implement the bi-quad and delay processing ASAP. With that, I can start experimenting with crossovers until I figure out what I want to do for FIRs.

Yes. DC-coupled outputs is a feature of the Motu, and I should expect the same phenomenon (energizing of DC-blocking caps in downstream equipment) when intentionally outputting DC. However, there should have never been DC on the output. I'm not using any custom software thus far. All I'm doing is routing analog inputs from my Denon AVR to analog outputs. I believe the Motu inputs do have DC blocking, and I am almost certain that my Denon AVR filters DC from its outputs. The DC is most certainly due to an electrical fault in the unit. I could try doing more testing to better characterize the problem, but unless Motu wants to pay me for doing their quality testing and troubleshooting for them, I'm not really interested messing with the unit anymore. And just to emphasize, the distortion I heard on music playback was severe and was always at least as loud as the actual music. Thus far, this third unit is the least useable of the three. I did send them an email last night describing the problems in detail and asking if I could return the third faulty unit instead of the second faulty unit. I also gave them the benefit of the doubt asked them *not* to rush ship me a fourth unit and instead to take their time to figure out the problems on their end. Receiving one defective unit is unlucky. Receiving two defective units in a row is *very unlucky*. Receiving three units in a row where the third unit was shipped to me with very expensive overnight shipping is a sign of something seriously wrong. Maybe someone just swapped the "good" units stack with the "bad" units stack by accident. I hope that's all it is, because I still really want to use this product and like this product. Nevertheless, I'm losing patience and confidence. I expect them to get this sorted out and send me a fully working unit within a week or two. Otherwise, I will be demanding a partial refund discount for unit #2 with the bad display and will actively warn others away from this product. It's just sad. I would happily pay $2000 instead of $1500 for the product if that's what it cost for them to implement some proper quality control. But I guess cheap reigns supreme in the marketplace, and in the end, we all pay for this unfortunate fact.

So Motu actually overnighted me my third A16 and made good on their promise to send a return shipping label. I hooked up the new unit and ran through my tests, using a patch cable to check for a signal on each output and the ability to sense a signal on each input. The display worked fine, and it looked like everything was great. Then I turned on my amp. The music I was playing sounded heavily distorted. The distortion sounded a bit like the digital noise you hear on crappy computer sound card outputs, but it was modulated by the signal and got louder as I increased the level of the signal input. More disturbingly, my center speaker was emitting this noise, even though I was playing music in 2 channel stereo. More testing confirmed that the center emitted this noise even with signal sent to my subs, which were turned off at this point in time. After playing with the patch cable, I was able to confirm that at least one output was producing a strong signal even when no signal was being routed to it. It gets even more fun. I turned it off and disconnected the cables in order to switch it out with unit #2, which at least has a working audio section. While working on getting all the cables situated inside my rack, I felt a sudden sharp prick in one of my fingers. Being a bit confused and thinking some kind of sharp piece of debris had found its way into my cabinet, I looked more closely and fished around with my fingers again. I felt another prick. Then I smiled, got up and fetched my multimeter. I measured like 2V across the hot and shield of one of the cables. Whoa! Never before have I felt an electric shock while handling line-level cables. Mind you, these were plugged into the *line inputs* of my OpenDRC units. After thinking for a bit, it occurred to me that I probably bridged the connection between the ring (cold) and shield with the edge of my finger while I was working and discharged DC blocking capacitors in my OpenDRC units that had been charged to *at least* 2V. So whatever electrical flaw exists in my Motu A16 #3 caused DC output at substantial levels. Awesome! I'm just a bit shocked (pardon the pun, but this really does describe my emotional state) right now and don't even know how to react. Should I demand a refund and start shopping for product from another vendor? This thing is/was to be the heart of my custom build, starting with the speakers I made. I don't even know of any comparable products on the market. If they won't give me a refund, I'll at least keep the unit with the broken display, but I am starting to worry about other problems cropping up with the unit that could damage other equipment connected to it. At least my OpenDRC units, despite their seemingly shoddy design, used DC blocking caps. Otherwise, I might have been smelling some smoke or something.

Yeah. One of the things I want to implement and use in my custom software is self-test and diagnostics. I really want my system to know within a reasonable margin of error how my system should respond in each channel in at least one key measurement location and to be able to give me a quick sanity check with a quick set of sweeps taken from that location. Right now, I have no way to easily assess whether the subs are properly balanced, or if a knob got bumped (damn cats) or simply lost its memory because it's just a cheap POS. Of course better quality equipment will help there too, but there's always opportunities for things to get knocked and nudged ... for faulty inputs to drop output in one speaker by 6 dB at random times, and so on. For large immersive systems with many channels, I believe this kind of functionality will be essential to the sanity of whoever has to maintain such a system. Apart from design work on surrounds, I also spent some time this weekend coming up with a scheme to use the last 2 outputs on my Motu A16. I'm thinking I want two sub channels up front as I have now and a total of four subwoofer channels behind my sofa. Part of my overall plan is to eventually use some or all of my subs to improve in-room accuracy of sounds above 100 Hz produced in other channels. What hadn't occurred to me until I gave it more thought is that subwoofers are large enough that the box shape matters a lot, especially for higher frequencies. And if there are multiple woofers with sufficient spacing relative to the wavelength being played, then the dispersion may become very directional and/or lobed. But wait, that's an opportunity in disguise. Imagine varying the phase difference between a pair of opposing woofers with respect to frequency. The consequence is a speaker whose directionality is made to vary with frequency in a way that's optimal for the room. In a sense, the sound can be more accurately "beamed" to where it's needed. Now try doing this with subs in multiple locations, and you can imagine how powerful this concept can be. It makes me really want to do 4 channels up front too, but I don't I'll have the 2U to spare. And of course, the cost of amps for 6 channels of sub is already astronomical. What would people think if I told them I had 48 kW of amp power (plus some spare change for the mains) in my living room?

To Motu's credit, the display issue is intermittent, which means it could have been missed by their QC. When I tried it this morning, the display worked fine for a while before reverting back to broken again. I got in touch with them and we are doing a cross-ship this time. My second replacement unit will be sent via expedited shipping and will include a return shipping label. So hopefully this mess will be resolved once and for all very soon.

Today, UPS delivered my replacement Motu A16. I got it plugged in and turned it on to something like this: That was actually snapped while I tried to update the firmware in hopes that it would fix the problem. Nope. It looks like the front display is bad on this one. On the upside, all the inputs and outputs at least appear to work. Still, receiving a replacement unit with a defect that any casual QC would have revealed does not inspire confidence. What really burns is that I can use this thing with the display broken, but it's a major nice-to-have because it has level meters for all 16 inputs and outputs. And of course, there's the principle of holding the company responsible for delivering a fully working product to me. I just hope they'll *at least* cover my side of the shipping this time around. On another note, I forgot to mention a few other things about the surround design. First, I modelled a ported version of the two TD6Hs, but I didn't like what I saw. These woofers seem to want a big box with a low tuning frequency, despite their small size. Otherwise the result is way too resonant and/or the ports can't be made to fit without unacceptable port velocities. Even in the large/low alignments, there's still some roll-off in the 100-300 Hz range due to the extreme over-damped nature of the driver. Where these woofers seem to excel is in providing a good balance of sensitivity and bass response in a very tiny box that's intended to be used in a car or other environment in which room gain makes up the difference. The other thing I worked out is that the acoustic interaction of the surround speaker with the ceiling is actually fairly simple to model by treating the ceiling as a mirror plane. Instead of thinking about reflections, I can instead consider the ceiling to be transparent and that a mirror image of the speaker exists above the plane of the ceiling. In other words, our modelled speaker is actually two horns high and four woofers tall. The mirror image is inverted, as would be expected for a mirror. In other words, the bottom of the real speaker is actually the top of the mirrored speaker. Using this analogy, I can understand how the ceiling reflections are heard by the listener by simply looking at how all the different sources (real and mirror) interfere with one another. Right away, I can say that I don't want two horns at different locations playing together. That creates a lot of nasty comb filtering. Hence, I'll want to "block out" the sound from the horn above the plane of the ceiling. I can do this by acoustically treating the part of the ceiling through which the sound from the mirrored horns would travel to the listeners. I want either good high frequency absorption or diffusion. The up-side to using diffusion to to preserve high frequency room energy and improve spaciousness of the sound. The down-side to using diffusion is that some sound still gets reflected to the listener, which blurs the otherwise sharp images produced by the images. For surrounds, the diffusion should work fine. I'm thinking 2D grid-style diffusers would look super cool and would be totally called for as a lot of HF energy hits that area of the ceiling. It gets even more interesting when I consider the four vertically-stacked woofers (two real and two mirror images). The presence of four woofers means we get double the bass output for free, which is of course what we expect with wall-ceiling corner loading. However, those four vertically-oriented woofers also have a pattern that extends to a much lower frequency, and indeed, the pattern at higher frequencies is both too narrow and full of lobes in the listening area. So we also need acoustic treatment to selectively absorb or scatter the mid-range, roughly down to a frequency at which the pattern is not so yucky. I haven't done the calculations, but my rough estimate suggests the center lobe will be -6 dB @ 90 degrees at a frequency at least as low as 500 Hz. Unfortunately, low frequency directivity like that isn't really desirable if the pattern widened somewhere between there and the crossover. I may have to want to absorb or break-up frequencies even lower than 500 Hz. That's kind of a bummer, because I'm not sure I can adequately scatter sound much lower than 1000 Hz or so. Instead, I may need to use absorption to get the best 250-1000 Hz response at the seats. Some 4" OC703 right on the ceiling would probably work great and would not absorb too much bass. One more thought is that if I could somehow take an SEOS horn and cut it perfectly down the middle, then I could mount the cut edge perfectly flush with the ceiling. The inverted mirror image would complete the half horn into a full horn again. If I could do this and opted for a different woofer layout (horizontal MTM?), I wouldn't have to treat the ceiling at all. I'd also benefit from a doubling of output across the entire spectrum. Not bad ehh? Of course, it's just a bit impractical due to what you'd have to do with the compression driver. You can't exactly cut it in half. I won't say it's impossible, but I'll leave that for someone else to experiment with.

Yeah, I'm planning an active crossover as I already have the extra amp channels. And yes, we'll see how the execution works out. However, I won't be committing to this design until I've had a lot more experience with my new front speakers. Hopefully (*fingers crossed*) my replacement Motu A16 will arrive today in full working order. Then I can get back to work on the more immediate project. Now I see the directivity plots for the B&C. I didn't arrow through the measurement plots far enough to get there. I'm not sure how to read their plots. The contours are done in solid colors, each of which are 6 dB apart, according to the colorbar. That makes for a lot of uncertainty, too much to ascertain what's really going on. If we assume the on-axis response is "0 dB" and the first color change (brown to red) is "-6 dB", then it does indeed look like the -6 dB is at +/- 40 degrees for both parts. Diffraction effects may be playing a role in making the woofer appear to hold pattern a bit lower, not that those should be discounted, but of course, the results there will depend on the baffle design. As for the the availability of the TD6H, I haven't yet enquired about it, but I take it as a given that anything ordered from Acoustic Elegance will have a long wait period. At least my TD12Ms were shipped approximately within the quoted lead time (i.e. 8 weeks), which actually surprised me!

The crossover is no wildcard. In my view, it must be considered as part of the overall design. In my calculations I found to my surprise that a driver playing even -24 dB from another driver influences its dispersion. Below 24 dB, it might not matter much, but for a crossover with 24 dB/octave electric+acoustic roll-off on each side, the crossover region where the drivers interactions are significant is full two octaves. The pair of 6.5" woofers in a vertical stack should yield pattern control in the vertical dimension to a much lower frequency than in the horizontal dimension. Indeed, the acoustic width of the woofer pair in the vertical dimension is very similar to the acoustic width of the side-ways horn, and in the crossover region, the woofer pair and horn exhibit similar vertical dispersion over a fairly wide range. Furthermore, the difference in distance between the woofer pair and the horn doesn't change much with vertical angle, so after time-aligning the drivers, the combined vertical directivity pattern should be essentially the same as for either source (the woofer pair or the horn) alone. That's good because both the horn and driver hold vertical pattern very well down to 800 Hz and exhibit some residual vertical pattern control to beyond 500 Hz. In the horizontal dimension, the crossover has a much more dramatic effect. Both the horn and woofer pair have widening dispersion that begins well above the crossover point. However, when you measure from different horizontal angles, the difference in distance between the woofer pair and horn changes a lot. At the most extreme angles, the two are about 8 inches apart, essentially the center-to-center spacing. This means that the sources interfere completely destructively (180 degrees out of phase) at a horizontal angle of 90 degrees (assuming an infinite baffle) and a frequency of 850 Hz. (This is because 8" is half the length of an 850 Hz wave.) At +/- 45 degrees and 850 Hz, the sources are .707 times as far away as at 90 degrees. This gives a phase difference of 120 degrees between the sources, and the SPL will be -6 dB down from on-axis. This means that if the crossover is at 850 Hz, the system will have 90 degrees horizontal dispersion at 850 Hz. That the horizontal directivity can be controlled to so low a frequency is not obvious. Meanwhile, the vertical dispersion is only just starting to widen there so is not much more than 90 degrees. Below that point, the dispersion will widen back out as the influence of the horn diminishes. As such, there is an advantage to being able to use shallow slopes below the crossover. Above the crossover point, dispersion will eventually narrow even more, but it may widen a bit before narrowing, again depending on what the crossover slopes look like. If the slopes are fairly shallow, then the loss of directivity as the woofer's influence diminishes will be counteracted by the gain in directivity as the dispersions of each source alone are also narrowing. I can also play a trick in which I use a slight offset in time-alignment between woofers and horn to steer the main lobe more toward the rear of the room than the front. I can't do that at all with a coaxial. I do like the looks of the woofer in that coax, but it wants a bigger box than I can give it. Another concern I have is that the woofer response starts looking ragged before it even hits 1 kHz, and it's designed to cross at 1.2 kHz. I'm not sure what makes you think it has 80 degrees dispersion from 500Hz-20kHz. On the low end, a piston radiator hits -6 dB @ 90 degrees when the wavelength equals the diameter. Assuming that 15" woofer has 12.5" of radiating surface, that frequency is 908 Hz. However, woofers with a concave cone tend to behave acoustically like they are smaller than they really are, so the real 90 degree point for this woofer may be closer to the recommended crossover point @ 1.2 kHz. If we model the concave 15" as an 11.25" piston instead (which gives 90 degrees dispersion @ 1.2 kHz), we see the dispersion widens to 120 degrees at 985 Hz, and 150 degrees at 875 Hz. It's practically omni at 500 Hz. In contrast, my design can keep the vertical dispersion inside of 90 degrees all the way down to 875 Hz and is able to hold some residual pattern lower still. My calculations suggest my woofers are 90 degrees out of phase at +/- 60 degrees vertical and 500 Hz, which is -3 dB from on-axis. The 11.25" piston is only -1.4 dB at +/- 60 degrees and 500 Hz. (Or -1.7 dB for a 12.5" piston, in case that mattered.) Of course, these calculations aren't perfect because the pair of woofers is treated as a pair of infinitely small point sources, but they probably still have a bit more acoustic width, vertically speaking, than the 15". In the 500-1 kHz range, the combined effects of diffraction and reflections from the wall and ceiling whose corner the speaker is installed into are of similar if not greater importance to the dispersion pattern than the woofer and horn directivity. In my view, the baffle and box design is the hard part and is where I need to focus more of my attention. It's job will be to essentially provide a clean transition between the SEOS-15 horn and the "horn" formed by the wall-ceiling junction itself while keeping the construction sane. That is no easy task!

Thanks for the additional info on the Radian. I wish we could learn more about what was adversely affecting the sound of his heights. How often to heights get used for voices anyway? The "harsh gritty edge" imparted to voices must have been pretty bad. I wonder if there was a box resonance problem or something. Or maybe the crossover had serious issues. We are all left to guess. Turning the SEOS on its side does "blast a whole of output vertically", but it's no different from using a coaxial because both have about 90 degree dispersion in that direction. The difference is that the side-ways SEOS will hold that pattern to a lower frequency and will provide a tighter horizontal pattern. Either way, I can take advantage of all that energy being directed to the ceiling by putting diffusers up there to scatter it around the room. Note that my setup is unique, and I believe it is unusually well-suited to use of a side-ways SEOS horn configuration. My room is set up with the MLP about 10 feet from the front wall and with the surrounds placed at 90 degrees from center, about 9 feet away. My sofa is a odd-shaped sectional with a very gently curve. The end seats are about 1.5 feet forward and about 5 feet to each side of the MLP. This places all the listeners roughly in line with each surround speaker, and my initial estimates suggest I can achieve a remarkably good level balance between the two surrounds at all of the seats by using a 90 degree dispersion pattern. (The sofa design also makes for a comfortable viewing angle to the TV from each seat). Both the coaxials and side-ways SEOS horns provide a 90 degree dispersion pattern across the listener seats, but as I said, the SEOS will control the front wall early reflections better. What I see as the downside to using the SEOS versus the coaxial in my application is that the SEOS has more direcitivity and so will provide less overall energy for the same SPL. For home theater, it is often argued that surrounds should have a more diffuse sound compared to the fronts, and less directivity definitely helps with that, but I don't think it's likely to make that big of a difference here. Edit: I forgot to add that most of the TD6H Klippel results are posted here: http://aespeakers.com/forums/topic/td6h-klippel-results-from-diymobileaudio-com/. It looks like the 6mm Xmax figure is legit. It also looks like its excursion is probably usable out to more like 8 mm, especially if used in a box that's smaller than Vas, at which point the air spring is more important than the suspension for determining the overall compliance of the system.

I hadn't read those threads yet. Thanks! I get the impression that bass addict wasn't so much dissatisfied with the 2 X TD6 speakers he built for "heights" as he was dissatisfied with the overall result, particularly the TD-15M + SEOS-12 + BMS-4550 builds he did. I can try to guess at what he might of done wrong, but it's largely speculation without some good measurement data. Another important point is that the design uses TD6M drivers. I don't know why anyone would try to use TD6Ms as woofer because they really are mid-range drivers. Notes from mtg90 on the DIYSG forum indicated that the mid-range was brought way down to deal with the bass roll-off resulting in a sensitivity around 92 dB/2.83V and 4 ohm impedance. That's worse than what I have now. For my design, I planned to use TD6Hs instead. They have a lot more excursion capability as well as more mass and motor strength for use as actual woofers. I sit about 3 m away. You would be correct to say that my build is a bit overkill for reference level playback. But that hardly puts me in strange company around here. I want something that'll keep up with my mains and deliver excellent sound quality. The TD6H pair with help from the corner can probably do about 3 dB less than my TD12M mains. I know it's typically argued that surrounds need not be as big, but I do want a lot of headroom. For one thing, I'll be sticking to 5 mains channels in this room, so the surrounds will have to handle sides and rears on 7.1 tracks. For another, Atmos tracks and tracks originating from Atmos theatrical mixes have a lot more bass in the surrounds because of the surround bass management that theatrical playback systems added for Atmos. Third, I want to use the surrounds to help with room EQ, even for sounds that originate in the mains speakers. As such, I really wish I could make a pair of 8s work, but that causes too many other issues in the design. Thanks a lot to you, Ricci, and 3ll3d00d for your suggestion. If I decide to go with a coax, I'll definitely lean heavily in that direction. I do wish they made a 4 ohm version so I could get a bit more power into them. Both this and the Eminence have sensitivity of 95 dBish. I expect a TD6H pair to yield 97 dB. Of course, the Radian probably has a bit more bass if I end up needing it, but I can't find an Xmax number for it. Can anyone help? The other reservation I have is that I'm pretty sure I want the tighter horizontal dispersion of the SEOS-15 horn turn on its side. I'm fairly certain it will help a lot with front wall reflections I have. I know because I've done a lot of acoustic measurements already with speakers that are already installed in the planned location.