"What I love the most about the line arrays is that anywhere I go, the sound is there. Stand up, sit down, side to side, other side of the room. I want play for the other side of the house and when I walk back to the listening room it doesn't murder my ears as I walk closer like my horns do."
This sounds great to me, a very definite positive appeal
"The location that worked best for imaging for me was 1/3rd of the way into the room."
But this doesn't! -no can do! - has to be the corners, although the left hand side is really more like mid back wall, the speaker front will be at a similar point to the pillar front.
Back to other practicalities, will treating the right side wall between the corner and the window, which covers most of the first reflection points, likely do most of what I need i wonder?
When you use Dirac, is it connected in loop full time or is it used to set up and then not necessary, like using REW as i did before.
This sounds great to me, a very definite positive appeal
"The location that worked best for imaging for me was 1/3rd of the way into the room."
But this doesn't! -no can do! - has to be the corners, although the left hand side is really more like mid back wall, the speaker front will be at a similar point to the pillar front.
Back to other practicalities, will treating the right side wall between the corner and the window, which covers most of the first reflection points, likely do most of what I need i wonder?
When you use Dirac, is it connected in loop full time or is it used to set up and then not necessary, like using REW as i did before.
If your arrays are built tightly into the corners, the first reflection points will be close to the corners, and you won't need very wide absorber panels to tame them. You don't need to treat the entire wall, just near the array where sound diffracts off the baffle edge, onto the wall and out to you. Without that treatment you will get nulls where that small path length distance equals half a wavelength. One hopes that the reflection from the opposite side wall is delayed long enough to add to the ambience and thus not require treatment.
That isn't to say there won't be reflections off of objects like the fireplace protrusion, which would also benefit from treatment.
Attached is a sketchup export of what I built - an array of 32 2.5" drivers with 4" thick absorbers butted up to its sides. The sides are asymmetrical to shift the aiming point . If you leave some space between the array and the absorber, you can shift/fine tune the aim without losing effectiveness against those first reflections. With TC9s, this would work very well for you at very reasonable cost. The array cabinet could extend all the way back to the walls to provide more volume for the drivers, although I chose not to do that myself because I had designed it to work with subs from 80 or so Hz and up.
That isn't to say there won't be reflections off of objects like the fireplace protrusion, which would also benefit from treatment.
Attached is a sketchup export of what I built - an array of 32 2.5" drivers with 4" thick absorbers butted up to its sides. The sides are asymmetrical to shift the aiming point . If you leave some space between the array and the absorber, you can shift/fine tune the aim without losing effectiveness against those first reflections. With TC9s, this would work very well for you at very reasonable cost. The array cabinet could extend all the way back to the walls to provide more volume for the drivers, although I chose not to do that myself because I had designed it to work with subs from 80 or so Hz and up.
Last edited by a moderator:
We are at fundamental odds on both of these points……and with good cause.
Not sure what musings you have read about crossovers….or lack there of. In a situation where there were a single, point Source full range driver in a well designed cabinet playing less complex music at ambient to low listening levels in the near to mid field……yes…..you’d have my attention as to viability. Once you add multiple drivers playing the same source, the argument falls apart as you’ve compromised the primary acoustic function of the point source……..
Bass……we’re not talking subsonics here and the word gets thrown around so far out of context on forums to make it nearly impossible to define. Moving or displacing air is essential……percussive live performances with midbass impact won’t happen with multiples of the tiny SD of 3.5” drivers…..there’s simply no displacement from the linear movement of the cone. Yes, you’ll measure bass response down to the nethers…..but it’s lifeless…..there‘s no startling midbass slam…..no startling pressurization of the room when Bonham hits that 26” kick.
Without being condescending, I think maybe you and Wesayso share an aesthetic more so than the performance……a minimalist approach so to speak. And that’s OK…..the visual is important to the overall performance factor.…..it’s just that you have to work extremely hard through what I find are far too many compromises to get there.
Last edited:
Let’s start with the unavoidable here
That’s the TC9 response……not unlike any other 3-4” driver…..cone breakup due to size is inevitable as is off axis behavior ………everything above 10k is a mess and demands extreme measures to correct. Of axis behavior?…….the graph says more than I ever could with words…….the rationale for a crossover less implementation falls apart completely after 5khz.
Again, combine four of these drivers in an MMTMM on a very slim free standing baffle with suitable bass support?……you have my attention…….a 4” wide and deep slim tower out 18” from the side and back walls?…….yep…..now we’re creating a soundstage with reasonable effort. Choose the right tweeter…….build it around a hybrid active/passive system with a .5 inductor rolloff on the outer TC9 drivers?……..yep. Cross the whole top over to a pair of 8” woofer per side with a Hypex sub amp at around 400hz?……..oooooooh…….now we’ve got some serious power handling out of just four TC9s and some wonderful midrange clarity happening……and a focused ‘line source’ too!……we’ve narrowed the vertical directivity within a passband that’s workable.
Build it artistically following linear minimalism?……..sure……….a tapered enclosure with a mild curve to the baffle to align the drivers acoustic centers?……oooh….now we’ve got form and function going on…..how bout that! A nice matte white finish with a charcoal baffle ……some nylon chord grills aka Sonus Faber?……..some eye candy.
That’s the TC9 response……not unlike any other 3-4” driver…..cone breakup due to size is inevitable as is off axis behavior ………everything above 10k is a mess and demands extreme measures to correct. Of axis behavior?…….the graph says more than I ever could with words…….the rationale for a crossover less implementation falls apart completely after 5khz.
Again, combine four of these drivers in an MMTMM on a very slim free standing baffle with suitable bass support?……you have my attention…….a 4” wide and deep slim tower out 18” from the side and back walls?…….yep…..now we’re creating a soundstage with reasonable effort. Choose the right tweeter…….build it around a hybrid active/passive system with a .5 inductor rolloff on the outer TC9 drivers?……..yep. Cross the whole top over to a pair of 8” woofer per side with a Hypex sub amp at around 400hz?……..oooooooh…….now we’ve got some serious power handling out of just four TC9s and some wonderful midrange clarity happening……and a focused ‘line source’ too!……we’ve narrowed the vertical directivity within a passband that’s workable.
Build it artistically following linear minimalism?……..sure……….a tapered enclosure with a mild curve to the baffle to align the drivers acoustic centers?……oooh….now we’ve got form and function going on…..how bout that! A nice matte white finish with a charcoal baffle ……some nylon chord grills aka Sonus Faber?……..some eye candy.
Last edited:
So, if its about cone area, then one off 12" is about 113 square inches, whereas a 3 1/2" is 8 1/3", so if there are 25 of them surely they can do the same thing with ten to spare at low excursion, actually i take your point and although i don't do Bonzo and this is a lowish volume environment, is this a real problem for me? by the way, with my 3 ways i rarely see significant bass unit excursion with for example a big pipe organ at the volumes i use in the bigger room. Is it about absolute concurrent displacement? - or are there other matters afoot?
That also all sounds appealing. No idea how to do that though, by the way please, what is "we’ve narrowed the vertical directivity within a passband that’s workable" - i am not uninterested in what you describe but am more likely to stay in the realms that others have already done. one assumes this has the same first reflection question?
So let’s address first reflections…….above 10khz…..who cares…..drywall and most flooring sucks it all up. Below…….and area rug does the trick. Ceiling reflections?……..that’s where the MMTMM design comes in….. no different than the array…..still line sources. Two drivers playing the same frequency on the same axis ( this case vertical) will be effected by comb filtering which narrows the beam. In the case of the MMTMM i‘ve outlined, that’s a physical array around 18” in length…..then expanded considering the expected directivity out to 20 degrees into the room……your talking about even coverage vertically of about 30” at the LP………center the tweeter at 36” or so off the floor and now your fully covered to 2 meters standing…….for those that need to dance around….silly stuff folks clamor on about.
The MMTMM……the adjacent MM to the T play the entire passband up to the crossover point……but if you use a narrow baffle, the low end will roll off as there’s no re enforcement from the baffle….that’s a loss. We can combat that and a few other things by adding the two outer MMs……but they need to be cut off lower with a simple 1st order passive filter……the added response below the baffle step compensates……and the lower cutoff eliminates combing due to the longer center to center distance. Power handling and efficiency benefit as well.
Your two adjacent middle MMs can play up pretty high at 3.5” before beaming starts……..you get to choose a small dome here as you don’t need to cross low AND you get the advantage of the wider directivity of the smaller dome. The SBAcoustics SB19 would be a great choice with a near flat polar response out to 13k or so.
The MMTMM……the adjacent MM to the T play the entire passband up to the crossover point……but if you use a narrow baffle, the low end will roll off as there’s no re enforcement from the baffle….that’s a loss. We can combat that and a few other things by adding the two outer MMs……but they need to be cut off lower with a simple 1st order passive filter……the added response below the baffle step compensates……and the lower cutoff eliminates combing due to the longer center to center distance. Power handling and efficiency benefit as well.
Your two adjacent middle MMs can play up pretty high at 3.5” before beaming starts……..you get to choose a small dome here as you don’t need to cross low AND you get the advantage of the wider directivity of the smaller dome. The SBAcoustics SB19 would be a great choice with a near flat polar response out to 13k or so.
Surface area is two dimensional displacement……..let’s examine the 3rd dimension of displacement…..and that’s the cones linear travel………take the 3mm of the TC9 and do the math…………now take a long excursion 12” of say 12-16mm linear travel……….same equations……..ooooops……..the compensated surface area argument falls apart pretty quickly. Your woofer is also located close to the floor……….boundary gain adds another helping hand.
I am benefitting enormously from all your inputs, whether agreeing or disagreeing, it is a good way to see the internet used.
I wrote recently I need to develop confidence that the results will be satisfactory, it's not blue sky development for the sake of it, hence I started this thread, and would add to that, my style of judgement and decision making involves gathering sufficient information to feel confident (whether justified or not!) and then taking a decision, to explain, if considering something that matters to me, i don't jump to conclusions. Hence i am unlikely to be a trailblazer in a complicated area, but more likely to refine solutions and constructions in areas where others are innovating.
Whilst i can visualise some of what Mayhem describes, its likely that my decision making style will stop me going down that sort of route this time. If ever I get to a point where exploring what are (for me) unproven alternatives then maybe that could follow.........I have done that with my tonearm over the last years.
So, although i don't have to understand this aspect please indulge me and explain more about the sound heard direct and first or dominant reflections. My mind is a mess over this.
The array is a whole series of sources behaving in some ways as one, however in the vertical plane they are not equidistant from the ear, in fact similarly less equidistant in the vertical plane than the first reflection route of the proposed corner mounted unit. If we are 4m away and the top transducer is 1m higher than ear height, the signal is travelling an extra 12 cms, (depending on the final design) the first reflection might be similar, in what way is the first reflection different.
I have a feel that with the signal spread over 25 units the effect of each will be reduced and so floor and ceiling reflections will not have one dominant type like from a single point/unit, and this can be expanded to include the horizontal routes from each unit. i imagine this aspect can be harnessed to argue that defects are less important and also that strong points are less emphasised, but that doesn't fit with the strong views of knowledgeable exponents.
Writers have talked of seeking to hear the direct wave only or predominantly so within a certain time scope, which implies to me this should all be at coincident times, whereas my (probably flawed) logic says that the array doesn't do that unless curved like some are, but most exponents don't do that either?
I wrote recently I need to develop confidence that the results will be satisfactory, it's not blue sky development for the sake of it, hence I started this thread, and would add to that, my style of judgement and decision making involves gathering sufficient information to feel confident (whether justified or not!) and then taking a decision, to explain, if considering something that matters to me, i don't jump to conclusions. Hence i am unlikely to be a trailblazer in a complicated area, but more likely to refine solutions and constructions in areas where others are innovating.
Whilst i can visualise some of what Mayhem describes, its likely that my decision making style will stop me going down that sort of route this time. If ever I get to a point where exploring what are (for me) unproven alternatives then maybe that could follow.........I have done that with my tonearm over the last years.
So, although i don't have to understand this aspect please indulge me and explain more about the sound heard direct and first or dominant reflections. My mind is a mess over this.
The array is a whole series of sources behaving in some ways as one, however in the vertical plane they are not equidistant from the ear, in fact similarly less equidistant in the vertical plane than the first reflection route of the proposed corner mounted unit. If we are 4m away and the top transducer is 1m higher than ear height, the signal is travelling an extra 12 cms, (depending on the final design) the first reflection might be similar, in what way is the first reflection different.
I have a feel that with the signal spread over 25 units the effect of each will be reduced and so floor and ceiling reflections will not have one dominant type like from a single point/unit, and this can be expanded to include the horizontal routes from each unit. i imagine this aspect can be harnessed to argue that defects are less important and also that strong points are less emphasised, but that doesn't fit with the strong views of knowledgeable exponents.
Writers have talked of seeking to hear the direct wave only or predominantly so within a certain time scope, which implies to me this should all be at coincident times, whereas my (probably flawed) logic says that the array doesn't do that unless curved like some are, but most exponents don't do that either?
I think I recognize the guy that spoke about that first wave front. 
I've corrected speakers with FIR filters at the listening seat, what that processing does is trying to reconstruct a perfect IR from any measurement made.
For that, Ive used DRC-FIR. Now to hammer down the IR towards that ideal goal would be a cure idea but doesn't work that well in practice. That's why there are many variables in DRC-FIR to guide its process along the way. You can set all kinds of filters that prevent over-processing of that signal. I've experimented with that for a long long while, just to study/learn from it. Processing the input IR with a simple IR reversal would re-create a perfect impulse from whatever the input was, so why all those variables? It is upon the user to determine what you want as the outcome. Which made me create filters that varied in length depending on frequency. I could use slightly longer filters at higher frequencies, as that's where most of the smearing of the signal would occur. At mid frequencies, where the driver to listeners ear differences in distance are way more favourable, the correction could be made shorter in time-span. And finally, the bottom end, where wave lengths are so long that they include whatever the room has to offer, I made those filters slightly longer (relatively speaking) to straighten out some of the room. Mind you, there's still all those extra filters at play. Like filters that prevent the FIR creation trying to boost a huge dip (can't be done after all). Many more decisions are present in that DRC-FIR-Package. If interested I'd say read the manual. What I've experimented with is what I could hear while playing with those variables. I've played extensively over a half a year period, to find boundaries and acceptable compromises (for me). The consensus I made is that I wanted the minimum amount of manipulation that still sounds good. So not forcing things back into submission, but more of a feathering.
Those same tools I have used with my simulations done in VituixCAD. So we could examine each of them without the room effects.
Here's what the IR looks like from an array as measured RAW:
And the SPL graph that belongs to that:
To get things into the window of using DRC-FIR, I'd have to apply some pré-EQ, just to make sure my output that I want to correct isn't filtered out because it is below a certain threshold of the filters I've set. So that changes the IR completely, because we boost the high frequencies by a lot.
The IR after pré-EQ:
And the belonging SPL graph:
That makes it pretty clear to see all those arrival differences, doesn't it? Mainly in the IR, we now see about a 0,7 ms hash in the STEP response, mainly consisting of high frequency arrivals fighting each other. The SPL graph shows the result of that, many dips and peaks on that top end. From 7 Khz and up the pattern shows us the combing.
After that, DRC-FIR does it's thing, again first the IR:
We see that top end is more gradual and arrivals are harder to see (but still there)
And the beloning SPL graph:
Dips aren't prevented completely, but the top end is corrected to nearly flat.
We do see quite a bit of movement in front of the peak in the IR though, that must be pré-ringing. So at what frequencies is that happening?
Let's filter the IR to look at what happens before the (official) IR start:
The blue parts are filtered out, we look at what happens before the peak. This gets us the following SPL graph:
This shows us the preliminary wiggles that are seen in that IR are dominated by high frequencies.
I never stopped there, after an automated edit like DRC-FIR applies I've always manually edited out the large boosts at high frequencies and also balanced the left and right channels manually as well. Just trying to show what happens at each step here. For these examples I just used what I had on hand about unfiltered arrays as far as VituixCAD goes.
This above is a quick example of an edit of the RAW measured IR towards one that has been shaped into a more desirable frequency curve. It isn't quite ideal and one of the reasons for me to change my approach a bit over the years to get cleaner results with less pré-ringing. I've mentioned that change in processing here on my web-page: https://www.vandermill-audio.nl/drc-fir-software/. The main differences are in what I tell DRC-FIR to do. I make it act like a minimum phase tool and correct any phase manipulation I'd like to apply with a tool like RePhase. In the end it provides a nearly pré-ring free end result, as can be seen in my thread more than once. An example of my current "filtered" arrays has been posted not that long ago and looks like this:
Any pré-ringing seen here is the change in sample rate from 44100 to 96000 that I apply inside JRiver. The wiggles after the peak are in fact room stuff and reflections. This is a sample of both arrays firing and measured at the sweet spot chosen as the listening spot. You can still see some of the arrival differences in that same 0 to 700 uS arrival times. But way less wild than it once was. And it's all behind the main pulse.
This has become long. I'll first press "Post reply" before I regret it.
I've corrected speakers with FIR filters at the listening seat, what that processing does is trying to reconstruct a perfect IR from any measurement made.
For that, Ive used DRC-FIR. Now to hammer down the IR towards that ideal goal would be a cure idea but doesn't work that well in practice. That's why there are many variables in DRC-FIR to guide its process along the way. You can set all kinds of filters that prevent over-processing of that signal. I've experimented with that for a long long while, just to study/learn from it. Processing the input IR with a simple IR reversal would re-create a perfect impulse from whatever the input was, so why all those variables? It is upon the user to determine what you want as the outcome. Which made me create filters that varied in length depending on frequency. I could use slightly longer filters at higher frequencies, as that's where most of the smearing of the signal would occur. At mid frequencies, where the driver to listeners ear differences in distance are way more favourable, the correction could be made shorter in time-span. And finally, the bottom end, where wave lengths are so long that they include whatever the room has to offer, I made those filters slightly longer (relatively speaking) to straighten out some of the room. Mind you, there's still all those extra filters at play. Like filters that prevent the FIR creation trying to boost a huge dip (can't be done after all). Many more decisions are present in that DRC-FIR-Package. If interested I'd say read the manual. What I've experimented with is what I could hear while playing with those variables. I've played extensively over a half a year period, to find boundaries and acceptable compromises (for me). The consensus I made is that I wanted the minimum amount of manipulation that still sounds good. So not forcing things back into submission, but more of a feathering.
Those same tools I have used with my simulations done in VituixCAD. So we could examine each of them without the room effects.
Here's what the IR looks like from an array as measured RAW:
And the SPL graph that belongs to that:
To get things into the window of using DRC-FIR, I'd have to apply some pré-EQ, just to make sure my output that I want to correct isn't filtered out because it is below a certain threshold of the filters I've set. So that changes the IR completely, because we boost the high frequencies by a lot.
The IR after pré-EQ:
And the belonging SPL graph:
That makes it pretty clear to see all those arrival differences, doesn't it? Mainly in the IR, we now see about a 0,7 ms hash in the STEP response, mainly consisting of high frequency arrivals fighting each other
. The SPL graph shows the result of that, many dips and peaks on that top end. From 7 Khz and up the pattern shows us the combing.After that, DRC-FIR does it's thing, again first the IR:
We see that top end is more gradual and arrivals are harder to see (but still there)
And the beloning SPL graph:
Dips aren't prevented completely, but the top end is corrected to nearly flat.
We do see quite a bit of movement in front of the peak in the IR though, that must be pré-ringing. So at what frequencies is that happening?
Let's filter the IR to look at what happens before the (official) IR start:
The blue parts are filtered out, we look at what happens before the peak. This gets us the following SPL graph:
This shows us the preliminary wiggles that are seen in that IR are dominated by high frequencies.
I never stopped there, after an automated edit like DRC-FIR applies I've always manually edited out the large boosts at high frequencies and also balanced the left and right channels manually as well. Just trying to show what happens at each step here. For these examples I just used what I had on hand about unfiltered arrays as far as VituixCAD goes.
This above is a quick example of an edit of the RAW measured IR towards one that has been shaped into a more desirable frequency curve. It isn't quite ideal and one of the reasons for me to change my approach a bit over the years to get cleaner results with less pré-ringing. I've mentioned that change in processing here on my web-page: https://www.vandermill-audio.nl/drc-fir-software/. The main differences are in what I tell DRC-FIR to do. I make it act like a minimum phase tool and correct any phase manipulation I'd like to apply with a tool like RePhase. In the end it provides a nearly pré-ring free end result, as can be seen in my thread more than once. An example of my current "filtered" arrays has been posted not that long ago and looks like this:
Any pré-ringing seen here is the change in sample rate from 44100 to 96000 that I apply inside JRiver. The wiggles after the peak are in fact room stuff and reflections. This is a sample of both arrays firing and measured at the sweet spot chosen as the listening spot. You can still see some of the arrival differences in that same 0 to 700 uS arrival times. But way less wild than it once was. And it's all behind the main pulse.
This has become long. I'll first press "Post reply" before I regret it.
Last edited:
The multiway mayhem is arguing for is not your best solution. It might be best for someone else or a different room but unless it's also a cardioid, its either going to need to be pulled out into the room or suffer from boundary interference. If you give it the large diameter woofer that we all would prefer absent space constraints, its going to have a fairly large footprint.
really, the only solution that can be nestled into a corner and sound excellent is a full range driver line array. You can have confidence that a line array will work well for you. It may not be the only solution but it is a good one and accessible to you. Someday you might make your line arrays more robust by hiding flat subs under or behind your chairs but even without them it will have good bass at reasonable levels. You could also improve them with shading, as did Wesayso his but only after multiple years of enjoying them thoroughly without shading.
Yes the discrete line array response is the sum of the responses of each individual driver, each response arriving with a slightly different delay than the ones next to it. This accounts for ripple in the frequency response that decreases in magnitude with the spacing between drivers and leads to audible combing if you listen too close (less than 2m or so). When you consider reflections, you have separate reflections from each driver that add or subtract from the composite response according to their time of arrival and thus tending to average out, leading to a nearly flat response after equalization for the -3db/octave HF roll off of a line array. You probably have found threads on the forum that analyze this. If you are so inclined, you could write a spreadsheet that does the response summing. You can also find that spreadsheet online here...
really, the only solution that can be nestled into a corner and sound excellent is a full range driver line array. You can have confidence that a line array will work well for you. It may not be the only solution but it is a good one and accessible to you. Someday you might make your line arrays more robust by hiding flat subs under or behind your chairs but even without them it will have good bass at reasonable levels. You could also improve them with shading, as did Wesayso his but only after multiple years of enjoying them thoroughly without shading.
Yes the discrete line array response is the sum of the responses of each individual driver, each response arriving with a slightly different delay than the ones next to it. This accounts for ripple in the frequency response that decreases in magnitude with the spacing between drivers and leads to audible combing if you listen too close (less than 2m or so). When you consider reflections, you have separate reflections from each driver that add or subtract from the composite response according to their time of arrival and thus tending to average out, leading to a nearly flat response after equalization for the -3db/octave HF roll off of a line array. You probably have found threads on the forum that analyze this. If you are so inclined, you could write a spreadsheet that does the response summing. You can also find that spreadsheet online here...
What that manipulation of the IR does in real life can be seen more clearly in the APL-TDA graphs. That is a tool, producing a graph as created by @Raimonds to show tiny timing differences.
After fine tuning my processing for unshaded arrays, my APL-TDA graph, as measured in-room looked like this:
You still see a sort of step-up above 10 KHz, but this graph here kind of functions as my "proof of concept" of that first wave-front arriving "in tact" at the listening spot. Prior to running APL_TDA I had never seen this in such detail. Mr. Raimonds Skurels had posted about his useful tool in the Full-range forums and I simply ran his Demo version of that tool to see what I got. That story is here from back in 2016: https://www.diyaudio.com/community/...can-we-do-with-dsp-power-now-availabl.284916/
I've made a comparison of that correction from 2016 to my more recent frequency shaded array here: https://www.vandermill-audio.nl/shaded-arrays-vs-unshaded-arrays/
While the differences can be seen and may seem large, in all honesty: yes there are noticeable differences but not as large as the graphs may make you think.
During this period in time I've also moved towards using sub-woofers to back up the low end and changed from the Vifa TC9 towards the Scan Speak 10F.
I've had a lot of fun with the unfiltered TC9 arrays and the later projects were more of a "Can I improve" or "Should I improve?" kind of thing than that is was born out of necessity. In my smallish room (about 4m x 8.3m, using half that as listening area) even the TC9 array all by itself provided all the bass I needed.
I remember a post from @koldby about bass in my room, which was definitely true, I did postpone my subwoofer build for a long time due to the excellent results of the arrays by themselves. The Scan Speak 10F I now use cannot do something similar! The trick was: the small room, close to back wall placement, enough enclosure volume and DSP.
I didn't arrive at that result in one day. I really had to learn how to get there, and wrote about it so others could benefit from my hard learned lessons. I've helped many others achieve something similar, both in the threads here as well as in many private messages.
On the subject of high frequency directivity in the room of an array of drivers. I just have to say a couple of things about it. I've spoken about my back-up plan by sticking a tweeter in the middle. I just never needed it. Want to know why? In my listening area the arrays are using toe-in that I determined in listening sessions. I simply chose the toe-in that did sound best to me. Wherever we sit (or stand) in our room to enjoy the speakers, we get satisfying high frequency results as our ears are never really further off axis than about 20 degree at the main in use positions. The further drop off-axis would have a larger effect if I had not used absorbing panels at first reflection points. Maybe then I'd hear/perceive that drop in output that the 3.5" drivers show in off axis graphs as posted here. But there's another que that's important to notice. Look at the graphs of directivity that VituixCAD produces for us:
In the left-middle graph look at that red solid line. That's the DI index. A prediction of directivity in-room. Due to the spray of the arrays to the ceiling, you know, what we called "lobes", the directivity stays somewhere between 10 and 15, which isn't bad at all. The most important part is that it climbs gradually when going towards the higher frequencies. That's how we perceive the speaker including that high end. We don't particularly hear the combing as something separate at higher frequencies.
I've already compared and shown the result of a single driver compared to an array. A single driver shows even more in-room combing than the array "as perceived" from floor and ceiling reflections alone!
So it isn't always that straight forward to say how it works until we take a closer look. I'd love to see @mayhem13 produce a DI curve for his proposal, just to see if he can arrive at an as smooth climbing curve as an array actually produces. But I'm not the one that's going to make such a comparison. Believe me, as the tools exist, we tried to include a single tweet (and expecting great results) but it sure isn't that simple. There's a good reason I never went that route. If you've seen what I've done over the years one conclusion might be appropriate: I wasn't afraid of the amount of work to not try a single tweeter. As I've done many time consuming projects if I was convinced it could mean I'd achieve a true improvement. I just never was convinced an added tweeter could give me that. While having a great set of ribbon tweeters on hand! I still might use those tweeters at one point, just not in the middle of the array.
Is this the single best speaker one can build? No, I don't think so. But it still is the best compromise I could come up with to suit my needs in my room. And the end result sure was rewarding and satisfying. My curiosity will probably never see me stop trying to improve, but I could have stopped all that back in ~2016 and be happy with the results too. I just like to tinker, I'm wired that way.
After fine tuning my processing for unshaded arrays, my APL-TDA graph, as measured in-room looked like this:
You still see a sort of step-up above 10 KHz, but this graph here kind of functions as my "proof of concept" of that first wave-front arriving "in tact" at the listening spot. Prior to running APL_TDA I had never seen this in such detail. Mr. Raimonds Skurels had posted about his useful tool in the Full-range forums and I simply ran his Demo version of that tool to see what I got. That story is here from back in 2016: https://www.diyaudio.com/community/...can-we-do-with-dsp-power-now-availabl.284916/
I've made a comparison of that correction from 2016 to my more recent frequency shaded array here: https://www.vandermill-audio.nl/shaded-arrays-vs-unshaded-arrays/
While the differences can be seen and may seem large, in all honesty: yes there are noticeable differences but not as large as the graphs may make you think.
During this period in time I've also moved towards using sub-woofers to back up the low end and changed from the Vifa TC9 towards the Scan Speak 10F.
I've had a lot of fun with the unfiltered TC9 arrays and the later projects were more of a "Can I improve" or "Should I improve?" kind of thing than that is was born out of necessity. In my smallish room (about 4m x 8.3m, using half that as listening area) even the TC9 array all by itself provided all the bass I needed.
I remember a post from @koldby about bass in my room, which was definitely true, I did postpone my subwoofer build for a long time due to the excellent results of the arrays by themselves. The Scan Speak 10F I now use cannot do something similar! The trick was: the small room, close to back wall placement, enough enclosure volume and DSP.
I didn't arrive at that result in one day. I really had to learn how to get there, and wrote about it so others could benefit from my hard learned lessons. I've helped many others achieve something similar, both in the threads here as well as in many private messages.
On the subject of high frequency directivity in the room of an array of drivers. I just have to say a couple of things about it. I've spoken about my back-up plan by sticking a tweeter in the middle. I just never needed it. Want to know why? In my listening area the arrays are using toe-in that I determined in listening sessions. I simply chose the toe-in that did sound best to me. Wherever we sit (or stand) in our room to enjoy the speakers, we get satisfying high frequency results as our ears are never really further off axis than about 20 degree at the main in use positions. The further drop off-axis would have a larger effect if I had not used absorbing panels at first reflection points. Maybe then I'd hear/perceive that drop in output that the 3.5" drivers show in off axis graphs as posted here. But there's another que that's important to notice. Look at the graphs of directivity that VituixCAD produces for us:
In the left-middle graph look at that red solid line. That's the DI index. A prediction of directivity in-room. Due to the spray of the arrays to the ceiling, you know, what we called "lobes", the directivity stays somewhere between 10 and 15, which isn't bad at all. The most important part is that it climbs gradually when going towards the higher frequencies. That's how we perceive the speaker including that high end. We don't particularly hear the combing as something separate at higher frequencies.
I've already compared and shown the result of a single driver compared to an array. A single driver shows even more in-room combing than the array "as perceived" from floor and ceiling reflections alone!
So it isn't always that straight forward to say how it works until we take a closer look. I'd love to see @mayhem13 produce a DI curve for his proposal, just to see if he can arrive at an as smooth climbing curve as an array actually produces. But I'm not the one that's going to make such a comparison. Believe me, as the tools exist, we tried to include a single tweet (and expecting great results) but it sure isn't that simple. There's a good reason I never went that route.
If you've seen what I've done over the years one conclusion might be appropriate: I wasn't afraid of the amount of work to not try a single tweeter. As I've done many time consuming projects if I was convinced it could mean I'd achieve a true improvement. I just never was convinced an added tweeter could give me that. While having a great set of ribbon tweeters on hand! I still might use those tweeters at one point, just not in the middle of the array.Is this the single best speaker one can build? No, I don't think so. But it still is the best compromise I could come up with to suit my needs in my room. And the end result sure was rewarding and satisfying. My curiosity will probably never see me stop trying to improve, but I could have stopped all that back in ~2016 and be happy with the results too. I just like to tinker, I'm wired that way.
Last edited:
I wish this were true, but it isn’t……the wavefront interactions that each drive unit have with one another are profound and shape the behavior of the array into something else.
two adjacent TC9’s will have a center to center spacing of roughly 100mm. We know from the work of others decades ago that the wavefront interference begins to shape and cancel the HF response at the 1/4 wavelength and increases with frequency……so around the full wavelength turn around 3300hz, the response of the drivers begins to roll off significantly.
If one were to build a model 2 way array, this would be the near ideal place to crossover to a line of tweeters (read very expensive). Instead with the full range solution, ……..the ‘solution’ becomes extensive equalization to flatten the overall response…………and this is where things get contentious rather quickly.
Things will get dicey here so stay with me all of you……I’m a live sound/recording/mix engineer by trade for over 30 years so it’s up to you to decide if that period of experience and observations matter to this discussion.………..
EACH APPLICATION of response modification via EQ effects phase…..and phase coherency matters in the translation of presentation throughout the entire chain. Have folks ever wondered why an engineer might use/choose one microphone on a source over another?…….very much like why we choose different drivers……it’s the response curve of the mic and how it closely matches what we want from the source………so we DON’T HAVE TO apply much EQ to shape the source response…..and this isn’t a time saving measure or being lazy……we HEAR the unmodified source much more clearly in the time domain within the mix and allows us to create separation between the sources via time based processing and compression much better…….when u see a video of a female singer in a studio into say a $5k U47 mic…..there’s a reason……it’s not the $5k price tag…….it’s the response curve. One of the greatest vocal performances of all time, Thriller my MJ was done with a $400 mic…..it was the response curve.
So while I can see the allure of the extensive toolbox of digital manipulations we have available, not everything is as it seems………and as a pragmatist, I REALLY can’t see building a speaker only to have to reshape the response extensively to make it work……core beauty, form and function……..sculptors rarely paint their creations.
Last edited:
And I would NEVER advocate for a single tweeter in line array……that’s a mess.…….a tweeter would have to be a near equal length line source to work in that case……and is quite expensive. But make no mistake…..a line source of properly crossed tweeters with narrow vertical directivity in respect to the line is something else and by FAR the best acoustic solution. I’d have chosen your path as well given your experience level and goals back in 2016.
My suggestion was for a LINE SOURCE solution and not an ARRAY….quite a different beast. We get an acceptable level of vertical directivity control with very wide horizontal directivity all from a compact package and negligent comb filtering requiring ZERO response manipulations outside of what is dictated by the crossover. In this use case of an MMTMM, the outer MM are essentially woofers…..and they’re crossed in with the most effective filter we have….the passive 1st order so no phase issues if we choose the crossover point and distance on the baffle properly.
Never knocking your accomplishment….I have no doubt it sounds wonderful!!!….and I’m being genuine here…..your contributions to the design have become the greatest source of research and info for what RR began so many years ago. We all applaud and thank you for the hard work and sharing!
BUT…..remembering my pragmatic nature……well……if I had to chose a 12 hour flight to Hawaii vs walking and a boat ride?…….well……while the experiences along the path might have been very special and educational……I’d have chosen the jet every single time! Lol.……and in all fairness to you….my 35 years of experience I wouldn’t trade for anything…..I haven’t really worked a day in my life.
The OP does need to make some distinctions though……coulda/shoulda/woulda journeys are much more consequential these days……….sorry to say the baby boomer generation weren’t better stewards of our society and future.….we GenX ers can only plug up some holes and paint over some graffiti….
Last edited:
I'd simply like to see what the MMTMM does . As far as in room behavior goes, in other words: that DI curve.
I can see the attraction of the Kef Blade 2 Meta concept. Even though coax drivers have their own drawbacks.
I also see the attraction of a MEH, even though that will become quite large to get the most out of it.
Curious to see more data than words on the MMTMM.
. As far as in room behavior goes, in other words: that DI curve.I can see the attraction of the Kef Blade 2 Meta concept. Even though coax drivers have their own drawbacks.
I also see the attraction of a MEH, even though that will become quite large to get the most out of it.
Curious to see more data than words on the MMTMM.
And some further reading
http://zaphaudio.com/ZA5/
but there’s lots of other examples of the vertical MTM and MMTMM…….but the most profound for you to understand would be the modern line array module…..it’s exactly an array of these modules and how that behaves is very well documented and researched…….FAR more R&D will always be injected into this sector and I’ve found there‘e essential parallels on tech to translate to home/recreational use. Both McIntosh/RR and Bose did extensive work here in the same time period……and yet it was the commercial application that saw the most R&D and where it went……creating an entire commercial market segment.
Last edited:
I've got a whole lot here to read and re-read, and maybe understand to some degree, many thanks.
I have learnt a little, and hope to learn some more, but still don't know what I don't know!.
I am somewhat pleased to see that Wesayso managed to recognise a reference i made, in my opinion that shows some real progress!
Also, i understand Mayhems wish that the OP should make some distinctions, however, and i outlined my decision making process for this reason, i had a quiet chat with him and find he cannot readily distinguish the questions that he needs to answer.
Where i can, i have.
One thing for sure, this is all about some fun, and if i ever start taking myself seriously in this please send the appropriate response.
Meanwhile, many thanks for all the contributions, please keep up the good work and the good advice coming!
M
I have learnt a little, and hope to learn some more, but still don't know what I don't know!.
I am somewhat pleased to see that Wesayso managed to recognise a reference i made, in my opinion that shows some real progress!
Also, i understand Mayhems wish that the OP should make some distinctions, however, and i outlined my decision making process for this reason, i had a quiet chat with him and find he cannot readily distinguish the questions that he needs to answer.
Where i can, i have.
One thing for sure, this is all about some fun, and if i ever start taking myself seriously in this please send the appropriate response.
Meanwhile, many thanks for all the contributions, please keep up the good work and the good advice coming!
M
Here is a line array paper I started but never finished and probably won't but in the distance that it does go, it shows quite clearly how multiple drivers combine to serve as one.
I agree with Mayhem that a tweeter in a line doesn't work. I proved that to myself a long time ago, then followed a chain of simulation/development that led to Dave Smith's (RIP) expanding array - which is a particular kind of MMTMM or WMTMW or WWMTMWW and, in effect, very much like a shaded array but where the shading is accomplished with driver diameter as well as amplitude and frequency response coefficients.
So we aren't really in disagreement overall so much as arguing for different sets of implementation details.
I disagree that equalization leads to phase issues. No doubt it does if you use a typical bank of sliders in a recording studio. It need not if you use FIR.
One thing I've learned in over 50 years of engineering is that its extremely rate to get multiple engineers to agree on what the best compromise is! Each of us evaluates tradeoffs differently so what one should do is listen to everybody, then choose what seems best to him.
I agree with Mayhem that a tweeter in a line doesn't work. I proved that to myself a long time ago, then followed a chain of simulation/development that led to Dave Smith's (RIP) expanding array - which is a particular kind of MMTMM or WMTMW or WWMTMWW and, in effect, very much like a shaded array but where the shading is accomplished with driver diameter as well as amplitude and frequency response coefficients.
So we aren't really in disagreement overall so much as arguing for different sets of implementation details.
I disagree that equalization leads to phase issues. No doubt it does if you use a typical bank of sliders in a recording studio. It need not if you use FIR.
One thing I've learned in over 50 years of engineering is that its extremely rate to get multiple engineers to agree on what the best compromise is! Each of us evaluates tradeoffs differently so what one should do is listen to everybody, then choose what seems best to him.
Last edited:
Here’s my in home work space….when I need to get things done in comfort……notice how my commonly shared principles are Implemented in my own space. Notice the very inexpensive, simplistic and yet elegant ceiling reflection mitigation method……that cloud is four eye bolts, around 2 hours to fabricate, weighs around 35 lbs and costs around $80.…..I have no 3d acoustic ceiling in this space.
The KEF Q150s are modified……but they’re still excellent point source and allow me to make very accurate decisions based on phase……I could NEVER listen to them recreationally in the nearfield BECAUSE they expose so many horrible phase decisions that are embedded into some of the great recordings……