The making of: The Two Towers (a 25 driver Full Range line array)

Given constraints of speaker and room with reflections, the phase wraps will always be present when viewing IR with long windows. With center to center driver spacing of about 90mm, and off axis response at 30 degrees to 10kHz the frequency response ripple above 3kHz is unavoidable. Smoothing used hides it. Shorter windows for FR remove many room reflections.

With line array 2.3m tall, measurements made at 3m, and about 1m from floor place top driver about 80cm further from microphone than closest driver. This has delay somewhat more than 2ms; implying use of correspondingly short window to only capture direct response of speaker. Quarter wave effects start at about 500Hz, speaker starts showing that it is not a single source.

Your waterfall plot:


uses 3.1ms window and shows FR ripple. Ripple in FR equates to ripple in phase response.

Even single 3.5" full range driver shows itself not to be single source at higher frequency. When I correct a response of such driver to flat within 0.2dB at measurement position, and then move microphone a few millimeters, ripple becomes evident.

Even single tall and narrow ribbon as line array will show ripple.

All speakers are compromise, some are easy to live with, like yours.
I believe you linked your own waterfall plot here :). But the point stays the same. My plot definitely shows the ripple.

Back in 2012 when I was still designing this project I made this drawing:

I've updated it to show the actual as build size. You can see the differences to the microphone in this sketch. If I'm correct it is a 0.629 ms difference (216 mm) between the top driver and the microphone compared to the one that is level with the mic.
At least that's what I used in my mind to figure out if this had a shot of working properly :D. I have the vertical (off axis) angles in there to figure out if I needed high frequency shading, an idea that I entertained at the time, making it behave more like an expanding array. I decided to try it as is first, after some PM's with other array owners here on the forum. Though it never left my thoughts I haven't found a good enough reason to try that shading idea yet as it brings on other issues.
Last edited:
Next steps to audio nirvana (I hope)

I've been playing with the DRC processing after I got an idea because of the comb filter I discussed a few posts ago.

To get a phase coherent solution I went out of the box in DRC to use bigger numbers than recommended to see if I could straighten out the phase. I discussed this earlier in the thread and showed a couple of beautiful step responses to show that case.
Last night I wondered about the straightening of the phase in the higher frequencies. As noted above the arrays have a time smear inherent in their design. If I try and correct that forcing it to be linear phase I could actually get worse upper frequency response. At least that's what I thought.

The EP window used up till now, 4 cycles at 20 Hz, about 2/3 cycles in mid frequencies gradually going to 9 cycles up top in a simulated waterfall looks like this:

But id I limit the EP window to 1600 Hz, making it full ~4 cycle to 20 Hz I get this simulated plot:

It's even a better improvement above 5000 Hz but as said before, I need to be proud of that first to actually show it. :D

In all the graphs I checked I see an improvement, in pré-ringing level, waterfall and FR response and the illusive phase wraps. I've listened to it today and it seems to make a worthwhile improvement. But I can only confirm it with new measurements. Usually the sims are quite good as a prediction to I have high hopes. It seems to have the biggest effect in the CSD plot at high frequencies, that was my goal.
Here are my graphs, all with a 4 cycle window in REW in various stages
The smoothing is automatic in REW, I would have liked a little less smoothing to really see what's going on.

How do you use cycles for windows in REW? All I can find is 'ms' for window settings :confused:

The smoothing is automatic in REW, I would have liked a little less smoothing to really see what's going on.

I think you can change the smoothing for each individual measurement by selecting a different smoothing in the EQ tab.

Thanks a lot for your replies & suggestions.

I am planning on some room treatment very soon & of course adding a ambience tweeter ( although I am not sure - front, up or rear):confused:
The latest REW has the ability to choose a window in cycles.
If you do that, the smoothing is automatic. At least I think so.
Choosing a different or no smoothing does not help change the graph.
A prior beta release had a fixed number of cycles with a smoothing based on 1/24 th.
If I up the number of cycles the smoothing comes trough but it's still different from normal gating.
I managed to lower the number of cycles of my windowing in DRC. Have not experienced any drawbacks yet. In fact, I have something back that I lost in my last experiments.
First impression is that it sounds definitely as good as before in the listening spot, but better than before way off axis.
I'm down to 6 cycles in bass frequencies, moving down to 4 cycles at mid frequencies and up again to 9 cycles at 20 KHz. The longer window up top is primarily because of the time differences of the upper and lower speakers compared to the one on axis.
I will see where the boundaries are for the higher frequencies but expect I can't lower it much more without some loss of focus there as I did run tests before that showed me this effect. But that was with the Excess Phase running full range. That isn't the case any longer as the Excess phase now only runs 4 cycle from 10 Hz to 1600 Hz. Leaving the top end as is after correction of FR.
Not that far off anymore from the short window correction I was after without experiencing mayor drawbacks. A lot of work but worth it.

In short: I'm happy. The goose bumps were back that were somehow missing last week. That could very well be me, as I didn't spot any big errors before but I'm glad this worked out. It will take a couple of days to know for sure if the way off axis sound really improved. By that time the next experiment has already taken place I'm afraid :D.

I hope to experiment with an artificial "Haas Kicker" soon. Basically I will create some virtual reflections from behind the listening spot. It probably will be an L - R signal which is attenuated, band passed and delayed between 15 to 25 ms, that's were the Haas name comes from. Decorrelated from the mains they will provide a fake (in my case) late reflection early enough not to be perceived as echo, as if it were coming from a wall further behind the listening position.

What this should do is give more sense of space. It's said to enhance the front stage and not to degrade it. It should provide a better sense of 3D imaging providing better depth. So lets find out! It is something I wanted to try and I guess my mains are playing good enough to start experimenting with something else.

I know, no pictures or new graphs, boring post ;).
Some background on the "Haas Kicker" can be found here:

Found in this thread: as used in the studio to mix the tracks we listen to.

An interesting quote from that thread:
The view on the Haas kicker did change some. A problem with the very strong reflections from using hard panels was that the material played back in the presence of the strong Haas trigger sounded better than the original. And that is of course not the goal. The goal is having an environment that reveals flaws and inaccuracies in the micing techniques, recordings, and mixing choices. Russ Berger, together with a colleague (Wrightson) tested the veracity of the hard reflections and their results concluded that the energy from behind should be at least 10-15dB lower than the direct energy. Reflective panels which was used at an earlier stage was found to have a less accurate perception of the image. The termination should arrive from diffuse energy.

Sounds to me this warrants enough reason to experiment. The bold part wouldn't form a problem for me, it is desired :D.
As I'll implement this with DSP and speakers faking that reflection I'll have more freedom to play with it in terms of attenuation, bandwidth, timing and direction (trough aiming).

I stumbled across the idea a long time ago on the DIYMobileAudio forums where the fake implementation is used to overcome the extremely small listening space a car provides. My listening area is by no means big. With a rear wall much closer than I would like. So up till now I've only used damping, not diffusion to get a good clean first arrival of the wave front. Now it's time to put some energy back in to simulate a bigger (more pleasing) listening area.
Last edited:
An example of the passive Haas Kicker as would be used in a studio can be found in a thread that ended up in the full range forum by mistake.
The thread was about trying to minimise group delay of a multi way speaker:
Jim has an awesome room available to listen to his music. Seeing his Spectogram with the Haas kicker immediately reminded me of the tricks used in Car's to make it sound bigger than it is.

Here's Jim's passive Haas kicker (hope you don't mind Jim1961)

Visit the thread on Gearslut from Jim1961 to see what kind of room treatments are necessary to achieve something like that! I just loved to look around in the measurements Jim posted, how often do we get to peak in a room like that! Here's the thread about Jim1961's room:

I hope this gives some of you guys some new inspiration too...
Is the Haas Kicker somewhat similar to ambiophonics?

Nice to know you are satisfied enough with the mains to move on to special effects. Maybe you have reached the pinnacle of tweaking!

Now, do your towers sound anywhere near a $18000 system? (it looks like one for sure, IMO)

The Haas Kicker is based on a psychoacoustic principle. Due to having the first ~ 20 ms reflections as damped as possible our mind gets busy searching for clues about room size. That's quite staining on the mind, or so it is said. By introducing a fake room queue, preferably by a diffused sound from behind, you set the mind at ease. Theory has it that this queue has to fall within 15 to 25 ms after the first arrival. In studio's this is done by clever routing of reflection and diffraction. In my case with speakers in the back. Making my version highly tuneable as to what sort of sound it plays. I'll play with different types and combinations, starting with a R-L ambience type of sound.

The back reflection makes the mind think we're in a much larger room and make it ignore the smaller queue's. It supplies creature comfort to enjoy the music and let the focus be on that. It has been described as making the sound prettier than it actually is. Not a bad idea for a listening environment :).

Hard to say if my speakers sound like a $ 18000,00 system. I was kinda hoping for more, actually (lol). Seriously, I don't have enough personal experience with expensive systems to tell. I tune it to my liking and than may very well mean it does not sound like the High End equipment at all! I'm after natural and pleasing sound. I can play a lot of different material and genre's and do well. Sure faults in recordings become more "visible". But that's usually the over use of compressors and limiters.
I'm probably too modest to make such a claim. I'd rather let the people that listen to it decide.

I'm pretty proud of what I achieved though, both in looks and sound. I tried to get every ounce of performance out of it and I'll keep doing that as long as I have new ideas on how to. But the truth is I am satisfied right now. If this is it, no worries from me.
Do you still have the Ambience tab turned on in JRiver?

If so, I'm afraid it might compound with the new effect and get pretty wild.

Or not.

Nothing like trying it!

I actually still have it turned on, but to soften the effect somewhat with the mid/side processing. We'll see if both principles collide. Easy enough to try.
The ambience effect does bring a bit of cross talk cancelation with it, making things more pronounced. Even though the description in JRiver says it makes the soundstage wider that is not my experience with the arrays. It is as wide without the effect, just a bit more subdued. Tuning it on makes the back ground vocals sound like actual people to give an example. Without it they sound thinner, less real. We'll see if these principles are conflicting or meant to be working together.
I can also try something like: V.I Stereo to 5.1 Converter VST Plugin Suite
Though from reading it, seems like it's doing the same tricks I already use.
According to the paper I linked it is advised the first ~20 ms be down at least 20 dB. The first 4 ms being even more important. 25 to 30 dB down is better.
In my living room I placed the damping panels to try and reach that goal.

There's still a peak at ~6.3 ms corresponding to the wall behind the mic, but it is down below the 20 dB limit. At about 20 ms I have another peak, that's where I want to time my Haas kicker for a first try. That would be down by 12 dB or less(!). The actual attenuation will be determined by listening. Down enough not to notice it's presence but loud enough to maximise the effect it gives.
A studio would have a bit more steep tail ending at ~300 ms. But I'm pretty proud of what I managed in a living room with just 3 damping panels (granted, 2 of them are huge but out of sight). Trying to create a reflection free zone in those first 20-25 ms is vital for this concept.
Last edited:
I was expecting to work on the enclosures for my rear channels over the weekend but sadly other things came up. But I did manage to puzzle together a design for a small box with the 10F as driver. Highly inspired by X's dagger design but I chose to go with a four sided shape:

Maybe I should call this enclosure the Nail, partly because of the resemblance but also because I hope to have nailed it's design ;).
It's a small sealed enclosure of ~0.7 litre giving a QTS in the 0.707 neighbourhood.
Bandwidth of interest will be about 200 Hz to 3.5 KHz bandpassed so this should be able to do the job. I plan on hiding it between wall and couch firing up. I might play with deflectors later if needed.

I also played with a few house curves for a bit, trying to find a nice middle ground for multiple genres. Here's the 1/3 octave smoothed plots I played with:

I gave all these curves a listen with a couple of songs to let me get used to them.
From top to bottom you'll see the curves inspired by what I've found online;
Sean Olive's 10 dB slope from 20 to 20000 Hz
JBL curve, more gentle later JBL curve (*)
JBL curve with quite a bit of boost down low.
Flat with bass boost

To me, in my room, the top two sounded most natural to me. The top one is what I've been using for a while, a gentle downwards curve. I also tried some variations on that curve, with less downward slope, not shown here.
The flat curve didn't do it for me. Though it sounds pretty natural, at ~85 dB I'm missing the bass. So that's why I tried flat with bass support (the green ones). The balance just wasn't right. Next I tried a JBL standard curve, 6 dB higher bass, 3 dB downwards slope up top. Better but not right.
After that I made a new Sean Olive curve, listened for a while and on a hunch made it slope down to ~200 Hz, And follow the slope of the Sean olive in the higher notes. That was most natural to me, though I still have a few EQ cuts right now after listening to more songs.
I later found a very similar JBL curve and that's why I called it a more gentle JBL curve. To me it does the most things right. I expect the SPL level having a great influence here. I need just a bit more bass, and a slight slope of high frequencies for a pleasant sound. Don't worry about the roll off at 13K, that's more of a by product of the 1/3 octave smoothing. It goes to about 17 KHz, as shown here:'

1/12 dB smoothing, making it harder to recognise all the curves...

So I need a bit of shaping to keep the bass and not have the higher frequencies become a bit too much. The gentle JBL curve as I call it goes a long way there. It's only a little different from the sloped 10 dB 20-20k I used for quite a while.
I've listened to a few these impulses over headphones and that does make the room stand out more. Part of the reason to do this experiment.

(*) Here's an example that made me name my own preferred curve a gentle JBL curve:

It seems there's is something to be said to do as much research as JBL. Though it is highly depending on the speakers interaction with the room. Your results may vary :D.
Last edited:
Changed my mind about the design in the last minute, figured I wanted a bit more internal volume but not much of a size increase. Started working on them over the weekend.

Feels great creating saw dust again :). Though these won't be as pretty as the arrays. They are meant to blend in with the environment.


First one almost finished, second one not far behind.

The enclosed volume is about 1,1 litre. Still wondering how I can diffuse the sound from these, they will be pointed up, at least at first. Hope to have some sound out of them soon.

Getting closer...

Now am I confident enough this is going to work and paint the enclosures before I test it? :spin::D

Very interesting shape. Fewer parallel sides to foster standing waves internally. I did something NO parallel sides after reading about Ed Schillings's Bucket Sub. The technique wouldn't be difficult to adapt to your project. See here: bucket sub, with a twist | Audiokarma Home Audio Stereo Discussion Forums

I'm thinking about building two more and implementing a distributed bass system in the room.
Hello Guy,

My shape was loosely inspired by xrk971's daggers. Though his dagger is a 3-sided shape without any parallel walls. I switched to 4 sides and the wider shape as their intended use is with the speakers pointed up between my couch and the (much to close) back wall.
I like that bucket sub!

On an entirely different note, in Legis' Horny Tales I stumbled over his recommendation of the new Infected Mushroom album: Converting Vegetarians II. I already enjoyed their previous work and decided to try this one. I agree with Legis! Try this album, highly recommended!
An externally hosted image should be here but it was not working when we last tested it.