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

Sd poker....See your 925 & raise you.... !!

Hiya Wesayso,

921....Hhmmmn....Not bad for a first attempt, but I'll see your 921 and raise you 159.....Then add in a pair of subs....550 Sd and 880 Sd !

A LOW & MID FREQUENCY 12SW1300Nd | Beyma
Plus a :
LOW & MID FREQUENCY SM-115/K | Beyma

Overkill by name overkill by nature!
Each one of my 4.5 inch BMR's (112mm chassis) has an Sd of 67.5 and I use 16 per side for stereo L / R. Thats an Sd of 1,080 per speaker....

Overkill by name overkill by nature!

:D
PS Thats just the two channel stereo....Got another 16 BMR's on the centre channel!
 
Last edited:
Hi Derek,

With a Sd of 925 for the 25 driver line arrays we've got that covered ;). Also well above the 75% floor to ceiling height. I agree that it seems to work quite similar to the Geddes multisub approach. With the advantage of spreading out the sources in the vertical direction as well. Geddes proposal spoke of a height difference as well, placing at least one sub at an elevated spot in the room.

Well, essentially, you do have bass loading from close to 4 corners, equalling having 4 subs in the room, and also all the positions in between .... :D

The late John Dunlavy made a strong statement for something similar related to his SC IV/a in the way that you have woofers at top and bottom, and this loads the low bass in the room in a way that the in-room response is significantly better than what the anechoic response + room gain in general would suggest. Now, the late Mr. Dunlavy knew a few things about loudspeaker design, so the thing of taking this into account is for sure relevant
(Sorry I don't have the link to this, and I really can't find this... but will look for it)

I don't really understand fully this effect that Mr Dunlavy wrote about in Stereophile, neither did John Atkinson; but maybe I will later .... I booked the updated version of the "Master handbook of acoustics" so maybe I will shed some more light onto this later :p
 
Last edited:
The late John Dunlavy made a strong statement for something similar related to his SC IV/a in the way that you have woofers at top and bottom, and this loads the low bass in the room in a way that the in-room response is significantly better than what the anechoic response + room gain in general would suggest. Now, the late Mr. Dunlavy knew a few things about loudspeaker design, so the thing of taking this into account is for sure relevant
(Sorry I don't have the link to this, and I really can't find this... but will look for it)

I don't really understand fully this effect that Mr Dunlavy wrote about in Stereophile, neither did John Atkinson; but maybe I will later .... I booked the updated version of the "Master handbook of acoustics" so maybe I will shed some more light onto this later :p

I've read it, actually I've read all I could find from/about Mr. Dunlavy. Including his posts on "rec.audio.*" from 1997 till 2001. He was a smart man, but not as good at marketing ;).
 
Made some measurements this week, still have to record louder than I would want to. I'll make sure to get some measurements with less ambient noise sometime soon. No distortion plots to publish yet as they are swamped with ambient noise and way to high in level.

Just wanted to show a couple of impulse responses during the process.

Impulse response right channel, no EQ:
right%20impulse.jpg


Impulse response after EQ:
right%20impulse%20with%20eq.jpg


Impulse response with EQ and DRC convolved signal:
right%20impulse%20with%20eq%20and%20convolver.jpg


EQ-ing before using DRC gave me much better results. I mainly used REW Auto EQ with some manual tweaks to prevent high Q's with excessive boost or cut. I'll redo this with some proper measurements but I am pretty pleased with the outcome so far. Some early reflections can be seen but I found the cause for at least one of them and that's an easy fix. (without cluttering up the room with damping panels)
Left channel is clean (-20 dB up to the first 20 ms) even though that is the side with the speaker up against the wall. So the damping panel works (as I had found out while listening, much wider stage).

First remark about DRC: it manages to take these huge line arrays down to really small size speakers. :eek: What I mean by that is that it don't sound that huge anymore, but I lost some of the ambience in the process. Pin point imaging but you wouldn't know you're listening to small speakers or big ones.
 
Last edited:
Always fun to take a peek! Just to see some main trends... I opened your first measurement which was done with the same MacBook I believe and it should look decent enough.

By the way, I played Amy Winehouse the other day. Rehab from her album Back in Black. What was that! So I figured to take a peek:
amy.jpg


Looks like I found a source for a square wave test file!
 
EQ-ing before using DRC gave me much better results. I mainly used REW Auto EQ with some manual tweaks to prevent high Q's with excessive boost or cut. I'll redo this with some proper measurements but I am pretty pleased with the outcome so far. Some early reflections can be seen but I found the cause for at least one of them and that's an easy fix. (without cluttering up the room with damping panels)Left channel is clean (-20 dB up to the first 20 ms) even though that is the side with the speaker up against the wall. So the damping panel works (as I had found out while listening, much wider stage).

First remark about DRC: it manages to take these huge line arrays down to really small size speakers. :eek: What I mean by that is that it don't sound that huge anymore, but I lost some of the ambience in the process. Pin point imaging but you wouldn't know you're listening to small speakers or big ones.

So you used DRC to generate correction filters for both eq'd and non-eq'd speaker impulse responses? In what way (or ways) were the results improved with the former approach? I would guess that some tweaking of the peak limiting stage within DRC is in order since the towers require so much correction in the freq domain. I wouldn't expect you to get good results using only DRC with the default settings.

As for your observation on DRC's effect on the perceived size of the speakers, I would think that not being able to tell the size of the speakers is a good thing. Perhaps you're noticing slightly less room ambiance and slightly more recording ambiance (where applicable). I think my (tiny) system is pretty well corrected and some recordings sound huge while others hug the speakers a bit more. I listen to a lot of classical music as well as some rock and one thing that surprised me when I first started to get the hang of DRC is how "small" and "flat" some recordings sound while others can be "big" and "deep".
 
I'll explain why I think this happened. I played with the maximum allowed boost/cut in DRC and it actually sounded better than the EQ-ed array by itself. But to get a reasonable measurement away from my noise floor the midrange was at a really high level, 111 dB at ~340 Hz after filtering. It basically shook up a lot in the room to get that measurement.
spl.jpg

After EQ the midrange was at ~96 dB level, so that measurement I trust way more as a base for the DRC to work on.
Like I said before, I still need to make some measurements with lower ambient noise around. But for that I need to have the house by myself either after working hours or on a quiet Sunday morning. But the DRC after EQ is generally more clear sounding.

I agree on the size thing. That's not necessarily a bad thing. I notice the effects you speak of, some recordings have more space definition embedded than others. The stage width is way beyond the speakers on several tracks. I.m.h.o. that is a good thing, the phase must be right to achieve that.

A lot of recordings, studio work, have nice stereo "effects" but not much natural ambience embedded in the songs. Some are almost plain boring mono sources. Others do open up real well. But there's none of that big image huge lips, bigger than life things often attached to line array reviews going on.
I played with JRiver's effects for a bit and using the Surround Field option gave more body to all the separate instruments/voices without adjusting the width of the apparent stage. Very enjoyable and close to a live feeling of being present AT the recording.
So I investigated a bit further and found out that Surround Field setting is based on the 1930's Blumlein Shuffler technique. Googling that came up with products like the BSG Technologies qøl Signal Completion Stage.
After reading a few reviews on that technology I agree with what was heard with the Signal Completion Stage most of the time. The separate instruments and voiced sounded more real. It isn't EQ you're hearing but a sort of reshuffle of the original stereo signal. Anyway, it works. And quite well I might add. It gives a life like quality to the performance/performers. Some life tracks don't need it. But even a general favourite I see mentioned all the time (probably because we all know it), The Eagles - Hotel Callifornia from Hell Freezes over works better with the effect added.
Close your eyes and you're there. At least that was the feeling I got. So to put it to the test I had my girlfriend take a listen. She was a bit disappointed. Not by the performance, she felt like being there. The audience at the end of the track gave away it was just a recording. That was her beef with the song.
After that she listened to a couple of other songs, for instance, she likes Robby Williams, so I played her Road to Mandalay. I know that song works well as a voice recording. She felt so close, almost as if she was sitting on his lap while he sang to her :D.
 
Last edited:
Don't get me wrong though, I'm not saying this is THE or MY solution. I haven't played with enough settings in DRC to know. I just played with different window settings and boost/cut factor so far. It was a fun discovery to use the JRiver effect though, and I liked it a lot! It shows the potential of the corrected speakers. I guess you need speakers with good phase behaviour to benefit. It even sounds nice on headphones, I have a German Maestro GMP 8.35D as a sort of reference.

I downloaded the free Voxengo plugin, which is also a Blumlein Shuffler like tool to be able to play with it. Haven't done it yet though.
Voxengo MSED free VST plugin
 
I'm still not sure if you eq'd your measurement after the fact or somehow used an eq'd sine sweep for your measurement. Anyway, it seems like your primary concern regarding measurements is a sufficient SNR in the bass and treble regions because of the reduced amplitude relative to the mids - is that right? The thing about recording a sine sweep and transforming it to a spike versus just recording the spike itself, is that there is an improvement in the SNR of the measurement with the former method and that benefit increases with longer sweep lengths. In other words, don't worry so much about the sweep level vs. the ambient noise level because that sweep will be converted to a spike whose level is much higher relative to the ambient noise than the sweep level was. If the ambient noise surrounding you is fairly steady in level, you may want to try a longer sweep length (60 sec) combined with a reduced but reasonably loud volume level (not enough to rattle the room). Then, with the appropriate peak limiting parameters in place, you can start tweaking.

I apologize if this info is not completely pertinent; I'm trying to better understand your thought process as well as your techniques so I'll be better able to advise. Based on my own personal experience, I think you'll have an easier time and get the best results doing this all with DRC even if the initial results aren't what you were hoping for.

As for stereo imaging, I find that it can often be distracting and generally appreciate recordings that have a strong center image. I like my Beatles in mono. However, with classical recordings I like the spaced omni sound more than coincident (I don't really care too much for the coincident/blumlein pair sound in general except when mixed to mono). I'm glad you're having fun with the shuffler though, and soon (if not already) you will have a (non-headphone) system with the transparency and neutrality needed to fairly judge these things.
 
I too, started down the path of room correction, and it can be a rabbit hole!

I looked up DRC, but it was a bit much just to get my feet wet!
So, I went and got the 15-day demo or ARC2, which is basically AudysseyXT. It pushed my setup beyond the basic EQing I was already doing. Sound opened up and it was good!

So, I'm sold! I will add room corrections to my system. Didn't settle on which one yet, as ARC2 is a bit expensive.

Thanks for sharing the Voxengo plugin, I will take a look.
In mixing and mastering, I've used a Brainworx plugin and it sound really good. Lots of tweaking too, including EQ on M/S "channels"

Good luck on the ongoing task! :)
 
I'm still not sure if you eq'd your measurement after the fact or somehow used an eq'd sine sweep for your measurement. Anyway, it seems like your primary concern regarding measurements is a sufficient SNR in the bass and treble regions because of the reduced amplitude relative to the mids - is that right? The thing about recording a sine sweep and transforming it to a spike versus just recording the spike itself, is that there is an improvement in the SNR of the measurement with the former method and that benefit increases with longer sweep lengths. In other words, don't worry so much about the sweep level vs. the ambient noise level because that sweep will be converted to a spike whose level is much higher relative to the ambient noise than the sweep level was. If the ambient noise surrounding you is fairly steady in level, you may want to try a longer sweep length (60 sec) combined with a reduced but reasonably loud volume level (not enough to rattle the room). Then, with the appropriate peak limiting parameters in place, you can start tweaking.

I apologize if this info is not completely pertinent; I'm trying to better understand your thought process as well as your techniques so I'll be better able to advise. Based on my own personal experience, I think you'll have an easier time and get the best results doing this all with DRC even if the initial results aren't what you were hoping for.

As for stereo imaging, I find that it can often be distracting and generally appreciate recordings that have a strong center image. I like my Beatles in mono. However, with classical recordings I like the spaced omni sound more than coincident (I don't really care too much for the coincident/blumlein pair sound in general except when mixed to mono). I'm glad you're having fun with the shuffler though, and soon (if not already) you will have a (non-headphone) system with the transparency and neutrality needed to fairly judge these things.

I'll try to word it different. Yes, I recorded the sine sweep that I used for my second DRC attempt using a signal with EQ in that path. That brings down the midrange peak the un-EQ-ed line array has. The graph I showed is the line array without EQ. While I do need boost in the low's I practically need no boost in the high frequencies when dividing the low point and high point of the graph I showed. That came as a bit of a surprise really.
I tried recording with REW, which uses a short burst, and with Audiolense. With the latter you have the opportunity to use longer sweeps. I did use that feature but the control function that gives a basic analysis of the recorded signal wasn't passing the measurement as "good" until the level was way up again. So that confirmed to me my high ambient sound level is messing with my measurement. It's not a steady level, my house (with a big window) is only 3 meters away from a busy road.

While my first attempt at DRC sounded good using an un-EQ-ed sweep, even better than the same sweep with EQ applied to the same sweep by itself, my second attempt with an EQ-ed sweep as the base for DRC sounded much better, much more clear and natural. The naturalness improved with the Blumlein Shuffling applied to the signal. I'm not Shure about that last step (Blumlein Shuffling) being a gimmick or if it really is that much better :D.
 
Last edited: