• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Murphy's Corner Line Array project

Ericglo

Senior Member
Joined
Apr 21, 2020
Messages
452
Likes
323
One more thing ppataki... for your own sanity. Get an SPL meter. It doesn't have to be expensive. Once you get distortion low enough, it's very easy to play (go) louder than you realize. You'll only notice it if you want to start a conversation ;).

To avoid that, I use JRiver as my mediaplayer and let that average the SPL level of the songs. It is using the R128 algorithm to make every song 'appear' to be equally loud. (once it has every song analyzed) Before I started using that I've ran the arrays at outrageous levels on some songs.
(without really realizing it)

My friend Mike Garrett of AVScience has a high efficiency Beyma horn tweeter and Audio Excellence woofers for his home theater. He says he regularly listens at SPL levels that I would find uncomfortable. He thinks the efficiency and low distortion allows him to listen at higher SPLs comfortably.

In that same vein. At Cedia one year, JBL played their Everests at very high SPLs to close the show. I stood in front of them and was amazed at the amount of SPLs coming out of them. It was like a waterfall of sound. They were still quite loud on the other side of the convention center. Interestingly, the high SPLs weren't uncomfortable.
 

Ericglo

Senior Member
Joined
Apr 21, 2020
Messages
452
Likes
323
At last the speakers are in their final place!
I have made some preliminary measurements but take it with a pinch of salt since the insulation of the cabinet is not yet finished, there is a pretty severe air leakage going on (that will be fixed either today or tomorrow)

Measurements were made at main listening position using UMIK-1 with REW (1/12 smoothing) at 75dB

Frequency response:
View attachment 181197

The blue one is the original, non-compensated curve - pretty horrid, as expected
The peach is the curve that Dirac Live 3 produced when I set the target curve to be totally flat - getting there
And finally the highlighted one is a result of tweaking the target curve in Dirac + using Pultec EQ (Waves PuigTech) to further fine-tune the sound
As you can see the highs have to be compensated very heavily, eating up like 15dB of amp power
I have some pretty nasty room modes between 50-60 and 70-90Hz - I can compensate those but at the expense of increased ringing as you will see below

T60M curves for the same

View attachment 181199

I will have to decide if I prefer the more flat sound in the lows with increased ringing vs less perfect low-end with significantly lower ringing

Waterfall for the linearized Dirac + Pultec measurement:
View attachment 181200

Wavelet diagram for the same:
View attachment 181208

Step response for the same:
View attachment 181203

Impulse response for the same:
View attachment 181204

IR zoomed in:
View attachment 181205

Distortion curves - these look pretty bad but I suspect that the air leakage might have some impact here:
View attachment 181207

And finally, a video that shows how it looks and how it sounds:


Here are some additional thoughts:
- dynamics are extremely good
- transients are excellent (really 'fast' sound)
- stage is wide, but precise but obviously not too deep
- phantom center is a bit less forward vs my previous setup
- sound is no less detailed vs my previous setup - both micro and macrodynamics are excellent
- sub is definitely not needed for music, there is plenty of energy down to below 25Hz
- even for movies I will not consider using a sub, however I need to say that the impact is a bit less compared to my previous setup where I had two 12" subs too (however the lows seem to be more 'detailed' with the line arrays)
- footprint is extremely small (the speakers are tall but they consume virtually no space since they are in the corners) - WAF just went over the roof, at least in my case :)
- very heavy EQ compensation is needed so take that into account when choosing an amp for such a project
- sound is kind of filling the whole room, anywhere you go, stand, sit, lie, etc. you hear almost the same
- my listening spot is 325cm away but even if I go to the other end of the room (720cm) volume does not seem to decrease (see above point)

The next steps will be to fix the insulation then I will post a new set of measurements in a few weeks' time once the drivers had their 'break-in'

Congratulations!!!

I have been eagerly awaiting your results. Now I am going to have to strongly consider doing my own MLCA.:eek:
 

fluid

Addicted to Fun and Learning
Joined
Apr 19, 2021
Messages
691
Likes
1,196
I didn't realize that you had a Linkwitz LX521. I have read a lot of owners consider that their end game speaker.
The LX521 are great speakers and there was some music that they were better with, but some music I could never be happy with. As an overall compromise I prefer the arrays more smiles and less grimaces. If it was practical to keep both I would have. They cost too much money and took up too much space for too little benefit to stay.
 

Ericglo

Senior Member
Joined
Apr 21, 2020
Messages
452
Likes
323
The LX521 are great speakers and there was some music that they were better with, but some music I could never be happy with. As an overall compromise I prefer the arrays more smiles and less grimaces. If it was practical to keep both I would have. They cost too much money and took up too much space for too little benefit to stay.

I went back and reread my original thread and you mentioned the Linkwitzs in there. Thanks for the extra info.
 

dziemian

Member
Joined
Jun 11, 2020
Messages
41
Likes
36
@ppataki
Hi. I wonder how you managed to use Dirac with your Line Array speakers if Dirac is meant to be used with typical speaker's design. I have asked the same question on Constant Beam Transducer (CBT) speakers thread here at Audiosciencereview. post #118.

Does anybody have an idea how to measure CBT36k (or any other cbt design) to be used with Dirac / Audiolense/ Acourate dsp/drc software? All of this software is based on an impulse measurement from a single driver. How to measure several drivers at once playing the same frequency, each delayed by some ms? Is that possible at all? Is the measurement not going to be flawed in general?

j_j - technical expert here replied: Doing this with a phased array is an extremely complex undertaking.

So whats your opinion on that?
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,219
Likes
1,360
Location
Budapest
I just used it as-is.
I usually measure the 1+8 positions forming a cube with a size of 80cm
Then I check what Dirac did and I correct its target curve to have a flat response at the listening position, you can refer to my earlier post here about the measurements.
Now I am trying other methods as well, like using CurveEQ to flatten the response at the listening position first and then use Dirac to further correct that - so far that approach has yielded the best results
I will post some more measurements once the air leakage issue is fixed
 

abdo123

Master Contributor
Forum Donor
Joined
Nov 15, 2020
Messages
7,425
Likes
7,941
Location
Brussels, Belgium
I just used it as-is.
I usually measure the 1+8 positions forming a cube with a size of 80cm
Then I check what Dirac did and I correct its target curve to have a flat response at the listening position, you can refer to my earlier post here about the measurements.
Now I am trying other methods as well, like using CurveEQ to flatten the response at the listening position first and then use Dirac to further correct that - so far that approach has yielded the best results
I will post some more measurements once the air leakage issue is fixed
Dirac just takes the average of the measurements for the amplitude response. If you want a flat response in the listening position then only take that one particular measurement.

It’s only the phase correction that is behind closed doors.
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,219
Likes
1,360
Location
Budapest
I have tried that several times and it has always resulted in elevated ringing and (strangely) a less optimized response
I forgot to mention that when I correct Dirac's response I leave it intact below 100Hz, I only fine-tune above 100Hz since - again if I mess around with it below 100Hz I end up with higher ringing (I highlighted that in my initial measurements)

btw. this is not just valid for the MCLA but to all my previous systems where I was using Dirac (in my living room - my office is another story, there I can manually flatten down to <25Hz without causing any increase in ringing - my living room has some really screwed up room modes)
 

hollis

Active Member
Joined
Sep 3, 2018
Messages
117
Likes
84
@ppataki
... Dirac is meant to be used with typical speaker's design....
I have built a slightly modified set of MCLA, mine have a bit less air volume to fit a tight room.

I have tried them both with and without Dirac, in three different rooms. Dirac I feel, is an excellent finishing touch. It has never taken anything away and only added to both my subjective impression and objective results.

The only caveat I would mention is that Dirac has maximum and minimum gains on its filters, MCLA in their raw response need more eq gain than Dirac can provide. PEQ upstream is a must. I use a minidsp 88d and max out the 10-bands of filters before I run Dirac.
 

Attachments

  • Capture.JPG
    Capture.JPG
    245.6 KB · Views: 71
  • Final FR.png
    Final FR.png
    148.9 KB · Views: 82

hollis

Active Member
Joined
Sep 3, 2018
Messages
117
Likes
84
Since you seem interested in their impulse response please see before/after attached below:

Left
1643499620537.png

Right
1643499584866.png


Please note, front wall and side walls are treated with ~12' of curtains. 1st reflection points are treated with 2" 2'x3' ATC panels and there is a desk center.
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,219
Likes
1,360
Location
Budapest
Happy to report that finally the cabinet air leakage has been fixed! (although the journey has not yet come to the end as you will see below)

We used these type of screws to pull the cabinet really tight together:
20220203_081009.jpg



I am not sure how these are called in English but they are generally used for heavy wood construction

They did a perfect job, you can see a video here where I knock the area of the cabinet without the new screws vs with the screws
The difference is day and night as you will see on the measurements below too


The screwheads are not visible since the acoustic elements are right next to the speakers:

20220206_200504.jpg



Some measurements where we can clearly see the difference before the screws vs after the screws

Distortion:

1644220428287.png


Step response:

1644220475990.png


and now the Frequency Response to explain the rise in distortion around 55Hz (due to a dip)

1644220581441.png



Pretty good results I believe

The cabinet vibrates no more (literally no vibration, even if I cranck up the volume listening to a 30Hz LF test signal)
BUT it actually revealed another issue: there are some drivers (luckily only in the left hand side cabinet) that do have air leakage...
So the next steps will be to try to fix those + to re-measure the whole system

I decided to share the whole signal path that I am currently using in Jriver - it might come in handy for people planning to build a line array (or any other one-way system for that matter)

So let's see the original frequency response vs the latest optimized response that I shared above:

1644220955870.png


As you can see we need to boost the highs by >20dB and there is a ~10dB boost in the lows

My first tool to use is Dirac Live 3 but even that cannot cope with such a huge compensation in the highs - it simply could not raise the levels adequately....

So here is what I am doing to circumvent that:

I am using CurveEQ to smoothen out the response based on a sweep measurement at the listening position

1644221158052.png


You can see the signal path too on the left hand side (the Param EQ is only used to set the volume of my rear speakers relative to the fronts)

Then I optimize the system with Dirac Live (Dirac will 'see' my system's response with the flattened curve fed from CurveEQ)

Then I further optimize Dirac with another curve in CurveEQ (this is needed to smoothen the response curve that Dirac creates)

1644221342584.png


And the last step is to apply a room curve to taste (I personally do not use any of those 'standard' room curves, I just set the highs and lows to my personal preference)

1644221417575.png


And all that results in the optimized frequency response you saw above!

And finally a few remarks after having listened to the speakers for a few weeks now:
  • they have still not broken in - sound signature still changes from time to time
  • soundstage is obviously extremely wide but really laser precise
  • soundstage depth is OK, you can hear which instrument is behind which one but not like meters apart
  • phantom center is a bit bigger vs with my previous system (12" one-way Audio Nirvana speakers) but it is still precise and it blends nicely into the whole soundstage (= it does not come more forward)
  • detail is superb, surprisingly so (especially with 'noises' like when a singer takes a breath or when a musician touches his/her instrument, etc.)
  • as a result of the wide soundstage instruments are not so densely 'packed' on each other so you can hear them very nicely separated and that I believe provides further details
  • bass goes to the basement, it is solid down to about 22-23Hz - for music you definitely do not need a sub
  • bass is also extremely detailed - I have never heard anything like this with any previous systems that I bought or built
  • for movies I can also live without subs, I can totally feel the couch moving under my ass - however to be fair I need to add that the impact sometimes is not as visceral as it used to be with my dual 12" subs in my previous system
  • SPL: I have measured the SPL at listening position (>3 meters) vs going to the back of the room (>7 meters) and the difference was only 2dB!!!. This is totally mind-blowing......I guess it is yet another advantage of the line array concept
  • The sound totally fills the room - really no matter where you go, sit, lie, etc. you hear pretty much the same sound (except for the lows below the Schroeder frequency). Another mind-blowing phenomenon for me (and actually for the whole family - they do appreciate that fact)
That's all for now; I hope you guys found this useful - any questions, comments are welcome!
 

kipman725

Active Member
Joined
Jul 6, 2020
Messages
255
Likes
224
20dB of boost in the digital domain is a bit worrying, especially as this was achieved by cascading EQs. Its certainly possible for this to be OK but I would carefully check that there is no clipping with test tones and looking at the DAC output.
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,219
Likes
1,360
Location
Budapest
Good point @kipman725!
There is no clipping since I use Jriver's 64-bit Internal Volume Control upstream in the chain - meaning that all the plugins receive the attenuated signal.
I still have approx. 30dB headroom left after all these boosts
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,219
Likes
1,360
Location
Budapest
Those are the actual correction points - CurveEQ is a curve (or spline) equalizer, meaning that it does not operate as a parametric EQ but it performs correction based on a series of points (eg. -2.1dB at 1200Hz, +3.6dB at 1856Hz, etc.)

How I came up with the correction curve:
- perform a measurement at the listening position at 75dB
- export the measured frequency response with 1/12 smoothing
- in Excel subtract the measured values from 75 (ie. 75 minus the measured values)
- save as CSV
- import into CurveEQ

For me this is a better method vs working with filters - and for me personally it sounds waaaay better too in contrast with for example using filter export in REW and convolute it in Jriver with its built-in convolution engine
Also CurveEQ natively supports multichannel which is very important for me. Unfortunately it has poor resolution under 100Hz but that's where Dirac comes to play.
 
OP
P

ppataki

Major Contributor
Joined
Aug 7, 2019
Messages
1,219
Likes
1,360
Location
Budapest
Just a tiny update: now the driver related air leakage issues haven also been fully fixed!
I have replaced one driver that had a strange mechanical noise and I have fixed the insulation for two other drivers
Now there is neither cabinet nor driver related air leakage issue at all

I have made a video where I play a 30Hz LF test tone at 80dB at the listening position (325cm), recorded at approx. 50cm
I can barely hear any self-noise - I may even say that the self-noise is less than that of my 12" Dayton subs



Next step will be to re-measure and re-optimize the whole system and post all the measurements again
And with that I am planning to conclude this project and turn my attention to the next one.
 

DanielT

Major Contributor
Joined
Oct 10, 2020
Messages
4,752
Likes
4,633
Location
Sweden - Слава Україні
I have replaced one driver that had a strange mechanical noise ....
As one person who builds line speakers said: Order at least one extra driver. Someone will be broken, measure incorrectly or you drive the screwdriver into someone during assembly. There will be something and you will be annoyed that you did not order one or more extra driver from the beginning.

By the way, as usual: Good job!:D
 
Top Bottom