Murphy Corner Line Array

[hollis]

I crossed out the unnecessary word. Yes I could and I did https://www.audiosciencereview.com/...-low-q-subs-quasi-plain-wave-radiation.24396/
Ppataky's room isn't ideal for DBA:
View attachment 221890

Agreed that that room would not work with DBA, but disagree on the crossing out of "line". Getting the front and rear aligned takes quite a bit of time/work. I would never recommend someone attempt a DBA with non matching front and rear bass sources. Even different models of sealed subs from the same brand have been trouble.

 

[Flaesh]

I would never recommend someone attempt a DBA with non matching front and rear bass sources

At the first time I used midbass 15" boxes as rear subs to try and then sealed rear with IB front. In both cases rears was equalised in near field to the same responce; it works.

 

[hollis]

My updated approach to equalization, requires a HTPC or DDRC-24:

1. Left and Right linked, first round of adjustments in mono
2. New, better calculated LT thanks to @Joecarrow and a greater knowledge of simulation software + empirical testing. This has far less unneeded boost below ~35hz (the working cutoff where I think MCLA need to dig down to)
3. 2x High Shelfs cascaded, one at 2,700hz Q0.5 +7dB, second 15,000 Q0.5 +16dB. These are to linearize the response as much as possible pre Dirac, so that Dirac has less work to do. When we make its job easier, it can work harder at what its best at (making left and right match very closely).
4. 2x PEQs to correct nearfield behavior, one at 6000hz Q8 -11dB, second at 13000 Q10 -6dB (optional). Based off nearfield measurements there are some driver anomalies that could use some extreme correction. That second one might not be audible, I'm still A/Bing bypassing it.
5. </end LR link>
6. 5x PEQs remain for room based behavior. My current methodology is to play pink noise through one ch at a time, add filters to both attenuate the midbass hump (100-300hz) and make L and R match as closely as possible. Alex was there last time I did this in his old studio. Joe, gotta invite you for my next MCLA tune. I like using the RTA in iOS Audiotools, though REWs RTA would be fine with some adjustments to the default settings.
7. Dirac. 3-6x measurements, then playing with target curve, maybe a family of target curves. The newest version, 3.3.3, has a sense of room gain and by default, does not attempt to correct room gain back to flat in the bass region. The seems to sound more natural to most users. In the case of the the line array in a studio-space last night, Dirac chose +6db as the lower shelf on the target curve. I had never seen it chose a value so high, but it sounded great.

I am still questioning how much I should bother with maxing out all the PEQ channels pre-Dirac. I'll also look into 100% FIR based eq at some point.

 

[ppataki]

My updated approach to equalization, requires a HTPC or DDRC-24:

1. Left and Right linked, first round of adjustments in mono
2. New, better calculated LT thanks to @Joecarrow and a greater knowledge of simulation software + empirical testing. This has far less unneeded boost below ~35hz (the working cutoff where I think MCLA need to dig down to)
3. 2x High Shelfs cascaded, one at 2,700hz Q0.5 +7dB, second 15,000 Q0.5 +16dB. These are to linearize the response as much as possible pre Dirac, so that Dirac has less work to do. When we make its job easier, it can work harder at what its best at (making left and right match very closely).
4. 2x PEQs to correct nearfield behavior, one at 6000hz Q8 -11dB, second at 13000 Q10 -6dB (optional). Based off nearfield measurements there are some driver anomalies that could use some extreme correction. That second one might not be audible, I'm still A/Bing bypassing it.
5. </end LR link>
6. 5x PEQs remain for room based behavior. My current methodology is to play pink noise through one ch at a time, add filters to both attenuate the midbass hump (100-300hz) and make L and R match as closely as possible. Alex was there last time I did this in his old studio. Joe, gotta invite you for my next MCLA tune. I like using the RTA in iOS Audiotools, though REWs RTA would be fine with some adjustments to the default settings.
7. Dirac. 3-6x measurements, then playing with target curve, maybe a family of target curves. The newest version, 3.3.3, has a sense of room gain and by default, does not attempt to correct room gain back to flat in the bass region. The seems to sound more natural to most users. In the case of the the line array in a studio-space last night, Dirac chose +6db as the lower shelf on the target curve. I had never seen it chose a value so high, but it sounded great.

I am still questioning how much I should bother with maxing out all the PEQ channels pre-Dirac. I'll also look into 100% FIR based eq at some point.

My current correction chain for my MCLA, just FYI (a bit of a different approach compared to yours)
I also apply the same chain to my rear truncated line arrays

1. PreEQ using Jriver's PEQ filters

2. A number of plugins in DDMF Metaplugin:

CurveEQ flattens the response like an iron:

But its resolution is not enough below 200Hz so I apply convolution using REW generated filters to iron out the lows too:

And finally CraveEQ to apply a curve to taste:

I used to use Dirac too but it could not make the response flat enough for me, not even with pre-EQ
And this way I have more freedom too in terms of compensating for the various dips

EDIT: and the results:

 

[hollis]

1. PreEQ using Jriver's PEQ filters

Ah! In Jriver you have Linkwitz Transform right allongside all the eq options. Definitely try that out sometime, I am told the LT has less negative effects on phase than cascasded LF boosts.

2. A number of plugins in DDMF Metaplugin:

I was considering BlueCat but that one looks promising as well.

CurveEQ flattens the response like an iron:

I need to try that one, I use Voxengo Span a lot, its my favorite visualizer.

But its resolution is not enough below 200Hz so I apply convolution using REW generated filters to iron out the lows too:

Why not just convolution full range?

EDIT: and the results:

Epic!

I think next for me will be to try Cavern + Equalizer APO with an obscene number of taps:

Cavern

Download a free room correction software and Dolby Atmos decoder for PC. Improve the sound quality of your home theater and enhance your listening experience.

cavern.sbence.hu

 

[fluid]

Ah! In Jriver you have Linkwitz Transform right allongside all the eq options. Definitely try that out sometime, I am told the LT has less negative effects on phase than cascasded LF boosts.

Minimum phase filters that create the same magnitude response produce the same phase response so unless they meant something else that doesn't make any sense.

Metaplugin is great and fairly cheap it really increases the flexibility of processing that can be done.

 

[ppataki]

Ah! In Jriver you have Linkwitz Transform right allongside all the eq options. Definitely try that out sometime, I am told the LT has less negative effects on phase than cascasded LF boosts.

I am using LT in my home office setup
The only caveat with that is that you definitely need a high-pass filter to combine it with, otherwise the driver will bump into over-excursion
@fluid the LT filter does have a different frequency curve vs the low shelves, I guess hence the phase difference

Why not just convolution full range?

I cannot provide an 'ASR-compatible' answer here
It just sounds better....when I use convolution for full range vs using CurveEQ, somehow the latter sounds more 'lively' (even though the frequency response is the same....)

Epic!

Thank you

I think next for me will be to try Cavern + Equalizer APO with an obscene number of taps:

I did try that too but since I could not get Equalizer APO to work with ASIO or WASAPI I ditched it (I am using Jriver's WDM driver to have a system-wide DSP)

 

[fluid]

I am using LT in my home office setup
The only caveat with that is that you definitely need a high-pass filter to combine it with, otherwise the driver will bump into over-excursion

That is down to the design and driver being used, it is perfectly possible to use a Linkwitz transform without the need of a high pass filter, but most people use it with huge boost and small drivers hence the excursion.

@fluid the LT filter does have a different frequency curve vs the low shelves, I guess hence the phase difference

My point is it makes no sense to describe the phase of a minimum phase filter as worse than another one, the phase follows the magnitude.
What matters is the combined acoustic and electrical responses, the phase of the individual filters might look bad but when combined be just what was intended.

A Linkwitz transform is great at realigning the free space response of a closed box, the likelihood that on its own gives you a great in room response is very low.

I did try that too but since I could not get Equalizer APO to work with ASIO or WASAPI I ditched it (I am using Jriver's WDM driver to have a system-wide DSP)

Equalizer APO works as an audio processing object inside the base windows audio system, so it cannot be used with drivers that access the hardware directly like ASIO and WASAPI, because you have cut off it's access to the audio hardware

 

[gnarly]

Hi guys, just a little 2c caution about LTs or low shelving on the lines...
I share ppataki's thoughts on needing a hpf with them.
Apart from what I think has to be a degradation of SQ from excursion modulation;
a bigger issue is I've had both driver coils and leads to the cones, come unglued/separated .....from what i'm pretty sure was plain old over excursion, over prolonged use...

 

[ppataki]

Hi guys, just a little 2c caution about LTs or low shelving on the lines...
I share ppataki's thoughts on needing a hpf with them.
Apart from what I think has to be a degradation of SQ from excursion modulation;
a bigger issue is I've had both driver coils and leads to the cones, come unglued/separated .....from what i'm pretty sure was plain old over excursion, over prolonged use...

and I guess this is not just valid for the line arrays but also in general

I usually check where the -3dB point is with all the filters and then apply a HPF there in linear phase mode to decrease the post-ringing introduced by the HPF filter
That works pretty well for me at least and I can see the cone's excursion is limited then (easy to spot even with my eyes from the sides)

 

[hollis]

Hi guys, just a little 2c caution about LTs or low shelving on the lines...

Thanks for that word of warning. I use a both high and low protective filters.

I actually added the HF LP first, since REW sweeps were making crazy noise/distortion at the end of sweep. Turns out REW starts and stops its sweeps before the points you set. So if you sweep 10hz to 20khz, it is actually sending 5hz to 30khz to the speaker, with my cascaded high shelfs that was a recipe for disaster. I currently have 23dB of boost at 20khz and a 20khz 12db/oct Butterworth got rid of the noise.

 

[Wesayso]

2x PEQs to correct nearfield behavior, one at 6000hz Q8 -11dB, second at 13000 Q10 -6dB (optional). Based off nearfield measurements there are some driver anomalies that could use some extreme correction. That second one might not be audible, I'm still A/Bing bypassing it.

Basing such EQ tweaks on a nearfield measurement isn't going to be re-presentable for normal listening distances.

Let's see an unshaded straight array, EQ-ed flat at 2.7 meter:

Nice and clean, the grey line predicts the (nice and flat) response (all floor and ceiling reflections turned down by 20 dB).

Now let's see how this looks more nearfield, say at 1 meter:

Not nearly as nice! Now the comb filtering is way more obvious and starts way earlier. But if your listening spot is at, say: 2.7 or 3 meter, you don't want to EQ up close!
So be careful what you do, and at what distance you do it. One thing I could advise if using FIR filters, tone down the over correction due to comb filtering, you see that straight line result I showed at 2.7 meter? It has peak corrections at higher frequencies that is only valid at a single listening distance/spot.

See the peaks at 9KHz and higher? Turning those down a little will even out the results around the sweetspot, making for a sweeter overall sound.
This is just a simulation, but these simulations can teach you a lot of valuable things about these arrays.

 

[hollis]

Basing such EQ tweaks on a nearfield measurement isn't going to be re-presentable for normal listening distances.

Of all the people to get schooled by, I cant pick anyone better. I will read and digest. And maybe respond with the data that prompted me to make those decisions. One thing I can say, 100% of the things I know for sure are only at nearfield, so 3m is a different world to me.

Thank you @Wesayso

 

[fluid]

Basing such EQ tweaks on a nearfield measurement isn't going to be re-presentable for normal listening distances.

Can't agree more line arrays are messy up close

See the peaks at 9KHz and higher? Turning those down a little will even out the results around the sweetspot, making for a sweeter overall sound.
This is just a simulation, but these simulations can teach you a lot of valuable things about these arrays.

I tried really hard to find a better sound by avoiding "overcorrecting" the upper end comb filtering, and whilst as general advice I agree, for whatever reason the best sound I get is when they are heavily corrected. There could be a ton of reasons for this and I don't have a good answer why. My thinking now is to try the options and see what you like. DSP corrections are pretty easy to AB without a significant time delay.

 

[ppataki]

for whatever reason the best sound I get is when they are heavily corrected.

I can totally confirm the same from my end too!

 

[Wesayso]

I leave pretty much all of it as is, but shave the top of the higher peaks above say 9 KHz with phase linear EQ and checking the levels of energy stays the same when smoothed somewhat. Another trick is to get left and right output balance as close as you can, when summing it it can really hit + 6dB over the entire span except maybe on the far ends of the spectrum.

Left, right and sum, where I used one of the array's strong points in bass to fix the hole(s) in the other side.
Right is weak at 30 Hz, left has a dip at 70 Hz. This probably is heavily smoothed 1/3th.
Phase is corrected to follow the frequency curve (minimum phase), resulting in this as a stereo sum at the listening spot:

This only works if the first reflections are absorbed by quite a lot though. Don't try it when there's bare walls left and right of the speakers or close to the back of the listening position.

 

[Wesayso]

Maybe this part deserves some extra explaining... I remove energy at 70 Hz at the left dip (seen as high distortion dip in measurement) and add an equal amount of EQ to the right side. Tweaks like this made with measurement gear in line to get the lowest overall distortion while summing up nicely. It only works at low frequencies though. One can do the same with a subwoofer and an array, use each other's strong points to make the sum work. Timing (phase) needs to be the same though.

 

[ernestcarl]

I leave pretty much all of it as is, but shave the top of the higher peaks above say 9 KHz with phase linear EQ and checking the levels of energy stays the same when smoothed somewhat. Another trick is to get left and right output balance as close as you can, when summing it it can really hit + 6dB over the entire span except maybe on the far ends of the spectrum.

View attachment 240884

Left, right and sum, where I used one of the array's strong points in bass to fix the hole(s) in the other side.
Right is weak at 30 Hz, left has a dip at 70 Hz. This probably is heavily smoothed 1/3th.
Phase is corrected to follow the frequency curve (minimum phase), resulting in this as a stereo sum at the listening spot:

View attachment 240886

This only works if the first reflections are absorbed by quite a lot though. Don't try it when there's bare walls left and right of the speakers or close to the back of the listening position.

What's the name of the program you're using to display your wavelet graph again? I've seen it before... but just cannot recall.

 

[fluid]

What's the name of the program you're using to display your wavelet graph again? I've seen it before... but just cannot recall.

APL TDA
http://aplaudio.com/conc2/products/tda/

 

[gnarly]

APL TDA
http://aplaudio.com/conc2/products/tda/

Have you guys seen/used APL TDA RT (real time)?
It's kind of hard to find on the website http://aplaudio.com/conc2/products/tdart

I bought it but couldn't use it very well, due to an old PC with not enough video card power. Its real-time graphics took down my old machine lol.
Upgraded the PC, but have been waiting for video cards to get back to $ reasonable....
Need to get a card...I'd like to help Raimonds promote it...what little time with it i tried, worked well.

It uses a continuous series of real short log chirps.
Interestingly, the "tornado" isn't the screen used for precise time alignment...it's more for overall general confirmation, (much like waterfalls and spectros imo).
The screen/tool for precise time alignment, as it chirps along....... is good ole real time phase trace alignment.