23 minutes ago, Keith_W said:

High end network cable from Jaycar  $20 for 10m. 

 

Do youse mean Cat 5/Cat 6 Ethernet cable, Keith?

 

 

I made sure I bought copper cable for that warm sound. 

I think you will find a Nordost Valhalla Ethernet cable makes a better speaker cable.

I was going to ask whats wrong with Andy's  cat 6 Belden speaker /ethernet cable but why not go for the Valhalla?

you only live once

12 minutes ago, frednork said:

you only live once

Perhaps, but you might spend an eternity in Valhalla

 

I once spent an eternity in Valhalla watching this movie (sorry, thats a Melbourne joke)

 

Edited February 8, 20242 yr by frednork

2 hours ago, Keith_W said:

I made sure I bought copper cable for that warm sound. 

 

Seems to be solid-core Cat6 - should sound good.

 

1 hour ago, Artnet said:

I was going to ask what's wrong with Andy's  Cat 6 Belden speaker / ethernet cable but why not go for the Valhalla?

 

Keith's cable has the RJ45 connectors already installed.  It is almost certainly PVC insulated - so won't sound as good as mine - but having the connectors already installed ... is a biiig plus.  

 

 

Who said it's for sound? It's a network cable! It's for connecting my PC to my 10TB NAS where all my porn music is stored!! 

@Keith_W what's the latest since room re-arrangement, are you contented with the sound and feel?

It is a massive improvement. I have pushed the subs into the corners. This has lead to a massive boost in the efficiency of the subs, as well as a ridiculously wide soundstage. I have roughly dialled in the DSP and I am still refining it. Amazing how much change a simple rotation in the room will produce. The room also looks a lot nicer. 

1 hour ago, Keith_W said:

It is a massive improvement. I have pushed the subs into the corners. This has lead to a massive boost in the efficiency of the subs, as well as a ridiculously wide soundstage. I have roughly dialled in the DSP and I am still refining it. Amazing how much change a simple rotation in the room will produce. The room also looks a lot nicer. 

 

Who would have thought.  Pushing the subs into the corners gives more bass 😂

The motivation for pushing the subs into the corners is (1) load the room so that I can gain subwoofer efficiency, and (2) take advantage of David Griesinger's theory of stereo subs. Of course it introduces other problems, which I am working out with DSP at the moment. I spent a few hours looking at curves, now I am taking a break. 

26 minutes ago, Keith_W said:

The motivation for pushing the subs into the corners is ... and (2) take advantage of David Griesinger's theory of stereo subs. 

 

Can you give me a link to this theory, Keith?

 

 

 

 

27 minutes ago, andyr said:

Can you give me a link to this theory, Keith?

 

Loudspeaker and listener positions for optimal low-frequency ...

davidgriesinger.com

http://www.davidgriesinger.com › asa05

 

Also:

 

Theory & Design of a Digital Audio Signal Processor for ...

Pearl HiFi

https://pearl-hifi.com › Audio_DSP_for_Home_Use

 

If you Google David  Griesinger etc, these and others are available.

 

Edited March 2, 20242 yr by parrasaw

1 hour ago, parrasaw said:

Loudspeaker and listener positions for optimal low-frequency ...

davidgriesinger.com

http://www.davidgriesinger.com › asa05

 

Thanks, Paul.  👍

 

@Keith_W

Fantastic thread - it's one of the reasons that I keep coming back to SNA. Re measurements: You're using MMM for measurements of the subwoofers and full speakers (with x/o filters in-place). What technique are you using for the measurements of the drivers that are used to linearise the drivers and then create the x/o for the speakers? Are these the traditional fixed microphone measurement set-up or are you also using MMM for these measurements? 

5 hours ago, zydeco said:

@Keith_W

Fantastic thread - it's one of the reasons that I keep coming back to SNA. Re measurements: You're using MMM for measurements of the subwoofers and full speakers (with x/o filters in-place). What technique are you using for the measurements of the drivers that are used to linearise the drivers and then create the x/o for the speakers? Are these the traditional fixed microphone measurement set-up or are you also using MMM for these measurements? 

 

I stopped updating this thread because I thought that SNA members were getting bored at looking at graphs. As Steff mentions, I recently rotated the system which meant redoing the DSP. I could have skipped some steps, but I decided to redo the entire thing from the ground up to see how close my new corrections resemble the older ones. And then there are new problems to solve. 

 

I do use MMM for measurements, but not for driver linearization. For those, I use a nearfield SPS - this is because MMM does not capture any time information, and I want to linearize the phase as well. I am sometimes not sure how far away from the driver to place the microphone - some say the mic should be so close to the driver that it is almost touching, some say the mic should be at a distance of 1.5x the width of the baffle to account for baffle step. And then there are the horns. I take measurements at various distances and look at the curves. For now I have settled for: tweeter 20cm, horn 60cm, and woofers 1m. I used to linearize the subwoofers, but I don't do it any more because a nearfield linearized subwoofer is meaningless, given that the response is dominated by the room. 

 

I use MMM for studying the "overall" speaker response and for "overall" response at the MLP. 

@frednork asked me how I determined speaker position, listening position, etc. in my new setup. Since it may be of interest to SNA members, this is how I did it.

 

1. I started by fixing the position of the subwoofers in the room. I wanted them in both front corners to provide corner loading (thereby increasing subwoofer efficiency) as well as Griesinger's stereo subwoofer effect. 

 

2. Next was to establish the plane of the speakers. OCA (owner of the Youtube Channel "Obsessive-Compulsive Audiophile") helped me by providing the following diagram*** illustrating ideal speaker placement which will result in maximally delayed reflections in my room. 

 

 

According to the diagram, placement of the speakers and MLP anywhere within that grey zone is OK. I want some area to walk behind the MLP, so I decided that the speaker plane should be minimum allowable in that diagram. I established the speaker plane at 1.2m from the front wall. 

 

3. The next step is to establish the equilateral triangle between two speakers and MLP.  I started by temporarily placing the speakers 4m apart as per the diagram, then I established where the MLP had to be. I knew that the y-axis of the MLP has to be exactly in the centre of the room, so I placed a tape measure on the y-axis and started taking sweeps: 

 

 

 

The reason for doing sweeps is to find the position with the fewest peaks and dips in the frequency response as possible. I overlaid all the sweeps in REW and compared them. I settled on a distance of 220cm from the rear wall. If you look at those sweeps, you will see that in every instance there is a dip in the right subwoofer at 40Hz; but at 220cm the dip isn't as objectionable as all the others. Incidentally, this is very close to the "rule of thirds" (the length of the room is 6m). 

 

4. Now that the position of the MLP is determined, and the position of the speaker plane is determined, we can set up our equilateral triangle. Given that we know the speaker plane, the MLP, and the angle that we want, it is a simple calculation: 

 

 

... so I needed to place my speakers 4m apart. I determined the central point on the speaker plane, and pushed the speakers there. I wasn't too happy with that position (I thought they looked funny, and conflicted with furniture), so I ended up with less than an equilateral triangle. But if I had no furniture constraints, they would be an equilateral triangle. Final decision: semi-equilateral triangle. Distance between speakers 3.2m, distance between MLP to speakers 3.5m, and speakers pulled further into the room. All positions are still within the grey zone calculated above. 

 

Although the diagram indicates that a smaller equilateral triangle (2.4m on all sides) is OK, I do not want to sit too close to my speakers. They are big horns with large distance between drivers, and they do not fully integrate if I sit too close to them. Not to mention, they are physically intimidating because the speakers are so big. 

 

Once the final position of subs, speakers, and MLP is established, the next step is to dial the DSP in. FYI, I have finished dialling the DSP in. This post is getting quite long already, so I will make a separate post later. 

 

*** If you want to run this for your room, go to this website, click on "web application", and upload this code (replace Lx and Ly with your room width and length dimensions in meters) : 

 

import contour;
import graph;
unitsize(3cm);
real Lx = 7.0;
real Ly = 5.9;
real ly = 1.5;
real lx = 0.5*Lx;
real mindelay = 6.0;
pair listener = (lx,ly);
real rt(real x1, real x2, real y) {
real dd = sqrt( (x1-x2)^2 + y^2 );
real dr = sqrt( (x1+x2)^2 + y^2 );
return (dr-dd)/0.343;
}
real rtT(real x, real y) { 
return rt( Ly-y, Ly-ly, x-lx );
}
real rtB(real x, real y) { 
return rt( y, ly, x-lx );
}
real rtR(real x, real y) { 
return rt( Lx-x, Lx-lx, y-ly );
}
real rtL(real x, real y) { 
return rt( x, lx, y-ly );
}
real rtmin(real x, real y) { 
  return min( rtT(x,y), rtB(x,y), rtL(x,y), rtR(x,y) );
}
guide[][] g=contour(rtmin,(0,0),(Lx,Ly),new real[] {mindelay});
fill(g[0],mediumgrey);
picture equiguide=new picture;
real L=sqrt(Lx^2+Ly^2);
path equilateral = (listener+L*dir(120))--listener--(listener+L*dir(60));
draw(equiguide, equilateral, dashed );
clip(equiguide,(0,0)--(Lx,0)--(Lx,Ly)--(0,Ly)--cycle);
add(equiguide);
draw( listener, marker(scale(2mm)*polygon(3),FillDraw(green) ) );
draw( new Label[] {Label("FRONT WALL",Relative(0.6),Center,UnFill(1bp))},
contour(rtT,(0,0),(Lx,Ly),new real[] {mindelay}), red+linewidth(2) );
draw( new Label[] {Label("REAR WALL",Relative(0.4),Center,UnFill(1bp))},
contour(rtB,(0,0),(Lx,Ly),new real[] {mindelay}), red+linewidth(2) );
draw( new Label[] {Label("LEFT WALL",Relative(0.15),Center,UnFill(1bp))},
contour(rtL,(0,0),(Lx,Ly),new real[] {mindelay}), red+linewidth(2) );
draw( new Label[] {Label("RIGHT WALL",Relative(0.15),Center,UnFill(1bp))},
contour(rtR,(0,0),(Lx,Ly),new real[] {mindelay}), red+linewidth(2) );
int n=5;
real[] c=new real[n];
for(int i=0; i < n; ++i) c[i]=i+1;
Label[] Labels=sequence(new Label(int i) {
return Label(c[i] != 0 ? format("$%0.0f$\,ms" ,c[i]) : "",
        Relative(0.49 - 0.01*i/n),(0,0), UnFill(1bp));
},c.length);
draw( Labels, contour(rtmin,(0,0),(Lx,Ly),c, nx=100, ny=100) );
draw( (0,0)--(Lx,0)--(Lx,Ly)--(0,Ly)--cycle, blue+linewidth(2) );
xaxis(Label("$x$ [meters]",Relative(0.5)),BottomTop,RightTicks,xmin=0);
yaxis(Label("$y$ [meters]",Relative(0.5)),LeftRight,LeftTicks,ymin=0);
path contour6 = g[0][0];
pair[] furthestpoints = intersectionpoints(equilateral, contour6);
pair p1 = furthestpoints[0];  pair p2 = furthestpoints[1];
draw( Label(format("$%0.2f$\,m", length(p2-p1)), LeftSide),p1--p2, dashed, Arrows );

 

Edited March 7, 20242 yr by Keith_W

@Keith_W

Some time ago I used a Win 10 computer running Audiolense to do measurements as well as create and execute FIR filters. It all worked well with integrated workflow and an OS with which I’m familiar but the stability of the computer ultimately made it untenable for use in our family system. I spent time optimising the setup (turning off updates, auto-resistant of apps…)  but still would find occasions when the system wasn’t working for the family due to some issue with the computer. Its tangential to the thread but it’d be good to understand your Windows computer setup as well as how you’ve addressed such issues.

Hi zydeco. My strategy is very simple: I have a backup. I regularly get stupid minor niggles, especially since i "upgraded" my wi-fi recently. But if anything catastrophic happens I just restore it like any other PC. I agree with you that the fragility of my solution is a concern, and should I ever decide that it's "enough" i'll move all the convolution to a SFF PC running Linux and JRiver. But at the moment I am doing a lot of experiments, and it is really convenient to do all the work on the same PC which does the convolution. 

@Keith_W
From memory, the approach used in Acourate is to use near-field measurements to build filters that linearise drivers and execute cross-overs before using measurements at the listening position to build filters that correct against an overall target curve.  Is this correct?  And, if so, then any idea as to the rationale for this approach rather than just doing measurements at the listening position that are used to deliver both the cross-over and overall target curve corrections? 

2 hours ago, zydeco said:

@Keith_W
From memory, the approach used in Acourate is to use near-field measurements to build filters that linearise drivers and execute cross-overs before using measurements at the listening position to build filters that correct against an overall target curve.  Is this correct?  And, if so, then any idea as to the rationale for this approach rather than just doing measurements at the listening position that are used to deliver both the cross-over and overall target curve corrections? 

 

Hi zydeco. The further you get away from a speaker, the more the measurement gets contaminated by reflections. Thus more non-minimum phase behaviour is introduced into the measurement. You can remove this with windowing for high frequencies and by analysis of the GD graph for lower freqs, but the result is still contaminated by non minimum phase behaviour. If I am taking a measurement to correct the driver, I would rather measure the driver only rather than attempt to correct non-minimum phase room behaviour as well. A bit of experimentation is required to determine where the "nearfield" is for any speaker, usually a minimum of one baffle width away to a maximum of 1m. 

1 hour ago, Keith_W said:

 

Hi zydeco. The further you get away from a speaker, the more the measurement gets contaminated by reflections. Thus more non-minimum phase behaviour is introduced into the measurement. You can remove this with windowing for high frequencies and by analysis of the GD graph for lower freqs, but the result is still contaminated by non minimum phase behaviour. If I am taking a measurement to correct the driver, I would rather measure the driver only rather than attempt to correct non-minimum phase room behaviour as well. A bit of experimentation is required to determine where the "nearfield" is for any speaker, usually a minimum of one baffle width away to a maximum of 1m. 

I get that nearfield measurements are needed to linearise drivers but not why this step is required when building the overall filters. Or, put differently, if sound that we hear is a combination of direct and reflected sound then doesn’t it make sense to design the crossover around this combined response at the mai listening position. I have, for example, never linearised subwoofer driver performance but rather just measured and corrected for the listening position.

17 minutes ago, zydeco said:

I get that nearfield measurements are needed to linearise drivers but not why this step is required when building the overall filters. Or, put differently, if sound that we hear is a combination of direct and reflected sound then doesn’t it make sense to design the crossover around this combined response at the mai listening position. I have, for example, never linearised subwoofer driver performance but rather just measured and corrected for the listening position.

There are different ways of doing this. Acourate/  DEQX uses this nearfield/farfield  approach but other approaches like Audiolense or focus fidelity or dirac only measure farfield. They use different techniques to determine the critical info required to develop an appropriate filter l. If you have read  any of the blogs and walkthroughs from Mitch Barnett you would see he shows how both approaches can be used successfully with some of these tools to get an excellent result. But these varying approaches may suit some some tasks better than others so it does depend on the use case as to which may be a better choice.