Login / Register
 
DIY Sound Diffusers—Free Blueprints—Slim, Optimized DIY Diffuser Designs (+Fractals)
New Reply
Subscribe
#181
13th December 2012
Old 13th December 2012
  #181
Gear nut
 
unqlenol's Avatar
 
Joined: May 2009
Location: Cape Town, South Africa

unqlenol is offline
Quote:
Originally Posted by Arqen View Post
Looks Great, Nolan!

Thanks for going through with this! Looks like your version worked out well with the 35mm wells.

I agree with you and Jens that the performance is good considering how simple the base shape is. One thing I noticed is that the overall design is not quite symmetrical. To achieve symmetry the second or sixth module could be flipped (reversing the well order). I don't know if this will make any significant improvement to the performance (it might even hurt the performance), but it's worth a shot if you've not yet tried it.
Pleasure Tim, thank you!

I will make more models. I am trying to scratch my way to a decision.
There are a number of issues in my head:
I am still not sure about EPS. I would buy low density for cost reasons. This may waste LF performance potential even with a very hard polyurethane paint job.
Your method with wood is very inviting and I think it may not be much more expensive then EPS (even if I use the 'inverted offcuts' from a block of EPS)....unless I use Birch ply or a fancier ply
Also, I prefer wood as a material and a stained wood grain can often be a lovely thing.

If I were to include inverted EPS panels, I am not sure how I would implement them in a practical way with this symmetrical design (bearing in mind the inverted panels are meant to save money as they would technically be an off-cut from a giant block of EPS).

Jens, as I understand it, you use the high density stuff. Do you think 16kg/m3 EPS is a waste of time LF performance wise? I am less worried about toughness as certain paints can be surprisingly hard after a few coats.
#182
13th December 2012
Old 13th December 2012
  #182
Gear nut
 
unqlenol's Avatar
 
Joined: May 2009
Location: Cape Town, South Africa

unqlenol is offline
Tim, something I am not clear on regarding the mod sequence is:

Why multiply 0 4 5 3 5 4 0 by 2 in the first place?
Is this simply to improve array performance compared to modulating based on the standard 0 4 5 3 5 4 0?
Arqen
Thread Starter
#183
14th December 2012
Old 14th December 2012
  #183
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by unqlenol View Post
Tim, something I am not clear on regarding the mod sequence is:

Why multiply 0 4 5 3 5 4 0 by 2 in the first place?
Is this simply to improve array performance compared to modulating based on the standard 0 4 5 3 5 4 0?

That's a very good question Nolan...

If you were to literally create a fractal while keeping all the proportions the same, you would have to multiply by 7 (because 7 modules is 7 times as wide as 1 module, you'd also multiply the depth by 7 so it's 7 times as deep). But this would mean the modulated diffuser would have a total operational depth of 7x5cm + 5cm = 40 cm. And that's too deep for most people!

So, I simply tried out different scaling factors to see what worked well, without being too deep. Multiplying by 2 had better performance than multiplying by 1, so I used that. Simply a trial and error approach without much math behind it. You could also try multiplying by any number you want, such as 2.5 or 3. Multiplying by 3 would take the total operational depth up to 20 cm, and this might be a good option for those who don't mind the extra depth. My reflex trial expired right after I tested this modulation, and I did not have a chance to try a scaling factor of 3.

So yes, you're right! The scaling factor was simply chosen to provide a good trade-off between performance and depth!
Arqen
Thread Starter
#184
14th December 2012
Old 14th December 2012
  #184
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
N = 17 optimized stepped diffuser

Well, it turns out that I don't have a well depth sequence for an N = 17 optimized stepped diffuser. Only N = 7 and N = 36 designs. These diffusers were designed by Trevor J. Cox in 1994.

Here's a link to the paper
Attached Images
File Type: jpg cox-opt-n7-array-diffuse-25deg.jpg (105.7 KB, 413 views) File Type: jpg cox-opt-n36-bem-diffuse-25deg.jpg (103.3 KB, 517 views)
#185
14th December 2012
Old 14th December 2012
  #185
Lives for gear
 
Joined: May 2012
Posts: 2,318

OpusOfTrolls is offline
You're fired!

Arqen
Thread Starter
#186
16th December 2012
Old 16th December 2012
  #186
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Hahaha. Being fired /abducted by aliens looks like fun, and perhaps a good innovation strategy. I bet those aliens could help us come up with some majestic diffuser designs. Like a diffuser that diffuses the listener, rather than the sound!
#187
16th December 2012
Old 16th December 2012
  #187
Lives for gear
 
Jens Eklund's Avatar
 
Joined: Oct 2009
Location: Stockholm
Posts: 4,688

Jens Eklund is offline
Quote:
Originally Posted by unqlenol View Post
Jens, as I understand it, you use the high density stuff. Do you think 16kg/m3 EPS is a waste of time LF performance wise? I am less worried about toughness as certain paints can be surprisingly hard after a few coats.

Assuming thoroughly painted so the surface gets hard, I wouldn’t worry too much about using lower density EPS. Especially if you’re not concerned about toughness.
#188
16th December 2012
Old 16th December 2012
  #188
Lives for gear
 
Jens Eklund's Avatar
 
Joined: Oct 2009
Location: Stockholm
Posts: 4,688

Jens Eklund is offline
Tim:

A quick question regarding FDTD: How long does it take to render a random incident normalized diffusion coefficient plot with a decent resolution (like at least 1/3 oct up to about 4-5 kHz) using FDTD? Depends on the number of elements naturally, so say 5 periods of your B2-Frac as an example.
Arqen
Thread Starter
#189
16th December 2012
Old 16th December 2012
  #189
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by Jens Eklund View Post
Tim:

A quick question regarding FDTD: How long does it take to render a random incident normalized diffusion coefficient plot with a decent resolution (like at least 1/3 oct up to about 4-5 kHz) using FDTD? Depends on the number of elements naturally, so say 5 periods of your B2-Frac as an example.
Good question Jens...

A long time, unless special acceleration techniques are used!

On page 54 (pdf page 60) of my thesis there is a table that gives simulation times, using an i7 2600K processor.

For the simple stepped diffuser designs, the simulation time on a domain just over 10 meters long (large enough to accommodate analysis over a 5m radius, based on the AES-4id standard) is 20 minutes for each trial (i.e., each angle of incidence).

For the fractal designs, I set up the problem so that the domain uses a smaller mesh size, therefore the time jumps up to 22 hours for each trial. If you were to do this for 36 different angles of incidence, that's 33 days spent on computation!

There are ways to speed up the simulation, and I've summarized these in Section 10.1 of the paper. One method is to use some programming tricks, such as the near field to far field transformation (NFFFT), which simulates scattering over a large radius without having to compute the results at every single mesh node inside the radius. Another way to accelerate it is to use a hardware solution: run the simulation on multiple graphics processing units instead of on the CPU.

In general, FDTD is SLOW compared to BEM, but it gives lots of extra useful insight!
#190
17th December 2012
Old 17th December 2012
  #190
Lives for gear
 
Jens Eklund's Avatar
 
Joined: Oct 2009
Location: Stockholm
Posts: 4,688

Jens Eklund is offline
Quote:
Originally Posted by Arqen View Post
Good question Jens...

A long time, unless special acceleration techniques are used!

On page 54 (pdf page 60) of my thesis there is a table that gives simulation times, using an i7 2600K processor.

For the simple stepped diffuser designs, the simulation time on a domain just over 10 meters long (large enough to accommodate analysis over a 5m radius, based on the AES-4id standard) is 20 minutes for each trial (i.e., each angle of incidence).

For the fractal designs, I set up the problem so that the domain uses a smaller mesh size, therefore the time jumps up to 22 hours for each trial. If you were to do this for 36 different angles of incidence, that's 33 days spent on computation!

There are ways to speed up the simulation, and I've summarized these in Section 10.1 of the paper. One method is to use some programming tricks, such as the near field to far field transformation (NFFFT), which simulates scattering over a large radius without having to compute the results at every single mesh node inside the radius. Another way to accelerate it is to use a hardware solution: run the simulation on multiple graphics processing units instead of on the CPU.

In general, FDTD is SLOW compared to BEM, but it gives lots of extra useful insight!
So I guess you used quite a lot of CPU/GPUs if you tried more than 260 million shapes. Even if each render only took 1 second, it would still take over 8 years to finish.

A bit OT perhaps, but can you say something about the system used? How many GPUs etc (or does it say in your paper ... haven’t looked yet. lazy I know ... I guess I need to read it someday)?
Arqen
Thread Starter
#191
17th December 2012
Old 17th December 2012
  #191
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by Jens Eklund View Post
So I guess you used quite a lot of CPU/GPUs if you tried more than 260 million shapes. Even if each render only took 1 second, it would still take over 8 years to finish.

A bit OT perhaps, but can you say something about the system used? How many GPUs etc (or does it say in your paper ... haven’t looked yet. lazy I know ... I guess I need to read it someday)?

That's assuming I did a brute force exhaustive search, which would be ridiculous! In reality, I used a form of artificial intelligence called evolutionary optimization to efficiently search the design space of 260 million possibilities. Optimization was only feasible because of the efficient algorithm used (a genetic algorithm), which simulates natural selection.

In a nutshell, here's how the genetic algorithm works: Multiple candidate designs (diffuser shapes) are generated, their performance is tested, the top performers breed and the weakest performers are killed. Next, the algorithm generates some new candidates (by modelling natural processes like mutation). These new candidates compete against the top performers and the 'offspring' of the top performers. The cycle repeats over and over, and if left to run for as long as possible it will converge to the best possible performer (i.e. the 'fittest' individual). In practice, the algorithm is generally stopped once it has run for a while without finding any new candidate that can beat the current top performer, and at this point a solution is declared. This means it has found one of the top performers out of the 260 million possibilities... perhaps it's found the very best of the best (called the 'global optimum'), but more likely it has converged to one of the other top performers (called a 'local optimum').

Typically there will be numerous designs that have excellent performance, rather than one design that stands head and shoulders above the rest. In the rare case that there is one design that dwarfs the rest, it tends to be easier to find, and less sophisticated optimization algorithms can be used.

In my case, the genetic algorithm searched the design space using FDTD simulation to test the candidate designs, and after about 1200 trials the algorithm converged to a solution (a top performer). Similarly, Trevor Cox and Peter D'Antonio relied on about 1000 scattering predictions when they were optimizing diffusers.

Also, each trial during optimization only took about 2 minutes, because I performed optimization at a receiver radius of 2.5 m instead of the 5 m standard (2.5 m seemed to be a more reasonable working distance for home studio owners). All this is explained in Chapter 7 of the thesis, which I don't expect many people to read, but it's there if you're interested.
#192
17th December 2012
Old 17th December 2012
  #192
Lives for gear
 
Jens Eklund's Avatar
 
Joined: Oct 2009
Location: Stockholm
Posts: 4,688

Jens Eklund is offline
Quote:
Originally Posted by Arqen View Post
That's assuming I did a brute force exhaustive search, which would be ridiculous! In reality, I used a form of artificial intelligence called evolutionary optimization to efficiently search the design space of 260 million possibilities. Optimization was only feasible because of the efficient algorithm used (a genetic algorithm), which simulates natural selection.

In a nutshell, here's how the genetic algorithm works: Multiple candidate designs (diffuser shapes) are generated, their performance is tested, the top performers breed and the weakest performers are killed. Next, the algorithm generates some new candidates (by modelling natural processes like mutation). These new candidates compete against the top performers and the 'offspring' of the top performers. The cycle repeats over and over, and if left to run for as long as possible it will converge to the best possible performer (i.e. the 'fittest' individual). In practice, the algorithm is generally stopped once it has run for a while without finding any new candidate that can beat the current top performer, and at this point a solution is declared. This means it has found one of the top performers out of the 260 million possibilities... perhaps it's found the very best of the best (called the 'global optimum'), but more likely it has converged to one of the other top performers (called a 'local optimum').

Typically there will be numerous designs that have excellent performance, rather than one design that stands head and shoulders above the rest. In the rare case that there is one design that dwarfs the rest, it tends to be easier to find, and less sophisticated optimization algorithms can be used.

In my case, the genetic algorithm searched the solution space using FDTD simulation to test the candidate designs, and after about 1200 trials the algorithm converged to a solution (a top performer). Similarly, Trevor Cox and Peter D'Antonio relied on about 1000 scattering predictions when they were optimizing diffusers.

Also, each trial during optimization only took about 60 seconds, because I performed optimization at a receiver radius of 2.5 m instead of the 5 m standard (2.5 m seemed to be a more reasonable working distance for home studio owners). All this is explained in Chapter 7 of the thesis, which I don't expect many people to read, but it's there if you're interested.
I actually had a quick look.

Yes, I can imagine you did some search optimization like Peter and Trevor but if FDTD is slower than BEM, I guess you still used a beefy system to complete the search process unless serious detail was omitted in the performance predictions, right? Did you only simulate normal incidence for the optimization? (sorry, you don´t need to answer if it´s in the paper; I´ll have a look later).

Did you figure out what was wrong with the calculation of the diffusion coefficient b t w?
Arqen
Thread Starter
#193
20th December 2012
Old 20th December 2012
  #193
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by Jens Eklund View Post
I actually had a quick look.

Yes, I can imagine you did some search optimization like Peter and Trevor but if FDTD is slower than BEM, I guess you still used a beefy system to complete the search process unless serious detail was omitted in the performance predictions, right? Did you only simulate normal incidence for the optimization? (sorry, you don´t need to answer if it´s in the paper; I´ll have a look later).

Did you figure out what was wrong with the calculation of the diffusion coefficient b t w?
Yeah, it's pretty beefy. Beefier than I'd like . The design system I built does a more extensive search of the available candidates compared with most other optimization techniques, because it used an intrinsically parallel search algorithm (the integer genetic algorithm). I performed raw, unsimplified FDTD simulations (to measure scattering over a 2.5 m radius), which is a truly high resolution, nonlinear physical model. But I only optimized for normal incidence, because this allowed a much more extensive search.

It's usual to make some simplifications to the problem when doing a huge search like this. For example, Cox performed optimization at I think 7-10 selected frequencies, rather than at all frequencies. I originally was optimizing for all frequencies, but this extra detail made it more difficult for the algorithm to differentiate between a top performer and an average performer, and so I later took a hint from Cox's approach and resorted to 10-15 selected frequencies (I forget the exact number, but it's in my paper, which I don't have handy as I'm traveling right now).

The optimization system itself was the most important part of this project. And in order for optimization to be effective, the problem had to be set up so that the algorithm finds clear, authentic patterns within an extremely noisy problem...

Which brings us to the diffusion coefficient calculation. This was done after the optimization was complete, and while I've not empirically verified the culprit I believe the problem was caused by the inexact sensor array (the array of virtual microphones). You see, the sensors were arranged on a rectangular mesh to approximate a semicircle. I believe the rounded coordinates caused numerical errors when computing the autocorrelation diffusion coefficient. Naturally, I've redone all those computation using Reflex. BTW, thanks for originally pointing out those wonky results.
#194
23rd December 2012
Old 23rd December 2012
  #194
Gear nut
 
unqlenol's Avatar
 
Joined: May 2009
Location: Cape Town, South Africa

unqlenol is offline
More model fun:
35mm well width.
12mm ply as the basis with the 0453540 ratio.
periods modulated by multiplying ratio by 3 (0 12 15 9 15 12 0)
Seems like some very nice low end performance.

I wonder what a safe listening distance would be for this fairly deep chap?
Attached Images
Attached Files
File Type: pdf New Model1_2012-12-23_10-21.pdf (203.5 KB, 136 views)
#195
23rd December 2012
Old 23rd December 2012
  #195
Gear nut
 
unqlenol's Avatar
 
Joined: May 2009
Location: Cape Town, South Africa

unqlenol is offline
As above, but half the height using 6mm ply instead.
Could be a nice low profile option?
Attached Images
Attached Files
File Type: pdf New Model2_2012-12-23_11-42.pdf (273.7 KB, 99 views)
#196
23rd December 2012
Old 23rd December 2012
  #196
Gear nut
 
unqlenol's Avatar
 
Joined: May 2009
Location: Cape Town, South Africa

unqlenol is offline
And, last thing for today, here is another 'half height' (6mm ply) as above, except with 30mm wide wells - which lets the array fit into a 1.5m long space.

No report as Reflex crashed.
Attached Images
File Type: jpg half height 30mm well mod x 3.jpg (70.6 KB, 327 views)
Arqen
Thread Starter
#197
26th December 2012
Old 26th December 2012
  #197
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by unqlenol View Post
As above, but half the height using 6mm ply instead.
Could be a nice low profile option?
Nolan, I've been away for a few days but what a pleasant surprise! Fantastic that you've tested out various versions, and I agree that this one looks like a nice option.

As for the deep diffuser, I think a safe minimum listening distance would be about 5 meters. For QRDs it's often recommended that the minimum listening distance be three times the longest wavelength diffused. wavelength = velocity/freq, where velocty = 344 m/s. So if we were to say 200 Hz is the lowest frequency diffused, going by this guideline you'd want to be seated 5.16 m away .... And if it's 250 Hz, you'd want to be seated 4.13m away.

Of course that rule is not black and white. For example, some diffusers scatter energy in all directions except that of the listening position, in which case they can be used at a closer distance. I believe the RPG skyline has this trait. So the minimum distance depends on the low frequency polar response of this particular diffuser, and if this diffuser tends to scatter low frequencies away from the listener, you might be okay seated as close as 3.5 meters (which is just over twice the wavelength of a 200 Hz sound wave), provided that you don't care about hearing the low frequency diffusion. Bear in mind that I'm just using intuition here, so you should take this with a grain of salt.

While shallower diffusers sacrifice low frequency diffusion, they allow you to sit closer without experiencing artifacts. And in that sense, I think the second diffuser you posted provides a nice tradeoff.
#198
26th December 2012
Old 26th December 2012
  #198
Gear maniac
 
Joined: Jan 2012
Posts: 193

pitchfork is offline
Quote:
I used a form of artificial intelligence called evolutionary optimization to efficiently search the design space of 260 million possibilities.

Meanwhile, real intelligence is being used by millions of ipod listeners to not give a damn.

note to self...file under..

HOW TO UTTERLY WASTE A LIFETIME
rez
#199
26th December 2012
Old 26th December 2012
  #199
rez
Gear addict
 
rez's Avatar
 
Joined: Jun 2011
Posts: 387

rez is offline
@argen: thank you for making this tread and for also suggesting not to build your optimized diffusors because of the problems discussed here.
we ended up building a cox optimized n36 diffusor as you suggested and are pleased with the results. the build was a bit hard, not because of the layout, but because of the materials used and the lack of experience in building such a complex wooden thing.

there are some pictures here: diy n-36 optimized diffusor if you want to have a look.
there are some unfinished other accoustic treatments also visible on that pictures, so we have no final measurements at the moment.

peace, rez
__________________
here is some of my own music and some tracks that i have mixed: https://soundcloud.com/blackcapsound
Arqen
Thread Starter
#200
26th December 2012
Old 26th December 2012
  #200
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by rez View Post
@argen: thank you for making this tread and for also suggesting not to build your optimized diffusors because of the problems discussed here.
we ended up building a cox optimized n36 diffusor as you suggested and are pleased with the results. the build was a bit hard, not because of the layout, but because of the materials used and the lack of experience in building such a complex wooden thing.

there are some pictures here: diy n-36 optimized diffusor if you want to have a look.
there are some unfinished other accoustic treatments also visible on that pictures, so we have no final measurements at the moment.

peace, rez
Looks great!

Fortunately, we've resolved the earlier issues with the shallow diffuser designs in this thread by mounting the modules at different depths (as shown in the images above posted by unqlenol and I). So, I should clarify that you CAN achieve great performance with the designs in this thread, and they are simpler to build than the big stepped diffuser you've built.

But in your particular case, I think the N = 36 stepped diffuser was a great choice because it fits in the space you have available better than my modular designs would have, and it has better low frequency performance because of the extra depth.

Glad to see you've built it!
Arqen
Thread Starter
#201
27th December 2012
Old 27th December 2012
  #201
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by pitchfork View Post
Meanwhile, real intelligence is being used by millions of ipod listeners to not give a damn.

note to self...file under..

HOW TO UTTERLY WASTE A LIFETIME
Haha, yeah you know it... I'm a geek who gets twisted enjoyment out of stuff like this, at the expense of my own hair. But fortunately, this project attracts many rad opportunities and clients, so in the business sense it was certainly not a waste.

And hell, the ipod would not exist if not for nerds drooling over algorithms behind the scenes. So, those who don't give a damn sometimes rely on the weirdos who do.
Quote
1
#202
28th December 2012
Old 28th December 2012
  #202
Gear nut
 
unqlenol's Avatar
 
Joined: May 2009
Location: Cape Town, South Africa

unqlenol is offline
Quote:
Originally Posted by Arqen View Post
Looks great!

Fortunately, we've resolved the earlier issues with the shallow diffuser designs in this thread by mounting the modules at different depths (as shown in the images above posted by unqlenol and I). So, I should clarify that you CAN achieve great performance with the designs in this thread, and they are simpler to build than the big stepped diffuser you've built.

But in your particular case, I think the N = 36 stepped diffuser was a great choice because it fits in the space you have available better than my modular designs would have, and it has better low frequency performance because of the extra depth.

Glad to see you've built it!
I did look at that N36 and think 'hell, the performance chart looks good, I should build that!'. But then I started thinking about the construction side and thought ' hmmm, in my case I would be willing to sacrifice some performance in the name of lower cost and construction complication'.
The n36 also would not fit in my space as is.

I will probably go with my low profile option for these reasons:
1. Very decent MF and HF performance
2. Especially cheap on material due to 6mm ply (not 10 or 12)
3. Minimum listening distance is very forgiving for smaller rooms.

If I were to make the N36 I would go about it in one of two ways:

1. I would plot the design on paper and take it to my local hotwire company and they would simply cut the whole thing out from one giant block of EPS. Then I would prime it well with an appropriate water based primer and finally paint it with many coats of polyurethane paint (safer to use water based polyurethane paint btw but not as hard as solvent based poly paint).

The above way is just drawing, measuring, paying and painting.

2. If I wanted wood, what I would do is: I would STILL cut the whole thing from a block of EPS. Then I would have all the wooden plates cut by a company that does CNC cutting. Then I would glue each plate on to the EPS with cold glue. A big blob of acrylic caulk on either side of each plate would keep the plate fixed to the eps while the cold glue dries.
Effectively the EPS would be like a docking station for the wood. The EPS measurements would have to also take the 6mm side plates into account!
This way is pricier, but I think it could be effective.

Either way I think its awesome that Rez built that. Looks great!
Arqen
Thread Starter
#203
29th December 2012
Old 29th December 2012
  #203
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by unqlenol View Post
I did look at that N36 and think 'hell, the performance chart looks good, I should build that!'. But then I started thinking about the construction side and thought ' hmmm, in my case I would be willing to sacrifice some performance in the name of lower cost and construction complication'.
The n36 also would not fit in my space as is.

I will probably go with my low profile option for these reasons:
1. Very decent MF and HF performance
2. Especially cheap on material due to 6mm ply (not 10 or 12)
3. Minimum listening distance is very forgiving for smaller rooms.

If I were to make the N36 I would go about it in one of two ways:

1. I would plot the design on paper and take it to my local hotwire company and they would simply cut the whole thing out from one giant block of EPS. Then I would prime it well with an appropriate water based primer and finally paint it with many coats of polyurethane paint (safer to use water based polyurethane paint btw but not as hard as solvent based poly paint).

The above way is just drawing, measuring, paying and painting.

2. If I wanted wood, what I would do is: I would STILL cut the whole thing from a block of EPS. Then I would have all the wooden plates cut by a company that does CNC cutting. Then I would glue each plate on to the EPS with cold glue. A big blob of acrylic caulk on either side of each plate would keep the plate fixed to the eps while the cold glue dries.
Effectively the EPS would be like a docking station for the wood. The EPS measurements would have to also take the 6mm side plates into account!
This way is pricier, but I think it could be effective.

Either way I think its awesome that Rez built that. Looks great!
I think you have a good approach for how one could build the more complex n = 36 design. And I have to say that your version of the 7-module, low profile diffuser really does hit a sweet spot for use at a listening distance typical in smaller rooms.

Also, the fact that it's modular means you can always reconfigure the setup if you want to later, and if you ever want to shrink / expand your diffusion system, you can use fewer / more modules.

Great work! I can't wait to hear how the build goes!
#204
30th December 2012
Old 30th December 2012
  #204
Gear nut
 
Rockenberry's Avatar
 
Joined: Apr 2012
Posts: 78

Rockenberry is offline
Amazing thank you!

Sent from my GT-I9300
Arqen
Thread Starter
#205
31st December 2012
Old 31st December 2012
  #205
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by Rockenberry View Post
Amazing thank you!
Hey, my pleasure!

It's actually very rewarding to see people building these and experimenting to find what configuration will best meet their needs.
LSP
#206
1st January 2013
Old 1st January 2013
  #206
LSP
Gear addict
 
Joined: Feb 2009
Location: Inman, SC USA
Posts: 447

LSP is offline
I love the looks of this design. I wonder though, for aesthetic reasons, could one use different length boards ?

What I mean is, let's say the far left and right outer boards will be 4', then the next ones in on both left and right could be 2', then the next pair could be 3', etc.

I ask because these look similar to the ones Gavin Haverstick did for a fellow slut here, on his basement studio design, and can be seen on his you tube link.

At any rate, do you think varying the length of the boards would hinder, or alter the diffusion
properties.

Also, would it alter the diffusion properties if I rounded the mm measurements up or down to the nearest inch?, or do I need to stick with the exact mm as indicated in the blueprint?

Thanks,
Marc
#207
2nd January 2013
Old 2nd January 2013
  #207
Lives for gear
 
AwwDeOhh's Avatar
 
Joined: Apr 2009
Location: State of Insomnia, sleepless USA
Posts: 2,547

AwwDeOhh is offline
Quote:
Originally Posted by LSP View Post
I love the looks of this design. I wonder though, for aesthetic reasons, could one use different length boards ?

What I mean is, let's say the far left and right outer boards will be 4', then the next ones in on both left and right could be 2', then the next pair could be 3', etc.

I ask because these look similar to the ones Gavin Haverstick did for a fellow slut here, on his basement studio design, and can be seen on his you tube link.

At any rate, do you think varying the length of the boards would hinder, or alter the diffusion properties.
Yes, it would change, and possibly hinder. To what degree would depend on the design.

Keep in mind that the size of the overall panel needs to be bigger than the wavelength(s) you're trying to diffuse. If you're taking away surface area just because it might look cool, that's not the best plan - IMO.
Quote:
Also, would it alter the diffusion properties if I rounded the mm measurements up or down to the nearest inch?, or do I need to stick with the exact mm as indicated in the blueprint?
Yes it would alter it.
Whether that is a good or bad thing depends on what you're trying to achieve with your diffusor(s) in your room.
Generally:
Smaller width slats have a higher High Frequency cutoff
Wider slats have a lower high frequency cutoff.
So, by using wider panels, you'll lower (in frequency) the upper range the diffusor will diffuse. This is just general stuff though. Minor changes to the design won't have a huge negative impact, assuming you're keeping all the ratios the same. You could possibly improve it.. but you should know why you're changing things. "looks" isn't a great reason to change design. Altering the plan to take into account the size of common wood or material stock in your area might be a better reason.
-------------------------
{EDIT}:
Though the stepped diffusor is a different beast than the QRD, there are some similarities. You may want to DL the free "QRDude" software even if you don't want to build a QRD, just to see what altering certain parameters of the design do the the response. (it's also useful for making QRDs)
http://www.subwoofer-builder.com/qrdude.htm

Last edited by AwwDeOhh; 2nd January 2013 at 04:24 AM.. Reason: adding information
Arqen
Thread Starter
#208
3rd January 2013
Old 3rd January 2013
  #208
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by LSP View Post
I love the looks of this design. I wonder though, for aesthetic reasons, could one use different length boards ?

What I mean is, let's say the far left and right outer boards will be 4', then the next ones in on both left and right could be 2', then the next pair could be 3', etc.

I ask because these look similar to the ones Gavin Haverstick did for a fellow slut here, on his basement studio design, and can be seen on his you tube link.

At any rate, do you think varying the length of the boards would hinder, or alter the diffusion
properties.

Also, would it alter the diffusion properties if I rounded the mm measurements up or down to the nearest inch?, or do I need to stick with the exact mm as indicated in the blueprint?

Thanks,
Marc
Hi Marc,

+1 to the things AwwDeOhh mentioned above.

You can round the dimensions, but depending how much you round it can have a significant effect on the performance. If you want to do some rounding but you're unable keep the proportions the same, here are some things to consider:
  • In general, I've found that increasing the total depth of the diffuser while keeping the well widths the same is usually better than increasing the widths while keeping the depths the same.
  • Likewise, Nolan's (user name unqlenol) findings suggest that decreasing the well widths while keeping the depths the same can yield excellent performance, but I expect that decreasing the depths while keeping the widths the same would tend to cause the performance to suffer.

So in summary, if you need to alter the depth:width proportions, you're most likely better off making them proportionally deeper than proportionally wider.

This makes sense, when you consider that if you were to make it 1000 times wider and show it to a giant, it would look like a flat wall (i.e. a reflector, not a diffuser). And if you were to make it 1000 deeper and show it to a giant, it would look like a wall of spikes... the ultimate sound shredder (see my video below).


LSP
#209
3rd January 2013
Old 3rd January 2013
  #209
LSP
Gear addict
 
Joined: Feb 2009
Location: Inman, SC USA
Posts: 447

LSP is offline
Quote:
Originally Posted by Arqen View Post
Hi Marc,

+1 to the things AwwDeOhh mentioned above.

You can round the dimensions, but depending how much you round it can have a significant effect on the performance. If you want to do some rounding but you're unable keep the proportions the same, here are some things to consider:
  • In general, I've found that increasing the total depth of the diffuser while keeping the well widths the same is usually better than increasing the widths while keeping the depths the same.
  • Likewise, Nolan's (user name unqlenol) findings suggest that decreasing the well widths while keeping the depths the same can yield excellent performance, but I expect that decreasing the depths while keeping the widths the same would tend to cause the performance to suffer.

So in summary, if you need to alter the depth:width proportions, you're most likely better off making them proportionally deeper than proportionally wider.

This makes sense, when you consider that if you were to make it 1000 times wider and show it to a giant, it would look like a flat wall (i.e. a reflector, not a diffuser). And if you were to make it 1000 deeper and show it to a giant, it would look like a wall of spikes... the ultimate sound shredder (see my video below).


That all makes sense, and I definitely don't want to alter the design if it lessons it's efficiency. However,...and this would be admittedly just for aesthetics,..but what if I were to change the length of the boards only,...not the width or the depth.

I realize it probably hard for you to visualize what I'm talking about, but I'll try again. For example, one of your diffusers is 4' high, where all of the boards used are the same length from top to bottom, obviously.

Would it change the effect of the diffuser if for instance, I made the outer boards (far left and far right) say... 3' long, the ones beside those maybe 4' (also L & R sides), the ones to inside of those 4', etc. ? The end result would be a diffuser that wasn't exactly 4' high, because some of the boards used were different lengths. I will reiterate that in this example, the "depth" and the "width" are exactly as your blueprint calls for.

Again, this is purely because I've seen some designed like this, and I really did like the appearance. Not necessarily better, just different.

Thanks for taking the time,
Marc
Arqen
Thread Starter
#210
3rd January 2013
Old 3rd January 2013
  #210
Gear addict
 
Arqen's Avatar
 
Joined: Jan 2012
Location: Victoria, BC
Posts: 306

Thread Starter
Arqen is offline
Quote:
Originally Posted by LSP View Post
That all makes sense, and I definitely don't want to alter the design if it lessons it's efficiency. However,...and this would be admittedly just for aesthetics,..but what if I were to change the length of the boards only,...not the width or the depth.

I realize it probably hard for you to visualize what I'm talking about, but I'll try again. For example, one of your diffusers is 4' high, where all of the boards used are the same length from top to bottom, obviously.

Would it change the effect of the diffuser if for instance, I made the outer boards (far left and far right) say... 3' long, the ones beside those maybe 4' (also L & R sides), the ones to inside of those 4', etc. ? The end result would be a diffuser that wasn't exactly 4' high, because some of the boards used were different lengths. I will reiterate that in this example, the "depth" and the "width" are exactly as your blueprint calls for.

Again, this is purely because I've seen some designed like this, and I really did like the appearance. Not necessarily better, just different.

Thanks for taking the time,
Marc

Good question. You can do it, but as AwwDeOhh mentioned, it will likely effect the low frequency diffusion performance.

When the wavelength is very large compared to the panel size, little or no sound will be scattered.

If you reduce the height of the centre panel, certain low frequency waves may behave as if you've taken the centre panel out all together. And if these waves are blind to the centre panel, they will tend to wrap around the diffuser (diffraction) instead of scattering back. However, if the diffuser is mounted on a reflective wall, the wall will still reflect the sound back.

These diffusers start working around 500 Hz, which has a wavelength of 0.7 m, or about 2' - 3". So if the panel is only 2' high, it is shorter than the wavelength of a 500 Hz wave, and you can expect the performance at 500 Hz to be reduced.

You can use the pattern you mentioned, but if your low frequency diffusion performance is critical you might want to avoid it.

Do you by any chance have a link to the video of the diffusers done by Gavin Haverstick?

Hope this helps,
Tim
New Reply Submit Thread to Facebook Facebook  Submit Thread to Twitter Twitter  Submit Thread to LinkedIn LinkedIn  Submit Thread to Google+ Google+ 
 
Topic:
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Similar Threads
Thread
Thread Starter / Forum
Replies
Ethan Winer / Bass traps, acoustic panels, foam etc
86
Riddler / Bass traps, acoustic panels, foam etc
18
Soundproof / Post Production forum!
6
DaleNixon / Low End Theory
14
nuendoness / Low End Theory
15

Forum Jump

SEO by vBSEO ©2011, Crawlability, Inc.