Diffraction

Status
Not open for further replies.
The article states that even horns have diffraction problems at their edges and benefit from acoustical absorption material.

The thing is that putting absorption on the baffle is not going to affect the horns mouth diffraction to any appreciable extent because the diffraction occurs before the sound wave even sees the absorption. The only thing that can help in this situation is a large radius of the mouth into the baffle, and then absorption doesn't do much because there much sound travling along the baffle.
 
The thing is that putting absorption on the baffle is not going to affect the horns mouth diffraction to any appreciable extent because the diffraction occurs before the sound wave even sees the absorption. The only thing that can help in this situation is a large radius of the mouth into the baffle, and then absorption doesn't do much because there much sound travling along the baffle.

I think the author specifically said that putting absorption material around the mouth of a horn will not help with diffraction caused inside the horn. I believe he was addressing only the diffraction caused at the edge of the mouth of the horn. He does show by measurements that absorption material at the edge of the mouth does have an effect on the frequency response of the horn.
 
More on absorption to reduce diffraction:
"A form of audible and measurable distortion arises when the sound energy radiated by a loudspeaker driver is diffracted from an external edge of a loudspeaker enclosure.

The level of this distortion becomes particularly serious when an offending edge is at a distance from a driver which exceeds about one quarter of a wavelength at any frequency within the intended range of the driver.

Such diffraction causes an audible smearing of musical transients, a degradation in definition and ambience, and generally alters the amplitude versus frequency properties of the system, either on-axis or off-axis.

It explains why many designers choose to arrange drivers in an asymmetrical grouping along the front panel of the loudspeaker, mistakenly believing that they are counteracting the effects of diffraction. However, while the scheme may be used to remove some of the bumps and drop-outs in the curve of amplitude versus frequency, it does not materially improve the ability of the system to accurately reproduce the fine detail of many musical passages. Perhaps even worse, such loudspeakers usually exhibit a grossly asymmetrical radiation pattern, one of the chief contributors to poor stereo imaging.

An obvious solution to the problem would be the elimination of sharp edges by rounding them in a manner reminiscent of early 1930 designs. A little rounding is not sufficient. Minimum radii of several inches are necessary to alleviate the problem.

A better and more cosmetically acceptable solution is to apply an effective sound absorbing material between the drivers and the relevant edges. Properly executed, the sound waves reaching the edges will be attenuated to a level resulting in inaudible and virtually immeasurable distortion, as in Figure 4.

Duntech has solved the problem by covering all relevant surfaces with a highly efficient, sound absorbing, natural felt.

The density, thickness and shape of the felt is matched to the specific requirements of the loudspeaker to reduce the diffraction to an insignificant level."
TECHNOLOGY
 

Attachments

  • Diffraction absorption.gif
    Diffraction absorption.gif
    92.5 KB · Views: 287
But do any of the current diffraction simulators model this type of driver properly ? Not that I'm aware.

As I noted before, the diffraction simulation code that I wrote is part of a much larger program that calculates box response, amplifier and EQ response, and that reads in loudspeaker measurements and allows interactive crossover design. It has a driver database and even has a component database, so that you can select vendor part numbers (such as inductors or capacitors) from the database and create a bill of materials. My original plan was to spin off the box and diffraction parts of the program as open source software. So if someone wanted to model some unique topology, they could take this code and modify it however they needed to. I'm not interested in spending more time on this model, so if I made it open source I could fold back in changes from other contributors.

However, as the design got more and more complicated I started to rethink that plan. The model generates 40 points around the baffle perimeter and points at 1/2" spacing inside the driver. So for a large driver you can easily have 50 points inside the driver. It calculates both real and imaginary components of each possible ray summed at given point. I'm using 500 frequency sample points, so every time the driver is moved or the baffle changes the code recalculates 40*50*2*500 rays, or about 2 million of them. when the baffle roundover is calculated, the number increases by a factor of 4 to 8 million. The obvious problem I ran into was that the program was "sluggish"--move a line and then you had to wait until the rays got recalculated. That sluggishness was unacceptable to me, so I wrote the code to be multithreaded and compute the rays as an interruptable background process. Now when the drawing is changed the code spawns 4 separate threads that run in the background, using 4 CPU's if you have them. That fixed the response problem, but it added a lot of complexity to the code.

I changed my plan to make this an open source project simply because the code got way too hard to follow and I didn't want the burden of helping people understand it. However, if you have a real interest in changing the model and are willing to do it without any support, I will send you the source code. It can be developed in the Microsoft Visual Basic Express development tool, which is free.

No, I think it's you that has it backwards 😉 The difference between full space and half space with an infinite baffle is exactly twice the sound pressure level, eg 6.02dB.

I'll let you figure this out. Hint: "infinite baffle" (used in this post) is not the same as a "very large baffle" (the term used in your prior post and what I responded to).
 
I think the author specifically said that putting absorption material around the mouth of a horn will not help with diffraction caused inside the horn. I believe he was addressing only the diffraction caused at the edge of the mouth of the horn. He does show by measurements that absorption material at the edge of the mouth does have an effect on the frequency response of the horn.

I am talking about the edge of the mouth as well. Yes, if you do not radius the mouth then absorption on the baffle will help because it will reduce the secodary diffraction from the box edges. But if you radius the mouth, this diffraction reduction will make the absorption almost moot because there will now not be any diffraction in the plane of the baffle where the material is. His test proves what I say for his situation, but it does not indicate anything about what happens when you radius the mouth. The mouth radius is the more effective solution of the two approaches.
 
I changed my plan to make this an open source project simply because the code got way too hard to follow and I didn't want the burden of helping people understand it. However, if you have a real interest in changing the model and are willing to do it without any support, I will send you the source code. It can be developed in the Microsoft Visual Basic Express development tool, which is free.

The code is in VB? Is it standard VB which can be upgraded to VB.Net? If so then I'd be interested in it. (I wouldn't want to go back a decade to VB6 or anything like that since VB.Net is so much more fnctional.) I would likely recode the number crunching in FORTRAN, which basically nothing beats for speed.
 
I am talking about the edge of the mouth as well. Yes, if you do not radius the mouth then absorption on the baffle will help because it will reduce the secodary diffraction from the box edges. But if you radius the mouth, this diffraction reduction will make the absorption almost moot because there will now not be any diffraction in the plane of the baffle where the material is. His test proves what I say for his situation, but it does not indicate anything about what happens when you radius the mouth. The mouth radius is the more effective solution of the two approaches.

I think that unless we make our own horns, then we are stuck with a flat flange around the mouth of the horn and no radius to speak of. I have not checked every available horn however (edit: but some do seem to have a radius now).... Thanks for your insight.
 
Last edited:
The code is in VB? Is it standard VB which can be upgraded to VB.Net? If so then I'd be interested in it. (I wouldn't want to go back a decade to VB6 or anything like that since VB.Net is so much more fnctional.) I would likely recode the number crunching in FORTRAN, which basically nothing beats for speed.

It is VB.NET. Using FORTRAN or Visual C may help a bit, because VB.NET does not have a complex number type. I had to write my own complex number class 🙁 and it may not be an optimal implementation.

I'm still making changes in this module but I may reconsider the plan to make it open source once I have time to do some additional clean-up. Finishing this code is on the list of stuff to do when I retire at the end of the year.
 
It is VB.NET. Using FORTRAN or Visual C may help a bit

In my experience it is more than a bit, at least with FORTRAN. Modern FORTRAN is a marvel when it comes to crunching numbers, which your code does as you mention. It does matrix multiplications as single calls, which are coded in assembly and can be complex, even quad precision. So if you set up the code properly as a matrix its excedingly fast. You can then also compile the code as parallel and it will parse out the work to different processors if they exist.
 
I am talking about the edge of the mouth as well. Yes, if you do not radius the mouth then absorption on the baffle will help because it will reduce the secodary diffraction from the box edges. But if you radius the mouth, this diffraction reduction will make the absorption almost moot because there will now not be any diffraction in the plane of the baffle where the material is. His test proves what I say for his situation, but it does not indicate anything about what happens when you radius the mouth. The mouth radius is the more effective solution of the two approaches.

Almost moot.. we deal in almost moots around here all the time! Seems to me that for mid/tweet horns, foam outside a roundover could help with artifacts below the passband. Almost moot, given sufficient high-pass filter slope and sufficiently designed horn? Probably. But 1-2 dB of ripple, 20dB down, still seems worth improving if you don't have to give up anything to get it.
 
Almost moot.. we deal in almost moots around here all the time! Seems to me that for mid/tweet horns, foam outside a roundover could help with artifacts below the passband. Almost moot, given sufficient high-pass filter slope and sufficiently designed horn? Probably. But 1-2 dB of ripple, 20dB down, still seems worth improving if you don't have to give up anything to get it.

I think that most acoustical foam sheet is most effective in the range of 1KHz and upward, with a NRC of 0.8 to 0.9+ in that region, depending on thickness.
 
I think that most acoustical foam sheet is most effective in the range of 1KHz and upward, with a NRC of 0.8 to 0.9+ in that region, depending on thickness.

For grazing incidence I think that any standard measure of absorption would be highly unreliable. Grazing incidence on sound absorbing materials is a unique case that is hard to predict from the usual random incidence or normal incidence values. The action is quite different.
 
For grazing incidence I think that any standard measure of absorption would be highly unreliable. Grazing incidence on sound absorbing materials is a unique case that is hard to predict from the usual random incidence or normal incidence values. The action is quite different.

OK. Do you have any data or articles to back that up? The article that I have cited says it works to a measurable extent.
 
OK. Do you have any data or articles to back that up? The article that I have cited says it works to a measurable extent.

Its fairly fundamental in the physics of absorption that it should go to zero for grazing incidence since then the particle velocity of the wavefront normal to the surface is zero, hence no energy can be removed from the wavefront. This doesn't actually happen as it turns out, because instead a different effect is now occuring wherby the wave is actually difusing into the boundary rather than impinging on it. I could, of course, give you a whole slew of references on this, but it is actually quite well know.
 
My own measures were quite conclusive with a foam collar on the "econowave" horn. Of course, as the good doc said, that's fixing a bad termination, not "enhancing" a well rounded-over horn.

We don't always have a choice about buying a horn with well rounded edges at the mouth, so I suppose I and others are left with other solutions, like foam. Can I assume that unless this "well rounded" horn is mounted on a baffle that also has "well rounded" edges, diffraction will still be a problem? In other words, Dr. Gedes approach of using both, as in his loudspeakers, is a requirement? (can't use the foam plug, it's patented)

Of course, I did not know about this effect before I purchased my tweeter horns, but that's ok. I can't know everything before making a commitment.
 
Is it worth creating a detailed 3D model ? Probably not. It would be an extreme amount of extra complexity for a modest improvement in accuracy. As I said, typical baffle step simulators are useful for what they are - an estimation that gives a good idea of where peaks and dips are likely to occur, but an estimation that must be followed up with measurements, especially in the case of more exotic drivers, where the simplified model really isn't that accurate.
While true, the existing simulators are helpful. The BDS, linked earlier, was possibly the most advanced software available to DIY for years. Along with Jeff's newer software, one can examine any axis and get a reasonable simulation. They do use generic driver geometry and while it does not model it in 3D, it is sufficient since, as you point out, measurements are needed afterwards. But since anyone doing any serious design is going to measure, it's a bit of a moot point.

Fortuitously, baffle diffraction simulators almost always make things look worse than the really are, and when they are in error the measured response is usually better than predicted, not worse. 🙂 Understanding the limitations of simulations is key to making good use of them and not ascribing more accuracy to them than what they really have.
Precisely.

I would never go through the effort to build a baffle and driver cutout with recess and rear chamfer before modeling the basic baffle response. My experience with the BDS and other diffraction modeling software is that it is reliable enough for most work. I would like to see software that could reasonably model nearby driver diffraction, I'm not aware of any doing this, but then I am one who will not construct a system without some amount of diffraction control for this, even with my current dipole system.

The problem with any diffraction of this sort, baffle and driver, is that even were one able to model to high degree of accuracy, one must still decide on how to handle it. If no diffraction control is used, in this case requiring felt or similar between drivers, what do you do? It's a global issue. Treat the source and it's largely controllable. Short of that, you get into the problem of polar response. What good is a model of inter-driver diffraction if no diffraction control is to be used? Conversely, what good is a model when the inter-driver diffraction can be largely controlled? The latter works on a polar basis. You build, add felt, measure and you're done.

I will throw in my support for offsetting drivers. It's not the imaging issue some seem to think. It does not eliminate diffraction, of course, but it does aid in reducing it's impact if one designs the baffle and layout with care and intention. What is occurring is simply re-distributing its effects in time and thus also the frequency domains. The results can at times be such that across a reasonably window in the polar response, the magnitude is reduced. If in addition you use diffraction control such as reasonable roundovers or better (IMO), sufficient felt, the fact that the magnitude is reduced makes the combination of offset and felt extremely effective.

The problem is, most people are not interested enough in the absolute response. Aesthetics take priority for the majority, I think.

Dave
 
Status
Not open for further replies.