Speaker cabinets and edge diffraction

I've been trying to read up on diffraction problems and cabinet construction.

I'd be lying if I claimed to actually know the truth about any of it.
However it would be nice to have a few questions aired. :)

I'm planing on a two-way build with focus on constant directivity.

Geddes is a pioneer and authority in the field. He recommends as large a edge radius as practical.
Linkwitz who is a pioneer and authority in himself as well stipulates that rounding the edges will only provide benefits when the radius is greater than 1/8 of the wavelength.
In a document found at Pi speakers you can read about the baffelwidth being a major issue and edge rounding not so much when it comes to lower frequencies.

I'm not sure what to make if it?
A waveguide might for an example exhibit a 90degree radiating pattern at the crossover frequency.
The woofer will probably match the directivity fairly well if done right.

Since the waveguid has such a focused directivity diffraction problems in the higher frequencies should be a non-issue.
When dealing with lower frequencies the wavelengths are so great that the rounding won't do much good anyway.

I'd very much like to hear it from someone who actually understands this stuff.
Please explain it to me in simple terms.
Does edge rounding really do much difference if crossing around 1kHz using a HF driver + CD waveguide and a large DR woofer?

How big does the rounding have to be to make a difference?

I'm just confused from reading to much.
 
As fastbike1 pointed out, Edge can show what square edged baffles effects are @ higher frequencies.

My guess is that a trial/error type of approach combined with measurement will reveal how much and what kind of edge treatment is indicated for a particular driver/cabinet combination.
 
Last edited:
In very general terms, edge rounding really doesn't have much impact until at least ~40 mm radius / 1.5".

You can always download The Edge and see for yourself.

This is incorrect, though it's oft-repeated. Even a relatively small radius can make a significant difference, depending on the rest of the design details.

Considering the shortest ray to the baffle from the driver's edge, you consider the scenario in which the rounding is least effective. Now consider a ray with a more acute angle of incidence- the roundover "appears" to be much larger than it is in the perpendicular angle (shortest ray) case.

This has been borne out experimentally but a quick search fails me on whose site it was.

The nice part is that the solution is easy. Do the biggest roundover or chamfer you can.
 
My understanding is that AF wavefronts behave the same as RF wavefronts. Did you ever wonder just why & how radio reception gets lousy when your deep within a valley.....the station fades out? Wavefronts ...I would say "stall out" at the top of a hill/mountain ...but the edge where it "stalls" regenerates & starts 'transmitting' again at that edge. Let's say the station puts out One megawatt of power...it gets to that edge & 'stalls'......It 'stalls' at the edge of a mountain with only 3 watts of power.........picture a phantom antenna, having only 3 watts of power now transmitting down the back slope of the mountain.
Same thing with AF.....A 'retransmitting' of the signal originating at the baffle edge. What frequencies, multiples of fundamentals, or phase shifting....I don't know the particulars.............It is of course easy to put a big a** radius on the edges to null out these diffractions.......consider it a basic rule of thumb......a must do on any design.



_________________________________________________________Rick...............
 

Attachments

  • Wave.JPG
    Wave.JPG
    33.5 KB · Views: 1,213
I don't want to be a poor sport but that's precisely the sort of answer that I'm questioning with this thread.
Rules of thumb and "everyone is doing it so it must be good" doesn't quite cut it.
I'm not saying you're wrong but I would like to understand how it works and how big an impact it has.
(I'm a mathmatician btw, numbers and research references are gold.)

I believe it's been proven that frequency irregularities below 1dB is at best incredibly difficult to percieve and most would say it's inaudiable.
Room acoustics will have a huge impact on the sound. How will a sharp edge on the cabinet compare to a framed painting on the wall next to the speaker? Or a bowl of fruit on the table in front of the listening position?

The angle of incidence argument would be interesting to hear more about since most people seem to work in the time domain when discussing diffraction and not so much in wave propagation and frequencies?

Since the driver isn't an ideal point source I'm sure real world impact will not conform to the most basic idealistic models. From what I can remember the larger the driver the lesser the problem but I tend to mix things up.

If I wanted easy I would just keep my mouth shut and solve my problems like everybody else but this is my hobby and I like to actually understand what I'm doing. :)
So, please don't be mad at me for questioning some of your answers, it's nothing personal and I'm just trying to understand.
 
Some things are generally misunderstood.

Rounding edges does not lessen diffraction. It smears it over a large frequency range as the edge is now seen as gradually fading. This is similar to off-setting the driver from the exact center.

The diffraction peak frequency or apparent edge will be the halfway through the rounding.

Chamfering will have a similar effect. Although two having 2 peaks at each end of the chamfer. In many cases a chamfer will have a better result than a rounding especially if the chamfer is relatively large.

In many cases it's also worth considering that since the total diffraction is always the same as you cannot stop waves from transferring from 2Pi to 4Pi unless you have them built into a wall, then maybe you can use the diffraction constructively instead of trying to minimize the effect.

Many, in fact most, drivers are not linear both in frequency response and dispersion, especially nearing the cross-over frequency since that is the main reason for using multiple drivers to cover the audible spectrum in the first place. Diffraction can with careful calculation and optimization be used to correct driver frequency response and off-axis dispersion.

Note also that since rounding or chamfering edges increases dispersion there are some cases where any kind of rounding or chamfering will have a decidedly negative effect.
 
Last edited:
Very interesting and that's approaching some of my own thoughts.

If my crossover frequency is 900Hz, the wavelength will be 15".
As it happens, the driver is 15" as well and from what I understand a typical DR woofer will have a roughly 90 degree beam at the frequency corresponding to a wavelength matching it's diameter.

A compression driver in a waveguide with a 90 degree dispersion will match this pretty good. Now, the HF driver will have a steep roll-off by its nature so I don't imagine it causing much damage in the LF.
From measurements I've seen, drivers in waveguides are at least 6dB (or more) down at 90 degrees.
A common tool for looking at the baffle step is imagining a mirror driver out of phase radiating at half acoustic pressure causing a 6 dB drop.
Add these togeather we should be seeing something like a -2.5dB influence which is a worst case scenario.

Running a sim in the Edge with a realistic baffle the ripple will be less than 2dB.
In fact the baffle step will be very close to the frequency where you transition from a reverberant field to the modal region in a common livingroom making directionality a moot issue. If anything the first diffraction node is additive and it will increase the woofer's efficiency up to the crossover region while the first depression (not even causing a negative response) will coinside with the crossover slope.
It's still not a huge ripple and I can't help wondering if we're chasing ghosts?
A small notch-filter should take care of the initial hump.

Research seem to indicate that the threshold for audiability and frequency irregularities are in the 1dB area. (Hopefully I'm remebering this correctly.)
Drivers themselves have less than perfect frequency responce despite what the smoothed curves in the data sheets tells us.

Maybe I'm just being ignorant but I don't see what the big fuzz is about?
 
Also note that Edge is technically incorrect (unless updated since I last used it). It's the apparent acoustic center of each driver that is the source, not the center of the driver. For typical dome tweeters and smaller home audio drivers there is not much difference between the two even if exotic filtering is used.

For large cone drivers, horns and wave guides the difference could be quite substantial. Note that filtering in the form of phase and time delay can shift the apparent acoustic center as well.
 
I'm not sure exactly what you're refering to? You set the number of pointsources to represent the source manually and they are distrubuted over the drivers surface?
The larger the number of pointsources the better accuracy you get.

Yeah. But they're only in the horizontal plane, right? Apparent acoustic center will be some distance behind the horizontal plane. The exact center will vary will frequency as well.
 
Nice article wesayso, it deals with some of the stuff I've been talking about.
Some of the major wajor weapons against diffraction problems.
- Recessing the drivers
- Felt on the baffle
- Placement of the drivers
- Radiused edges
- Controlled directivity

I'll admit that the radiused edges appear to have considerably less ripple in the pictures shown. It's still hard to compare one method against another since all is trying to do the same thing but in different ways.

Foam and felt does indeed look to to a decent job. Looks a mess though. :/
 
You forgot the best and most obvious way to combat diffraction:

- to know what it does and incorporate it into the design.

It is in essence much the same as with cross over design. If you know your filter types and how they interact with real loudspeakers and not static resistors then you can start designing optimum filters that does not need impedance and phase corrections.