Edge Diffraction: Real World Examples

[Flint]

I recently completed a set of Lil Joker satellite speakers which are based around a Dayton Audio Coaxial speaker. They sound pretty good, but while making gated measurements (pseudo-anechoic) I couldn't help but notice some prominent comb filtering caused by baffle edge diffraction and more.

Here's a photo of the face of the baffle with dimensions written over it:


As you can see, the driver is positioned on the center width, but offset on the height dimension. There is also a small flare on the tweeter which extends from the edge of the dome's voice coil by about 1/4" all around. The edges of the baffle are rounded over with a 1/4" router bit to give the look a softer edge and to reduce the impact of edge diffraction.

Still, below a certain frequency there will be edge diffraction as any sound which travels along the face of the baffle (90 degree, or perpendicular, to the driver) will be create diffraction with it encounters the edge.

Using real world measurements, I am able to show the impact of this effect in real life.

I took a frequency response measurement about 19" from the speaker directly on axis then at 7.5 degree off axis until I reached 90 degrees, or directly to the side. All of those charts are shown below from 500Hz to 20,000Hz.



I normalized all of the responses to the directly forward, 0 degree, on axis measurement so each subsequent measurement shows how the response changes from the direct response. As the mic moves to the side, the distance from one side edge gets shorter and the distance from the other side edge gets farther away. This demonstrates how the edge diffraction changes the comb-filtering frequencies as the mic moves off center.

Below about 1,600Hz the diffraction is at it's maximum as the rounding over has no impact on the diffraction. However, in the octave above that. about 1,600Hz to about 3,200Hz, the amplitude of the diffracted audio gets softer from the rounding over of the edge until it completely goes away above 3,200Hz and the response is smooth. This is also the range where the directionality of the tweeter's output is narrowed because of midrange cone acting as a wave guide to it. However, above about 11,000Hz you can start to see diffraction artifacts again from the little wave guide flare on the tweeter itself inside the woofer.

This effect can also be demonstrated with the Polar Plot of the same frequency response data in the above chart:

The most red areas are the loudest and the blue to black areas are where the least about of sound is arriving at that angle and frequency.

You can clearly see the clear bands of energy below about 1,600Hz in highly separated red vertical bands. Then above 1,600Hz those bands get more blended and less defined until at about 3,200Hz where the orange/red area is smooth with not clear delineation. However, starting just above 10,000Hz you can again see some definition of banding.

This is all what happens when you have edge diffraction on a speaker, and in this case I made at least some effort to reduce it.



What does this mean to the sound?

The impact is rarely an audio frequency response issue, but instead all that comb-filtering is caused by delayed acoustic reflections off the edge of the baffle which arrive at your ears milliseconds after the original sound. Your brain has no ability to interpret such closely following reflections as "echo" or "ambience" and instead assumes all of that energy is the source signal. That makes the signal more "smeared" or lacking some detail in a way which is very hard to put your finger on.

Many often talk about the incredible resolution they hear from good headphones, and this is one reason for that. With very good headphones there are no early reflections like this to muddy up the sound, so you get a more accurate acoustic representation of what is on the source signal than when a speaker introduces diffraction.

Add other early reflections in an untreated room, and you get even more cluttering up of the original signal to make it less refined, clear, resolved or detailed.

 

[Flint]

In another thread I posted the response curve of my cheapo Sony 2.5" full-range drivers in a little cabinet I built from scraps. I foolishly forgot about diffraction when solving for a different problem and put a perfectly centered ring around the speaker to move it forward in the cabinet. Here's a photo:



In this case, the audible impact was both the muddied up clarity I wrote about above, but also a huge change in the frequency response.

To demonstrate the impact on the frequency response, I made a measurement of the speaker as you see it above, and then placed some wood and foam around the edges of the ring to reduce the possibility of diffraction occurring. Here's a photo of what I did:



The results were these two frequency response curves:



The black curve is the baffle as I built it with the perfect ring edge around the driver and the red curve is from the same driver but with the added foam and wood to vastly reduce the edge diffraction.

Notice the massive difference here. Because the ring is perfectly round and thus the edge of the baffle occurs at exactly the same distance from the driver all around, then all of the diffraction is impacting the same frequencies across the board. There is a huge null at 4,800Hz and some cluttering increase in output below that. Clearly this was a bad choice for me to make and really these speakers could sound better if I'd done something better when designing the baffle. Someday I may work on a do-over, just because I hate shoddy design.

 

[Flint]

One more chart... this time for a high quality mainstream slender German-made tower loudspeaker with little to no effort made to address edge diffraction. The metal done tweeter clearly has a strong off-axis output, as does the metal cone midrange.

In this case, the mic was about 0.75 meter away from the speaker, on axis, at the same height as the tweeter. The gating was on to eliminate room reflections down to about 800Hz - we will ignore everything below that.



You can clearly see the comb-filtering across the tweeter's operational range, but there is less in the midrange (though it is still present). This is audible, more because of the shimmer and sheen it gives to the sound caused by the increase in high frequency energy reaching you ears, but also from the distortions which result from combining so many reflections.

 

[Flint]

Holy crap!!! So, I've been putting a ton of thought and energy into trying to understand value and control diffraction, and as I've mentioned a million times on this forum, when it comes to the problem of diffraction, the worst of all possible worlds would be:

• A round baffle with the speaker placed in the center of the disk-like baffle
• Sharp edges on the baffle which create the strongest and loudest diffraction signals

Well, here is a "high end" speaker which incorporates the worst of both of these.



I mean, literally, you could not make a speaker with more terrible issues with diffraction than this. And, I am willing to bet all I have that it sounds accordingly bad.

 

[Dentman]

Holy crap!!! So, I've been putting a ton of thought and energy into trying to understand value and control diffraction, and as I've mentioned a million times on this forum, when it comes to the problem of diffraction, the worst of all possible worlds would be:

• A round baffle with the speaker placed in the center of the disk-like baffle
• Sharp edges on the baffle which create the strongest and loudest diffraction signals

Well, here is a "high end" speaker which incorporates the worst of both of these.

View attachment 8899

I mean, literally, you could not make a speaker with more terrible issues with diffraction than this. And, I am willing to bet all I have that it sounds accordingly bad.

Yea but the $2,000.00 speaker cables and wire lifts help.
Amazing!

 

[Randy]

Yea but the $2,000.00 speaker cables and wire lifts help.
Amazing!

And, it is very Avant Garde, and I don't mean the speaker company. My guess is they are pretty pricey.

 

[Flint]

I am working on a private product for a company which utilizes a 6" mid/woofer in an enclosure which will slide into a much larger cabinet. Since its intended final operating location is in a cavity, I cannot really fix diffraction issues with rounded over edges and huge & odd shaped baffles. So, I made a test box with a 8.5 x 14.5 inch baffle with the mid/woofer placed offset on the long length and centered on the short length.


Photo: 6" mid/woofer in enclosure being measured outdoors in an anechoic environment

When measuring in an anechoic environment, you can clearly see the diffraction in from the hard edges and narrow baffle in the response:


Photo: Anechoic (outdoor) response of mid/woofer on 8.5 x 14.5 inch baffle measured with the mic at 0, 15, 30, and 45 degrees

As you can see, below about 1.6kHz there is a clear comb filter effect to the response. This is because the output from the 6" driver is omnidirectional off the cone below about 1kHz. Also of note, the extreme fluctuation in the ripple of as much as 5dB SPL below 450Hz. That is huge and would be very audible when listening to music.

A note with a primary pitch of 210Hz will be 5dB louder than a note with a primary pitch of 280Hz. Now, most musical instruments make a sound with a huge amount of complex harmonics won't be so obviously louder or softer at different notes, but the timber of the instrument will change depending on which note the musician was playing. I am sure you have all experienced this effect, especially with bass guitars, where one note will seem to be resonate and deep and another note from the same instrument will seem to fade into the music and be difficult to isolate in your mind.

This is another example of the acoustics of the baffle and the resulting edge diffraction causing issues in the performance of a speaker.

For this design, however, the speaker will be in a much larger cabinet where the effective size of the baffle will be at least 4 ft. wide and 2.3 feet tall. So, the impact of this issue as measured will be vastly reduced, if not completely eliminated.

 

[TKoP]

So what are you testing? volume enclosure? Shape?

But I like the information...

 

[Flint]

So what are you testing? volume enclosure? Shape?

But I like the information...

I am measuring the performance of the mid-woofer and tweeter in order to design a crossover which will provide the performance we are after. Also, since these are prototypes, I am confirming that the enclosure parameters create the loading we need to get the results we desire. Both of these things can only be accomplished effectively in an anechoic or pseudo-anechoic environment where the performance is not impacted by acoustic reflections.

 

[Flint]

So, as a comparison I created a chart showing the anechoic frequency response of two different speakers placed on different baffles.

One is a 6" mid-woofer centered on a baffle on one dimension and offset at the 2/5ths point (ideal) on the longer dimension. This speaker has significantly stronger edge diffraction induced comb-filter affects due to it being centered on the baffle on one dimension.

The other is a small 3" driver which is offset at the 2/5ths point on the long dimension and the 3/7ths point on the shorter dimension. This creates less impact from comb-filtering created by edge diffraction.



The red curve is the smaller driver on the more ideal baffle while the black response curve is from the less ideally placed 6" woofer.

As you can see, the edge diffraction starts impacting the output at about 1,200Hz for the less ideal setup and has more extreme peaks and valleys in the midrange - especially between 400 - 1,000Hz. The smaller driver placed as ideally as possible for the small baffle it is on doesn't exhibit any diffraction until about 820Hz and even at its worst isn't as extreme as the 6" speaker.

This isn't about the speaker's size, but the placement of the speaker on the baffle. In fact, the smaller speaker is omnidirectional clear up to nearly 4kHz, so the presence of identically match edge diffraction frequencies would be visible in the response up to that frequency, while my the placement is such that edge diffraction doesn't show up above 900Hz.

Meanwhile, the 6" woofer is omni to a lower frequency, like 2.2kHz, yet it exhibits the negative impact of edge diffraction as high as 1,200Hz.

That's significant in this case because our ability to hear these things is diminished as you move away from our most sensitive listening frequency range of about 2,000 - 4,000Hz.

 

[Flint]

I am a member of a few online speaker builder groups and today someone posted a speaker project in progress:



Clearly this guy has very impressive wood working skills and this speaker is going to look very attractive. Those things are impressive.

However, can anyone spot what I am going write about that speaker baffle?

That's right, the design, while sexy, is going to create terrible diffraction issues. The rounded baffle at the top and bottom in which the mid-woofer is perfectly centered creates nearly infinite number of edge diffraction points which are going to impact the exact same frequency range, which means the comb filtering will be incredibly audible. If the crossover is well below the frequency of the comb-filtering, then it may not be terrible, but if I assume this is a 6" woofer and he is smart enough to set a crossover point where the dispersion is starting to beam (which is usually most ideal), then I believe he will experience very extreme comb-filtering from that edge which will be audible.

Comb-filtering is easiest to show on a chart and this easiest to understand, but the real impact of a slightly delayed out of phase reflection, which is what edge diffraction is, will smear the midrange performance and make the audio less detailed and less revealing. This slightly delayed reflection makes all the difference when considering why some speakers are incredibly revealing versus those with a nice balance but somehow lacking detail.

The tweeter is also dead center on that baffle and as such will experience significant, though not as significant diffraction which will also muddy the sound.

Now, this speaker could very well have a nice, balanced sound and good dynamics if he gets everything else correct. But its ability to resolve detail and reveal the fine content in the material will be limited.

If this design had the drivers offset, the round top and bottom not perfectly centered on the midrange/woofer and perhaps varying edge shape from one side to the other (and built as a mirrored stereo pair), this issue would be massive reduced.

 

[Flint]

Since I have been speaking so much about edge diffraction (mostly to explain my less-common baffle designs and use of offset driver placement on those baffles), I was sent this video which was supposed to contradict what I am espousing.

In the description of what edge diffraction is, the presenter is pretty close to spot-on given the short summary he is using. So, good on him.

In the summary of the results, he is completely wrong - 100% wrong!

So, he makes two "baffles" on which he mounts identical tweeters then measures their polar plots within a 45 degree window. Right off the batt it is obvious he is not comparing two real world examples given the size of what he calls a traditional baffle. In real life, I don't think you'd see a baffle that small since there is clearly no room for a mid-woofer, much less the area to make up the front part of a decent enclosure. So, where would one see that baffle??? Also, the baffle is perfectly square, which inherently results in baffle reinforcement very specific frequencies.

The "Pod" however, is interesting. Yes, this is similar to a round baffle, like you see on the high end B&W speakers and is proven to be beneficial. Yes, it also has the tweeter sunken into the baffle to form a small wave-guide like you'd see from JBL, Paradigm, and others. And, yes, it is rounded over which dramatically reduces the energy levels of any diffraction wave which may occur. However, it is also perfectly round and the round-over is minimal which increases the scale of the diffractions.

So, let's look at his data which I grabbed as screen-shots.

Here's the polar plot of the tweeter on the "traditional baffle":


As you can see, the off-axis output drops significantly above 4.3kHz, which I would argue is a product of the size of the baffle, not edge diffraction. Also, I cannot tell what offset location he chose - if he chose to offset the tweeter by 1/3 or 1/4, this is exactly what one can expect. Furthermore, while there is some clear bands of comb-filtering (the obvious sign of the negative affects of edge diffraction), those are not constant and covering the entire output range of the tweeter. In fact, the only major comb-filtering anomalies are centered around 1.9kHz to 2.6kHz. There is plenty of comb-filtering present in the low output off-axis range above 8kHz, but it also is weak and is more likely due to the fact his test baffle had a sharp edge at the back of the chamfer. In real life the back of the chamfer would be married to the sides of the enclosure making that angle 45 degrees rather than 135 degrees and this baffle has.

So, here's the same polar plot measurement made for the tweeter pod:


It doesn't take me writing a few sentences for the comb-filtering to be blatantly obvious! Clearly there is very significant and impactful comb-filtering occurring across the entre operating range with extreme peaks and nulls. This is due to the tweeter being mounted dead center in a circle which is has a small-radius round-over. To me this is an extreme example of why edge diffraction is bad and much be dealt with.

If the larger problem with edge diffraction is NOT the frequency response comb-filtering effect, but rather the smearing of the acoustic waveform by adding a short delay with inverted polarity to the original tweeter output (which is what edge diffraction is), then anytime you see strong comb-filtering that means you have tons of edge diffraction which means the audio is more distorted through smearing from the delayed energy. This is, my friends, what separates a highly detailed and accurate sounding speaker from another which may have the same timber, output, dynamics, and dispersion but lacks the detail and resolution.

The presenter also shows the frequency response differences and suggests that edge diffraction is playing a role in why his Pod design has a flatter overall response to the traditional baffle, but that isn't the case. The reason there is such a difference is because of reinforcement from the baffles is different.

I think one day I will make a similar test comparing various baffle designs and driver placements like this one, but I will focus on how to get more clarity out of a speaker, not how to get a flatter frequency response from a tweeter.


I think this was intended as a rhetorical test. It is clear he is proud of his CAD designed and probably CNC machined "pod" and the horns he makes. They look really nice and I am very impressed with his production quality. I imagine a small horn super-tweeter made like that sounds pretty nice, which is what he said he typically makes from that general shaped pod.

But, for a direct radiator dome-style tweeter, his beliefs and tests are actually in opposition from his conclusions.

 

[Flint]

By the way, he does another video on edge diffraction which is MUCH better at discussing the issue.

He discusses the nature of edge diffraction relatively accurately, even getting into the issue I refer to as "smearing". However, his conclusions are, once again, false or confusing. Some of his diagrams are inaccurate, such as when he overlays the "diffraction" signal at 4.5kHz - there is no way the amplitude would be the same as it would be way quieter (like 10dB lower) and it should have been phase reversed. But, his point was made. Also, I am 100% convinced that there is no way on this planet that there was any edge diffraction at 18kHz. The output from that full-range speaker at 18kHz cannot possible propagate directly to the side of the driver. The difference has got to be one of test setup, even a very minor difference in the distance of the mic to the point-source of the driver could cause that.

He says the results of edge diffraction on dynamics are:

• "The quietest parts of the music are masked."
  (I really don't like his terminology, but I completely agree with this)
• "Transient impact (Slam) s reduced because of cancellation effects"
  (I would never use "Slam" to describe it because the edge diffraction doesn't take place at frequencies we associate with "Slam", but the concept of how it impact midrange and treble is correct)
• "Music sounds flat and veiled with no driving rhythm."
  (I agree the music sound flat and veiled, but what the hell does he mean by driving rhythm. Rhythm occurs in beats per minute where edge diffraction is occurring in milliseconds. I don't we he associate millisecond timing as a "Rhythm".)
• "Will not sound convincingly live or real."
  (100% exactly correct!!!! Whoop!!!)

This guy is smart and I think he really understands much of the science behind this stuff. I fear he cannot translate the science into how music actually sounds. He may hear the improvements of removing or reducing ne effects of edge diffraction, but they way he describes it is misleading his viewers.

Still, I love it that at least a few people are discussing it.

I prefer my solution for eliminating the ill-effects of edge diffraction over his, but his approach is well-accepted in the industry.

 

[Flint]

I have some data of another example of significant edge diffraction which will hurt the sound quality of the speaker. This is a DIY speaker with some very costly raw drivers.



As you can see, the drivers are centered on the baffle and the round-over on the edge is so small it is mostly cosmetic. This is an amazing VOLT dome midrange - perhaps one of the finest midrange domes on the planet.


These are the measurements made by the owner of just the dome midrange at tweeter level both on axis and 30 degrees off axis.

It is clear there is a serious comb filter issue which starts at 800Hz, balanced out a little bit in the 3 - 6kHz range, then returns in the higher treble. There is nothing that can be done in the crossover to reduce this issue. The only truly effective solution is rounding over the baffle and offsetting the driver appropriately.

 

[TKoP]

Is there a fix for something like that save redoing the box?

 

[Flint]

Is there a fix for something like that save redoing the box?

Really heavy acoustic absorption material on the face of the baffle, something which absorbs more than 3dB of acoustic energy at frequencies as low as 800Hz and up to the highest treble (the ribbon tweeter will have a worse issue with in this design).

Another decent and more cosmetically acceptable solution would be to build a grill cloth frame designed to act as a wave guide, of sorts, which presses against the baffle as has sharp angles flaring outward at an angle greater than 45 degrees. That would work, but it would also have an effect on the frequency response that would need to be addressed in the crossover.

 

[Flint]

More data for the same speaker above, this time including the tweeter and woofer...



Once again, the diffraction issues start at about 800Hz, then get their worst in the 1kHz to 5kHz range, then go away in the treble.

Since humans are most capable of hearing the most detail in the range of 800 - 7,000Hz, this is the worst place to have issues like diffraction.

 

[Flint]

Whilst doing some research today I came across a speaker builder who was demonstrating his new design. It uses very good drivers from reputable manufacturers and a cabinet with the current trend in edge diffraction reduction, chamfering (cutting a 45 degree angle off the edge of the baffle, typically at odd angles to make the chamfer narrow at one edge and huge at another).



As you can see, the tweeter and mid-woofer are still dead center on the baffle.

Here is the finished, unfiltered response of the speaker:


For a home designed speaker, I am impressed with the flat response and extended bass. The dip at 110Hz is from a floor bounce, but otherwise the overall performance is very good.

That said, it is very clear to see that starting at just over 1,000Hz there is extremely strong comb filtering from the edge diffractions which remain prevalent in the tweeter output top end of the response. This is not insignificant. The swing in output is well over 5dB and the nulls deep enough to potentially make specific synthesizer notes (pure tone notes) to be relatively inaudible in the music.

I would love to see this speaker's unfiltered response with a felt or foam tweeter ring or placed on a larger baffle with extremely rounded edges.

 

[Flint]

I helped another speaker builder on crossover design and he recently posted the off-axis frequency response measurements.



I started helping him after the cabinet had been built and while I pointed out the inherent issues with diffraction, he wasn't keen to design and build new enclosures. But, you can see clearly that the tweeter (not the woofer) is the root source of the diffraction. This is because the woofer is directional, ie. it doesn't beam midrange frequencies to the side, in the crossover range.

The tweeter, however, does beam very well to the sides and as a result the cabinet is contributing acoustic output from diffraction starting around 2,100 Hz. As with most speakers, the comb filtering is significant and would be clearly audible if compared side by side with an otherwise identical speaker where edge diffraction was treated and removed.

None of this means the speaker won't sound good. It probably sounds very good! However, this problem is inhibiting the speaker sounding its best. How can one describe the difference? Well, if the edge diffraction issues were removed, the sound would be more detailed, less distorted and "smeared" and offer a more precise stereo image. The separation between instruments would be greater and the listener would get more of the sense that the speakers have disappeared and the music is suspended in front of them.