Is baffle shape really important? What about these ideas?

[mhardy6647]

View attachment 289665pure concrete

Pure concrete is an amusing turn of phrase -- since concrete is an admixture of cement (itself made of multiple components) and gravel with some sand (aggregate)

 

[Rick Sykora]

As tends to happen when the focus of a thread is around one aspect of speaker design, the importance of system design gets buried in my opinion. Baffle shape can be be very important, but there are designs that can make it less of an issue. That said, many other considerations are in play. As this is the topic of entire books, only going to touch on some other keys ones:

1. Driver dispersion characteristics
2. Cabinet size and shape including potential use of waveguides
3. Shock and vibration mitigations
4. Crossover design for multiway speakers
5. Overall SPL and power handling goals

The baffle shape may be driven by these as well as practical considerations such as fabrication complexity. While one hopes form follows function for the most part, some high end designs can be more about form. However, I think it is important to emphasize the baffle as part of the overall cabinet structure. For more considered designs, the cabinet design plays a major role in dealing with the stress and vibrations that often result from the movement required by drivers to produce the primary speaker output. As much of Amir’s measurements show, controlling various resonances is a key aspect of better speakers.

 

[JRS]

I think it originally came from Olson's Acoustical Engineering. But it may predate him.

Certainly where I saw it first--in his monograph. Used to have a copy I could have checked, but thinking (and its been decades since I read it) it was his work.

 

[JRS]

Great example of form following function. The Aussie speakers above are much too primitive for my taste but seem motivated by exactly what you ask for.

 

[JRS]

Honestly most speaker designers default to the rectangle shaped baffles because that's the easiest one to make, lets not fool ourselves. I see no reason not to make interesting smooth shaped baffles to fit the radiation pattern that you want.

Everything less than the ideal spherical shape is a compromise. KEF understands this well with their LS50 and Blade series.

I really wish the DIY community moves towards 3D printing. but that seems to be taking longer than I expected.

Right? By now I figured one could just submit the dimensions of the desired waveguide to some guy on the internet specializing in such and a few hundred $$ and you'd have it in your hands within 3 weeks. And that was a decade ago. When I last looked all I saw were a few waveguides available for a handful of some very popular tweeters. No custom fab at all.

 

[JPA]

I think it originally came from Olson's Acoustical Engineering. But it may predate him.

Yes, it is from Harry F. Olson's "Acoustical Engineering".
@sigbergaudio, there's no reason to doubt the graphs. Olson was twice as smart as anyone on this forum.

 

[kemmler3D]

No custom fab at all.

From what I can tell from Augerpro's thread on DIYAudio (prob the source you're talking about), designing a waveguide is pretty tricky, and you actually need to test the tweeter in the waveguide to make sure it worked properly. Sims only take you so far, I guess.

There are numerous services that will take the 3D model upload and send you the waveguide. But designing the 3D model for the waveguide is an acoustic engineering task, even if it's not at the bleeding edge of the field, not that many people know how to do it properly.

If you could just punch the tweeter specs and waveguide parameters into a simulator and get a good waveguide out of it, you might be able to get a custom one in the way you said, but from what I've seen out there, unfortunately it is not so simple with today's software.

 

[617]

A baffle can essentially be thought of as a waveguide, which boosts the output of shorter wavelengths while doing nothing to support frequencies with wavelengths larger than the baffle.

In between these frequencies there is a transition region which may be abrupt or smooth, having many ripples and off-axis irregularities or not. Rounded edges are marginally smoother but unless your round-overs are in the 50mm radius size I would say the effects are pretty modest. Genelec does a good job with their die cast enclosures.

In the earlier days of my speaker-design experience, people would use tools like Edge or various excel spreadsheets to simulate baffle step, which lead to many people favoring off-axis tweeters, since this creates a more linear direct measurement. Nowadays we can measure and simulate the baffle condition in 3D.

As others have pointed out, more directive transducers such as larger woofers and tweeters in recessed waveguides will tend to be less prone to baffle edge issues since the amplitude of the sound they produce is attenuated at 180 degrees.

To the question 'is diffraction important' the answer is, not really; it is contributes to the off-axis sound field which in turn contributes to the DI curve. Some people think diffraction is in effect an early reflection which impedes clarity and intelligbility, but if people really cared about that they would be doing a lot more than making speakers with curved edges.

@Duke I always felt Olsen was exagerating those diagrams to make a point by using a small transducer, which I think is fine, but it is much more revealing to look at a directivity sonogram, which would have been very difficult for Olsen to produce.

 

[sigbergaudio]

Yes, it is from Harry F. Olson's "Acoustical Engineering".
@sigbergaudio, there's no reason to doubt the graphs. Olson was twice as smart as anyone on this forum.

As always, context matters. So if the graphs are calculations based on a situation with a 1" driver on an 100cm baffle, it's not super applicable to the real world - at least not without that added explanation.

So as it was presented, it was easy for a layman to conclude that normal rectangular speakers would have pretty severe issues, which they typically do not. Hence actual reason to doubt the graphs in the context they were presented. That has nothing to do with how smart Olson was.

 

[tmuikku]

Yeah its all related to wavelength and physical dimensions of transducer and the object its attached to. Wavelength is physical size of sound and relates how it interacts with objects of various sizes. Its all very important, size and shape of the baffle and the structure in total, sizes and positions of the transducers and how they all relate as one unit and how it plays with the room so that you the listener gets maximum pleasure. That is what we listen to, pressure variation in the room and the speaker structure is responsible for it.

I suggest everyone who cannot currently imagine in their heads how it plays out to experiment with VituixCAD diffraction tool. Its real time and very much idealized and simplified showing the underlying phenomena quite clearly. Its easy to build intuition with it and one general rule of thumb that emerges from it is that flat baffle area needs to be minimized to minimize interference, and that any roundover helps to reduce it even further. Least interference when big roundover starts right at the edge of transducer, aka sphere. Minimal square baffle with no roundover works almost as good and is much easier to manufacture, it approximates a small sphere. One could also spread the edge as much as possible, narrow and tall baffle, which spreads the edge diffraction in time. One might make the baffle big enough so that edge diffraction is delayed more than early room reflections.

But, of course it all affects system design. Making minimal baffle woofer most likely requires another bigger box to get some bass, so its a multi-way speaker. Keep on going choosing compromises and you'll land on something. If it has to be a two way bookshelf speaker with woofer and tweeter on same baffle, just slant / cut all the flat area away and its best you can get with the constraint.

By "best" I mean simplified situation, where edge diffraction related sound source makes so little interference with direct sound that it is possible to get frequency response within +-1db, or even +-3db. If you have flat baffle box, or other physical structure that makes strong diffraction, like in 99% speakers out there, it is not possible to get performance within the bracket other than on one observation angle. Even if the speaker sounded fine it wouldn't sell well for cult of single number quality metric

The interference pattern from diffraction is just indication of relative magnitude and bandwidth of edge diffraction, which makes secondary sound source and whose fingerprint the interference is. Interference doesn't directly show how audible edge diffarction is, and I don't know how audible it is or how to listen to it, I don't know yet how to perceive to actually evaluate if its a problem or not. Minimizing amplitude and bandwidth of diffraction interference makes most sense to me, which also smoothens the frequency response to all directions, the classic metric for goodness.

 

[JRS]

From what I can tell from Augerpro's thread on DIYAudio (prob the source you're talking about), designing a waveguide is pretty tricky, and you actually need to test the tweeter in the waveguide to make sure it worked properly. Sims only take you so far, I guess.

There are numerous services that will take the 3D model upload and send you the waveguide. But designing the 3D model for the waveguide is an acoustic engineering task, even if it's not at the bleeding edge of the field, not that many people know how to do it properly.

If you could just punch the tweeter specs and waveguide parameters into a simulator and get a good waveguide out of it, you might be able to get a custom one in the way you said, but from what I've seen out there, unfortunately it is not so simple with today's software.

Interesting and thanks for the info. I was under the impression that the modelling was quite good, and that particular radii based on the transducer size could be counted on to provide the desired results (provided one could get the waveguide and transducer to physically mesh properly). It seems as if with damn near everything else in audio, we have a significant gap between theory and practice.

 

[617]

Interestingly, the advantages to an off-center tweeter or irregular baffle shape are similar to the advantages of using an elliptical voice coil or off-center voice coil. Drivers with these characteristics are esoteric, but ScanSpeak makes the former and Tang Band makes the latter.

If you trace a line from the center of the driver to the edge of the baffle, the length of that line basically dictates the frequency at which the 6db baffle-step drop-off occurs at that point. If you trace a bunch of these lines, say in 1 degree increments, and average them, you will have simulated the baffle edge diffraction for your measurement point.

If the baffle is circular and your driver is in the center, every measurement will have that 6db drop-off at the same frequency, which means that there won't be a bunch of different responses averaging out - you will have the crazy ripple you see in Olsen's diagram. If it's a square, with the driver in the center, it's a bit better, but you still have four sides which are equidistant, so not a ton of variation. If you make the baffle rectangular, with the driver in the middle, it's a bit better, and if you make it rectangular with the driver offset from the center both in the X and Y axis, you will have created a condition where a great many different path lengths are averaged together.

In the case of elliptical/off center drivers, the idea is similar. By making a variety of path lengths, in this case from the edge of the voice coil to the edge of the cone, you are able to dissipate various standing waves over a frequency band or eliminate them altogether.

 

[kemmler3D]

I was under the impression that the modelling was quite good, and that particular radii based on the transducer size could be counted on to provide the desired results (provided one could get the waveguide and transducer to physically mesh properly).

I also would have assumed that, but I think you might need even more robust simulation tools than Hornresp or akabak or whatever to do this. I think if you have a modern engineering-grade COMSOL setup then the modeling is probably aces. However, the free consumer-grade tools are not there yet, or at least Augerpro does not treat them that way. As far as I know, 1. he is the #1 creator of free waveguide models on the planet and 2. He actually measures everything after sims and prototypes, doesn't just sim-and-go.

I think part of the reason is that there are fine details of the tweeter's shape and surround that have significant impacts on performance >10khz that are very hard to put into the sim, because they are physically very small and difficult to measure or model accurately in the sim. The surrounds can be <1mm wide, but the way they interact with the edge of the horn throat apparently matters.

 

[617]

I also would have assumed that, but I think you might need even more robust simulation tools than Hornresp or akabak or whatever to do this. I think if you have a modern engineering-grade COMSOL setup then the modeling is probably aces. However, the free consumer-grade tools are not there yet, or at least Augerpro does not treat them that way. As far as I know, 1. he is the #1 creator of free waveguide models on the planet and 2. He actually measures everything after sims and prototypes, doesn't just sim-and-go.

I think part of the reason is that there are fine details of the tweeter's shape and surround that have significant impacts on performance >10khz that are very hard to put into the sim, because they are physically very small and difficult to measure or model accurately in the sim. The surrounds can be <1mm wide, but the way they interact with the edge of the horn throat apparently matters.

From what I've read over the years, the closer you are to the transducer, the more critical the shape of the horn is. Geddes's OS equation for example asks for the compression driver exit angle, which is not a specification provided by compression driver manufacturers.

 

[sarumbear]

So as it was presented, it was easy for a layman to conclude that normal rectangular speakers would have pretty severe issues, which they typically do not. Hence actual reason to doubt the graphs in the context they were presented. That has nothing to do with how smart Olson was.

But why would a “layman” design a speaker and sell commercially? Isn’t it expected for manufacturers to know their stuff?

 

[sigbergaudio]

But why would a “layman” design a speaker and sell commercially? Isn’t it expected for manufacturers to know their stuff?

I was thinking more about the readers on this forum. And also about how it was presented here in the thread, not by Olson.

 

[sarumbear]

It seems as if with damn near everything else in audio, we have a significant gap between theory and practice.

I disagree. It’s about understanding the theory. Audio is a complicated discipline and few can understand it.

 

[tmuikku]

I also would have assumed that, but I think you might need even more robust simulation tools than Hornresp or akabak or whatever to do this. I think if you have a modern engineering-grade COMSOL setup then the modeling is probably aces. However, the free consumer-grade tools are not there yet, or at least Augerpro does not treat them that way. As far as I know, 1. he is the #1 creator of free waveguide models on the planet and 2. He actually measures everything after sims and prototypes, doesn't just sim-and-go.

I think part of the reason is that there are fine details of the tweeter's shape and surround that have significant impacts on performance >10khz that are very hard to put into the sim, because they are physically very small and difficult to measure or model accurately in the sim. The surrounds can be <1mm wide, but the way they interact with the edge of the horn throat apparently matters.

There is

Its still a bit tricky but easier than ever. Anyone can get better results than whats available commercially within few weeks with it, and its free. And for any application which is the greatest benefit. Any size and shape simulated with any source, including detailed shape of dome or various wave fronts from compression driver phase plug. Incredibly powerful tool.

Plenty of very very good waveguides shown by many in the associated diyaudio thread during last few years.
Augerpro waveguides seem great so if they fit to application its a no brainer, no need to come up with your own.

 

[kemmler3D]

There is

Its still a bit tricky but easier than ever. Anyone can get better results than whats available commercially within few weeks with it, and its free. And for any application which is the greatest benefit. Any size and shape simulated with any source, including detailed shape of dome or various wave fronts from compression driver phase plug. Incredibly powerful tool.

Plenty of very very good waveguides shown by many in the associated diyaudio thread during last few years.
Augerpro waveguides seem great so if they fit to application its a no brainer, no need to come up with your own.

Thanks, ATH was the tool I was trying to think of. I think augerpro uses ATH but my observation is that it still sometimes takes a few tries (and measurements) to dial in the final design, although it seems like the sims will get you 90% of the way.

TBH I haven't tried it, I am just basing this on observations of augerpro's thread.

 

[kevinh]

Geddes Loudspeakers were 2 way speakers that had a Horn with various size woofers the top of that line used a 15" Woofer with a horn tweeter one of the goals was preventing diffraction distortions. see the rounding of the cabinet and how the Horn was integrated into the cabinet. They measured very well for their time. '