Interesting.
I'm no expert on this stuff, but what you describe seems to imply that a brace that ISN'T rigid is preferable?
IE, in constrained layer damping, there's a damped core sandwiched in between a rigid material like wood, metal, or carbon fiber.
But your post seems to imply that you could get comparable benefits by using an enclosure with rigid walls and a brace in between that is damped.
I plan to test this with some mockups this fall. Do you have any guidelines on how much overlap there should be between both braces? I'm thinking of overlapping about 90% of the distance from wall to wall.
Damped braces look like a sensible idea in practice.
Theoretically this is less effective on modes with nulls (nodes) near the brace.
So a brace in the middle of a panel kills the lowest resonance but not the even harmonics, and so on.
Did you "have a multiplicity"* or just focus on the main resonances?
A "matrix" style presumably would work fairly well.
Possibly useful to supplement CLD with braces optimised for dominant resonances.
Best wishes
David
*I did work in the Patent Office.
There is a patent I saw at some point along those lines:
US20060231327A1 - Loudspeaker enclosure with damping material laminated within internal shearing brace
- Google Patents
It may have been this one. Is that what we're talking about?
US20060231327A1 - Loudspeaker enclosure with damping material laminated within internal shearing brace
- Google Patents
It may have been this one. Is that what we're talking about?
The properties of the viscoelastic material are strongly nonlinear as shown in the nomogram someone posted earlier. This nonlinearity means the modes are different at different frequencies. The only practical way to obtain a forced response for a full speaker cabinet over the frequency range of the lowest few modes without a supercomputer is with modal analysis which requires linearity over the frequency range of interest.
So the task is to find/implement a method of nonlinear modal analysis. This will inevitably require more computer resources than linear modal analysis but hopefully nothing like lots of direct solutions in physical coordinates. There is more than one approach but they are not widely implemented in FE codes. The one I did find turned out to have been bolted on by a third party and not the main developers of the code. It appears to support too few features of the code to be useful for realistic geometries rather than test geometries like single panels.
Damping panels have their pros and cons (and I cannot believe any patent for a damping panel will stand up in court). They tend to add mass to the vibrating walls and limited stiffness which is not usually helpful although not usually particularly unhelpful either. Unlike CLD which, if effectively implemented, will damp all motions involved in radiating sound a damping panel will only effectively damp the modes with shapes which shear the pair of panels. They are more like discrete dampers than broadband damping. However, as KEF demonstrated, if you have one large problematic mode, you know it's shape and are able to place the panel ends where they will be effective then you can put a great deal of damping material into shear to create a substantial damping force. This also means the properties of damping material can be a bit less critical than with CLD.
The greater the overlap the better.
Studies of enclosure radiation show that it is the lowest modes and the largest panels that radiate the most. In my speakers the baffle and back panel were the largest and both are CLD. Then the whole structure in braced internally with a cross that is highly damped. Yes, this is most effective for the lowest modes where the panel move as a unit - by far the most efficient radiator and the one least affected by CLD. So all in all this technique is highly effective.
I still think that you are confusing a nonlinear with frequency with nonlinear with displacement. Materials will be nonlinear in displacement, but the displacements of the panels are extremely small and it is highly unlikely that they would ever reach a displacement such that it would be nonlinear.
All modes have shear, the higher the mode the more shear for a given amplitude (which never happens.) But it is also true that sound is radiated from a panel almost exclusively at the modes - where it resonants. So damping the modes IS the most effective method.
Last edited:
I can see nothing to suggest I am. I explained why the nonlinearity was an issue for the FRF of a large model and, not surprisingly, it has been widely studied in academia which is why there are a number of different ways it can be tackled. Having failed to access an FE code with a usable implementation over the last few weeks my task has now become how to implement one so that I can obtain the FRF for a range of prospective cabinet designs for my mid+tweeter unit using my desktop computer.
Now you would appear to be confused about the nature of the nonlinearity of viscoelastic materials and where it causes problems. The problem lies with an inability to use modal analysis to obtain a FRF and not an inability to obtain a result at a fixed frequency using physical variables and a large amount of CPU. Too much CPU to perform lots of single frequency simulations for a full model of a speaker using my desktop computer.
Like Earl, I think that there is some confusion here, let's try to sort it out.
What nomogram has the data that you think shows non linearity?
Can you provide references to the academic studies in non linearity that you think are applicable?
There is some non linearity in any material but I estimate it is practically irrelevant here.
Best wishes
Davi
Last edited:
Any nomogram for a viscoelastic damping material. The ones in the paper you posted are typical.
At a single frequency you can pick out a storage modulus and a loss modulus and solve the motion using physical or generalised coordinates (modal analysis). You can do likewise at a different frequency using a significantly different storage and loss moduli but the modes will be different. The problem is that this nonlinearity means we have lost superposition and can no longer add the motion at different frequencies as we can with a linear system. For realistic sized FE models it is a requirement to transform from physical to generalised coordinates and retain just the few energy containing modes in order to reduce the computational size sufficiently to be able to generate a Frequency Response Function (FRF).
Yes but a user's guide for a commercial viscoelastic vibration toolbox might give a better overview of the type of thing used by engineers in industry today. Implementing a similar approach would require working at the element level but there are more approximate approaches that can work by iteration and corrections at the matrix level and are likely to be OK for CLD-type speaker walls and damping panels. Damping from other mechanisms like joints means that adding large amounts of computational overhead for one particular form of damping present in the structure is not cost effective.
The nonlinearity of viscoelastic materials is strong and handling it in a reasonable fashion for large practical FE models requires dedicated software which is absent in many, though not all, FE software packages.
I haven't posted any paper. Someone else mentioned a nomogram but none have been posted by anyone AFAIK. What post do you mean?
Thanks for the educational read but it doesn't seem to support your claim, there is a mention of non linearity in section 1.3.2 on p.19 but the import is that it shows up at levels that are so extreme that the viscoelastic material starts to heat up and confuse the issue.
This is not reasonable for a DIY speaker.
The other mention is if the material is preloaded, as in a machinery isolation mount, also not reasonable for a speaker.
What section do you think is relevant?
This paper uses the term "non linearity" in reference to non linearity with respect to amplitude.
This is consistent with my use of the term, and Earl's too I think.
Perhaps there is a translation or nomenclature problem and you mean non linear variation of a property as frequency varies.
Can you explain in what sense you mean "non linear"?
Best wishes
David
Last edited:
Apologies for my poor recall. The paper was in post #92 when we were in discussion but it was actually posted by twinter. However, any nomogram of any viscoelastic material whether measured or a sketch/diagram will show the strong nonlinearity of the material with respect to temperature and motion/frequency.
I think I will have to leave that as a decision for you and other readers. Having said the same thing in the previous 3 posts and linked to software used by industry to handle the modelling of viscoelastic materials I think that is enough when chatting on a forum like this. The confusion would appear to be arising from a lack of background knowledge on the subject of computational methods for vibration and the modelling assumptions involved. The user's guide was not intended to be a teaching text book but an example of how industry handles the simulation of viscoelastic materials.
I would like to add that the failure to get across that viscoelastic materials are nonlinear has been interesting to me and is clearly something that would need to be addressed to some extent in any article aimed at the audience here. I must confess it is not something I would have thought to include prior to this discussion and so thanks for that.
Perhaps but are you sure if the damping material is thin like green glue? This is another aspect that should be present in an article. In truth, I almost certainly would have included it anyway because discussing how energy enters and where it goes is at the heart of understanding cabinet vibration. Nonetheless gathering/confirming issues that would be of interest is useful so thanks for the suggestion.
The practical answer given in the 3 previous posts is that linear modal analysis cannot be used to generate FSFs for structures which include viscoelastic materials. Clearly this hasn't been understood. It is a point that would need explaining to engineers using computational methods to simulate the vibration of cabinets but to a bunch of speaker DIYers wanting to understand the physics behind cabinet vibration it is likely to be an unnecessary mathematical detail even if it is one that has and will continue to soak up a lot of my time.
Perhaps I can answer the question with a question given I have decided not to keep repeating myself. If a material behaves very differently at two frequencies how can it be classified as linear? Frequency analysis requires the same linear system at both frequencies in order for superposition to hold and for motions at different frequencies to sum. At best one can claim a reasonably linear behaviour of a viscoelastic material at constant frequency (and temperature) but this isn't much use if you need to work with generalised coordinates instead of physical coordinates for reasons of computational efficiency in order to generate FSFs. I am repeating myself. The point requires an understanding of how modal analysis works and is used in practical simulations. If you don't have this already as background knowledge (I assumed Earl did but perhaps it has got a bit rusty with age) then it is best addressed by reading a text on modal analysis or talking to someone you trust with knowledge about the subject. A few statements from anonymous posters on a forum is unlikely to help (particularly if they are getting a bit irritated at their failure to get things across!).
I am in accord with Earl that you have confused some issues.
The definition of linearity is, simply stated, that the output varies by factor X if the input varies by factor X.
Temperature is not an input variable.
For example transistors are quite non linear wrt temperature but an audio power amplifier can be almost perfectly linear, it's not a temperature amplifier.
Similarly an equaliser can have quite a non linear frequency response, that's the whole point, but can be essentially perfectly linear, it creates no distortion.
Yes, you have repeated the same claim but simple repetition doesn't support your claim any better.
Neither does the software manual, rather the reverse.
Perhaps if you tried to explain it rather than simply repeat the claim?
Some of us speaker DIYers have university level qualifications in maths, I think we can cope
It may help you.
I find that when I try to explain an issue then I am forced to clarify the subject in my own mind.
Claims that it's too complicated for simple DIYers looks like an excuse for your inability to explain yourself.
I would be happy to be proved mistaken.
Best wishes
David
I finally sat down and did the analysis for optimum constrained layer thickness.
It turned out to be not so difficult if I made a few simplifications that seemed reasonable for speakers - thin constrained layer, sheet panels with no beams, treat as a 2-D problem( no round panels and damn Bessel functions)
The hard bit is the experimental data for loss factors at various frequencies.
I made a few estimates and the numbers look plausible, optimum thickness for a plywood panel with a span of 300 mm, fundamental mode, is about a millimetre for butyl rubber.
It increases with the square of the panel span so it can end up thicker than I had expected.
Indeed. And so if that input consists of more than one frequency for a system that is nonlinear with frequency...
I think that may rest on the level of understanding of the reader.
Having spent the last few weeks off and on studying the alternative methods in the literature for implementing nonlinear modal analysis in order to determine which one/s are likely to be the easiest to add to my own FE software or, hopefully, third party FE software I would like to think I am reasonably on top of the basics although probably not the low level details which should build after implementation and testing.
Reread what was posted. I was referring to the article I am writing on the physics of speaker cabinet vibration for DIY folk (rather than engineers wanting to use computational methods) to whom low level details about computational methods will be irrelevant whether they can understand the maths or not. I will almost certainly now mention something about nonlinearity and superposition but it will only be a short aside.
Why do I need an excuse? I have supplied information in a simple form which you have chosen not to pickup. Fair enough. I could persevere but have chosen not to given that nonlinear modal analysis of viscoelastic materials is almost certainly of no significant interest to anybody on this forum apart from me.