Announcement

Collapse
No announcement yet.

The real world importance of cabinet bracing is?

Collapse
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • #91
    Originally posted by andy19191 View Post
    In order to isolate effectively the resonant frequency of the mass (driver) on the spring (gasket) ideally needs to be about a decade below the lowest frequency in the driver's passband. If the spring is too stiff (often the case) then you are adding an (often strong) resonance into the passband of the driver and making things worse. Grommets allow softer springs than gaskets and so are usually a better choice.

    Even if you can isolate a driver should you do it? Isolating a driver means the magnet moves more in reaction to the acceleration of the cone/coil assembly and this distorts the motion of the cone with respect to the electrical signal. The size of this distortion follows from the relative mass of the air/cone/coil versus frame/magnet at least below the resonance of the frame/magnet on the baffle. This leads to something like isolation being beneficial for tweeters, not for woofers and with midranges being debatable.
    I have never isolated a driver in anyway. I too have questioned
    the methods and effectiveness. I was just mentioning another
    potential use for the newly found joys of 3D printing.
    Guess xmax's age.

    My guess: 15. His grammar is passable. His trolling is good.

    Comment


    • #92
      Perhaps a bit more about the free engineering FEM software might be helpful. The subwoofer mesh shown above was created with Salome/CodeAster plus some supporting python scripts which for me helps speed up making lots of small parametric changes to geometry and grid. All the geometry and grid could have been created interactively directly in the Salome software. Salome is a pre and post processor for the CodeAster FEM solver.

      The grid is a coarse one for initially sorting out stuff (currently working on damping models) but the results are sufficiently reasonable to guide large modifications if not necessarily small detailed ones. The grid consists of 3568 Hexa8 elements (not a suitable element for anything other than sorting out): 160 steel, 504 aluminium and the rest MDF. The magnet is little more than a heavy mass to which the reaction force from the coil moving in the magnetic field will be applied and so it doesn't need resolving with a mesh. The frame on the other hand does resonate within the frequency range of interest and needs to be reasonably well resolved with a mesh.

      To assemble and calculate the resonant frequencies and mode shapes took roughly 5 seconds elapsed time (23 seconds user+system CPU) running on a dual Xeon system that is 2 or 3 years old. If I understand the output correctly (it's in French) the minimum memory required was 65 MBytes, the optimum 192 MBytes but with plenty available the peak actually used was 3585 MBytes.

      The picture below shows the Salome application in post processor mode. The other main modes are geometry, mesh and solver plus a few minor ones for doing this and that. The main view shows an animation of the 7th mode shape (the model is unconstrained and so the first 6 are solid body motions at 0 Hz). This is the lowest non-trivial mode with a frequency of 421 Hz and shows the driver twisting around an axis from top left to bottom right and the cabinet vibrating in sympathy. The 8th mode is the twin at 421 Hz with the driver twisting around the top right to bottom left axis.

      PS How do I insert a small thumbnail that you click to see a large image?

      Click image for larger version

Name:	SubMode7_coarse.jpg
Views:	1
Size:	328.1 KB
ID:	1399288

      Comment


      • #93
        Originally posted by danmarx View Post
        Did you end up tweaking any of the constants?
        Hi Dan.
        I used the same values as yourself. My biggest hiccup was realizing the calculation was dimensionless and so adhering to SI units was a must for cabinet dimensions. You really gave me a head-start to hit the ground running!

        Originally posted by andy19191 View Post
        A standard finite element assembly requires the elements to butt face to face and with the same nodal distribution on each face. Things can be done to get round this but it needs to be explicitly specified with care and understanding of how the relevant degrees of freedoms are handled. Your elements do not match face to face at the edges of your box but only, apparently, line to line.
        Noted with thanks. This makes sense intuitively, but is easy to gloss over when moving as fast as I was.

        Originally posted by andy19191 View Post
        ... and one would use a significantly finer mesh for one or two accurate simulations of the final design.
        Are you aiming to quantify the theoretical reduction in cabinet noise transmission? Otherwise I'm failing to see the importance of resolving a fine mesh. My interest lies in identifying the cabinet resonant modes' geometry simply as a way to identify where damped braces should be located. It seems this should be possible even with the crudest of models.

        Originally posted by andy19191 View Post
        Perhaps a bit more about the free engineering FEM software might be helpful. The subwoofer mesh shown above was created with Salome/CodeAster plus some supporting python scripts which for me helps speed up making lots of small parametric changes to geometry and grid. All the geometry and grid could have been created interactively directly in the Salome software. Salome is a pre and post processor for the CodeAster FEM solver.

        The grid is a coarse one for initially sorting out stuff (currently working on damping models) but the results are sufficiently reasonable to guide large modifications if not necessarily small detailed ones. The grid consists of 3568 Hexa8 elements (not a suitable element for anything other than sorting out): 160 steel, 504 aluminium and the rest MDF. The magnet is little more than a heavy mass to which the reaction force from the coil moving in the magnetic field will be applied and so it doesn't need resolving with a mesh. The frame on the other hand does resonate within the frequency range of interest and needs to be reasonably well resolved with a mesh.

        To assemble and calculate the resonant frequencies and mode shapes took roughly 5 seconds elapsed time (23 seconds user+system CPU) running on a dual Xeon system that is 2 or 3 years old. If I understand the output correctly (it's in French) the minimum memory required was 65 MBytes, the optimum 192 MBytes but with plenty available the peak actually used was 3585 MBytes.

        The picture below shows the Salome application in post processor mode. The other main modes are geometry, mesh and solver plus a few minor ones for doing this and that. The main view shows an animation of the 7th mode shape (the model is unconstrained and so the first 6 are solid body motions at 0 Hz). This is the lowest non-trivial mode with a frequency of 421 Hz and shows the driver twisting around an axis from top left to bottom right and the cabinet vibrating in sympathy. The 8th mode is the twin at 421 Hz with the driver twisting around the top right to bottom left axis.
        Impressive work!
        You've been sly not letting on that you've been working on this behind the scenes this whole time.

        So are you just going to identify resonant modes, or are you also going to apply a reactive force to the magnet to see what happens? I'd be interested to see how one would apply this? How do you resolve the linear motion of the voice coil?

        As this is for a sub, do you intend to brace to move resonances up beyond the passband of the driver?

        Comment


        • #94
          Originally posted by gregrueff View Post
          Are you aiming to quantify the theoretical reduction in cabinet noise transmission? Otherwise I'm failing to see the importance of resolving a fine mesh. My interest lies in identifying the cabinet resonant modes' geometry simply as a way to identify where damped braces should be located. It seems this should be possible even with the crudest of models.
          The project as a whole is/was to see what could be done with free 3D software to simulate speakers and speakers in rooms. It started and then was put on hold last summer because of moving to a new house which requires a great deal of work. I have done enough to discover that in many areas some free software is fine (most isn't) although more time consuming to use than good commercial software (there's plenty of bad commercial software) but there are one or two areas that are poorly handled or missing. For this project:
          - efficiently handling viscoelastic damping models with properties that vary with frequency is missing
          - efficient and reasonable models for damping acoustic waves in porous materials like foams, fibreglass, furniture,... seems to be missing
          - BEM models use very low order elements which seriously limits what can be resolved
          - grid generation tends to be unsophisticated requiring a lot of time
          When I restart the project I will probably write software to address the first two, perhaps the third but live with the fourth.

          A reasonably fine mesh is required for a reasonably accurate simulation. This is of significant interest for everything I will actually build but of less interest for the tens perhaps hundreds of configurations that are evaluated on the way to the final design/s.

          In order to reliably determine the mode shapes your model needs to include everything that is significant and to resolve it sufficiently so that the actual physics is not swamped by unphysical numerical errors. Converged 3D solutions often look plausible to the inexperienced (and sometimes the experienced!) even when they include gross errors of the kind in your example because everything hangs together. I posted the grid and example above to give you an idea of what you need to do to get a reasonably reliable first coarse grid with which to start sorting things out. Without the drivers hanging off the baffle your mode shapes will be largely irrelevant because the worst modes for actual speakers almost always involve the driver bouncing or twisting on the baffle.

          Originally posted by gregrueff View Post
          So are you just going to identify resonant modes, or are you also going to apply a reactive force to the magnet to see what happens? I'd be interested to see how one would apply this? How do you resolve the linear motion of the voice coil?
          The forced response including the cabinet radiation is perhaps the most useful information. That is, SPL vs frequency at a point such as the listening position, SPL at a particular frequency over a sphere around the cabinet in free space or a room,... I have run simulations to generate such plots but so far with incomplete models. I am not happy with the unphysical but simple damping models, have changed my mind about how best to handle the internal air pressure loading on the cabinet, not happy with BEM accuracy for a room,... It is normal to be unhappy with the accuracy of large 3D simulations but these haven't crossed the line of being sufficiently developed to be worth taking forward or showing.

          The reaction force on the magnet is simply a body force on the magnet from the mass of the air plus coil and cone assembly. This is straightforward to determine for a rigid motion at a particular frequency (i.e. the same way as SPL is determined in speaker design programs) but high accuracy isn't particularly vital since it is the difference between the SPL of the cone and the cabinet that is of interest.

          The driving force on the inside of the cabinet wall from the air pressure within the cabinet is usually less important and is close to constant in space at low frequencies (easy to apply) but varies at high frequencies (fiddly geometric calculation for boundary conditions or include the simulation of the internal air).

          Originally posted by gregrueff View Post
          As this is for a sub, do you intend to brace to move resonances up beyond the passband of the driver?
          Yes which will then require only a minimal amount of damping. They are also going to be initially pressed into service as woofers for the main speakers which raises the required minimum frequency for the lowest resonance considerably. In addition, they are primarily intended to be used in substantial numbers to absorb sound to actively control the room response of a modestly sized badly shaped room rather than generating deep low frequency booms as is common with most subs. So they need to be small enough to fit on shelves and in various nooks and crannies around the room which puts a lot of pressure on avoiding thick walls.
          Last edited by andy19191; 01-04-2019, 05:57 AM. Reason: Typo

          Comment


          • #95
            My original post was based on an incorrect, intuitive hunch about the nature and sonic significance of cabinet vibration. The input since then makes me believe that cabinet bracing and damping do have audible sonic effects that are worth considering even in relatively small speakers. Much of the recent input suggests that optimum cabinet design is complex and requires engineering skills far beyond those of many DIYers. Fortunately, I am shooting for decent vibration control, not optimum, state of the art results.
            So, based on what I have gleaned from the input that is not way above my head, I would like to throw out for comment my low tech plan for cabinet bracing-damping in my next build. The speaker will be a 2 way with interior dimensions of 6" W X 10.5" D X 13.5" H out of 3/4 MDF. Using 100% silicone glue I plan to glue 4" squares of ceramic tile to the four interior walls and top ( this piece 4" X6"). The squares will be placed in such a way as to not be centered in any panel. I then plan to glue in a side to side and front to back brace centered on the tiles. These braces will be 3/4 X 1 1/4 plywood. Each brace will consist of two pieces, overlapped 4" and glued with a 1/8" thick layer of silicone glue as shown in post #56 to load the flexible glue in shear. That is it.
            I don't have drawing or sketch up skills, but hope the description is adequate to describe my plan. I know this is far, far from optimum, but think that it may be a practical way to incorporate some of the excellent ideas that have been put forward in this thread. Input for any simple, easy to change improvements are invited.

            Best,
            Jay

            Comment


            • #96
              That sounds like a solid plan and better than 90+% of both DIY and commercial speakers. Sent from my SM-G960U using Tapatalk
              Electronics engineer, woofer enthusiast, and musician.
              Wogg Music
              Published projects: PPA100 Bass Guitar Amp, ISO El-Cheapo Sub, Indy 8 2.1 powered sub, MicroSat

              Comment


              • #97
                I'm with you on the brace construction, although I'm not sure about the 1/8" thick damping layer. The first time I was exposed to this idea was by Geddes and he said thickness is actually not your friend. I would just apply evenly to the braces just like you were gluing them up.

                Can someone expound on the ceramic tile idea?
                ~Brandon

                Soma Sonus
                DriverVault

                Comment


                • #98
                  Originally posted by joeybutts View Post
                  Thickness, span, material all play a part. If that "U" is connected through the middle you really have some effective bracing. I sometimes use battons to stiffen, though not eliminate, panel resonance. Just depends on the project.

                  If I truly am building out a project, I try to not have over 1 sq ft of panel not broken up by some kind of bracing to somewhere near the middle of the panel. And then sometimes I just go overkill but take it into account with cabinet volume.

                  This helped create a dead cabinet (and completely and utterly overkill):




                  On that note, it puts me at peace of mind that I have created the best product I could, instead of leaving a little doubt if it would could sound any better. And knowing what I know now, I'm sure you could obtain the same thing using sticks throughout, but I really wanted to push the envelope since I knew I would have those cabinets a long long time (7 years now and still satisfied).

                  Mustang, corvette, porsche, lamborghini, bugatti.....at what point are you satisfied.

                  I had a speaker in a cardboard box for poops and giggles and I grabbed it while it was playing one day and noticed an incredible improvement in the sound quality of midrange and down. Extreme example but I think illustrates the point.

                  PS. someone posted testing data of unbraced areas of panels, though I have zero recollection of results or conclusions.
                  Porsche.
                  Guess xmax's age.

                  My guess: 15. His grammar is passable. His trolling is good.

                  Comment


                  • #99
                    Originally posted by Drjay View Post
                    Much of the recent input suggests that optimum cabinet design is complex and requires engineering skills far beyond those of many DIYers. Fortunately, I am shooting for decent vibration control, not optimum, state of the art results.
                    In that case you would implement a cabinet with effective CLD walls since this will damp all the motions of the cabinet. You would not use shear panels, as you intend to do, because shear panels only effectively damp part of the motion of the cabinet and you don't know which motions you want to damp. As mentioned earlier, the tile business would work to some extent if the tiles were on the outside of the structural layer but on the inside they are unlikely to do much with something soft as a damping layer. Then again, the little they do is likely to be positive and so why not if you want to do it.

                    But is audible cabinet radiation a bad thing for audiophiles? In the UK these speakers have become fairly popular with their "exciting" sound using the "technology" described on the concept page.

                    Comment


                    • Sometimes experts disagree:

                      " For making speaker enclosure I think that just using some form of CLD is far more important than worry about if it is "optimum". Don't let seeking "the best" be the enemy of "the better".

                      I also find damped cross bracing to be very effective, maybe more so than CLD, although the techniques are similar."
                      __________________
                      Earl Geddes Gedlee Website

                      As Augerpro points out, Geddes does suggest that the adhesive layer joining the two halves of the cross braces should be thin rather than thick as I was planning, so that makes actual building a bit easier.

                      Comment


                      • Originally posted by Drjay View Post
                        Sometimes experts disagree:

                        " For making speaker enclosure I think that just using some form of CLD is far more important than worry about if it is "optimum". Don't let seeking "the best" be the enemy of "the better".

                        I also find damped cross bracing to be very effective, maybe more so than CLD, although the techniques are similar."
                        __________________
                        Earl Geddes Gedlee Website

                        As Augerpro points out, Geddes does suggest that the adhesive layer joining the two halves of the cross braces should be thin rather than thick as I was planning, so that makes actual building a bit easier.
                        In my experience science/engineering experts almost never disagree because science and engineering is built on knowledge that is required to be always true. This is not the case with the soft sciences like socal science, economics and the like where the fundamental building blocks/rules are not required to be always true (e.g. demand decreasing with increasing price). As for media/internet personalities...

                        Why do you consider Geddes to be an expert? A fair amount of what he posts is sensible but he also comes up with some off-the-cuff stuff which is nonsense which he then won't correct or explain when questioned. You can usually spot genuine experts by their being able to say why and not just what and in words you can understand and not complicated jargon.

                        In order for CLD to be effective it has to dissipate mechanical power (force * velocity in direction of force). In order to do that the damping layer must be significantly deformed (shear being significantly more effective than compression) and the internal forces must be significant (it must be a stiff lossy material and not a soft lossy material). The problem with your tiles is that the soft glue is loosely hanging the tile off the wall creating something more like an isolation system and not a CLD system. This is why it should work on the outside by preventing the external air experience the movement of the structural layer but not on the inside.

                        Comment


                        • Originally posted by andy19191 View Post
                          This is why it should work on the outside by preventing the external air experience the movement of the structural layer but not on the inside.
                          Andy!
                          You make some excellent points, but I've been giving this a good deal of thought.
                          As long as we impart shear strain in the damping layer, we should achieve the damping we are after.
                          I don't believe it should matter which side or which face of the constraining layers are imparted with the excitation source or resonance. So whether the ceramic tile is on the inside or outside of the speaker enclosure should not matter. What matters is that it is sufficiently stiff enough to force shear deformation in the damping layer by sandwiching it between another sufficiently stiff medium (the MDF speaker box in this case). The resulting composite structure will still resonate, it will just now be damped.

                          Just as with a simply supported beam undergoing simple harmonic motion given an initial displacement, it does not matter whether you push the bar up or down, it will still be damped and come to rest faster than an undamped beam.

                          In keeping with the conservation of energy, if the damping layer is deforming, heat is created and so then displacement and pressure must be reduced.

                          I believe this is how KEF is achieving their results, because they are essentially just "putting tiles on the inside of the box" where their tiles just happen to be a rigid X-brace.

                          Appreciate the high level of discourse you bring to this discussion.

                          Comment


                          • Originally posted by gregrueff View Post

                            Andy!
                            You make some excellent points, but I've been giving this a good deal of thought.
                            As long as we impart shear strain in the damping layer, we should achieve the damping we are after.
                            I don't believe it should matter which side or which face of the constraining layers are imparted with the excitation source or resonance. So whether the ceramic tile is on the inside or outside of the speaker enclosure should not matter. What matters is that it is sufficiently stiff enough to force shear deformation in the damping layer by sandwiching it between another sufficiently stiff medium (the MDF speaker box in this case). The resulting composite structure will still resonate, it will just now be damped.

                            Just as with a simply supported beam undergoing simple harmonic motion given an initial displacement, it does not matter whether you push the bar up or down, it will still be damped and come to rest faster than an undamped beam.

                            In keeping with the conservation of energy, if the damping layer is deforming, heat is created and so then displacement and pressure must be reduced.

                            I believe this is how KEF is achieving their results, because they are essentially just "putting tiles on the inside of the box" where their tiles just happen to be a rigid X-brace.

                            Appreciate the high level of discourse you bring to this discussion.
                            The shear layer must be inside the constrain layer to work.
                            craigk

                            " Voicing is often the term used for band aids to cover for initial design/planning errors " - Pallas

                            Comment


                            • Would a bowing or bending inner penal against a more ridged outer penal, with a viscous bond between them, be considered a shearing force ?
                              In-turn, an energy absorbing inner wall.
                              I'm hoping so. This post just happen to come up as I was in need of some internal volume for a center channel speaker. Several 3mm MDF sheets bonded together with silicon sealer (RTV).

                              Comment


                              • Originally posted by gregrueff View Post
                                As long as we impart shear strain in the damping layer, we should achieve the damping we are after.
                                As I mentioned earlier, the stiffness of the damping layer is important. If it was as stiff as the structural and constraining layers (theoretically since high loss and high stiffness don't go together) then it would deform only a small amount which we don't want. If it is significantly less stiff than the structural and constraining layers but still fairly stiff as a material it will be deformed close to the maximum amount and the damping force will still be large enough to do something about the motion of the stiff speaker walls which is what we want. If it is a soft material as suggested above it will be deformed the maximum amount but the size of the damping force generated will be too small to significantly affect the motion of the stiff strong cabinet walls which will continue to resonate with only a small amount of damping.

                                Originally posted by gregrueff View Post
                                I don't believe it should matter which side or which face of the constraining layers are imparted with the excitation source or resonance.
                                We want the layer that moves least to be on the outside to radiate sound and this is normally the constraining layer. In addition we would also like the structural layer on the inside so that bracing is straightforward to apply without cutting holes through the constraining and damping layers. This is the normal CLD arrangement. (I wanted to link to a section through the old AE1 monitor but can no longer find it on the web. Anyone? Or an alternative?).

                                If you want the constraining layer on the inside for cosmetic reasons (?) then an effective CLD arrangement (i.e one that creates a significant damping force) will likely be a bit more complicated and a bit noiser but possibly not by much. An ineffective one (i.e. one that creates a small damping force) will perform little better than an undamped structural wall. However, it may still perform reasonably well with the "constraining layer" on the outside if it places a barrier that moves little between the air and the largely undamped motion of the structural wall. The physics of this is different since it is more isolation than CLD and isolation needs a soft not a stiff material to be effective.

                                Originally posted by gregrueff View Post
                                I believe this is how KEF is achieving their results, because they are essentially just "putting tiles on the inside of the box" where their tiles just happen to be a rigid X-brace.
                                Yes but their damping material will be relatively stiff and not soft or else it wouldn't create the results shown.


                                Comment

                                Working...
                                X