My bending wave driver is designed to have as close to no pistonic movement as the membrane is constrained at the periphery. So TS parameters don't apply. I have no way to measure x-max precisely but it likely is below a few 100 µm. The voice coil height is around 5mm, so it would (barely) count as underhung but then again I tried higher voice coils which could be considered overhung but that did not change the low frequency behavior and had a small negative impact on high frequencies.
I guess that the membrane has a diameter of some 15cm or so? So what is about the typical SPL, as there is no pistonic movement, that you can produce at different frequencies with your setup shown in your first post?
a field coil acts as a huge shorting ring which can lower distortion. has most likely not much to do with the magnet material
I have some time today to write more about the membrane and surrounding damping structure.
My goal is to make a driver which covers most of the frequency range without the need of filters or compensation. Low frequencies can cause intermodulation distortion (Doppler distortion) and so I aim for a frequency response going down to 150 to 250Hz and the upper limit needs to be above hearing threshold so past 20kHz. I have not seen any design of bending wave drivers which go much below 150Hz so that just maybe the lower limit of this driver design. One advantage having a crossover frequency at that range is that the integration of woofers (or even subwoofers) becomes easier as the required distances are achievable (1/4 wavelength). Presently I am using no high pass filters on my bending wave drivers and only a lowpass for the woofers.
Membrane shape and size:
I decided to use flat membranes because they yield good lateral dispersion. As I mentioned before corners lead to strong resonances (uncoupled from the membrane, ringing in their eigen modes) and are impossible to control. I made various shapes without corners like this one:
But none of these shapes were better than a round disk which is much easier to produce with reasonable precision so I concentrated on that form. The diameters of the round disks are between 90 to 120mm, presently I am using 110mm. I tried larger disks but these gave no improvement in the low frequency range and had a decreased high frequency (180mm only to 8kHz). Smaller disks work well in the high frequencies but have much less low frequencies (I tested a 70mm disk which went to only 450Hz).
Membrane construction:
The membrane should have sufficient stiffness (that is in contrast to the Manger membrane) but also enough self damping. There are not many materials which have both high stiffness and high damping, bone is one of them (knock on your head), but that is better achievable in composite structures. So I am using a sandwich structure with central core and upper and lower composite layers.
I tried different core materials like Nomex honeycomb or foam like Rohacell but none of them gave me the required results. I am using a core of lighter endgrain balsa wood (130kg/m3) at 3 to 4mm thickness. That core has good stiffness but also good damping properties. As fabric for the upper and lower composite layer I tried many different materials over the years. Glass fiber, Innegra, flax or kevlar did not perform too well as they lack stiffness and damping. Zylon and carbon fiber work well but it is easier to work with carbon fiber. I also tried TeXtreme carbon/zylon spread https://www.ebay.com/itm/122032681305 and that had some minor advantages over pure carbon fiber. The fibers with the best damping properties is Vectran but that material is very difficult to process as it is extremely hard to cut. It is also very difficult to get it as fabric (200gsm) but Vectran is what I am using now for almost all my membranes.
To laminate the core to the fabric I am using solvent free high strength laminating epoxy. To avoid excess resin I cover both sides with porous teflon coated fiberglass and filter paper to absorb the surplus resin. That hole sandwich is between 2 plates clamped together during curing.
Surround:
That is the most challenging and frustrating part. It should absorb all vibrational energy at the rim of the membrane. It is necessary to find a material with no rebound i.e. has viscoelastic behavior. Most sources do not understand the difference between materials with elastic behaviors, which means that the energy is reflected back and damping/viscoeleastic materials where the energy is absorbed. It is easy to illlustrate the difference with the toy balls which were fashionable years ago, one version is super elastic: when ball is dropped it almost returns back to the same height and the other which lands just on the floor without any rebound. I tried several polyurethane (PU) resins and gels and several NVH damping materials and sorbothane but all were only partially fulfilling the requirements. I had a sample of a Microsorb sheet which was a PU gel with dispersed viscous micelles (micro droplets). These micelles get deformed by vibrational energy which is converted into heat. But Microsorb has been discontinued for a very long time.
What works quite well and is also used in the version shown in my first post is Koyosho Vibration absorption sheet (Z8006B), a PU material 5mm thick (the 3mm blue version works much less well for my purpose). I cut it in strips and attach it to the base plate with thin NVH tape. Recently I discovered silicone base Taica gel tape (gt-5) which has even better performance but it comes only 3mm thick. My latest speakers use that material.
My goal is to make a driver which covers most of the frequency range without the need of filters or compensation. Low frequencies can cause intermodulation distortion (Doppler distortion) and so I aim for a frequency response going down to 150 to 250Hz and the upper limit needs to be above hearing threshold so past 20kHz. I have not seen any design of bending wave drivers which go much below 150Hz so that just maybe the lower limit of this driver design. One advantage having a crossover frequency at that range is that the integration of woofers (or even subwoofers) becomes easier as the required distances are achievable (1/4 wavelength). Presently I am using no high pass filters on my bending wave drivers and only a lowpass for the woofers.
Membrane shape and size:
I decided to use flat membranes because they yield good lateral dispersion. As I mentioned before corners lead to strong resonances (uncoupled from the membrane, ringing in their eigen modes) and are impossible to control. I made various shapes without corners like this one:
But none of these shapes were better than a round disk which is much easier to produce with reasonable precision so I concentrated on that form. The diameters of the round disks are between 90 to 120mm, presently I am using 110mm. I tried larger disks but these gave no improvement in the low frequency range and had a decreased high frequency (180mm only to 8kHz). Smaller disks work well in the high frequencies but have much less low frequencies (I tested a 70mm disk which went to only 450Hz).
Membrane construction:
The membrane should have sufficient stiffness (that is in contrast to the Manger membrane) but also enough self damping. There are not many materials which have both high stiffness and high damping, bone is one of them (knock on your head), but that is better achievable in composite structures. So I am using a sandwich structure with central core and upper and lower composite layers.
I tried different core materials like Nomex honeycomb or foam like Rohacell but none of them gave me the required results. I am using a core of lighter endgrain balsa wood (130kg/m3) at 3 to 4mm thickness. That core has good stiffness but also good damping properties. As fabric for the upper and lower composite layer I tried many different materials over the years. Glass fiber, Innegra, flax or kevlar did not perform too well as they lack stiffness and damping. Zylon and carbon fiber work well but it is easier to work with carbon fiber. I also tried TeXtreme carbon/zylon spread https://www.ebay.com/itm/122032681305 and that had some minor advantages over pure carbon fiber. The fibers with the best damping properties is Vectran but that material is very difficult to process as it is extremely hard to cut. It is also very difficult to get it as fabric (200gsm) but Vectran is what I am using now for almost all my membranes.
To laminate the core to the fabric I am using solvent free high strength laminating epoxy. To avoid excess resin I cover both sides with porous teflon coated fiberglass and filter paper to absorb the surplus resin. That hole sandwich is between 2 plates clamped together during curing.
Surround:
That is the most challenging and frustrating part. It should absorb all vibrational energy at the rim of the membrane. It is necessary to find a material with no rebound i.e. has viscoelastic behavior. Most sources do not understand the difference between materials with elastic behaviors, which means that the energy is reflected back and damping/viscoeleastic materials where the energy is absorbed. It is easy to illlustrate the difference with the toy balls which were fashionable years ago, one version is super elastic: when ball is dropped it almost returns back to the same height and the other which lands just on the floor without any rebound. I tried several polyurethane (PU) resins and gels and several NVH damping materials and sorbothane but all were only partially fulfilling the requirements. I had a sample of a Microsorb sheet which was a PU gel with dispersed viscous micelles (micro droplets). These micelles get deformed by vibrational energy which is converted into heat. But Microsorb has been discontinued for a very long time.
What works quite well and is also used in the version shown in my first post is Koyosho Vibration absorption sheet (Z8006B), a PU material 5mm thick (the 3mm blue version works much less well for my purpose). I cut it in strips and attach it to the base plate with thin NVH tape. Recently I discovered silicone base Taica gel tape (gt-5) which has even better performance but it comes only 3mm thick. My latest speakers use that material.
Last edited:
I do the measurements in my listening roomroom at close distance to avoid as possible the effect of room acousticsIf the microphone is very close (1-2 cm) any small movement gives very different response which should be expected with a vibrating membrane. At 0.3m I see almost no difference in response if I move the microphone parallel to the membrane so that is my preferred measuring distance. To avoid early reflections from the floor I have 2 wedges on the floor to avoid floor reflections to the microphone positions. The ceiling is very well treated (microperforated acoustic panels). If I measure 1000Hz at 2.83V at 1m I get 82-86dB, so low but acceptable (one reason I played around with different magnet structures).
wouldn't even a slight (and damping) surround help distortion behaviour at 150/200hz?
ever tried aluminium foil as sandwich cover?
Aluminium is my choice with good results concerning resonances:
also measurements are to be found
ever tried aluminium foil as sandwich cover?
Aluminium is my choice with good results concerning resonances:
also measurements are to be found
What an appealing maze of interesting trial-and-error! Your ongoing project is evoking questions and ideas inside of me, and also the growing motivation to dive myself into this subject. Therefore, let me ask some more questions: While steadily experimenting over time, did you eventually already try ...
- some kind of 60°-angled plywood compound glued together out of 3 sheets of 1mm balsa each, instead of using a uniform 3mm-sheet? And if so, then also slightly varying the "plywood-angles" in the idea to further spread potential resonances or to give the membrane a distinct bending pattern to eventually optimize the match with the non-circular, linear/rectangular geometry of the voice coil? Thus "plywooding" e.g. at 50°, 60° and 70°?
- a balsa "quasi-only" membrane, without the two outer composite layers spreading completely from the center to the periphery? E.g. only with some smaller and possibly irregularly shaped reinforcement layers in it's central part in order to attach the voice coil? E.g. such as a star-like reinforcement layering such as in the Mangers?
- used two sand filled ,"floppy" (e.g. non-shrinked, "native" shrink tube) circular tubes as a membrane fixing surround to sqeeze the membrane periphery between them?
- ...
- ...
Last edited:
I tried years ago laminated standard balsa wood to make ply as you suggested but end-grain (in German: Hirnholz) balsa performs much better with much more even frequency response. In general end-grain wood (crosscut) is much stiffer than longitudinal cut and is close to isotropic (there are some small variations in density of the balsa blocks the end grain sheets are made from). End-grain balsa is often used as a core material for composite laminates but I found that the density of the wood used makes a significant difference. Too light versions have too little stiffness and too dense add to much weight.
I tried several modifications of an isotropic membrane to change the behavior of the membrane by adding damping material at different positions (obviously I tried moon gel there too), perforations or slots loaded with damping compound (even some filled with tungsten powder) but all of these tempt resulted in a restricted and uneven frequency response. Here are some pictures of my trash pile:
I tried several modifications of an isotropic membrane to change the behavior of the membrane by adding damping material at different positions (obviously I tried moon gel there too), perforations or slots loaded with damping compound (even some filled with tungsten powder) but all of these tempt resulted in a restricted and uneven frequency response. Here are some pictures of my trash pile:
Questions, questions ...
What is the merit of a balsa core?
I guess you eventually tried Vectran solo, or twofold in a 45°-"plywood"-configuration as a non-balsa-sandwiched membrane?
The single or the two layers soaked/hardened/bonded with e.g. a lowishly polymerized (10:1 or 10:2) PU or a "softened" epoxy resin?
What is the merit of a balsa core?
I guess you eventually tried Vectran solo, or twofold in a 45°-"plywood"-configuration as a non-balsa-sandwiched membrane?
The single or the two layers soaked/hardened/bonded with e.g. a lowishly polymerized (10:1 or 10:2) PU or a "softened" epoxy resin?
Montverdi,
I'm very excited to read the details here of your membrane construction. Thanks so much for sharing! Where did you find the end grain balsa?
Thanks,
Eric
I'm not sure how well it will work for your membrane, but so far I have found that Poron 92 is very good. I'm not sure how it compares to the material you found.
See my testing here :
A single layer of Vectran or carbon fiber composite is not very stiff. A flat sheet can be very easily bent
left carbon fiber right Vectran.
If one wants to develop of soft membrane similar Manger's MSW a single or double (maybe partial double) layer of Vectran would be a good material to try (which I did for a few attempts).
But I am interested to use a stiff plate for a bending wave driver and to get a stiff plate with Vectran one needs multiple layers (one also could combine it with carbon fiber and/or Zylon) but that would be much more mass than a composite construction with a low weight core. End-grain balsa has (as any other wood) high stiffness perpendicular to the cross cut direction but it can break relatively easy parallel to the fibers (think of splitting wood). Laminating with a high tensile strength fabric generates a very stiff structure (like a 3d I-beam only with microscopic honeycomb structures, i.e. longitudinal oriented wood cells). Compared to honeycomb Nomex balsa yields much more stiffness and damping (also much easier to laminate).
End-grain balsa is widely available from stores supplying composite materials but not in the thickness and weight range which worked for me. My source is: https://specializedbalsa.com/products/end_grain.php
One can specify the weight class (costs a little more) mostly I use 5/32" thickness.
left carbon fiber right Vectran.
If one wants to develop of soft membrane similar Manger's MSW a single or double (maybe partial double) layer of Vectran would be a good material to try (which I did for a few attempts).
But I am interested to use a stiff plate for a bending wave driver and to get a stiff plate with Vectran one needs multiple layers (one also could combine it with carbon fiber and/or Zylon) but that would be much more mass than a composite construction with a low weight core. End-grain balsa has (as any other wood) high stiffness perpendicular to the cross cut direction but it can break relatively easy parallel to the fibers (think of splitting wood). Laminating with a high tensile strength fabric generates a very stiff structure (like a 3d I-beam only with microscopic honeycomb structures, i.e. longitudinal oriented wood cells). Compared to honeycomb Nomex balsa yields much more stiffness and damping (also much easier to laminate).
End-grain balsa is widely available from stores supplying composite materials but not in the thickness and weight range which worked for me. My source is: https://specializedbalsa.com/products/end_grain.php
One can specify the weight class (costs a little more) mostly I use 5/32" thickness.
I tried several PU foam/gel version but that one I will look into!I'm not sure how well it will work for your membrane, but so far I have found that Poron 92 is very good. I'm not sure how it compares to the material you found.
See my testing here :
Eric