After working my way through 4 pages and 80 posts, I get the feeling that, in mentioning FEM and BEM and specifically COMSOL, Andy is 19191 is adressing a completely different aspect of loudspeaker design than what is ususally the target of even the most advanced and versed do-it-yourselvers.
COMSOL and the likes thereof are intended to make sims for fundamental driver design and fundamental and in depth enclosure/structure vibrational analysis. Virtually no SME or/and hobbyist loudspeaker system designer has access to such advanced FEM and BEM tools:
1) We are dependent on what is available on the market, both driver and enclosure materials alike: most here are not researchers. The the focus is rightfully on proper on and off axis SPL performance. That is good enough for 97% of any loudspeaker.
2) COMSOL costs a house and takes 3-5 years to master. Furthermore it has no added value for the average diy-er given point 1).
3) DIY accessable software is getting more advanced by the year and thus the gap becomes smaller: 4 ATH has already been mentioned
That being said, the focus indeed seems on x/o modelling plus optimizing and relatively little on enclosure behaviour, but recently more attention is given to diffraction and enclosure shapes.
On the the other side of the spectre: Genelec uses relatively cheap Peerless drivers in some of their systems: I do not think Genelec does any COMSOL modelling. I do not know of the Neumann sourcing, but Neumann certainly does not make all its drivers in house. Idem as per use of COMSOL.
Only very few manufacturers take a truly holistic overall approach requiring FEM or BEM tools: KEF and some of the high end Harman design come (JBL M2) to mind. The others basically focus on some aspects of the design and tend to neglect others.
As a hobbyist I would very much like to have acces to easy to use free FEM or BEM tools and affordable scanning hardware in order to be able to analyse cone and dome behaviour, but here only a small crowd seems interested in such topics.
COMSOL and the likes thereof are intended to make sims for fundamental driver design and fundamental and in depth enclosure/structure vibrational analysis. Virtually no SME or/and hobbyist loudspeaker system designer has access to such advanced FEM and BEM tools:
1) We are dependent on what is available on the market, both driver and enclosure materials alike: most here are not researchers. The the focus is rightfully on proper on and off axis SPL performance. That is good enough for 97% of any loudspeaker.
2) COMSOL costs a house and takes 3-5 years to master. Furthermore it has no added value for the average diy-er given point 1).
3) DIY accessable software is getting more advanced by the year and thus the gap becomes smaller: 4 ATH has already been mentioned
That being said, the focus indeed seems on x/o modelling plus optimizing and relatively little on enclosure behaviour, but recently more attention is given to diffraction and enclosure shapes.
On the the other side of the spectre: Genelec uses relatively cheap Peerless drivers in some of their systems: I do not think Genelec does any COMSOL modelling. I do not know of the Neumann sourcing, but Neumann certainly does not make all its drivers in house. Idem as per use of COMSOL.
Only very few manufacturers take a truly holistic overall approach requiring FEM or BEM tools: KEF and some of the high end Harman design come (JBL M2) to mind. The others basically focus on some aspects of the design and tend to neglect others.
As a hobbyist I would very much like to have acces to easy to use free FEM or BEM tools and affordable scanning hardware in order to be able to analyse cone and dome behaviour, but here only a small crowd seems interested in such topics.
Times are changing and elaborate FEM/BEM methods, with optimizers and all, etc will become more wide-spread. Currently, I would think any relevant percentage of drivers with such a design foundation, if any, is found in the PA pro audio segment because that's where the money is ;-)
They all have access to free open source software like salome, codeaster, acousto, calculix, opencfd, elmer, spice and many others and have done for decades. Free software obviously doesn't come with commercial support but there will be documentation and usually forums similar to this. Not using it seems odd to me. There is a growing use of BEM but even here it is odd in being a limited cut down commercial code rather than an open source one like acousto, bempp or similar that can be more easily integrated with a design workflow.
Simple 0D/1D design methods can address most of what is needed to put together a loudspeaker but I would have expected some of the more experienced DIYers to want to go further and raise the quality of their designs by learning to use tools that allow the details to be refined and optimised. It's a hobby though and so if it looks a bit like school work then perhaps it is understandable. Don't know it just seems odd to me.
There's little to no interest in this thread. Do you think there's interest elsewhere?
I started a design software project a few years ago on a different forum when Jeff Bagby started to throttle back on his software and was looking for people to continue. It gathered no interest and taught me a lot about what not to do. Firstly, don't start a project when you have only a modest level of interest in what the software is doing. Secondly, don't start a project that uses a wide range of languages and libraries appropriate to each task if you want people to join. Thirdly, don't put up outlines and incomplete code and documentation because most won't get it and will be put off. Fourthly,...
As it happens I have started a project a few months ago to discuss more of the detailed stuff not covered by conventional speaker design programs. Have a few articles on cabinet construction (surprise!), simple linear modelling (required for boundary conditions), on wall speakers, and a list of ten or so further topics which I am adding to. Writing software to address what is needed but missing in the freely available CAE software is hard work but when I get over this initial barrier progress should be swifter and more enjoyable. Threads like this have been informative about likely levels of interest within the speaker DIYer community and has rather shifted my target audience and the organisation of the content.
My interest will be full on if there is proper evidence that the contribution of those cabinet resonances is very significant.
Without a baseline it's still extremely vague and all in theory, even with BEM/FEM.
I asked a few times for clarification, but that seems to be just ignored.
So sorry, but than I also lose interest if people don't answer those questions. 🤷♂️
Without a baseline it's still extremely vague and all in theory, even with BEM/FEM.
I asked a few times for clarification, but that seems to be just ignored.
So sorry, but than I also lose interest if people don't answer those questions. 🤷♂️
Loudspeaker manufacturers used to even share distortion plots as well.Under recognized? U kiddin’? The only complication being a lot of enthusiasts seem to like resonances
Anyway, Vifa delivered data sheets with CSD plots from the early nineties. They stopped doin it somehow.
We can all guess the "somehow".
These days they don't even stay at the recommended AES guidelines.
Speaking only of the numerical methods 3-D stuff, yeah the audience is necessarily limited because of the divergence in education, interests and preferences. People will use what they're comfortable with--from the methods all the way down to the toolchain. Also want to echo that once the insights are gained, you don't repeat those particular deep-dives because they're not worth it on a hobbyist level. For years, I periodically looked-in on 5 of the 7 of those you listed (plus several more) but was already spoiled by dayjob code by then. Not saying that there's not value in what you're discussing--the contrary--there's always a few things to look at if a person has any curiosity at all.
I say scratch your own itch, leave breadcrumbs for the next person down the line, and that's as good as it can get and not become drudgery. If you still have the chops+energy+time, you might be the right person born at the right time. The right person can move things quite a bit as we see around here daily.
That's a problem, isn't it? Even if we exclude human nature, like being a terrible scientist, or a masochist tendency to negatively train your hearing to "endure" rather than enjoy the sound (because of some ideology that X 'ought' to sound better than Y...) there's still a multi-dimensional space affecting sound quality.Under recognized? U kiddin’? The only complication being a lot of enthusiasts seem to like resonances
I was looking to buy some speaker drivers from a selection that seemed quite similar to each other on the surface, and couldn't quite figure out the catch between different levels of impedance 'flatness' they had on offer. If one driver could have a ruler-flat impedance that goes from 7 ohm @ 500Hz to 10 ohm @ 20kHz, why should another only stay flat up to 5kHz and then curl up to 20 ohm @ 20kHz?
AFAICT, the answer lies in balancing the THD-reducing properties of copper shorting rings against possible downsides like excessive braking or frictional loading. Similar to the way RF reflections can occur if a terminating impedance is too high OR too low, I realised that an excessive frictional load at the voice coil position could cause returning 'ripples' (from the cone edge back to the middle) to bounce right off, causing standing waves. So it would be a very similar phenomenon as my long-time argumentation against an excessively high damping factor, except that the implementation is electro-mechanical and baked into the hardware, rather than electrical and user-adjustable.
As with the never-ending controversy with tube amp THD vs subjective sound quality, it seems there may be situations where one type of more objectionable distortion (eg break-up modes) is traded for higher THD but also a tighter CSD & impulse response when real-world vibrations exit the speaker and reach the microphone/ear.
Usually it’s the human that’s the problem. Data entry is arduous and boring so mistakes are common. There is a large learning period with the archaic, crude, free simulation software. It’s often quicker for me to make a physical model in the shop and get actual measurements than sit for several evenings messing around with a simulator that produces odd results because the speaker manufacturers data is overly optimistic and my fingers are stupid. People work in different ways. I find measurement systems to be very useful, simulation systems are only as good as the algorithms and data. Also, simulations are good to try out new wacky ideas before cutting material, but the system isn’t done until it’s built and measured in an actual listening room.
I, too, see some people using a crossover simulator to decide whether drivers will work together.
It's not really the question though, to see whether you can straighten out a response trace with some components, and I usually forgo this step. Their acoustic compatibility requires some thought as to their sizes, where is breakup, baffling etc.. Some would say it begins with a need dictated by the room and you have to fit the design to it.
It's not really the question though, to see whether you can straighten out a response trace with some components, and I usually forgo this step. Their acoustic compatibility requires some thought as to their sizes, where is breakup, baffling etc.. Some would say it begins with a need dictated by the room and you have to fit the design to it.
As far as electrical simulation, there's no problem with the accuracy of any of the crossover sims in general. At audio frequencies the calculations are straightforward.
It could be. Parasitic inductance and resistance are regularly measured, or at least anticipated when building most equipment. It is especially a concern at radio frequencies, but also in audio amps where much higher frequencies are able to cause issues, like oscillation.
Much higher frequencies have virtually no effect on speakers and there's no necessity to worry about it. However on those components where for example ESR (resistance) of a capacitor will cause some variation in response, it can be simulated, as can inductance. You just don't see it factored in most of the time.
Much higher frequencies have virtually no effect on speakers and there's no necessity to worry about it. However on those components where for example ESR (resistance) of a capacitor will cause some variation in response, it can be simulated, as can inductance. You just don't see it factored in most of the time.