??? Large manufacturers use commercial CAE software developed by software companies like COMSOL or equivalents for detailed design. One or two might have their own TS design method software knocking around but most are likely to use commercial design software like that from Peter Larsen or equivalents. They will have the free software around and some of the one man in a garage setups may even rely on it but the larger companies won't.
Yes. The theory that one must build a structural FEM in order to design a low signature speaker cabinet is not supported by evidence... evidence in this case being the hundreds of successful speaker designs produced over the last couple of decades which did not utilize a structural FEM. The sound transmission and structural resonances of a speaker cabinet need not be optimized, or reduced to the lowest possible level, they simply must be below the audibility threshold of the end user / customer.
A very illuminating example
Boundary element methods (AKABAK/ATH4, COMSOL, others) are very useful in designing a horn or waveguide. These tools can provide insight into the outer moldline cabinet shape and its effect on radiation, for example the effect of cabinet depth, the effect of edge radius on the rear of the cabinet, or for predicting the interference that a woofer cone causes to the tweeter radiation. However, once those insights have been discovered, it is hardly necessary to do BEM analysis on every new cabinet design. Just like it is not necessary to work through the calculus of Maxwell's equations everytime we use an inductor.
If one wants to use numerical methods to the maximum extent possible when designing a speaker project, one should do so, by all means yes. But lets be honest... this person is doing this because they want to... not because it is the only path to success.
j
I don't know if a single negative about using a crossover simulator. The problems they solve, of helping to simultaneously balance frequency, impedance, phase and power across multiple devices is just so much better than the alternatives.
Of course, there are always limitations but those are usually limitations of the user not applying the software correctly, or failing to take something into account, such as off-axis performance or in-room bass gain and interactions. These tools are partners in achieving your goals but that ultimate goal is up to the user, not the software.
Also, we need to separate out the tools for measuring speakers such as REW, OminMic and DATS or whatever the more expensive versions are, vs. the simulation software.
There's at least three stages here. Measurements of drivers, components and boxes, simulation of alternatives, and construction. XSim and Virtuix are in the middle. Lots of opportunity for the speaker maker to do things wrong before or after that have nothing to do with the simulation tools.
Of course, there are always limitations but those are usually limitations of the user not applying the software correctly, or failing to take something into account, such as off-axis performance or in-room bass gain and interactions. These tools are partners in achieving your goals but that ultimate goal is up to the user, not the software.
Also, we need to separate out the tools for measuring speakers such as REW, OminMic and DATS or whatever the more expensive versions are, vs. the simulation software.
There's at least three stages here. Measurements of drivers, components and boxes, simulation of alternatives, and construction. XSim and Virtuix are in the middle. Lots of opportunity for the speaker maker to do things wrong before or after that have nothing to do with the simulation tools.
@andy19191
I don't follow the point you're trying to make?
Yeah, some (big) companies spend a lot of time and money in modelling stuff.
Some companies bought a whole Klippel NFS system, just to use it once every 3 years or so.
Sometimes even concluding the wrong results.
Some companies are really good in it, some companies totally suck at it.
Some companies found a great niche were these things are very usefull.
Some companies use it just as a bad excuse to bump up their prices.
Some companies go way overboard with simulation models.
In the end of the day, tools are only as good as the skills and knowledge (or the lack thereof) of person who has to work with it.
If you every follow Erin's Klippel NFS measurements, there is very little correlation between what fancy simulation software (and budget) companies use and how well the speakers perform.
I even dare to say that this correlation is quite the opposite in some cases.
The original question of this topic is what the pros and cons are of simulation software.
Obviously that is within the context of diy'ing.
I don't follow the point you're trying to make?
Yeah, some (big) companies spend a lot of time and money in modelling stuff.
Some companies bought a whole Klippel NFS system, just to use it once every 3 years or so.
Sometimes even concluding the wrong results.
Some companies are really good in it, some companies totally suck at it.
Some companies found a great niche were these things are very usefull.
Some companies use it just as a bad excuse to bump up their prices.
Some companies go way overboard with simulation models.
In the end of the day, tools are only as good as the skills and knowledge (or the lack thereof) of person who has to work with it.
If you every follow Erin's Klippel NFS measurements, there is very little correlation between what fancy simulation software (and budget) companies use and how well the speakers perform.
I even dare to say that this correlation is quite the opposite in some cases.
The original question of this topic is what the pros and cons are of simulation software.
Obviously that is within the context of diy'ing.
Any of these methods work with the garbage in = garbage out principle.Yes. The theory that one must build a structural FEM in order to design a low signature speaker cabinet is not supported by evidence... evidence in this case being the hundreds of successful speaker designs produced over the last couple of decades which did not utilize a structural FEM. The sound transmission and structural resonances of a speaker cabinet need not be optimized, or reduced to the lowest possible level, they simply must be below the audibility threshold of the end user / customer.
A very illuminating example
Boundary element methods (AKABAK/ATH4, COMSOL, others) are very useful in designing a horn or waveguide. These tools can provide insight into the outer moldline cabinet shape and its effect on radiation, for example the effect of cabinet depth, the effect of edge radius on the rear of the cabinet, or for predicting the interference that a woofer cone causes to the tweeter radiation. However, once those insights have been discovered, it is hardly necessary to do BEM analysis on every new cabinet design. Just like it is not necessary to work through the calculus of Maxwell's equations everytime we use an inductor.
If one wants to use numerical methods to the maximum extent possible when designing a speaker project, one should do so, by all means yes. But lets be honest... this person is doing this because they want to... not because it is the only path to success.
j
Or in other more practical words, one ALWAYS has to verify the results with measurements.
As I mentioned before, this can be a VERY big deal when having to a design a VERY expensive mold.
They cost around 20k-30k per mold. So hell yeah that you want to take no weird changes that will be totally off.
The very vast majority of even the very expensive loudspeakers are just wooden boxes.
So in the end it's just the path of diminishing returns.
It takes a lot less time and effort to just spit out a whole bunch of prototypes.
Especially when CNC machines or even prototyping companies are so cheap these days.
Even for waveguide/horn design the variations are very limited.
Most of the time we just want a constant directivity.
So like you said, once the insights are discovered (plenty of literature about that), it's about making a few iterations.
For a commercial product this is very often determined by costs and aesthetics.
IF (!!) companies already go that far, because most of them just don't even bother.
At this point it's nothing more than just splitting hairs in the context of loudspeaker making.
Even more so from a DIY point of view.
I feel I need to address another point of disagreement. I have asserted that the measurement process is integral to the simulation process. I am not saying they are the same, but both must be done correctly. Simulation software usually demands that measurements be made in a certain way, and the designer must be cognizant of this. The designer must be fully aware of the limitations of the measurements (accuracy, precision, measurement uncertainty) so that the simulation process does not produce misleading results.
I can not imagine any industrial engineering situation where numerical analysis does not begin with data which was empirically determined... i.e the simulation uses measured data.
I can not imagine any industrial engineering situation where numerical analysis does not begin with data which was empirically determined... i.e the simulation uses measured data.
I am chatting about the pros and cons of engineering simulation software which most here are unfamiliar with.
A bit on the pro side but relatively little on the home audio side. Home audio speakers are low tech and strongly marketing rather than engineering orientated to the extent that nearly all large companies that want to use their engineering capabilities to add value have pulled out of the sector. Some like Sennheiser are still around on the pro side. It leads to some pretty weird stuff.
This is a measurement system not simulation software. It's pros and cons are also misunderstood by many DIYers who lack an engineering background in industry.
I don't recognise this. Can you give some examples?
What have Klippel measurements got to do with fancy simulation software? I don't understand the link. How well the speakers perform in terms of profit or by some other criteria?
PS If I include @b_force will this get round the problem of not being able to quote the previous post?
Seriously?
The link is that fancy and expensive tools don't (always) make a good product.
Good in terms of technical performance, not in how much money you throw at marketing.
Still I am missing the practicality of your arguments?
Quite literally, what is the point you're trying to make in the context of loudspeaker development?
A good portion of people IS familiair with simulation software?
The other portion doesn't need all this information for just designing some home hifi loudspeakers?
So again, I am missing the point.
At this stage it's splitting hairs.
The fact that you don't recognize that companies don't go overboard with simulations, totally blows my mind.
I can draw a direct correlation with the size of the companies I worked with and how much time and money is wasted.
I can't give any direct examples without disclosing some IP or NDA thanks to the IP and ego trolls. (read: we make people believe we do stuff, but we actually don't do anything except selling expensive stuff)
This is basically a known fact for anyone I know in the industry as well.
I don't understand this. Are you saying an engineer would start by constructing a prototype and measuring it before performing simulations (which seems unlikely). Or perhaps measuring a set of drivers and then simulating with that? Or the stress-strain relationships of a material approximated by a continuum? Or something else?
One aspect of CAD that has not been mentioned here is crossover emulation. To my knowledge, LspCAD and SoundEasy are able to do this. That is, to hear how your crossover will sound before building it. SoundEasy calls it a Digital Filter, and I have used it a fair amount of times and I can report that it is accurate.
From SoundEasy website:
May I remind you that we are talking about low tech constructions (your words)?
So, stress-strain relationships, fancy beforehand simulations? eh?
Again, maybe if you want to go super fancy.
But after countless designs I made and consulted for, I am not gonna bother for just a average loudspeaker design.
Which 90% of the industry is (if not more)
Mostly because dimensions are most of the time already a constraint and not a variable anymore.
Otherwise one could do some bafflestep simulations at best.
The rest is mostly just relying on (third party) measurements of the performance and (off-axis) response of the loudspeakers.
Like @markbakk rightfully mentioned, there are things that are WAY more significance than those minuscule details.
Iterations or mistakes can be much easier done in practice.
Any preliminary research and calculation can be often be done with ballpark figures from the classical acoustic science and engineering (in combination with experience).
But again, if you have done so many designs, you're not gonna bother with those things anymore since they are so extremely predictable in an average loudspeaker design.
A free program like LTSpice can do this as well, by just importing and exporting a wave file.
Or just use your favorite DSP or even just PC (Equalizer APO or VST) for FIR filtering.
All free or very cheap options that do the same thing and people have been using for years.
I was speaking to a more general case. All numerical methods begin with data which was measured by someone. The modulus of elasticity of some material, the coefficient of thermal expansion, the density of air, etc... all of these and hundreds of other datum must be measured, and how it is measured matters... it matters a lot.
So in terms of speaker design, yes, a driver must be measured in order to begin simulating the system. And by simulation in this context, I mean something like VituixCad, which allows us to simulate the effects of crossover, driver spacing and rotation in all 5 axis, and baffle shape. The primary aspect of speaker performance is the on/off axis frequency response, and VituixCad is an engineering tool which handles this quite well. A structural FEM of a cabinet would help in dealing with cabinet signature, which is 2nd or 3rd order importance.
Actually Andy, I don't really expect you to understand this. Over the years I worked with many excellent FEM experts. These people were at the top of their game, but many of them had a huge blind spot when it came to other aspects of engineering the product. They often felt that the customer requirement for a full scale test of a prototype was a waste of time and money... afterall, the FEM shows the structure to meet the requirements. And sometimes the testing revealed nothing, but most of the time there were surprises. The FEM guys were often shocked when analyzing field failures because the failures revealed unknown load paths. The FEM guys tended to be fairly insulated, so they were not exposed to the full range of engineering expertise that had to be applied to make a successful product. They typically, albeit naively, thought that their field of study (FEM) was complex, difficult, and important, while all of the other engineering activity was trivial and simple. All engineers are prone to this conceit, but it seemed a bit worse among engineers who more insulated, which in a complex project the FEM folks tend to be.
1+ for LSPcad. 15 years ago, when I made first contact, LSPcad was a breakthrough:
- real-time parameter/value changes have immediate effect, playing music or whatever through the engine
- there is a very capable parameter/value optimizer
However, like others have said, even with the best tools you cannot avoid the occasional garbage in --> garbage out incident and smoking drivers ;-)
Numerical methods begin with scientific laws like the conservation of mass, momentum and energy. These are expressed mathematically without empiricism (long ago measurements were performed to check these observations held). It is the introduction of assumptions like a continuum that introduce empirical coefficients (if we calculated every particle we wouldn't need them). Many empirical coefficients are sufficiently accurate that it is common to use them with direct numerical simulations to calibrate measurements systems (e.g. hot wires and laser anemometers to measure turbulence).
So a company would not start with the concept of a speaker design, run simulations to size and optimise the drivers, go and get drivers with those design parameters, build a speaker, measure, iterate the design,... They have to start with physical drivers they can measure? Or do you mean DIYers who are not interested in simulation need to start with the measurements of a driver? (This I would agree with).
Quickly sorting out the overall configuration is the purpose of design methods using fairly large assumptions.
Optimising details of the design is the purpose of design methods using low levels of assumptions. This is where CFD, FEM, CAA, BEM, etc... come in. If the increase in quality this can bring is not considered worth the cost then it can be omitted.
Hmmm... I think this needs some context given I have never come across anyone that has suggested not checking a relatively complex device after it has been manufactured. I briefly worked in an aerospace stress office in the late 70s and they were aware of their relatively low status particularly as internal software development was turned off in favour of commercial software. I think after I left they were reorganised with many people placed in other departments to act as semi-technicians providing local stressing services. Some of the tricky nonlinear stressing would have remained together.
What I have observed though is the decline of the ebullient cut metal and measure approach to be replaced by a more considered one based on simulation with substantially reduced prototypes and measurement. When I first started the process was well underway in high tech engineering and was causing noticeable friction in my first company between the military side (smaller cheaper and mainly cost plus funded) hanging onto the old ways with the civil side (larger more expensive and more competitive) being forced to adopt more efficient and effective ways. In the late 80s and 90s I observed medium tech industries like automotive going through a similar process. Today the larger more engineering-orientated companies in low tech industries like loudspeakers are following. It is optional though in an industry like loudspeakers where engineering adds relatively little value. Indeed a fair few companies have little internal engineering design capability using external consultancies with their internal resources focused on manufacture, marketing and such which contribute more to a profitable product.
@andy19191
It all stays very vague, can you maybe provide an example why those FEM/BEM simulations were absolutely essential that could not be done by just practical hands-on prototyping or predicted beforehand with just general knowledge and understanding of (general) acoustics?
In the context of loudspeaker design obviously.
It all stays very vague, can you maybe provide an example why those FEM/BEM simulations were absolutely essential that could not be done by just practical hands-on prototyping or predicted beforehand with just general knowledge and understanding of (general) acoustics?
In the context of loudspeaker design obviously.
I gave an example earlier: the sound radiated from a speaker cabinet. This is difficult for a DIYer to measure directly because it masked by the sound from the drivers. One or two have had a go at creating rigs but with little success. It can be done with a laser vibrometer, a lot of time and some maths (or an accelerometer and a huge amount of time) but I am not aware of any successful DIY attempts only ones in labs.
A FEM/BEM simulation of a cabinet with it's sound radiation is straightforward with commercial software. It is reasonably so with free open source software although I am not aware of one that properly supports efficient modal analysis for viscoelastic materials (one makes the claim but the implementation isn't general). A short while viewing animations of the modes, how they radiate and how they are driven will reveal a significant amount about the physics of what is going on. A few parametric changes to stiffness and damping will soon reveal what is effective and what is not. It is a lot more useful for sorting out a cabinet design than a set of measurements (assuming access to a scanning laser vibrometer) because it quantifies all the physics involved everywhere not just at a few quantities at a few places. It is also a lot quicker (assuming one has learnt how to use the software effectively) since various cabinets don't need to be manufactured and measured.
A FEM/BEM simulation of a cabinet with it's sound radiation is straightforward with commercial software. It is reasonably so with free open source software although I am not aware of one that properly supports efficient modal analysis for viscoelastic materials (one makes the claim but the implementation isn't general). A short while viewing animations of the modes, how they radiate and how they are driven will reveal a significant amount about the physics of what is going on. A few parametric changes to stiffness and damping will soon reveal what is effective and what is not. It is a lot more useful for sorting out a cabinet design than a set of measurements (assuming access to a scanning laser vibrometer) because it quantifies all the physics involved everywhere not just at a few quantities at a few places. It is also a lot quicker (assuming one has learnt how to use the software effectively) since various cabinets don't need to be manufactured and measured.
And the flip side, lots of commercial speakers where the speaker cabinet is not nearly as good as what DIYers do, because we have time on our hands and don't have to cut costs to make a profit.
I don't know what his degree was, but an undergraduate degree in economics in Europe usually gets through differential equations pretty quickly but very little math for an undergraduate in the US outside of the top 20 programs. A graduate degree in economics requires quite a bit more math than engineering. (Although, not necessarily the kind of math useful for speaker building in some cases, like optimal control theory, stochastic optimization, real analysis and topology.)
I agree that VituixCad can be a nice tool. I would not call it a pure "simulation" tool, though. Most of the things it does is to process measured data from single drivers and calculate the behaviour of a loudspeaker system.
VituixCad may be free as in "free beer", but not free as in "free speech". Vituix is closed source, and it's hard to really know what the software is doing under the hood. The manual is very extensive, but is often lacking the precision and details needed to fully understand the functionality of the software. I had some lengthy discussions about this with the VituixCad author when it came to the method of calculating the power response. In the end I wrote my own code (open source) following published techniques. The output was similar to VituixCad, but not identical, although the VituixCad author claimed the method implemented in Vituix was the same.