In which case why aren't DIYers using similar simulation software to the engineers at KEF, Genelec, Neumann, etc... in order to understand and design speakers with comparable technical performance? Simulation has a massive con for DIYers in that learning how to use it competently requires learning a substantial amount of science and maths (engineering) which most quite understandably don't want to do when it comes to a hobby interest. So they substitute measurements (an easier but less useful alternative when one is unable to apply the science) along with simple design rules (e.g. TS lumped models) supported by empiricism to design speakers. It works but limits the quality of the loudspeaker designs.
Also, these days there is no need for expensive software anymore.
They all work fine, reliable and accurate.
So you can start with REW (I personally find the user interface absolutely horrible).
Or use something like ARTA.
For simulating the crossover filter, use whatever you like.
There are so many free choices these days.
VituixCAD is great because it can do almost everything. The learning curve can be a little steeper, although that also has to do with a user interface that is very confusing at moments.
They all work fine, reliable and accurate.
So you can start with REW (I personally find the user interface absolutely horrible).
Or use something like ARTA.
For simulating the crossover filter, use whatever you like.
There are so many free choices these days.
VituixCAD is great because it can do almost everything. The learning curve can be a little steeper, although that also has to do with a user interface that is very confusing at moments.
Ehm?
Most or at least a lot DIY'ers ARE using such software programs?
For many designs there is also no or very little math required.
There is no substitution for measurements, unless you count copying and rebuilding and existing design.
Almost no DIYer uses CAE software to design their loudspeakers whereas the more engineering-orientated loudspeaker manufacturers all do.
Understanding the assumptions/maths in the software is required because without it the user doesn't know what is correct and what is incorrect in the simulations (or measurements in many cases). This assumes quality in a real sense matters to the user. If not they can simply accept everything or reject everything depending on how they are feeling about things at the time.
Simulation is the primary tool in engineering because it can do a lot more than measurements in terms of quantifying all the physics involved not just the values of a few quantities. It also doesn't require what is being studied to exist which is significant when considering the large numbers of optimisations/trade-offs involved in a high quality design. Measurement is a secondary tool for checking simulations and for getting a handle on things less amenable to simulation such as sound perception. Of course, if the user opts not to use simulation software then pretty much everything has to be measured with a lot of hand waving and guessing involved due to the lack of information.
@andy19191
My experience as a DIY'er as well as a professional is totally different.
A big portion of DIY'ers I know do actually use CAE software.
Not only that, but their approach, way of working and knowledge about these software programs is way better than quite some (big) professional companies.
I have seen quite a few of those companies in the last 15 years.
Understanding physics and maths behind certain aspects most certainly helps, that is 100% true.
But it's not always needed to be able to work with the programs and measurements we are talking about.
That's why just a good steady measurements protocol is such a good fallback method.
Everyone can do this, even in a small appartement.
Unless you are going to design super unusual designs, there won't be anything (or very little) out of the ordinary.
But without proper knowledge you can't really approach these designs anyway.
Also we are not talking about any BEM/FEM methods here, so "simulating" here is extremely basic to begin with.
In this case we are mostly talking about simulating the crossover filters with actually measured responses.
So can you explain what you mean with "everything has to be measured with a lot of guessing"?
Because I don't see a lot of that for a average loudspeaker design.
My experience as a DIY'er as well as a professional is totally different.
A big portion of DIY'ers I know do actually use CAE software.
Not only that, but their approach, way of working and knowledge about these software programs is way better than quite some (big) professional companies.
I have seen quite a few of those companies in the last 15 years.
Understanding physics and maths behind certain aspects most certainly helps, that is 100% true.
But it's not always needed to be able to work with the programs and measurements we are talking about.
That's why just a good steady measurements protocol is such a good fallback method.
Everyone can do this, even in a small appartement.
Unless you are going to design super unusual designs, there won't be anything (or very little) out of the ordinary.
But without proper knowledge you can't really approach these designs anyway.
Also we are not talking about any BEM/FEM methods here, so "simulating" here is extremely basic to begin with.
In this case we are mostly talking about simulating the crossover filters with actually measured responses.
So can you explain what you mean with "everything has to be measured with a lot of guessing"?
Because I don't see a lot of that for a average loudspeaker design.
The most useful feature of a sim program for an average DIYer is to be able to come up quickly with a workable crossover filter and a set of graphs that one can trust to be ok, assuming proper simulation setup. Better than that is a stash of passive parts to try listening to any conceivable sim proposal and not just stop at some point, based on pretty simulated graphs.
I can evaluate the proper spl summing at a crossover frequency by the way an mls signal sounds.
I can evaluate the proper spl summing at a crossover frequency by the way an mls signal sounds.
The simulation can be no better than what you simulate with, i.e. the measurements.
Use a decent analog measuring mic (no USB) and a dual-channel sound card.
Previously (almost 20 years ago!) I was using the LspCAD package, it is convenient in that it contains its' own measurement tool.
Then switched to VituixCAD and ARTA. VCAD is the most complete package IMO, and once mastered the results are 99% as simulated.
REW and HOLM IMpulse are convenient for final checks as well.
One does have to dedicate some time in order to learn (whichever SW one chooses).
Use a decent analog measuring mic (no USB) and a dual-channel sound card.
Previously (almost 20 years ago!) I was using the LspCAD package, it is convenient in that it contains its' own measurement tool.
Then switched to VituixCAD and ARTA. VCAD is the most complete package IMO, and once mastered the results are 99% as simulated.
REW and HOLM IMpulse are convenient for final checks as well.
One does have to dedicate some time in order to learn (whichever SW one chooses).
We are blessed, for the speaker often isn’t the weakest link. If one would start from scratch designing a ‘perfect’ sound reproduction system, one would incorporate the complete environment of the listener. We stop mostly at the boundaries of the speaker system. You don’t need to simulate a lot to get that part of the equation right.
Big companies use CAE more often than not to reach certain levels of cost efficiency. The ‘real’ technical performance seldom is the selling proposition. That even is true for Genelec and Neumann, since even a lot of studio buffs believe (a lot) more than they know (I shall exclude the forum members here, ok). Well, they know how to make music that sells, but that rarely requires a perfect sound reproduction system.
Big companies use CAE more often than not to reach certain levels of cost efficiency. The ‘real’ technical performance seldom is the selling proposition. That even is true for Genelec and Neumann, since even a lot of studio buffs believe (a lot) more than they know (I shall exclude the forum members here, ok). Well, they know how to make music that sells, but that rarely requires a perfect sound reproduction system.
Designing a speaker is something of a bucket list item for me. I bought a mic 
of some repute several years ago, not calibrated and made by Behringer . Software for measuring was a bit of an unknown. Time and other factors meant this has just sat there doing nothing.
More recently started looking at software mainly measurement and found REW which can run on Linux. One thing I want to be able to do is measure spl accurately so bought a calibrated USB mic. I do have a cheap USB interface that will take the Behringer but haven't tested the performance of that yet. I have tested my soundblaster usb card. Looks fine apart from the HF end being a touch short right at the top but using an old software package that appears to have it's origin in pentium days. It will run on wine It isn't supported any more. I wonder about latency etc but will be upgrading my PC anyway and it looks like I will be running a low latency version of Linux, The new machine also has win11 on it but I'd rather stay with Linux.
Next, box design. Only option appears to be winsd. It will run under wine on Linux. So have designed a woofer following suggestions of letting it do it's thing then increasing the box volume and tuning the port. This extended the bass response but with a 1.5db drop. The box shape is determined by where it needs to fit. First question is how accurate the box software is. Also the problems with measuring low frequencies but I also have my ears. The speaker will be in our living room and used with a pair of commercial DML speakers I have had for rather a long time. They were used for rears in the past. I fitted them along with their commercial woofer several months ago. No current need for rears. It will be interesting to actually measure their performance. I do know what my ears tell me.
So a run down from a complete tyro who is somewhat aware of the problems but has little interest in spending lots of money on a project like this. Cross overs - there seems to be a number of options on software for that. Face that when I get there. I have wondered about building an amp for testing speakers, A decision to buy one should speed things up, Fosi Audio V1.0G way more powerful than I need but quicker than designing and building a really high quality low powered amp but hopefully adequate.
of some repute several years ago, not calibrated and made by Behringer . Software for measuring was a bit of an unknown. Time and other factors meant this has just sat there doing nothing.More recently started looking at software mainly measurement and found REW which can run on Linux. One thing I want to be able to do is measure spl accurately so bought a calibrated USB mic. I do have a cheap USB interface that will take the Behringer but haven't tested the performance of that yet. I have tested my soundblaster usb card. Looks fine apart from the HF end being a touch short right at the top but using an old software package that appears to have it's origin in pentium days. It will run on wine It isn't supported any more. I wonder about latency etc but will be upgrading my PC anyway and it looks like I will be running a low latency version of Linux, The new machine also has win11 on it but I'd rather stay with Linux.
Next, box design. Only option appears to be winsd. It will run under wine on Linux. So have designed a woofer following suggestions of letting it do it's thing then increasing the box volume and tuning the port. This extended the bass response but with a 1.5db drop. The box shape is determined by where it needs to fit. First question is how accurate the box software is. Also the problems with measuring low frequencies but I also have my ears. The speaker will be in our living room and used with a pair of commercial DML speakers I have had for rather a long time. They were used for rears in the past. I fitted them along with their commercial woofer several months ago. No current need for rears. It will be interesting to actually measure their performance. I do know what my ears tell me.
So a run down from a complete tyro who is somewhat aware of the problems but has little interest in spending lots of money on a project like this. Cross overs - there seems to be a number of options on software for that.
Face that when I get there. I have wondered about building an amp for testing speakers, A decision to buy one should speed things up, Fosi Audio V1.0G way more powerful than I need but quicker than designing and building a really high quality low powered amp but hopefully adequate.
@markbakk We only have to look at Erin's measurements to see what companies
In that case "more often than not" can rather be seen as "more not than often" in some cases.
Worked in the field for so long you will be really surprised.
Maybe even more in the cases that some companies have ALL the resources and still make one big mess.
Worst of all, that is unintentionally in some cases. (aka; we have no clue what we are doing)
In that case "more often than not" can rather be seen as "more not than often" in some cases.
Worked in the field for so long you will be really surprised.
Maybe even more in the cases that some companies have ALL the resources and still make one big mess.
Worst of all, that is unintentionally in some cases. (aka; we have no clue what we are doing)
??? Engineering companies use CFD/CAA/FEM/BEM/etc... because that is what enables the details in a high quality loudspeaker design to be optimised. A configuration can be quickly put together using design methods based on the TS model and equivalents (that's what engineering design methods are for) but a high quality design requires addressing the details of the cabinet, nonlinear effects, sound diffraction, absorption, and much more.
A typical DIYer's cabinet is a good example of what happens when simulation is not used (OK(ish) open source FEM is freely available but not used). The absence of information on the physics of vibrating cabinets leads people to guess and follow what others do if it seems plausible. This has lead to things like large scale bracing in the belief it makes the cabinets quieter, sticking felt on the inside of walls,... When measurements are attempted using rigs the level of wishful thinking and guessing makes the conclusions more misleading than illuminating. In fact it is possible to perform measurements of cabinet sound radiation for low cost but it is very time consuming and requires knowledge of maths to integrate the many required measurements. Only aware of one DIYer that started it but nothing appeared.
Quickly and BEM/FEM can't exist in one sentence. lol
You seem to have experience with these programs?
Last edited:
I tend to agree with Andy19191 with respect to cabinet construction. Trouble is even quite flimsy looking but thoughtful sculpted cabinets can work out good. The industry isn’t helping though, with absurdities such as a lot of high end companies like Wilson present us. Or Troels advocating the use of felt linings. If you follow quite simple rules, enclosures are not that complicated and as I said before, they sure aren’t the weakest link.
Quickly was referring to TS design methods which have to be quick to support parametric studies. Yes I have used numerical methods like FEM, BEM and others as part of the day job since the days of punch cards and paper tape.
I couldn't agree less.
I'd even argue that you don't have to know much of a math to be able to competently design a loudspeaker - especially with help of good simulator programs. You can even simulate Sean Olive's preference rating in Vituixcad which has known positive correlation with subjective impression of how a loudspeaker sound
Dutch&Dutch 8C (arguably the worlds best standmount active loudspeaker in existence) is designed by Martijn Mensink who has no engineering background but a degree in economics and business. 8C begun as a diy project on this forum as a product of love for a hoby and grew to top notch studio monitor that it is today. Not saying that Martijn isn't good at math, but it was not his profession.
Large manufacturers use their own software because they developed it much before REW and Vituixcad existed. And i'm not mentioning Marcel's ATH4 because i'm positive that such a software never existed before. So, thanks to Marcel Batik (ATH4), John Mulcahy (REW) and Kimmo Saunisto (VituixCAD) we now have what was once a privilege for large companies.
Last edited:
That is even an understatement.
We were talking about just generic loudspeakers, not the fancy complicated stuff (as I mentioned before).
For when everything has to be integrated in specially molded design, like cars, intercrom systems, sound distribution systems, yeah sure, go for complicated models.
For just a normal stereo loudspeaker system, there is not such a thing.
Even if there was such a thing, often it's faster, cheaper and less complicated to just build the damn prototype instead of spending days on end modelling stuff.
Those last words are not my own, but come from Tom Magchielse, one of the highest respected engineers in the field and respected member of AES.
So in the end it feels a bit like heavily complicated things.
I am saying that also from my own experience working with FEM/BEM solutions.
As a friend of Martijn, I can say that isn't his strong point
lolUnderstanding and having a good vision about what is important or not, most definitely is!
Keep in mind that a lot of those things rarely develop in a vacuum.
The origin of many of those projects were just a result of many ideas, discussions and debates.
Plenty of very similar topics can be found as examples here on diyaudio.
I personally really miss that time back than.
The discussions were very different, very much more technically focused.
Very constructive and different tone.
Also EVERYONE was measuring back than, papers and literature were going back and forth.
Ironic, because the tools we have these days are cheaper and much easier to get and use.