How accurate is speaker design software? Is it just a starting point?

jaminjames

Member
2020-01-28 1:17 am
I'm a novice acoustic designer - and one thing any pro will tell you, is that acoustic modeling software is just the starting point. It will give you a very rough idea of what's going to happen. For instance, XYZ absorber might have an absorption coefficient of .8 at 65Hz in the software. Once it's in the room, it could be .8, or it could be .5. There are too many factors to make more than a rough estimate.

Is loud speaker design software the same, or is it a more accurate type of modeling?
 

conanski

Member
2013-03-31 3:53 am
Software will produce an accurate representation of the data it is given, if that data is incomplete or the end user implies conclusions beyond the scope of what the software was designed for the software isn't at fault. If the input data is correct and complete the output will be accurate for a speaker in an anechoic environment, that means outside and well away from all boundaries including the ground.. unless otherwise specified. Indoor spaces impart numerous effects on a sound systems response including boundary loading, reflections, constructive and destructive interference, and distortion(noise) from sympathetic resonances.. things vibrating, none of which speaker design software takes into account.
 
Last edited:
It would help if you could spell out your question in a more general technical terminology.

I think in your example of XYZ absorber, you really are suggesting that (1) the software library design parameters of devices deviate from actual production samples, or (2) design parameters of devices are just static estimates that do not reflect actual dynamic values during real use.
 
Most of the more known software is rather precise when used right, but the input data is very important, so measure your drivers as the published datasheets are averages. There are minor variations within the same model, and they can (but mostly won't) have an influence. If you measure your speaker, you have the exact data of that driver so you can model it right.

When you model with the data of timesheets, you will have a good impression what it will be, but with that data, it won't be very tuned to the driver you are using. idem with datasets from sites like Loudspeaker Database or collections of files from the net.

So measure your drivers if you want to be very precise. It can be done with inexpensive tools like the Dayton DATS (150€) & REW (freeware) and a measurement mic like the Dayton Audio UMM-6 USB (80€). Or you can use the expensive and complex pro audio tools (i'm not very familiar with it) who are even more precise (but hard to learn). I use DATS V2 and REW and a measurement microphone and usb inteface i already had, and since i use that, my speakers are way more predictable on outcome.
 
Are you talking about crossover Sims or box Sims? I believe both are typically quite accurate with measured parameters. Boxes are quite close generally even with manufacturer's parameters. I've always found my tuning frequencies to end up very close, even with manufacturer's parameters. Even the one horn sub I designed using hornresp measures almost exactly as the sim predicted
 
Think of the decision process this way: when you are done, in what ways will you be better off sim'ing or not? Are the sim'ed predictions useful or just fun to see? Are rules-of-thumb adequate? And all in light of the fact that your listening chair measured results will be barely related to the sim prediction.

It depends on what you are developing. For sure, you can't do back-of-an-envelope calculation for a horn or of a bass-reflex cabinet, or at least I can't (but then I'd never have such a think in a quality audio system).

Crossover should be done on the back of an envelope so that you have a good intuitive feel for it.

b.
 
Crossover should be done on the back of an envelope so that you have a good intuitive feel for it.

b.

Crossovers have a complex job to do, and certainly shouldn't be done on the back of an envelope.

It's not just taking the driver's nominal impedance, picking an arbitrary frequency, and banging those numbers through the equations.

You've got to account for the frequency response and phase angles of the drivers at the chosen crossover frequency (which itself should be a function of relative driver power handling, directivity, etc).

My current HiFi speakers used something like 15 components per crossover, but it was a difficult design to implement: aluminium-cone 8" midbass to a compression driver on a horn. The results, however, are excellent - there's no hint of the strong breakup peaks that typically show up with metal-coned drivers, and the HF horn is nice and smooth.

I'm not sure there's anyone in the world that could come up with a circuit for those drivers without some decent simulation software and some good measurements.

Chris
 
Crossovers have a complex job to do, and certainly shouldn't be done on the back of an envelope.

It's not just taking the driver's nominal impedance, picking an arbitrary frequency, and banging those numbers through the equations.....

My current HiFi speakers used something like 15 components per crossover, but it was a difficult design to implement: aluminium-cone 8" midbass to a compression driver on a horn. The results, however, are excellent - there's no hint of the strong breakup peaks that typically show up with metal-coned drivers, and the HF horn is nice and smooth...

I appreciate your reasoned and courteous tone.

But, the simple fact is that there is no agreement on the right curve, slope, or any other aspect of XO preference. So you may know how to calculate with great precision all your factors (and I don't dispute your physics), somebody else can come to an entirely different preference. So as much as you like your model, others don't*.

Ever since the famous BBC mini-speaker L3/5a (or do I have it backwards), manufacturers have been putting lots of elements into their XOs. They (and you?) do it based on listening and measurements, with fine-tuning not finally based T/S parameters that are sometimes imaginative from driver manufacturers or based on home-brew measurements that are impossible to do with much precision.

So I'd say, do FR runs on the drivers and pick your XO points with good judgment, buy a DSP, try your XO point and slope*, and test and listen. No sim needed.

B.
* Can't say as I've ever been able to hear any difference between Butterworth and any other high-falutin' curve that makes me think one pattern is better and the other worse. Easy to do what-if with a DSP.
 
Last edited:
If we talk about crossover networks, whether passive or active, the correct thread would be Multiway, not subwoofers.

Or are they talking about the Linkwitz transform?

But, the same question arises:


To demand a precision of the TS superiors to the one that averages the manufacturers, does not bring the problem of crossing networks and different tuning for speakers L and R? At least in tuned cabinets, and then, if we use sealed cabinets, why worry? Because those designs admit a volume tolerance that exceeds the differences that can be found in individual TS measurements of the speakers ...
I think, it seems to me ...