How to Design Loudspeakers without Performing Measurements

[LouC]

Re: How to Design Loudspeakers without Performing Measurements

Designing speakers without measurements:
In my opinion, the notion of developing a quality speaker design without making even one set of measurements is silly, and its validity is wishful thinking...

I agree to a point. I think you can get pretty close with plain simulations. And probably use them for "feasibility" purposes. But individual lots of drivers vary, and, IMHO, there's no reason to believe anybody else's measurements. If some folks don't trust the factory, why trust somebody else's measurements in their garage? (Me included).

I've had good sims produce marginal results. Also had bogus measurements, defective drivers,.... Sometimes, I just go back and start over. I've also knocked out a simple two way in a few hours including measuring and crossover design. In the box measurements save many hours of fiddling since all the baffle, offsets, and dimensional conditions are captured in the measurements.

I think a lot depends on expectations. If it's $30 worth of buyouts playing hyper compressed MP3 death metal, then folks can be happy with good enough.

Can you do a high quality design without measuring? Possibly. Having Jeff's skills probably increases your chances.;) Or having fun tweaking and changing may be it's own reward. Better quality drivers will usually behave a bit more predictably too. If you have time and no money, it's probably justified.

[[email protected]]

Re: How to Design Loudspeakers without Performing Measurements

I'll put the tutorial on the long list of things I'd like to work on someday. The existing User Guide inside the Response Modeler already explains the basic steps though, and that's been there for quite a while.[/QUOTE]

yes, please do. im still having trouble with the proper order of steps in rm. a tutorial would be great.

[DDF]

Re: How to Design Loudspeakers without Performing Measurements

Thanks Dave - If I use SplView and normalize again, I'll definitely use Excel - but for now I'm using Jeff's tools and finging the process to be much simpler.

Don, no problem,I'm glad you found an easier way forward that works for you!

[donradick]

Re: How to Design Loudspeakers without Performing Measurements

Don:
1. With the method I show, take the Unibox file and put it through SPL viewer so it has the same the frequency min/max and frequency step sizes as the other files (see slide Prepare Woofer FRDs).
2. My technique uses excel, not the FRC.

Jeff,
I felt Unibox was a bit more accurate for niche designs, based on data we've compared in the past, but for 95% of uses, I'm sure there is no difference.

Part of my reason for choosing these tools is my having verified them personally and perhaps a preference with the UIs. I'm sure your tutorial would be a big hit, and arrive at a more straight forward technique. I'd say go for it!

Dave

Thanks Dave - If I use SplView and normalize again, I'll definitely use Excel - but for now I'm using Jeff's tools and finging the process to be much simpler.

[donradick]

Re: How to Design Loudspeakers without Performing Measurements

Thanks guys! Soon after I posted that message, I gave up on FRC, and found Paul Carmody's excellent tutorial using Response Modeler linked in the sticky "Bible" thread. Wow - much easier using your sw, Jeff! I only have one question remaining - how to "dial in" Baffle Step Compensation in PCD?

[Jeff B.]

Re: How to Design Loudspeakers without Performing Measurements

Don:
Jeff,
I felt Unibox was a bit more accurate for niche designs, based on data we've compared in the past, but for 95% of uses, I'm sure there is no difference.

Have you compared to the current version of WB&CD? I believe they are based on the same math. I have put them through some very extensive comparisons and results appear to be essentially identical on all of them. I set up a file that allowed me to dump results from both into it and plot the differences. I considered Unibox the reference and used it to gauge how mine was doing. The results showed both to be the same to high level of precision. Mine doesn't do the line resonance, but I kinda question that part of Unibox anyway.

I'll put the tutorial on the long list of things I'd like to work on someday. The existing User Guide inside the Response Modeler already explains the basic steps though, and that's been there for quite a while.

[DDF]

Re: How to Design Loudspeakers without Performing Measurements

I'm working an a budget MT design using this guide, and have come to a sticking point on Slide 11 - using the Frequency combining worksheet and Splview normalizing on page 10.
1: While most FR plots cover 20Hz to 20KHz, Unibox puts out FRDs that end at 1000Hz. Is this a problem? Should I choose 20KHz to normalize the Unibox file also?
2: for the FRC worksheet, there are a number of sections that do not coorsespond directly to the DDF explanation on Slide 11. Can you tell me which section to load each source FRD into? (Driver base is obviously the manu frd, but which section for BDS transform - "Baffle Measure" and Unibox frd - "Box Measure"?)

Thanks again for the guide. I never really understood all the tools and the data flow through them as input to PCD.

Don:
1. With the method I show, take the Unibox file and put it through SPL viewer so it has the same the frequency min/max and frequency step sizes as the other files (see slide Prepare Woofer FRDs).
2. My technique uses excel, not the FRC.

Jeff,
I felt Unibox was a bit more accurate for niche designs, based on data we've compared in the past, but for 95% of uses, I'm sure there is no difference.

Part of my reason for choosing these tools is my having verified them personally and perhaps a preference with the UIs. I'm sure your tutorial would be a big hit, and arrive at a more straight forward technique. I'd say go for it!

Dave

[Jeff B.]

Re: How to Design Loudspeakers without Performing Measurements

I'm working an a budget MT design using this guide, and have come to a sticking point on Slide 11 - using the Frequency combining worksheet and Splview normalizing on page 10.
1: While most FR plots cover 20Hz to 20KHz, Unibox puts out FRDs that end at 1000Hz. Is this a problem? Should I choose 20KHz to normalize the Unibox file also?
2: for the FRC worksheet, there are a number of sections that do not coorsespond directly to the DDF explanation on Slide 11. Can you tell me which section to load each source FRD into? (Driver base is obviously the manu frd, but which section for BDS transform - "Baffle Measure" and Unibox frd - "Box Measure"?)

Thanks again for the guide. I never really understood all the tools and the data flow through them as input to PCD.

I know Dave put a lot of work into his tutorial, but he chose a much more diffcult path for modeling. I'm not sure why, I suspect he just went with what he was familiar with. However, there is a much easier way to do it, that's more intuitive too.

I designed the Response Modeler in an effort to consolidate most of these steps in a simple, but accurate format. I never use SPL trace, but I do some modeling or drivers that I don't have. I use the features in the upper section of the RM to progressively add detail and extend the response to create a reasonable facsimile of a driver's response curve - some with a lot of detail.

Next I model the box response, adjust its level to match, and splice it on. While doing so, I model the impedance curve. I spent a lot of time developing the math to very accurately reproduce the affects of Le that changes with frequency like in real loudspeakers. You will find my modeled impedance to be capable of very accurate results.

After the box response I model the cabinet diffraction and add that to the response. Finally, I save this out as an frd file and the impedance as a zma file and then extract the minimum phase from them.

It's simple and it works very effectively.

I am not sure why Unibox was chosen for this process. It is an excellent box modeler, but my WB&CD uses the same math an produces results that match for most modeling. However, my WB&CD can be adjusted for any frequency range, even 10-20kHz and includes my impedance algorithm for a more accurate Le. I'm not really pleased with the way it is described on the webpage linked above. I think I need to ask Charlie to change that. I don't like it being thought of as a "Subwoofer designer". That's why I changed its name a few years ago to remove Subwoofer from the title. It models subwoofers, but it can model tweeters too.

I guess I need to create a tutorial for using my software together, as I intended it to work, and how I use it myself. More than one view can sometimes be helpful.

Jeff B.

[donradick]

Re: How to Design Loudspeakers without Performing Measurements

I'm working an a budget MT design using this guide, and have come to a sticking point on Slide 11 - using the Frequency combining worksheet and Splview normalizing on page 10.
1: While most FR plots cover 20Hz to 20KHz, Unibox puts out FRDs that end at 1000Hz. Is this a problem? Should I choose 20KHz to normalize the Unibox file also?
2: for the FRC worksheet, there are a number of sections that do not coorsespond directly to the DDF explanation on Slide 11. Can you tell me which section to load each source FRD into? (Driver base is obviously the manu frd, but which section for BDS transform - "Baffle Measure" and Unibox frd - "Box Measure"?)

Thanks again for the guide. I never really understood all the tools and the data flow through them as input to PCD.

[raiderone]

Re: How to Design Loudspeakers without Performing Measurements

Every system I've designed for the last 25 years has been based on acoustic measurements. I've taken thousands, in and out of anechoic chambers. I've contributed to IEEE standards on it.

But not everyone is that fortunate, either in dollars, or time to learn all the ins and outs. I put this together for those without access to measurement gear, so they can still get well within voicing range, enjoy the hobby and make themselves a good sounding design they can be proud of. No shame in that.

Dave

Thanks. And even if one does have the measurement equiptment, most do not have the resources to buy and measure every driver contemplated for a build. This resource is also useful in selecting potential drivers for a build.

[DDF]

Re: How to Design Loudspeakers without Performing Measurements

Every system I've designed for the last 25 years has been based on acoustic measurements. I've taken thousands, in and out of anechoic chambers. I've contributed to IEEE standards on it.

But not everyone is that fortunate, either in dollars, or time to learn all the ins and outs. I put this together for those without access to measurement gear, so they can still get well within voicing range, enjoy the hobby and make themselves a good sounding design they can be proud of. No shame in that.

Dave

[fntn]

Re: How to Design Loudspeakers without Performing Measurements

Designing speakers without measurements: http://www.youtube.com/watch?v=crPp7LOduJM

In my opinion, the notion of developing a quality speaker design without making even one set of measurements is silly, and its validity is wishful thinking.

You either need to measure to design, or you need to measure in order to validate the simulations. But if you don't measure, you'll never know if your design came out right, and might be caught with your fly unzipped at a DIY event or when a more well-traveled audio enthusiast stops by your place with some of his own source material for a listen. Every serious designer needs to have access to a means of making FR magnitude measurements and impedance measurements (a mic and a WT3 suffice), and they are as essential to diagnostics as to design, so it makes sense to consider making the investment.

That said, with the kind of accuracy one can manage with a simulation, and with the difficulties in setting up a space in most rooms that can make good measurements of a speaker, I'm leaning heavily toward the second option these days: measure to validate. It's less work overall.

My $.02

+1

[fntn]

Re: How to Design Loudspeakers without Performing Measurements

Thanks Gowa and Rick. Rick, I'd be happy to field any questions at the next DIY.

fntn, use the approximations shown in the slides. They're close enough. If you have other driver sizes to worry about, measure where the voice coil attaches to the cone and use that as the approximate recess behind the baffle for teh driver acoustic center. It will get you close enough. If it doesn't, it means that your xover has really tight lobes with nulls near on axis, never a great thing anyway (ie you'd have bigger issues in that case).

:D I was just joking about the thread title - "without performing measurements". I use impulse recordings to get precise measurements of raw drivers in the baffle.

[Taterworks]

Re: How to Design Loudspeakers without Performing Measurements

Designing speakers without measurements: http://www.youtube.com/watch?v=crPp7LOduJM

In my opinion, the notion of developing a quality speaker design without making even one set of measurements is silly, and its validity is wishful thinking.

You either need to measure to design, or you need to measure in order to validate the simulations. But if you don't measure, you'll never know if your design came out right, and might be caught with your fly unzipped at a DIY event or when a more well-traveled audio enthusiast stops by your place with some of his own source material for a listen. Every serious designer needs to have access to a means of making FR magnitude measurements and impedance measurements (a mic and a WT3 suffice), and they are as essential to diagnostics as to design, so it makes sense to consider making the investment.

That said, with the kind of accuracy one can manage with a simulation, and with the difficulties in setting up a space in most rooms that can make good measurements of a speaker, I'm leaning heavily toward the second option these days: measure to validate. It's less work overall.

My $.02

[DDF]

Re: How to Design Loudspeakers without Performing Measurements

Thanks Gowa and Rick. Rick, I'd be happy to field any questions at the next DIY.

fntn, use the approximations shown in the slides. They're close enough. If you have other driver sizes to worry about, measure where the voice coil attaches to the cone and use that as the approximate recess behind the baffle for teh driver acoustic center. It will get you close enough. If it doesn't, it means that your xover has really tight lobes with nulls near on axis, never a great thing anyway (ie you'd have bigger issues in that case).