Simulated Measurements Part 2

Home > DIY Speaker FAQs > Simulated Measurements Part 2

So how well will these simulated measurements correspond to the real thing?

Let's find out....

I just so happen to have a cabinet the same size as the one I've been modeling, and with a baffle routed for the RS28 and RS150--just a coincidence, probably. I mounted the drivers in the cabinet, put it on a stand in the middle of my office and measured the drivers on-axis at a distance of 1 Meter. I used JustMLS to make the measurements, and used the free version of Praxis to display the graphs. The measurements are smoothed at 1/12 octave.


First, let's look at the tweeter's Frequency Response. The top graph is our simmed FRD file, the bottom is my own measurement.

Not bad, if I may say so. The response is very similar in the driver's passband (let's say, between 1500 - 20,000 Hz). The only noticeable difference is that the drooping top end in my measurement is not as pronounced as it is in the simulation. But in reality, the drooping upper treble response of the RS28 is audible; and would probably show up in a measurement if I moved the microphone a few degrees off-axis.

Next up is the Frequency Response of the Dayton RS150S woofer. The simulation is on top; measurement is on the bottom.

Again, a lot of similarities. First off, notice that the cone's "signature" breakup mode (the 3 mountainous peaks) is about the same in both the simulation and measurement. This tells us that there is a good level of consistency in the cone material from the manufacturer's spec unit to my own. Next, let's look closely at the response of the driver from 200-3000 Hz. The simulation very closely approximates the response of the actual measurement; the only discrepancy is that the baffle step roll-off is steeper in the measurement than the simulation. But still, it's within 1-2 dB difference. In fact, the only noteworthy difference between the two graphs is that the simulation actually accounts for the bass response, where the measurement simply rolls off to infinity below 200 Hz. This goes back to how I was saying that it's very difficult to measure bass frequencies "anechoically" below 200 Hz. (There are tricks to measuring the anechoic bass response such as "ground plane" and "nearfield splicing," however these are advanced topics in measurement). In other words, the simulation has the advantage here, because my measurements are useless below 200 Hz.

OK, moving on. Let's look at the Impedance graphs. First, the Dayton RS28. Traced on top; measurement on the bottom.


Seems very close to me. Give Dayton good marks on driver consistency once again, as my measurements are extremely close to the manufacturer's spec.

Finally, let's look at the woofer's Impedance in an 11 Liter enclosure. Simulation is on top; measurement on the bottom.

Overall, we're still pretty close. There are two points to quibble over. First of all, the vented box tuning--shown by the double humps in the lower frequencies--does not quite match up between the simulation and the measured box, as mine is tuned slightly higher. This is my own fault, however, because I cut the vent 1/2" too short and didn't bother to fix it (hey, no one's perfect!). Second, the inductance rise on the simulation is steeper than the actual measurement. This is caused by the fact that drivers actually exhibit different inductance values at different frequencies, and--believe it or not--you can actually compensate for this in Response Modeler if you want to by using the "Overlay Imported ZMA function." For additional information on how to do this, view the included User Guide in Response Modeler (if you haven't already).

Conclusion

As you can see, this method of "simulating measurements" does work to a good degree of accuracy. Is it perfect? No. Can it get you up and running, designing and modeling your crossover, without an investment in measurement gear? Yes. Will it allow you to design a better crossover than some "off-the-shelf" or "textbook" crossover? Oh My God, yes!

I've said it before, I'll say it again. Designing a loudspeaker is not easy; measurement is a whole another skill. All too often I see eager neophyte speaker designers who want to try their own hands at designing a speaker becoming overwhelmed, stymied, and frustrated when they try to set up their measurement gear and are suddenly faced with a long string of errors, problems, and various unexplainable phenomena. It's sad, because what they really want to do is design and build a speaker, not mess around with device drivers, gain stages, Fourier Transforms, testing jigs, and just plain "unexpected results."

Measurements are a necessary part of speaker design; let's face it, we couldn't simulate these "artificial measurements" without something to start with (manufacturer's specs or 3rd party tests). But measurement can also be a pain in the ass. There, I said it. If you're one of those people who's posted all over the forums, asking, "What am I doing wrong with my measurements?" or "How do I get my measurement gear working?" and no one's responding it's because people either don't know why, and maybe do know why but don't want to deal with it. Most designers who measure regularly have a system that works for them; and once it works--reliably--they stick to it. The rest of the populace is still futzing around with their mic, cables, mixer, sound card, inputs and outputs and getting weird results just like you. So I say, for the time being, go ahead and sim up some "artificial measurements" and go make something! That's the fun part! Then, sometime when you have free time (or you have the masochistic desire to use a driver for which there are no measurements), you can go back and read those user manuals and try and troubleshoot why your setup isn't working. Until then, don't let technical difficulties stifle your creativity.

by Paul Carmody | this page was last updated December 22, 2020