Audio OmniMic V2 Precision Measurement System
As a speaker designer, getting high resolution measurements has always been an issue for me. I have done a variety of measurements and evaluation methods, ranging from listening only to utilization of RTA based measurement (I used a Behringer 8024 and the 2496 equalizer to perform RTA functions with pink noise), impedance testing, and a variety of others as appropriate, but I never got over the hump of getting proper software, calibrated microphone, microphone preamplifier and having measurements that could be published. I could get a very solid idea of how a device was performing, but only publish graphically if I plotted points manually, and was limited when it came to what resolution I could measure at (which allows for much finer tuning of a device). Given my role as a DIY audio writer here at Enjoy the Music.com, this had to change; enter the Dayton Audio OmniMic V2 Precision Measurement System ($399.99) – an all-in-one software and calibrated USB microphone package.
Before we delve into the details, let me say that my utilization of digital modeling, impedance measurement via the WT3 and utilization of the RTA testing swept across different locations and careful listening was effective to tune loudspeakers to a reasonably high level. There were no big surprises when I began using the Omnimic V2 Precision Measurement System, but the higher resolution was very helpful in refining crossovers and getting things really locked down to the fraction of a dB level. If you want to design loudspeakers, an acoustic measurement suite isn't a full prerequisite; one can design to a fairly high level with theory, modeling, and fairly broadly available simple tools, but if you want to get things done faster and better (more refined), a full measurement rig is needed and if you are wired for simplicity in some things, as I am, OmniMic is a godsend. I'm not averse to getting my hands dirty, but there are only so many sub-disciplines I have time to chase around, and so simplicity in measurement is what I prefer.
Setup was very simple; the software is reasonably intuitive and the more major challenges for you to overcome will be in understanding your test methodology and getting measurements that truly show what you're looking for. Test tones come from either a CD or DVD source (which I ripped to my hard-drive playback system for convenience- very handy when having to repeat a track over and over). You'll use specific tones for each measurement- the software will indicate which tracks are needed. The unit comes with the software disc, the CD and DVD, a miniature tripod (which I haven't found a need for), a USB cable and an included full-size microphone stand, all for about $300 at the time of this article. I also got a longer USB cable and alternate mic stand, neither of which wound up being all that necessary.
Measurement in general can be very error prone, particularly if you don't have the benefit of a large quiet space to work in. Reflections off the walls, floor, or other in-room surfaces can cause severe peaks and dips that are not actually being reproduced by the speaker, as can noise from appliances or even traffic. This is particularly true in the bass, where rooms are extremely resonant and there's a surprisingly large amount of ambient noise in most modern areas. The Omnimi manages the difficulties of testing with the ability to test simple frequency response using gated, ungated, or hybrid approaches. It is fairly common to use gating at a reasonable (one to two meter) microphone distance for at some frequencies and combine that with nearfield measurements for the bass where there are different issues to manage. Gating is important to testing as it allows one to test while minimizing later reflections and their contribution to frequency response; it measures the impulse and cuts off after some amount of time so that there's not sufficient time for reflections to return to the microphone. A hybrid approach is typically optimal, as the shorter the gating window, the less resolution is possible at low frequencies.
In addition to pure frequency response on a given axis (typically one sees on-axis, 30 degrees, 60 degrees, or some similar set of off-axis data), one can do a series of measurements to create a polar plot- this is exceedingly handy in setting up a full picture of how a device disperses into three-dimensional space.
The ability to measure the off-axis effects in a more formal way is very helpful. The preferred method (or most familiar, to me) of looking at it was with polar plots, with the color showing SPL relative to the maximum (Red is maximum) and the vertical center being on-axis, with above and below that design center showing the change in SPL as you move away towards 90 degrees off-axis (perpendicular to the faceplate). To measure off-axis, one must utilize a turntable (similar to a lazy susan) or other method of rotating the speaker about the measurement point in fixed increments; getting measurements with a turntable setup can be tricky. I wound up making a large indexed reference card with the device under test clamped to a frame attached to the back of a chair. This setup got my tweeter isolated for measurement, but there are also baffle mounting setups and other ways to go about it; just be aware that you'll need to rig up something to allow you to turn the speaker incrementally and take measurements at each increment, if you want to work with polar plots. 10 degree increments are generally a minimum with 7.5 or 5 degree preferred for maximum resolution. Rotating the microphone is not optimal though it will give you enough information to correct gross issues, and it's also important to note that the axis of rotation should be the acoustic center of the device- if the distance to the device under test varies with rotation, you will have introduced additional variation into the measurement and the polars will be less accurate. I know the below looks crazy, but the tweeter was centered on the axis of rotation and the cardstock under the chair allowed me to track the incremental change.
As an example, I had a supertweeter to build for a friend who was using some modified Fostex FF225k 8" full-range drivers. The modified drivers peter out at about 10kHz and I'd previously used a ribbon tweeter to fill this in, but those took a walk and so a new solution had to be developed. I'd started out with trying to press my previous tweeter design into service.
This solution didn't work- the constant directivity response "lump" is in direct opposition to trying to filter for just the top octave, unless you want to use relatively steep and/or high Q filters. A ton of testing of various crossovers went in but I just couldn't get it to have the right combination of response shape and reasonable impedance for use with the 45 SETs my buddy uses. Because of the need for high efficiency, I decided to go a different route, using the same Apex Jr. soft dome (again with ferrofluid removed, and careful tweaking and alignment of the voicecoil within the magnetic gap). A lossy "waveguide" would be used to match the narrow directivity of the Fostex (or any 8" driver) within the upper treble, and this waveguide, because the directionality is imposed by absorption rather than with rigid walls and the retained energy, the axial response remains relatively flat. Two variants on this lossy waveguide were tested along with the tweeter mounted without any baffle or waveguide, the three polar plots are below. First is with a regular round faceplate, free air- bad for diffraction and you can see that in the roughness of the response. Second is with a phase 1 prototype lossy collar – you can see the dramatic reduction in off-axis energy, and improved smoothness of the contours. Third is the final variant- a different lossy waveguide, with smoother contoured shaping and more material thickness. As you can see, the device exhibits some diffusion around 4500Hz, but the polar envelope from 10 kHz to 20 kHz is exceedingly smooth, the overall energy off-axis is dramatically lower, and the diffused energy is more than 10dB down from the nominal axial sensitivity. This variant was chosen as the best match for the Fostex; which is very beamy in the upper treble (as would be expected) so the narrow beam of the supertweeter setup is optimal to mate to that. Doing this by ear would have been very challenging, so the clean visualization of my optimal performer simplified the process, despite the need to measure at dozens of increments along the way.
Also shown are the "normal" polars for the finalized supertweeter- you can see the features of the polar in a more familiar style for many readers. Note how closely the responses in the 10 kHz to 20 kHz range track as you move off-axis – this is exceedingly smooth response for this frequency range. Also visible is a deep dip in the off axis response centered around 7 kHz, with some variation based on angle. This deep notch is a very nice filter behavior, and because it's only off-axis that it's present, there's not the phase distortion that normally comes with deep notches. The smooth response is what I hinted at in my "Foaming at the Mouth" article- speakers should always have either very thorough roundover termination, or lossy materials utilized to ensure that diffraction is minimized. These techniques have gone out of fashion in some designs due to cost, complexity, and aesthetic concerns, but diffraction is a serious issue and a major limiting factor for many loudspeaker designs.
The Dayton Audio Omnimic V2 Precision Measurement System doesn't stop at frequency response, there are many useful other functions particularly distortion. This is a terrific feature for pinpointing problem areas and optimizing crossovers. Moving coil loudspeakers almost invariably have a U-shaped distortion profile, where the lowest and highest frequencies have the most distortion due to the movement demands of lows and piston breakup of highs, with a trough in the middle where the driver is in its happy place. This allows a significant amount of leverage in system optimization, where a couple hundred Hz shift of XO point can lower system THD by 10dB or more at those frequencies. There are features for displaying waterfall (decay) graphs, ETC curves, specialized bass decay functions, spectrum analysis and oscilloscope features as well- a broad set of functions that I've only begun to tap. There's definitely a learning curve; you'll have to understand how a loudspeaker functions, and how rooms and loudspeaker interact (in general terms) to set proper gates, and have a feel for what may be an errant measurement so that you don't wind up with measurement artifacts within your results.
Joseph D'Appolito wrote a well-regarded book on measurement, Measuring Loudspeakers, which may help users get the background needed to utilize the advanced features of this platform. There is also a lot of information online, and if you understand acoustics reasonably well you'll be in good shape to start playing and finding out just how well your system performs. I was able to rock and roll with the unit pretty quickly and can recommend it for anyone who's serious about designing loudspeakers but has been put off by the complexity or other issues associated with getting serious measurement capability. This is a wonderful addition to my toolbox and I'm thrilled that projects going forward in these pages will have a more robust description of their performance.
Dayton Audio OmniMic V2 Precision Measurement System