Microphone Frequency Response - The Window
Most microphone manufacturers quote frequency response numbers somewhere on their spec page, and it's usually something like "20 - 20 kHz" (or "30 - 15 kHz"), but what does that really mean, and how does it relate to what you hear?
For that you'll need to know how to read a frequency response curve, add in what they "don't tell you", and understand the amount of deviation possible between identical units. But before we can do that, you need to know how microphones are measured.
Even though computer measurements have replaced a lot of the mechanical measurement systems, companies (like B&K) still provide precision microphone test equipment, consisting of a frequency sweep oscillator, synched to a chart recorder, and a ruler flat test microphone.
Basically, you feed the oscillator signal into something that will generate the sound, hook up the mic you want to test, and the calibrated mic, then sweep the entire audible frequency range while you chart the "difference" between the calibrated flat mic and the mic you're testing. The resulting chart is the frequency response of
that one microphone.
Calibration mics usually come in two flavors: direct measurement mics (on-axis), and diffused field mics (usually 90° off-axis). Direct measurement mics are used in anechoic chambers where there is no sound bouncing around so the mic can be designed to be absolutely flat on-axis (i.e. pointed straight at the sound source). As you aim the mic away from the sound source, the high end response of the microphone drops off dramatically.
Diffused field microphones are used in normal type rooms where pointing the mic directly at the speaker will pick up unwanted reflections. When making measurements with diffuse field mics, they're usually pointed 90° off-axis (towards the ceiling, the floor, or one of the side walls. Diffuse field microphones are flat 90° off-axis, but they have a large rising frequency response on-axis.
So we now measure our mic, using one of the two methods described above and we look at the chart that was produced, but that only tells us about that one mic. . Here's the mic curve for "our mic":
We'll need to run a batch of the same mics to see how much they'll vary from this one mic we just tested. To make it easy to compare the frequency response, we'll adjust the level so that each mic is set to the same level at 1,000 Hz (although we'll keep track of how much the level needed to be adjusted for each mic). Let's say we test 50 mics. We lay out the 50 charts and we also have a blank piece of chart paper in front of us.
We find the lowest frequency (20 Hz) on each chart, and look for the highest signal level (loudest), and the lowest signal level softest) we measured at 20Hz. We put two marks (shown in red)on our blank piece of chart paper at 20 Hz. We do the same thing at each line, peak or dip on the chart, until we have an upper and lower row of dots that represent the maximum and minimum range of frequency responses from this batch of mics. Here's the curve for "our mic" and the variations we found in testing 50 mics:
We then connect all the upper red dots, and we connect all the lower red dots (with the final curves shown in grey):
We can then draw a line (the blue curve) exactly centered between the upper and lower dots and that's our "typical response curve" that we submit to the marketing department. Understand, at this point, the curve could look fairly flat, but individual mics can vary by 5dB or more from the "average curve", and still be considered "normal". (Remember we also adjusted the output level for a constant 1,000 Hz signal from each mic? That will thow off the results even more and be critical when it comes to finding a matched pair).
Well, our sample (in black) isn't too far off the average (in blue), but we might find some mics in that batch that are better in the bottom end. How tight to hold the "deviation from average" window is a judgement call by the company and then carried out by the quality control department. At companies like Neumann, they use a 4dB window, which means that all mics must fit within a +/- 2dB window (4 dB overall) of their published curve. B&K test mics may use a window as small as +/- 1/10th of a dB variation from their published curves.
But our "average curve" may still look "too jagged" for public consumption" from the marketing department's point of view, so the curve can be "smoothed" by averaging some of the jagged peaks, or slowing down the pen speed on the chart recorder (so it doesn't move as fast up and down and makes the curve look smoother by simply ignoring all the little jagged short bursts). These are usually marketing decisions, so that "our curve" look similar to "other companies' curves":
And there we have the final "respectable" frequency response curve that is published in the advertizing literature.
Now, here's another "gotcha" for most pressure gradient mics: the frequency response will change, depending on the distance from the sound source, or the angle to the mic. Some manufacturers will actually show the "proximity effect" on the frequency response chart, showing how the bass is boosted as you get closer. Some will also show the frequency response at different angles (usually 0°, 30°, 60°, 90°, and 180°), like this:
When you look at a number like "Frequency Response: 20 - 20k", look at the published curve to see what the "usable response" really is, and remember that the curve you see is "averaged and smoothed. Unless the deviation is shown (either as a gray area or a line above and below the curve, or a number like +/- 3 dB), you really don't know what
your mic is really doing. That's why it's so hard for the average person to tell what a mic might sound like, judging from the frequency response curve, or just reading the specs.
Any questions so far?