Moulton Laboratories
the art and science of sound
Acoustical Measurements For The Rest of Us
Originally published in Recording, approx. April 2001
by Dave Moulton
April 2001
2. The frequency response of your room

How to measure the room acoustics in your home studio.

< 1 2

Measuring the Frequency Response of Your Room

First, Measure The Nearfield Response Of Your Speaker.

You use pink noise for these measurements. Set your speaker up on a stand or something as far away from reflecting surfaces as you can reasonably get it (outdoors, facing up from the ground, is an excellent alternative). Send pink noise to the speaker at a modest level, say 75 dB SPL. Place your measurement mic about 2 feet from the speaker, on axis (line it up with the tweeter). Send the mic signal through its preamp, through an EQ (octave band graphic or parametric) and to some sort of level meter. Turn all EQ bands all the way down. Turn one up to +12 at 1 kHz. (if it’s parametric, set the bandwidth at 1/2 octave). Adjust the level to the meter so that it is reading approximately 10 dB below full scale (if you are using a conventional VU meter, this would be -7 VU). Write it down! Don’t touch the levels again for the rest of the measurement. If you have a test CD with octave bands of noise, you can skip the EQ and take a heap less time and trouble.
  
Figure 2: Flow chart of two possible test setups, one using an equalizer, the other using a test CD with noise bands.

Next, change the frequency of the EQ for each octave of interest. Write down the level you observe on the meter. You’ll have trouble, probably with the bottom two octaves, and maybe with the top octave. You probably have to do a lot of estimation. If the speaker gives any signs of distress at frequency extremes, don’t push it. In any case, the resulting set of levels for all the octaves will be the so-called “baseline” for the subsequent room measurement.

Next, Measure The General Response Of Your Room.

Repeat the above measurement, with the speaker in the room facing away from the test mic. The test mic should probably be at your favorite listening position. The speaker should be at least 5 feet away, but not right at a wall, and especially not in a corner. In this case, particularly at high and mid frequencies, the levels you observe will be primarily derived from reflected sound.

Once you have collected the raw data for this test, you have some arithmetic to do. First, you will “normalize” your room data. To do this, find the difference in level between the room measurement and the baseline at 1 kHz. In my hypothetical example below, the room level at 1 kHz is 6 dB below the baseline, so to make them the same, I would increase the room level by 6 dB. Apply that correction to all of the other the octave bands (i.e. in my example increase all room levels by 6 dB). The resulting level is “normalized.”

Octave Band:32631252505001K2K4K8K16K
Speaker “Baseline”: -25-12-2+8+40-3+3-6-15
Room Level: -30-20-2-2-4-6-9-10-18-25
Room Level Normalized: -24-14+4+4+20-3-4-12-19
Net Room Response: +1-2+6-4-200-7-6-4
Table 1: Octave-band levels for a hypothetical speaker and room. The Baseline measurement is the nearfield speaker measurement, the Room is the reverberant room measurement, Room Level Normalized is the room level increased to match the baseline level at 1 kHz., and Net Room Response is the Room Level Normalized level subtracted from the Baseline Level. You can take these numbers and easily make them into graphs.

Find the difference between the two levels at each frequency. In my example, for instance, at 63 Hz. the normalized level is -14 dB, while the baseline is -12 dB. The net value is -2 dB.

The resulting curve is the net room frequency response. Due to the way the measurement is set up, it should always be 0 dB at 1 kHz. If not, you’ve made a mistake somewhere.
  
Figure 3: Graph of Net Room Response.

In this case, there is a peak in the 125 Hz. range, and a dip around 4 kHz. This suggests (a) some sort of room mode buildup due to wavelengths around 9’ long (1130/125 = 9) and (b) a significant amount of absorption in the 4-8 kHz. range.

Measuring For Critical Distance.

Set up your monitors normally. Generate noise at a moderate level in one monitor. Measure the SPL at 2 feet from the speaker. Write it down. Measure the level at 4 feet from the speaker. Write it down. Is it approximately 6 dB less than at 2 feet? If it’s significantly less than 6 dB down (say only 3-4 dB), you’ve found the critical distance. If it’s approximately 6 dB down, measure again at 8 feet. If you are still within the Critical Distance, the level should be another 6 dB down. Keep doubling the distance from the previous measurement (this won’t take long!) until you get to the point where the level is only 3-4 dB below the previous measurement. That point is the Critical Distance.

Often, for small, dead rooms, the Critical Distance is beyond the walls, which is OK. Mainly, you want to make sure that your listening position is within the Critical Distance. Happily, it usually is. This is a measurement that is very handy in concert halls and other large rooms.
  
Figure 4: Graph showing direct vs. reverberant sound for distance from source in reverberant rooms.

Measuring For Buzzes.

Using the sine wave generator at a modest level (ca. 65 dB), slowly sweep across the spectrum. Listen for buzzes in the room. When you hear one, stop sweeping. Note the frequency, and find the vibrating offender and damp it so that it stops buzzing. Continue sweeping. When you’ve gotten across the spectrum, repeat at a slightly higher level (say 70 dB) just to confirm that you’ve gotten rid of everything. Don’t bother trying to debuzz your room at very high levels, because (a) you probably can’t, (b) you will drive yourself buggy trying, and (c) buzzes that occur only at high levels usually either aren’t much of a problem or else they’re such a huge problem you’ve long since fixed them!

Measuring And Logging Room Modes.

This is a test that uses a gated swept sine. If you don’t have the ProSonus CD, you need to have a VCA or something with a Low Frequency Oscillator that you can use to turn a signal on and off continuously. Set the on/off rate at about 1/2 second for a complete on-off cycle. Send the sine-wave through this to the speakers. Sweep the sine frequency slowly from say 50 Hz. to about 500 Hz. You will hear the sine wave switch on and off. More importantly, you will hear, during the “silences,” the sound of the reverberance of the room at each frequency. At frequencies where the room is resonant, the reverberation will swamp the sound of the gate switching, while at frequencies where the room is heavily damped, you will hear the sound of the gate clearly, as clicks.

As you sweep, note the frequencies and relative levels of the various major peaks and dips. From this measurement, you can get a fairly clear picture of the resonance that is actually occurring in your room. Log the placement of the mic, and repeat in other locations. Compare the patterns for an even more detailed view. You can even begin to map, in a crude way, the resonant behavior of your room. Whoa, dude!

Measuring Reverb Time.

Using the same gating device, send octave-band noise to a single speaker. The actual reverb time will be too short to measure directly with a stop watch. Instead, speed up the rate of the on-off gate until the reverberant sound definitely merges with the next on-cycle of the gate. Slow it down now until you can hear the reverb begin to die away to about half its loudness, which is the subjective equivalent of approximately -10 dB. Now, use a stop watch to time 16 cycles, for instance. Because each cycle is half reverb, divide that time by 32, and you’ve got the Early Decay Time for that frequency band. Multiply by 6 (remember, you measured only the first 10 dB of a 60 dB decay) for the Reverb Time.

Repeat with each octave band of interest. In general, times are taken for the octave bands from 125 Hz. to 4 kHz. I find the 63 Hz. octave interesting, if I can get a measurement (it depends on the speaker), and both 8 and 16 kHz. are interesting, too, for small rooms. However, they may be too short to reliably measure.
  
Figure 5: Acoustical level over time for resonant and non-resonant gated signals.

Measuring The Noise Floor

The problem with measuring noise floors is that the microphones and meters we’re using just aren’t quiet or sensitive enough to measure many of the levels we’d like to know about. Our ears, however are. So, if we can establish a known signal level, and then can attenuate the level by known amounts until it just becomes inaudible, we will have determined the threshold of audibility for that sound in our room, which is generally equivalent to the noise floor for that sound. This is best done by octave band.

Establish a known Sound Pressure Level with the source signal (say 1 kHz. noise at 50 dB SPL) run through a module of your console at maximum undistorted level and controlling gain with the monitor level pot. Calibrate by experimentation a range of -3 dB increments on that module, hopefully down from your known level as much as 40 dB (which would mean you are measuring to 10 dB SPL if your reference is 50 dB SPL). Once you’ve got everything set, start reducing the level of the noise in 3 dB increments until it is no longer audible. Write down the level 3 dB above that inaudible level as the level of the noise floor for that band. If you get to your lowest calibrated level and can still hear the signal clearly, note that level with the prefix “<“ (means “less than”). Do this for each octave band of interest.

When you are done, you will have a response curve for the noise floor of the room. Typically, low frequencies will seem to present a problem, because you won’t be able to hear them at any kind of low or even moderate level. That’s OK, because we don’t hear very well at low frequencies. So, those levels will seem high (like 50 dB SPL at 63 Hz., for instance), which is probably the truth.

To create more meaningful and relevant data, you can apply A-weighting, which corrects for the way we hear at low levels (i.e. levels equivalent to 30-40 dB SPL at 1 kHz.). Use the following corrections for each octave band (yup, you subtract 33 dB in the 32 Hz. band!).

Octave Band:32631252505001K2K4K8K16K
A-Weighting in dB: -33-21-11-4+2+5+6+6+4-2

Table 2: A-weighting for each octave band. Apply the weighted value to the measured level to obtain A-weighted levels in each octave band.

The resulting curve is the A-weighted noise floor spectrum, which should give you a reasonable idea of how your room is behaving. You can have lots of fun by turning various equipment pieces on and off and seeing how that changes the floor. You may be surprised!

What Does It All Mean?

Hearing is complex. Acoustical behavior of reverberant rooms is complex. These tests should give you an insight into these complex behaviors in a way you can use to go beyond the marketing hype. Using these tests, you can both hear and understand a little better what is happening in your studio and control room and what will happen with your recordings in other rooms.

Happy trials!

Dave Moulton updated this article from a really sleazy one he wrote in 1994. You can complain to him about anything at his website, moultonlabs.com
< 1 2

Post a Comment



rss2

rss atom