Search Gear
 

Understanding Gear Specs: Frequency Response

June 30, 2014
share

What frequency-response specs don’t tell you

When shopping for professional analog audio gear, we typically turn to frequency-response specifications in the accompanying documentation as a trusted measure of performance. But here’s our industry’s dirty little secret: The frequency response can be made to look better than it really is. All it takes is manipulating the test conditions under which the data are derived.

There’s no industry standard for the test conditions used to derive frequency-response specs, and most pro-audio manufacturers don’t reveal their methods. That makes comparing the frequency response of two like pieces of gear an apples-to-oranges exercise.

In this article, I’ll show you how the frequency response of a pro-audio reference monitor can be measured in various ways to obtain wildly different results. As you’ll see, it’s easy for manufacturers to cite a frequency response in a way that suggests excellent performance but is, in fact, misleading or even meaningless.

Fig. 1. Metric Halo SpectraFoo Complete shows the transfer function of a reference monitor at continuous resolution, revealing every peak and dip in frequency response.
Zero Tolerance To demonstrate my points, I used Metric Halo SpectraFoo Complete 4—a comprehensive software suite of analysis and metering tools for audio—to test the frequency response of a popular, inexpensive studio monitor. The manufacturer states the monitor’s response to be 57 Hz to 20 kHz, with no tolerances provided. That is, the spec doesn’t cite how far from flat the response deviates throughout its ostensible 57Hz to 20kHz range. There could theoretically be peaks and dips in response measuring 12 dB, in which case a more rigorous statement of the spec would be 57 Hz to 20 kHz, ±12 dB. Without that “±” bit, the spec is totally meaningless.

My test of the same monitor’s frequency response is shown in Figure 1 on page 70. (Figure 1’s graph shows the monitor’s transfer function in SpectraFoo Complete, with traces for coherence and phase versus frequency hidden to simplify the view. The averaging rate was set very high to eliminate ephemeral spikes and dips.) The horizontal axis plots frequency, while the vertical axis shows power (amplitude) in decibels. With the resolution set to continuous mode (revealing detail as fine as 2/3 Hz wide), you can see several problems with the monitor’s frequency response. These include a series of peaks and dips between 140 and 280 Hz (4.5 to 5.5dB swings in response) and a 12.5dB dip at 370 Hz, 7dB dip at 490 Hz, and 6dB peak at 740 Hz. To the manufacturer’s credit, the response is only about 3 dB down at the 57Hz and 20kHz limits cited in their spec. It’s the wild swings in between 57 Hz and 20 kHz that the spec sidesteps.

New Gear’s Resolution Even when “±” tolerances are provided with the frequency-response spec, you can’t always trust their veracity. A common way to favorably skew the spec is to average the transfer function’s amplitude across relatively wide slices of the audio spectrum, thereby flattening any narrow (but potentially deep) peaks and dips.

Fig. 2. The same monitor’s transfer function is displayed at 1/3-octave resolution in SpectraFoo Complete, averaging the amplitude in each frequency band.
Figure 2 illustrates this averaging technique. This figure shows the same monitor’s frequency response derived with 1/3-octave resolution in SpectraFoo Complete. That is, the monitor’s response is averaged in 1/3-octave slices across the spectrum. At this resolution, the series of peaks and dips between 140 and 280 Hz appear to flatten out a lot, completely hiding the dips and suggesting only a 2.5dB deviation in response. The 12.5dB dip at 370 Hz looks to be only 3.5 dB deep. The swings at 490 and 740 Hz look to be only one third as severe. Using 1/3-octave averaging, the ostensible frequency response can be stated to be 57 Hz to 20 kHz, ±3.5 dB. Trusting this spec implicitly, you would have no idea that the response varies -12.5 dB (at 370 Hz) and +6 dB (at multiple frequencies) across this monitor’s useful range.

Figure 3 shows what happens when the frequency response is averaged in one-octave bands. The response looks to be 57 Hz to 20 kHz, ±3 dB—a further improvement in overall deviation from flat. More deceptive, the severe and discrete dips at 370 Hz and 490 Hz appear to be a smooth, conjoined trough only 1 dB deep. The 6dB peak at 740 Hz has all but disappeared, inferring flat response in that band. The overall response appears to deviate only 4 dB (+3, -1 dB) between 90 Hz and 20 kHz. Impressive! But false.

Fig. 3. Displayed with 1-octave resolution, the monitor’s transfer function appears to be much flatter and smoother than it actually is.
Reality Check
Even highly trained ears can have difficulty discerning shallow and narrow deviations in a monitor’s frequency response. It can therefore be argued that continuous-resolution frequency-response charts are too detailed for many pro-audio buyers to find useful.

The rub is that some manufacturers take extreme license with their frequency-response documentation, and the other companies feel compelled to follow suit in order to prevent their products from looking poor in comparison. In the words of an insider at a company that makes reference monitors, frequency-response specs and charts are “pure marketing.”

Your best recourse when evaluating gear for purchase is to trust your ears or, if a product audition isn’t feasible, a credible review in a resource such as Electronic Musician. The frequency-response spec should be taken with a truckload of salt.

Michael Cooper is a recording, mix, mastering and post-production engineer, a contributing editor for Mix magazine, and the owner of Michael Cooper Recording in Sisters, Ore. (myspace.com/michaelcooperrecording).

You Might Also Like...

Studio Basics
Show Comments

These are my comments.

Featured

Reader Poll

Do you spend more time producing or playing?


See results without voting »