What Ever Became of Frequency Response? (Jun. 1991)

by ROBERT LONG

[Once the touchstone of audio, this important spec is on the endangered list. ]

It's my fault. And yours. We just haven't been paying attention. As the reams of manufacturers' specs pour in each year, we glance across them, questioning a number here and there, perhaps, and even ruminating on the other guy's sleaziness from time to time. But that's life, right? We all have our ideas about how audio products should be measured and specified, and we don't always agree--particularly when dollars are at stake, unfortunately.

It must have been at least a decade ago that the specs on a cassette deck I was testing caused me to blink in disbelief. The frequency response traces were exemplary--as flat as those shown in the sales brochure through virtually the entire frequency range. I don't today remember all the specifics (let alone the model number), but it was well within ± 1 dB from, I believe, below 100 Hz to at least 15 kHz and was down by 3 dB at around 35 Hz and 19 kHz.

In the manual, the frequency response was listed as ±3 dB from 20 Hz to 20 kHz. And if you were to shift the whole frequency response curve upward by 3 dB, so that its flat section ran at +3 dB instead of acting as the 0-dB reference, the-3 dB points would indeed fall at or even a hair beyond the specified frequencies.

Didn't they realize, I wondered, that a spec of ±3 dB implied much worse performance than one of +0,-3 dB, given the slight difference in specified bandwidth that would result? Even keeping the same frequencies and calling it "+0,-6 dB" would—correctly--suggest the admirably flat response, which "±3 dB" does not.

It was easy to see how this state of affairs had arisen. The ability of some high-end decks (Nakamichi's name comes immediately to mind) to resolve very high frequencies, through a canny combination of narrow head gaps and compensating EQ, had set them well apart from the crowd. High-frequency response was where it was at--competition, that is--and the also-rans were scrambling to cover the coveted specification.

As I soon noticed, company after company was doing the same thing: Listing frequency response within ±3 dB, no matter how flat or otherwise the trace, in order to claim a still higher possible frequency as within the deck's grasp. Shades of the battle, a generation ago, over amplifier power specs with a "±3 dB" slop factor! It certainly is no news to anyone in audio (or Audio) that differences of only a fraction of a dB can be perceived, at least in the "working" part of the frequency range, and that such differences create what we usually call "coloration." For this reason, the degree of flatness of the frequency response is of fundamental importance in audio products and has been recognized as such since audio (and Audio) began. Whether or not this flatness is maintained to extreme frequencies is of much less importance for two reasons: Some of the frequencies being bandied about in this context are demonstrably beyond the hearing acuity of most adults, and most music makes relatively little demand on these extremes.

In fact, as properly defined, frequency response has nothing directly to do with frequency; it is the range of level variation (that is, amplitude) in response that is to be expected of the product. The frequency range, or bandwidth, over which this variation is of interest should be implied or specified. For many products, a range of 20 Hz to 20 kHz is assumed; to be precise, however, the bandwidth should be spelled out.

So when we say that a device has a "frequency response of ± 1 dB, 20 Hz to 20 kHz," the "± 1 dB" is the frequency response, and "20 Hz to 20 kHz" is merely the bandwidth over which that frequency response applies. The current habit of saying "20 Hz to 20 kHz, ± 1 dB" gets the cart before the horse, but at least it contains all the necessary information.

More and more, however, one sees statements like, "Frequency response is 20 Hz to 20 kHz." Recently I did some writing for an audio publication (not Audio) that included manufacturer-supplied product specs. In the "frequency response" listings of the portions I examined, bandwidth figures outnumbered true frequency response by at least two to one.

The harder it is to characterize the true frequency response, the more pressure there is to go for bandwidth instead, perhaps. Loudspeakers are particularly pesky, for example, because their response can vary so much in space. Change the measurement position even slightly, and the response curve can look very different.

Furthermore, many speaker response traces leave one at a loss to determine what the reference 0-dB level should be. So it has become almost de rigueur to issue ratings like "60 Hz to 22 kHz" for speakers.

But the point is that this is not frequency response and should not be labeled as such. It is rated bandwidth.

We might argue over whether bandwidth is as useful as true frequency response data (were it available), but I see no room for argument about which is which.

The accuracy that is presumed inherent in digital audio systems makes the distinction more important than ever. The bandwidth of all CD players is substantially the same in the sense that it is limited by low-pass filtering that puts the -3 dB point up around 20 kHz, give or take an inconsequential kilohertz. But the actual response, particularly in the treble, can vary from "perfect" by as much as 1 dB (and occasionally more), a difference that can alter how music sounds.

(source: Audio magazine, June 1991)

Also see:

Impedance, Cables & Treble Loss (Nov. 1991)

= = = =