Down With Flat! AHC Responds

An article in response, written by Anthony H. Cordesman, appeared in the following issue, Vol.8 No.5:

Down With Flat? With Flat What?
Readers who regard reviewers as sources of revealed truth should read no further than as the point of this article is to take issue with J. Gordon Holt's editorial in the last Stereophile ("Down With Flat"), thereby revealing that differences of opinion exist even within such a highly structured cultural monolith as Stereophile. Worse, the gist of my remarks deals with differences literally at the "high end." The following material is strictly for consenting adults.AHC

The Case Against Mindless Flattery:
There are points of agreement; I agree with JGH's observation that flat frequency response is not an inherently desirable characteristic in high-fidelity equipment. He failed to point out, however, that "flat" is anything but easy to measure, even using simple steady-tone measurements. Granted, we can easily measure flat frequency response using sustained signals in amplifiers and tuners.

Such measurements are far less reliable, however, with devices like cartridges, tape recorders, and speakers. One key measurement problem with cartridges is whether a cartridge that is "flat" at one signal level will also measure flat at significantly higher and lower signal levels. The equipment simply isn't there to let us know. Test records are cut at roughly the same signal levels and cutting angle, levels chosen so that virtually all cartridges can track them—thereby minimizing the differences between cartridges. Frequency anomalies at very low signal levels are difficult to measure as is varying frequency response at very high tracking levels.

Some cartridges also exhibit different frequency response if one conducts tests using sine waves, white noise, and pink noise, although others do not. The Shure V-15 series, for example, shows a much wider range of difference between white and pink noise than do most other cartridges. Although the differences between standard test records (the CBS and JVC) are not significant—they do not explain the significant differences in curves run on the same cartridges, as Noel Lee of Monster Cable suggested in Vol.8 No.4—there are also minor differences between test records.

Tape recorders present serious frequency response test problems because of low level noise, bias limits, and head room. Most tape recorders lose their flat response at low signal levels near their noise floor, and lose flatness as the record level rises above -10dB. In fact, curves of tape recorder frequency response are of little help if they don't include 0, +5, and +10dB. There is considerable controversy about what is flat in the upper octaves using the standard cassette frequency test curves, and it is extremely difficult to determine how Dolby and dbx-equipped recorders should be judged since they are generally flat only within a comparatively limited dynamic range.

Speakers, the specific subject of Gordon's editorial, are the worst of all; we are dealing here with sound variations in excess of 5dB. And relatively small differences in distance between measuring device and speaker represent significant differences in perceived and measured power. For example, when talking about measured flat response in speakers in the last issue, JGH did not mention whether he was measuring 1 meter from the speaker, 2 meters, or at the listening position.

These differences may not seem important, but they make a tremendous difference in measuring top-octave speaker performance. Theory might dictate that a speaker should have flat response at the listening position, but this normally produces an upper octave balance that is unbearably bright. If seeking flat response at a normal sitting position is wrong, however, what reason is there to think that flat response at 1 meter, 2 meters, or any other "on-axis" distance correct?

Speakers also differ radically in the way in which they generate high and low frequencies. The highs from speakers with highly directional on-axis treble radiation sound different from bipolar speakers, as do those from omnidirectional and broadly radiating speakers, and do so even when equalized to measure exactly the same at the listening position.

I would also argue that low frequencies are virtually impossible to measure consistently. I don't know of any chamber that is "flat" in the deep bass, and I don't know of any test procedure that produces fully accepted results, even under laboratory conditions. Further, such tests don't produce results that will tell you what will happen in a normal listening room.

Tests in normal listening rooms, however, produce sharp variations according to the particular speaker/room interface. If you keep a constant wattage of power to a speaker, and then start varying the test tone using sine and square waves, warble tones, white and pink noise, and single tones, one-third-octave tones, and one-octave tones, you will get amazing apparent differences in frequency response.

Differences of a few dB in average bass level will often radically change the standing wave and cancellation phenomena in a room, as will shifting the position of the speaker by even a few inches. A speaker equalized to be flat in the bass at one average listening level will normally cease to be flat the moment major changes take place in average level. Since these changes in dynamic level are characteristic of most music, "flat" bass is literally unheard of.

I will be interested to see if any speaker designers care to comment, but I have found that measuring bass energy is also complicated by the radiating nature of the bass driver. Flat planar speakers are notoriously subject to weird bass suckouts. Rear-facing woofers often produce major bass humps. Forward-facing woofers need to be raised 18" or more off the ground to prevent peak-and-valley bass.