Down With Flat! Letters
Editor: Since you are interested in comments from loudspeaker designers, I thought I would make some brief comments and control my usual compulsion to develop a doctoral dissertation.
In the ages-old literature, you can find the answer to why a small monitor loudspeaker with flat extended highs sounds much too bright. If you were to study the work done in the 1920s and 1930s, it might be disturbing to you to realize that so much has been forgotten and reinvented.
For a loudspeaker to sound balanced, the geometric center of its bandwidth must be somewhere between 500Hz and 1kHz. This varies somewhat for different people. We have a balance point to our hearing system: if the lows are cut off, the highs must also be cut off in order to sound balanced. The design principle for the loudspeaker is that high-frequency rolloff times low-frequency turnover must equal 500,000. I won't swear this is the right number, but it will serve for this example. If the low frequencies cut off at 50Hz, the highs mustn't go above 10kHz. If the lows go to 25Hz, highs should go to 20kHz. The product of both these pairs of numbers is 500,000: 50 x 10,000; 25 x 20,000.
The perception of flat is related to our aural balance point, and varies with loudness. Variation with loudness is why music sounds more real when we play it at the level we would have heard it "live": the frequency response of our hearing matches the frequency response set by the musicians.
Instead of cutting off the highs to match the limited lows, I prefer to extend the highs and set the balance by ear, using pink noise. Once the trend of the frequency response is set by ear, the actual response can be refined to eliminate peaks and dips, relative to the trend. There are definite relationships regarding the bandwidth, the position of the bandwidth relative to our balance point, and the slope of the actual frequency response Mathematically, it is something like setting the center of gravity, or perhaps the moment of inertia.
Some people forget that constant power input to a loudspeaker bears very little relationship to constant acoustic output. Most of the input power is not converted to acoustic output, and this varies drastically with frequency.
There's another interrelationship here. The signal processing of our hearing system discriminates between the bass output of the loudspeaker and the bass buildup in the room. We mentally adjust to our room acoustics. If we then adjust the loudspeaker for flat bass response in the room, it will sound weak.
In its steady-state response at low frequency pressure leads the input signal by 90 degrees, leading to the belief that pressure is proportional to the acceleration of the cone. From the transient-response viewpoint, low-frequency components cannot lead the high-frequency components. What appears to be a 90-degree lead is actually a 270-degree lag.—David H. Blasetti, Upper Darby, PA
We are well aware of the so-called "500,000" law but it does not explain why a system that's essentially flat to 30Hz still needs a gradual rolloff above approximately 3kHz to sound "flat" (ie, musically natural) at the high end. The exact amount of rolloff required seems to be related to the HF dispersion characteristics of the system, and can only be determined by listening to music through the system. Pink noise will not do, as the ear has no frame of reference for what sounds "right" in terms of HF balance using pink noise.—JGH
Up With Down!
Editor: JGH's editorial, "Down With Flat!", in Vol.8 No.4 strikes me as the most important piece I have read in my nearly 40 years as a high-fidelity hobbyist.
If he is right, and loudspeaker response must be contoured to sound flat (that is, real), then everything else we are doing is almost irrelevant. We are like naturalistic painters who pay meticulous attention to the details of color and texture, but don't come within a country mile of the actual shapes of objects. We are, in essence, creating unwitting aural abstractions or caricatures, instead of attempting to recreate the original shape of the sound.
If Gordon is right—and I think he is—is not this the area urgently in need of investigation? Could he not begin by giving us a generalized graph of what a speaker's high end should be like?
I have tried rolling-off a speaker's top, and for the first time I feel I am beginning to recreate the live musical experience. Please, please follow up on this.—Kenneth La Fleur, Hingham, MA
Most "high-quality" loudspeakers I have tried equalizing seem to do best with a gentle rolloff starting at around 2kHz (-1dB), with about 4dB of loss at 10kHz, then flattening out again. (This can be accomplished between the preamp and power amps, via two resistors and a capacitor per channel.)—JGH