A History of the X Curve

Surround Professional, January 2000, Volume 3, Number 1, by Tomlinson Holman

The X curve celebrates nearly a quarter century of helping interchange in the industry.

In the history of multichannel sound, the standardization of the electroacoustic frequency response for monitoring film stands as one of the most significant developments. It was standardizing the monitor frequency response at the ear of listeners that provided for better interchangeability of program material, from studio-to-studio, studio-to-theater, and film-to-film. Work started on formal standardization of the monitor frequency response for large rooms for film in 1975 on both the national and international levels. The work resulted in the standards ANSI-SMPTE 202 in the U.S., the first edition of which was officially published in 1984, and ISO 2969 on the international level. Actually, the standardized response was in use for some years before the formal standards were adopted.

The X Curve: The measured electro-acoustic frequency response presented to the ears of listeners in a dubbing stage or motion picture theater. The curve is to be measured under specific conditions, and is to be adjusted for room volume as specified in the standards referenced in the text.

The background behind this work began with Texas acousticians C. P. and C. R. Boner, who established in the 1960s that a "house curve" was a needed concept. They showed that a flat electroacoustic frequency response in a large room sounds too bright on well-balanced program material. This was subsequently found to be correct by other researchers, such as Robert Schulein and Henrik Staffeldt, as well. While Boner's practice was for speech reinforcement systems that did not require theater-to-theater uniformity in the same way that film does, nonetheless the concept of a house curve traces back to them. This development paralleled the introduction of 1/3 octave room equalization, since there would be little point in establishing a house curve if sound systems could not be adjusted to it.

Ioan Allen of Dolby Laboratories realized that the idea of a house curve was a valuable one after applying Dolby A-type noise reduction to optical soundtracks and extending the bandwidth of the track. While we think of Dolby A as principally noise reduction of between 10 and 15 dB depending on frequency when used in a tape context, in the case of the application to optical soundtracks, most of the advantage in dynamic range was taken to extend the bandwidth. The ordinary mono Academy-type soundtrack had sufficiently low noise for its time only by imposition of a strong high-frequency roll-off that made the effective bandwidth of soundtrack reproduced in theaters about 4 kHz. If such a track was reproduced with a wide-range monitor, the noise was excessive. By extending the high-frequency bandwidth of the monitor, and applying Dolby A NR to tame the noise, a very useful extension of bandwidth from about 4 to 12 kHz was achieved, while lowering the noise a fairly small amount.

Then came the question of the best frequency response for the monitor. In an English dubbing stage, Allen did an experiment with a nearfield, flat hi-fi loudspeaker vs. the farfield film monitor loudspeaker, a VitaVox. He adjusted the frequency response by equalizing the film monitor until the balance was similar, although the monitor loudspeakers of the day only extended to about 8 kHz before giving up the ghost. The electroacoustic response curve Allen found measured with a microphone was flat to 2 kHz, then down 1 dB per one-third octave, to -6 dB at 8 kHz, and falling beyond. This was named the X curve, for eXtended response, whereas the older Academy curve got dubbed the N curve, for Normal response (although one wouldn't consider it normal today).

When extended-range compression drivers and constant directivity horns became available around 1980, the question became, "How should the X curve be applied to this new development?" The new systems had a full octave of high-frequency bandwidth over older systems, but delivered nearly the same output response across a range of angles, rather than concentrating the response on axis as frequency went up as the older driver-horn combinations did.

One theory floated in the middle '70s was that the need for a house curve was based on an artifact of the method of measurement rather than a real need for sound to be rolled-off at high frequencies in large spaces. This was because the quasi-steady-state pink noise stimulus measured by a real-time analyzer in a room is time blind, lumping the direct sound, reflections, and reverberation together indistinguishably. If the different soundfields had different responses, the pink noise stimulus plus RTA could not sort out the differences and would basically average all the responses. Since the microphone is in the farfield of the loudspeaker where reverberation is dominant, then the response with a collapsing directivity horn vs. frequency could be expected to be rolled-off at high frequencies, since the contribution of all the off-axis angles would dominate over the direct sound. Nevertheless, in this condition, the direct sound could be flat, and we might respond to the flat direct sound and ignore the later-arriving response as listeners.

If we then were to change to a constant directivity horn, with its output more constant over all angles within its coverage, and the system is tuned to a "house curve," then it might be expected to sound duller than the older horns, at least on axis at a distance. That's because, under these conditions, both the direct sound and the reverberant sound would be rolled-off and on the same curve. So one of the first experiments I did on this combination was to play conventionally mixed program material over constant directivity horns equalized to the X curve to see if the sound was too dull. It was not; in fact, with the bandwidth extension from 8 to 16 kHz, it actually sounded somewhat brighter, but this was due to the extended compression driver response instead of having to do with the equalization curve.

So what's going on here? This was later explained by Dr. Brian C. J. Moore, author of numerous refereed journal articles on psychoacoustics and the book, An Introduction to the Psychology of Hearing. The rolled-off house curve has a good basis in psychoacoustics, because a soundfield originating at a distance is "expected" to be more rolled-off than one originating nearby. It is a little like optical illusions in vision that show, despite occupying the same area on the retina, pictures look bigger on a larger screen, even when a small screen is closer and takes up the same horizontal and vertical angles. As it turns out, both spectrum and level are affected by the perception of the size of space you are in, and "getting it to match" perfectly from large to small room in physical sound pressure level and response does not result in sounding the same.

With the additional octave of high-frequency extended range of more modern drivers and horns came the need to calibrate the X curve to the highest audible frequencies. Later editions of the SMPTE and ISO standards show the roll-off to 8 kHz as originally standardized, but added rolloff from the extended curve in the bands above 8 kHz. Some users don't employ this additional roll-off, staying on the original X curve to 16 kHz, but in an experiment I did at USC, I found that following the letter of the standard was an improvement in high-frequency balance and interchangeability of program material. This was done in a very sensitive experiment, reported earlier in Surround Professional, that involved playing trailers in a large theater exactly as they sounded in the dubbing stage, with the agreement from the people who had supervised their mixes that they sounded correct, and this involved eight trailers mixed in a variety of studios. Both level and response standards had to be perfect to accomplish this, and just a 1 dB error over several octaves that crept in during setup was heard, and had to be corrected.

Another development of the X curve is how it should vary with room volume. Although a variation in the response with room volume was written into the original standard, further work shows that the response should be "hinged" at 2 kHz, and turned up at high frequencies in smaller rooms. Curves that extend the range out to higher frequencies before breaking away from flat do not seem to interchange as well.

Today, the major factors affecting interchangeability no longer have to do with the target curve, since the X curve is very well accepted, but rather have to do with how the curve is to be measured and adjusted electroacoustically. The standard calls for such needed items to make good measurements of quasi steady-state noise as spatial averaging, temporal averaging, and the proper use of measurement microphones. The largest variations among different practitioners are in the use of microphones. The problem is that the soundfield seen by a microphone in a large room is a mixture of direct sound, early reflections, and reverberation. Standard measurement 1/2-inch microphones demonstrate very different high-frequency response when measured anechoically on axis and with a diffuse field. Differences are on the order of 6 dB in the top octave between the two, and response in rooms is highly affected by the differences between these two. Only by the use of small, low-diffraction microphones, such as 1/4-inch or smaller diaphragm mics, are the differences kept small.

The best usage of measurement microphones today is to calibrate small ones for grazing incidence across the diaphragm rather than perpendicular to the soundfield, because, this way, the microphone will demonstrate the most similar response for the direct sound (across the diaphragm) and reverberation (a diffuse field). One of the primary ways in which problems show up in this area is in the difference exhibited between sound originating from a more-or-less point sound screen channel vs. a surround array: 1/2-inch microphones make serious errors between these two because the soundfields generated under the two conditions are so different.

The X curve now has nearly a quarter century of use and has absolutely acted to help interchange in the industry. Combined with level standards, and de facto industry standards such as speaker directivities, the whole film industry has benefited without a doubt. Problems linger in applying the standards uniformly due to different methods of measurement. Also, when heard over a modern flat loudspeaker in a small room, program material balanced on an X curve monitor sounds overly bright. That's because the original experiment that set the curve was made many years ago, without the frequency range available from today's components. This is not too important because, so long as everyone agrees to use the same curve, then the response sounds the same to the mixer on the dubbing stage as to the audience member in any auditorium. Interchangeability of X curve material with home video can be handled with a simple re-equalization. The ATSC television standard recognizes the differences, sending a flag that tells receiving equipment whether the program material was balanced on an X curve monitor, or on a flat monitor in a small room, and home equipment can take appropriate action to re-equalize the program accordingly.