Tech Topics: Modern Recording - An Overview of Crossovers

handslog.gif (9956 bytes)

An Overview of Crossovers
By John Murphy and Jim Ford

Reprint Note:

This is the first part of a two-part article I wrote for Modern Recording & Music back in 1980. It is primarily concerned with active crossovers and preceded a series of product reviews I performed on various active crossovers that were available at that time.

jlm

This month we are dedicating the "Hands-On Report" to the subject of loudspeaker crossovers. In upcoming months, we will be reviewing a number of currently available electronic crossovers, and we felt that it would be appropriate to first give our readers a discussion of the factors important to the selection and use of crossovers. We also feel an obligation to point out that many of the currently available units share a fundamental design flaw which prevents the complete loudspeaker system from delivering a flat frequency response through the crossover region.

The discussion will begin with crossover basics where we'll explain the difference between active and passive crossovers and look at the advantages of biamplification. Then we'll examine the problems that arise when you try to recombine the frequency spectrum after the crossover has divided it. Having pointed out the flaws in the most popular designs, we'll then talk about some well-behaved crossover types and ways of employing the "problem" crossovers to get the least frequency coloration.

With this information under your cap, you should be able to select a good crossover, establish an appropriate crossover frequency and interconnect the components of the loudspeaker system in such a way as to obtain the best performance of which they are capable.

Crossover Basics

The frequency range of human hearing spans from about 20 Hz to 20,000 Hz (20 kHz); likewise, the music we produce and enjoy contains audio information over approximately the same range. Now, in order to accurately reproduce this musical information it is necessary for the loudspeaker system (and the other components in the signal path) to reproduce the full spectrum.

Unfortunately, most loudspeaker drivers are incapable of accurately reproducing the entire 20 Hz to 20 kHz spectrum. This is because the characteristics that make a driver good for reproducing one frequency extreme make it unsuitable at the other extreme. For example, good low-frequency drivers need to be rather large in order to move a lot of air; but high-frequency drivers need to be small to maintain a wide radiation

pattern at the highest frequencies. This means that good low-frequency reproducers tend to perform poorly on highs and good high-frequency reproducers tend to perform poorly on lows. For high quality audio reproduction we are left with hardly any choice other than to use separate drivers to reproduce the frequency extremes.

In combining several loudspeaker drivers into a system, our goal is to obtain an accurate (that is, "flat") response across the complete audio spectrum. Although it is possible to use any number of drivers in a loudspeaker system, the most popular designs employ two, three or sometimes four drivers to make what are called two-way, three-way or four-way systems, respectively. Once it is known how to combine two drivers to produce a wide range response it is a simple matter to repeat the procedure and combine this two-way system with one more driver to produce a three-way system and so on. Considering that more complex systems are just an extension of the techniques used to create a two-way system, we will restrict our discussion to two-way systems. Now let's see how two drivers can be combined into a system.

The simplest multiple driver system would employ two drivers: a high-frequency driver (referred to as a "tweeter") and a low-frequency driver ("woofer"). Our first inclination might be to apply the full spectrum signal from the amplifier to both of the drivers by merely wiring them in parallel, but this approach would have some serious problems. First, in the range where the frequency response of the two drivers overlaps, there would be too much output from the system. That is, the system would have a "peak" in the overlap region. A second problem is that, in general,

the drivers will have different sensitivities. This means that one driver would play louder than the other. Since high-frequency drivers tend to be more sensitive than low-frequency drivers, the tweeter would probably play louder than the woofer. To make matters worse, the full-spectrum power-handling capability of most tweeters is quite limited. This is because the lower frequencies drive the tweeter's "cone" through very large displacements and can damage the unit at any significant power levels. Based on the power handling consideration alone it would not be acceptable to drive the tweeter with the full-spectrum signal. The only acceptable solution is to filter out the low frequencies from the signal applied to the tweeter. The loudspeaker crossover performs this function while controlling the overlap in response of the two drivers.

Multiamplification

There are two distinct ways to implement a loudspeaker crossover. The simplest and most often used approach is to perform the crossover function between the power amplifier and the loudspeaker drivers as shown in Figure 1. This type of crossover employs only passive components (resistors, capacitors and inductors) and acts directly on the speaker level signal from the power amplifier. Its main advantage is that it allows one power amplifier to drive a complete full range loudspeaker system.

The alternative to a passive crossover is multiamplification (also referred to as biamplification, triamplification, etc.). In a multiamplified loudspeaker system there is one power amplifier for each driver and the crossover filtering is performed between the preamp (or mixer) and the power amplifiers as shown in Figure 2.

The crossover consists of a pair of electrical filters which modify the frequency response of the signals applied to the drivers. The full-spectrum signal passes through a "high-pass filter" on the way to the tweeter while the signal to the woofer passes through a complementary "low-pass filter." The high-pass filter passes information above a selected frequency, referred to as the "crossover frequency," and attenuates the components of the signal which fall below the crossover frequency. Similarly, the low-pass filter passes information below the crossover frequency and attenuates frequency components above the crossover frequency.

An easy way to understand the crossover is to imagine an ideal pair of filters which direct all the musical information above the crossover frequency, say 1 kHz, to the tweeter, and directs all the information below 1 kHz to the woofer. When full-spectrum music is applied to the system, the woofer would reproduce everything below 1 kHz and the tweeter would reproduce everything above 1 kHz. Because the overlap in the response of the drivers has been eliminated, the system will exhibit a smooth transition from the woofer to the tweeter. The difference between real crossovers and this ideal crossover is that a real high-pass filter passes some of the information below the crossover frequency to the tweeter. Likewise, a real low-pass filter will pass some of the information above the crossover frequency to the woofer. The further a frequency is from the crossover point the less information is passed to the "wrong" driver.

Crossover filters have the characteristic that for each octave we go in frequency beyond the crossover point the response of the filter is reduced by some fixed amount. This characteristic is referred to as the slope of the filter. Typical crossovers have slopes of 6, 12 or 18 dB per octave. In other words the filters reduce response by 6, 12 or 18 dB with each octave change in frequency beyond the crossover point. (Remember that either a doubling or halving of frequency represents one octave.) As an example, consider the response of a pair of filters with a crossover at 1 kHz and filter slopes of 18 dB per octave. At the crossover point the response of each filter is down 3 dB. Above 1 kHz the

crossover will reduce the response of the woofer at a rate of 18 dB per octave. This means that at 2 kHz the response of the woofer will be reduced 18 dB and at 4 kHz the response will be reduced 36 dB compared to the response at the crossover frequency (-3 dB). Likewise, the response of the tweeter will be reduced 18 dB at 500 Hz and 36 dB at 250 Hz. In comparison, a 6-dB-per octave crossover would reduce the response of the woofer only 12 dB at 4 kHz and the response of the tweeter would be down only 12 dB at 250 Hz. Because of tweeter power handling considerations, the sharper crossover slope of 18 dB per octave would allow the use of a lower crossover frequency than either 6- or 12-dB per-octave filter slopes. An alternative view is that for a given crossover frequency, the greater the filter slope, the greater the tweeter power handling capability.

Multiamplification offers many advantages over systems employing passive crossovers. One very serious problem with passive crossovers is the fact that the filters are terminated by loudspeakers rather than simple resistors. This means that the driver is an integral part of the passive filter and that the driver's electrical characteristics strongly affect the response of the filter. This is not the case with a multiamplified system where the active filter crossover terminates its own filters. The response of the active filters can be established quite precisely and is totally independent of the loudspeaker driver's electrical characteristics.

Just as the loudspeaker driver affects the response of the passive crossover filter, the passive filter has an effect on the response of the driver. Consider a woofer sub-system which is powered through a passive lowpass filter. Ideally, the filter would only serve to attenuate the high-frequency response of the woofer

above the crossover frequency. However, because of the series resistance of the low-pass filter the low-frequency response of the woofer will also be affected. The unwanted series resistance, even though it might be rather small (maybe 1 ohm), will effectively reduce the damping factor of the amplifier and therefore raise the total "Q" of the woofer sub-system. Unless the loudspeaker designer has made an allowance for the additional resistance in the initial design the result could be a peak in the woofer's low-frequency response. Also, some of the amplifier power will be wasted in the crossover resistance. In a multiamplified system the drivers are connected directly to the amplifiers and these problems are avoided.

Some of the other benefits of multiamplification are the result of using separate amplifiers for each driver. For example, high-frequency distortion products from the woofer amplifier are not reproduced by the tweeter. The woofer system can actually be driven into significant amounts of distortion and, as long as the tweeter system is not overdriven, listeners will perceive clean sound (up to a point at least). This characteristic of multiamped systems makes them especially attractive when high sound pressure levels are required (such as for concert sound systems, P.A. systems or studio monitoring). Also, because each of the amplifiers handle only a portion of the complete frequency spectrum, the amplifiers intermodulation (IM) distortion will be reduced.

A proposed advantage of multiamplified systems that is not well agreed upon among audio professionals concerns a power advantage gained over single amplifier systems using the same total amount of power. One group examines the power available to

mrovfig1.gif (13115 bytes)

mrovfig2.gif (15442 bytes)

reproduce signal peaks and declares a definite power advantage for the multiamplified system; another group performs an analysis based on average signal levels an d says there is no power advantage. Until someone performs a rigorous analysis of the two approaches taking into consideration the dynamic nature of music the question will remain unanswered. Considering that music is quite dynamic with a high ratio of peak levels to average levels, we are most inclined to think that there is some effective power advantage with multiamplified systems. The one point that there seems to be little disagreement on is that multiamplifled systems sound better.

table1.gif (4785 bytes)

Crossover Frequency Selection

In order for the outputs of two drivers to be combined to produce a flat wide-range response it is first necessary for each driver to have a smooth response and that there be a generous amount of overlap in the responses of the two drivers. For the best results each driver should have a smooth response for at least one octave beyond the crossover frequency and for two octaves if possible. Less overlap than this will result in a rough response through the crossover region. Based on overlap considerations it is best to set the crossover frequency at the center of the overlap region. It is also necessary to consult the manufacturer's "lowest recommended crossover frequency" for the tweeter, as operation below this frequency will probably give poor results.

Because loudspeaker drivers start to become directional with increasing frequency it is necessary to establish an upper usable frequency for woofers and midrange drivers. In Table 1 we have fisted "conservative" and "highest recommended" upper frequency limits for various driver diameters. Using most drivers above these limits will result in rather narrow radiation patterns (i.e., "beaming").

Now, after introducing you to several design problems and techniques and the above-mentioned table of parameters, we will leave you-until next month-to mull over the crossover information. Next month we will describe some of the problems with present crossover designs and their possible solutions.