Crossover Basics  9

The main factors in speaker cable selection are therefore series resistance and inductance. If these parameters are less than 100 mΩ for the round-trip resistance and less than 3 μH for the total inductance, any effects will be imperceptible. [7] These conditions can be met by standard 13-Amp mains cable (I’m not quite sure how the equivalent cable is labelled in the USA). This cable has three conductors (live, neutral, and earth) each of 1.25 sq mm cross-section, made up of 40 x 0.2 mm strands. Using just two of the three conductors, a 100 mΩ round-trip resistance allows 3.7 metres of cable. The lowest cable resistance is obtained if all three conductors are used, normally by paralleling the neutral and earth conductor on the cold (grounded) side of the cable; the maximum length for 100 mΩ is now 5.0 metres, which should do for most of us. This three-conductor method does give what I suppose you might call an “asymmetric” cable, which could offend some delicate sensibilities, but I can assure you that it works very nicely.

The loudspeaker cables that I have in daily use are indeed made of such 13-Amp mains cable, bought from an ordinary hardware shop nearly 40 years ago. Should a passing audiophile query the propriety of using such humble cabling, I usually tell them that with so much passage of time in regular use, the electrons have been thoroughly shaken loose and move about with the greatest possible freedom. I do hope nobody reading this book is going to take that seriously.

The Advantages and Disadvantages of Active Crossovers

Here I have tried to put down all the advantages and disadvantages of the active crossover approach.

Some of them may not be very comprehensible until you have read the relevant chapter of this book. My initial plan was to attempt to put them in order of importance, but this is not an easy thing to do.

The order here is therefore to some extent subjective, if I may use the term . . .

The Advantages of Active Crossovers

The advantages of active crossovers are:

 1. The over-riding advantage of an active crossover is it offers ultimate freedom of design, as virtually any frequency or phase response that can be imagined can be used.The filter slopes of the crossover can be made as steep as required without using large numbers of big and relatively expensive components.Any increase in passive crossover complexity means a significant increase in cost.

 2. The design of passive crossovers is restricted by the need to keep the loading on the power amplifier within reasonable limits. With an active crossover, correction of the response for each driver is much simpler, as it can be undertaken without having to worry about the combined load becoming too low in impedance for the average amplifier.

 3. The design of passive crossovers is further complicated by the need to keep the power losses in the crossover within reasonable limits. The losses in the resistors and in the inductors (because of their inevitable series resistance) of a passive crossover, especially a complex design employing highorder filters or time-delay compensation, can be very serious. In a big sound-reinforcement system the losses would be measured in tens of kilowatts. Not only does this seriously degrade the power efficiency of the overall system by wasting power that could be better applied to the drive units,

10  Crossover Basics

but it also means that the crossover components have to be able to dissipate a significant amount of heat and are correspondingly big, heavy, and expensive. It is far better to do the processing at the small-signal level; the power used by even a sophisticated active crossover is trivial.

 4. If one of the power amplifiers is driven into clipping, that clipping is confined to its own band.

Clipping is usually less audible in the bass, so long as there is no intermodulation with high frequencies. It has been stated that an active crossover system can be run 4 dB louder for the same subjective impairment. This is equivalent to more than twice the power but less than twice the perceived volume, which would require a 10 dB increase in sound pressure level SPL.

 5. Delays can be added to compensate for differing acoustic centres for the drive units quite easily.

Passive delay lines can be built but are prodigal in their use of expensive, lossy, and potentially non-linear inductors, and as a result have high overall losses.

 6. Tweeters and mid drive units can have resonances outside their normal operating range, which are not well suppressed by a passive crossover because it does not put a very low source impedance across the voice coil. The presence of a series capacitor can greatly reduce the damping of the main resonance, [8] and it is also possible for a series capacitor to resonate with the tweeter voice coil inductance, [9] causing an unwanted rise in level above 10 kHz or thereabouts.

 7. Drivers of very different sensitivities can be used, if they happen to have the best characteristics for the job, without the need for large power-wasting resistances or expensive and potentially nonlinear transformers or auto-transformers. If level controls for the drive units are required, these are very straightforward to implement in an infinitely variable fashion with variable resistors. When passive crossovers are fitted with level controls (typically for the mid unit or tweeter, or both) these have to use tapped auto-transformers or resistor chains, because the power levels are too high for variable resistors, and so control is only possible in discrete steps.

 8. The distortion of the drive unit itself may be reduced by direct connection to a low-impedance amplifier output. [10] It is generally agreed that the current drawn by a moving-coil drive unit may be significantly non-linear, so if it is taken from a non-zero impedance, the voltage applied to the drive unit will also be distorted. This may be related to out-of-band tweeter resonance;

Jean-Claude Gaertner states that tweeters can have increased distortion below 1 kHz. [11] I do not know for sure, but I very strongly suspect that when drive units are being developed they are driven from amplifiers with effectively zero output impedance and that linearity is optimised under this condition. Any other approach would mean guessing at the source impedance, which, given the number of ways in which it could vary, would be a quite hopeless exercise.

 9. With modern opamps and suitable design techniques, an active crossover can be essentially distortion free, though care must be taken with the selection of capacitors in the filters. It will not however be noise-free, though the noise levels can be made very low indeed by the use of appropriate techniques; these are described later in this book.

Apassive crossover contains inductors, which if ferrite or iron-cored will introduce distortion. It also contains capacitors, often in the form of non-polar electrolytics, which are not noted for their linearity or the stability of their value over time. I haven’t been able to find any published data on either of these problems. Capacitor linearity is very definitely an issue because they are being used in filters and therefore have significant voltage across them. It is possible for capacitor distortion to occur in active crossovers too, but the signal voltages are much lower and one can expect the amount of distortion generated to be much less. See Chapter 15, where using the worst sort of

Crossover Basics  11

capacitor increases the distortion from 0.0005% to 0.005%, with a signal level of 10 Vrms. In contrast, the distortion from a passive crossover can easily exceed 1%.

10.With the rise ofAV there is more experience in making multi-channel amplifiers economically.

The separate-module-for-each-channel approach, where each module has a small toroid mounted right up at one end, while the input circuitry is at the other, is more expensive to manufacture but can give an excellent hum and crosstalk performance. The main alternative is the huddle-around- the-big-central-toroid approach, which has some serious and intractable hum issues.

11.If a protection system is fitted that is intended to guard the drive units against excessive levels, then it can be closely tailored for each drive unit.

12.Voice coil heating will increase the resistance of the wire in its windings, reducing the output. This is known as thermal compression. It also increases the impedance of a drive unit, and if it is part of a complex passive crossover, the interaction can be such that there are much greater effects on the response than that of thermal compression alone. In one set of tests conducted by Phil Ward, [12] the voice coil temperatures of four different loudspeakers showed a maximum of 195°C and a rise in resistance of 176%. That sort of variation has got to cause interaction with almost any sort of passive crossover.

13.It has been proposed that active crossovers can allow the modelling of voice coil heating by calculations based on signal level, frequency, and known thermal time-constants. Thus the effects of thermal compression (the reduction in output as the voice coil resistance rises with temperature) could be compensated for. It does however imply relatively complex computation that would be better carried out by digital processing rather than in analogue circuitry. There would have to be A to D conversion of the signal and perhaps D to A conversion of the control parameters, even if the actual crossover function was kept in the analogue domain. Controlling the active crossover parameters with analogue switches or VCAs without compromising signal quality is going to be hard to do. Modern volume control chips have excellent linearity, but they are not really adapted to general control, and using a lot of them would be rather expensive. If you are undertaking this sort of complex stuff then it’s probably going to be best to do all the processing in the digital domain.

Clearly this plan can only work if the crossover is programmed with the thermal parameters for a given loudspeaker and its set of drive units; this information would have to be provided by the

loudspeaker manufacturers, and once again we see the need for the active crossover to be matched to the loudspeaker.

14.Drive unit production tolerances can be trimmed out. It has been said that changes in driver characteristics due to aging can also be trimmed out, but since aging is not likely to be an absolutely predictable process, this would ideally require some sort of periodic acoustic testing. For a reference loudspeaker in a laboratory or monitors in a recording studio this is entirely practical, but it is less so in the home environment because of the need for an accurate measuring microphone, or, more likely, one whose response deviations are sufficiently predictable for them to be allowed for. Extra electronics are of course required to implement the testing procedure.

15.The active filter crossover components will have stable values.The inductors in a passive crossover should be stable (though Isuppose turns could shift under heavy vibration), but the non-polar electrolytic capacitors that are often used have a bad reputation for shifting value over time. The stability of these components has improved in recent years, but it is still a cause for concern. It has been stated that electrolytics in high-end passive crossovers should be regarded as having a lifetime

12  Crossover Basics

of no more than ten years. [13] Plastic film crossover capacitors such as polypropylene show better stability but are very expensive. A fashion has grown up recently for bypassing big passive crossover capacitors with smaller ones—whether this has any beneficial effects is very questionable indeed.

16.The active filter crossover components will not change in the short-term due to internal heating. In a passive crossover the capacitors will have large voltages across them and large currents through them; dielectric losses and ohmic losses in the ESR (equivalent series resistance) may cause these capacitors to heat up with sustained high power and change in value. Non-polar electrolytic capacitors (basically two ordinary electrolytics back-to-back) are considered particularly susceptible to this effect because their relatively small size for a given capacitance-voltage product means they have less area to dissipate heat, and so the temperature rise will be greater.

17.The relatively small capacitors used in active crossover filters can be economically chosen to be types that do not exhibit non-linear distortion—polystyrene and polypropylene capacitors have this useful property. Non-polar electrolytic capacitors when used in passive crossovers are known to generate relatively large amounts of distortion.

18.No inductors are required in active crossover circuitry (apart perhaps for a few small ones at inputs and outputs for EMC filtering). Inductors are notorious for being awkward and expensive. If they have ferromagnetic cores they are heavy and generate large amounts of non-linear distortion. If they are air-cored distortion is not a problem, but many more turns of copper wire are needed to get the same inductance, and the result is a bulky and expensive component. Martin Colloms has stated [14] that if an inadequate ferromagnetic core is pushed into saturation by a large transient, the resulting sudden drop in inductance can cause a drastic drop in the impedance seen at the loudspeaker terminals, and this sort of thing does not make life easier for power amplifiers.

19.Passive crossovers typically use a number of inductors, and it may be difficult to mount these so there is no magnetic coupling between them; unwanted coupling is likely to lead to frequency response irregularities. (It should be said that some types of passive crossover use transformers or auto-transformers, where the coupling is of course entirely deliberate.)

20.When a passive crossover is designed, it is absolutely not permissible to treat a drive unit as if its impedance was simply that of an 8 Ω resistor. The peaky impedance rise at resonance and the gentler rise at HF due to the voice coil inductance have to be taken into account to get even halfway acceptable results. This naturally complicates the filter design process considerably, and

one way of dealing with this is to attempt to compensate for these impedance variations by placing across the drive unit terminals a series-resonant LCR circuit (to cancel the resonance peak) and an RC Zobel network (to cancel the voice coil inductance rise). [10] This is often called “conjugate impedance compensation”, and while it may make the crossover design easier, it means there

are at least five more components associated with just one drive unit, and they all have to be big enough to cope with large signals. There may also be changes in impedance due to changes in acoustic loading across the drive unit’s passband. In an active crossover system the drive unit is simply connected directly to its power amplifier, and assuming that amplifier has an adequate ability to drive reactive loads, the details of the drive unit impedance curve can be ignored.

Determined use of conjugate impedance compensation (also called “conjugate loading”) can turn a peaky but essentially 8 Ω impedance plot into a more or less flat 4 Ω plot at the terminals. In my view, artificially reducing a loudspeaker’s impedance to make its plot flatter is a poor idea, as all amplifiers, so far as I know without exception, give more distortion with heavier loading.