Author: Mark Wheeler - TNT UK
Written: March-June, 2006
Most loudspeakers in domestic audio are "passive" transducers. their function is to convert an electrical signal, 'arriving' at its terminals, into an acoustic signal to arrive at our ears. In theory this loudspeaker should add nothing in the process. Indeed with conversion efficiencies typically in single figure percentages (typically 1-4%), passive loudspeakers succeed in wasting most of that electrical energy, so expensively produced by the amplifier. There is no gain in a passive loudspeaker, only varying degrees of loss.
The term "active loudspeaker" can now mean any loudspeaker system that contains its own amplifier (e.g. those for computers or i-pods), but previously only referred to loudspeaker systems that divide the frequencies for each drive unit before power-amplification, using an active crossover. Wharfedale launched the "Active Diamonds" some years ago, that contained an amplifier and passive crossover within the loudspeaker cabinet, but before then the term "active system" universally meant one where the crossover network divided the frequency bands at line-level voltage and separate power amplifiers were employed for each driver.
That's right, a separate amplifier channel for each frequency band (or even each drive unit if the system has only one bass-unit, one midrange and one tweeter per channel). This replaces the familiar crossover filter network between power-amplification and drive-units. In this two-part article we will consider how dividing the frequencies in the millivolt domain (between pre-amplifier and drivers) is better than in the whole volts domain (between the power amplifier and drivers).
In the pro-sector the active crossover has ruled the PA and monitoring speaker market for several years from quite modest quality levels upwards. Given that pro-gear has to earn its living and justify expenditure to an accountant, the implication is that this is a cost effective loudspeaker solution.
The inherent problems of passive crossovers are compounded by the priorities that have become a dominant hegemony since the 1960s. In the early days there was more diversity of design and more cross-fertilisation between the professional and domestic markets as production runs were so small that it made sense to be able to sell in both.
When I first became absorbed in this hobby I wrote to every manufacturer and importer I heard about, asking for sales literature, usually by pretending to be older than my 13 years. I often received white papers and patent documents among the sales-blurb and this is where I began learning. I did notice even then that the big American pro-sector companies, like JBL, talked about how their designs reduced distortion (just like modern amplifiers at that time) and talked about their innovations in the motor (magnet, spider & voice-coil) assembly of their products.
Some of the British manufacturers tended to boast of their expertise at reducing cabinet and cone colouration by lossy methods. Others wrote of of the rigidity of their cones and linearity of surrounds, while others of their phase coincidence at the crossover point! I soon reached that realisation where I wanted one manufacturer's motor assembly with another's cones and surrounds, in yet another's cabinet, and my DIY days were born.
One major difference in emphasis between UK and US priority was adherence to flat frequency response. Americans tended to mention "frequency range" in a vague promise of power-bandwidth, whereas the British offered tightly defined small-signal frequency response with plus & minus limits clearly advertised. Unfortunately the drive to achieve this with passive crossovers resulted in complicated multi-way speakers with ever increasing filter complexity in some brands resulting in poor transient response and wild phase-shifts across that millpond-flat amplitude response. I owned a pair of 4-way monitors that so obsessively sought a ruler-flat response to beyond 20kHz that the crossover board was worth big-money for its scrap copper value alone. The sound was mush. Some of the old parts illustrated here came from those crossovers.
Let's get one hoary hifi dispute out in the open straight away: "Back EMF". Whether or not back EMF (back electro-motive-force or voltage) affects sound, whether or not bi-wiring reduces the problem (as claimed), passive crossovers will be implicated in enabling and compounding the phenomenon. If cone-mass momentum continues to move the coil in the magnet gap for a few microseconds after the signal changes, there will be a voltage generated, because the coil is connected at both ends. This voltage will be generated and be divided in proportion to the resistances in the local circuit. The local circuit is the passive crossover, where the back EMF, if it exists, will combine with the input signal in a completely non-linear relationship. This is not even-order distortion, this is not odd-order distortion, this is not even high-order distortion, this is DISORDER DISTORTION (to coin a phrase). Passive bi-amping (using separate amplifier channels for each drive unit, but with a passive-crossover leg to each driver) would be a half-way measure, but spurious back EMF (voltage) even then adds to the amplifier signal voltage across the inductors, resistors & capacitors of the passive crossover causing non linear distortion. It will only be eliminated by eliminating the passive crossover.
In the active configuration the output devices or output transformer are directly connected to the drive-unit terminals (via the connecting wire), thus the so-called 'damping factor' (more accurately, impedance ratio) actually has the desired effect, just like at line-level where the low source impedance matched by a high input impedance, for low-noise and wide even bandwidth.
Big inductors in the passive crossover are in series with the bass unit. Smaller ones may be parallel to the treble unit and there may be even more both series and parallel to the midrange unit where fitted. Quite apart from the linear resistance and inductance of this component, there are more damaging non-linear distortions. Whether air-cored, ferrite cored, laminated silicon-iron cored, flat-wired, litz-wired silver wired, or coated in pixie dust, it is still nothing more than a BIG COIL OF CONDUCTOR. How do we make electric heaters? With a big coil of wire, that's how. As conductors heat up their resistance changes, so an underspecified inductor is a primitive time-delayed compressor at speaker voltage levels, at very small signal levels an inductor is probably more linear than capacitors, but the conditions inside a speaker cabinet are much more hostile.
When transient high signal levels are passed, something worse happens. The magnetic field generated can saturate a ferrite core, causing a loud 'crack' sound on really heavy peaks, and a fizz on smaller peaks, which is surprisingly easy & fun to determine by experiment. Air cored inductors do not suffer from this, but they need many more turns of wire to achieve the same value, which increases the DC resistance, and makes the heating problem worse (I have measured over 2 ohms in a commercial 8 ohm speaker, whose voice coil R was 6 ohms, thus a third of the amplifier power was being consumed to heat up the crossover). So the resistance and power handling can be improved by larger gauge wire, hence the great big bobbins common in better quality crossovers.
The magnetic field generated by every inductor will induce a voltage in every other inductor on a similar plane on the x-over board. This adds more non linear distortion, and if the crossover can vibrate, which it must inside the speaker cabinet, yet more non-linear distortion is added. Hence no loudspeaker with any pretence at quality should ever house a passive crossover inside the bass chamber of a loudspeaker.
Resistors? Many of the same problems as inductors apply to resistors, plus parasitic capacitance, and some are more inductive than others and their temperature gradient varies...etcetera, ad infinitum.
Capacitors have long been the topic of debate in audiophile circles. At large signal levels in loudspeakers there are even more problems. Try connecting a crossover capacitor across the output of an amplifier (making sure it will not damage the amplifier) and play some music through the cap. Hear it? the capacitor is making a sound. This just adds to the non-linear distortion of the passive filter.
The massive output stage of the universal solid-state power amplifier is mostly needed to drive the passive crossover. The over-specified iron of the popular single-ended triode amplifier needs to be able to handle steep-slope phase shifts of passive crossovers (unlike its brethren destined to drive single-driver horns) and the complicated feedback arrangements of push-pull tetrode and pentode amplifiers are needed to pull-down the output impedance low enough to interface linearly with reactive multi-way filter networks. Amplifiers could be much simpler (and cheaper) if only asked to drive one speaker drive-unit each.
With that catalogue of problems, it is a wonder anyone ever though passive crossovers belonged in anything more elaborate than a portable Dansette. With cheap digital amplification lower in cost per channel than the lumps of electronica in the passive x-over, the time has surely come to bury the passive crossover with fibre needles and cassette-tapes.
"many loudspeakers suffer from ringing, caused by the crossover chokes, stored energy which is released slowly, completely destroying the character of the original sound, low frequency are extended at the expense of distortion, all percussive bass instruments are made to sound as if they emanate from a large cardboard box", says Bill Dyer, pro-audio grandee and founder of Dyer Digital Audio Systems.
3-way speakers have even more complicated passive crossover filters so benefit even more from multi-amping between crossover filters and drivers. Steep-slope filters (higher order filters) have more components than 1st order, so benefit more too. The simplest 2-way 4th order passive crossover network cannot have less than 4 capacitors and four inductors in each speaker and the simplest 2nd order three-way cannot be made with less than the same, assuming the impedance and the sensitivity of every driver is the same and linear over its passband. The real world is far more complicated than this, and I have seen a first order crossover with a very complicated filter network to integrate two drivers.
The disadvantages of overly-simple passive crossovers (the single cap to the tweeter for example) are equally detrimental in different respects. It is almost impossible to design a perfect driver mechanically, so the filter is useful to prevent the driver being driven by high frequencies that will merely excite resonances, while producing no linear output (at the top of bass-mid driver ranges) and a simple 2nd order filter will improve midrange colouration on most speakers. The mechanical alternative to the electrical filter is effectively a mechanical filter of various doping and damping compounds that add mass to the moving parts of the driver, and are temperature variable too. Low-order tweeter high-pass filters do not block enough of the higher-amplitude lower-frequencies, causing intermodulation distortion at best, then overload distortion, before melting with sustained high levels.
The overlap region where both bass and treble driver are both making output is much wider with a gentle slope crossover. Thus there is more interference between them over a wider band and at higher amplitude. A first order crossover is only 6dB down an octave away from the crossover point. Thus for a 2kHz crossover in a 90dB speaker, the bass unit is still emitting 84dB at 4kHz, which interferes (either constructively causing a peak or destructively causing a notch). Likewise the treble unit is still struggling with an input of 84dB at 1kHz, which may be perilously close to its principal resonant frequency. Merely removing complexity from passive crossovers is not enough.
The frequencies a couple of octaves either side of the crossover region are again subject to big phase shifts through the passive crossover, demanding that amplifier current and voltage are not supplied in the nice cosy relationship expected with a well behaved linear resistance of the nominal 8 ohms. It is hard to decide which gets more upset here, the amplifier trying to deliver an evenly distributed power bandwidth or the drivers whose swinging impedance (mirroring their resonances) are not being well matched by a low amplifier output impedance. Indeed the driver is being mismatched by a fluctuating output impedance that is the sum of the amplifier and the x-over.
All amplifiers then have to be designed to cope with the whole gamut of possible wild gyrations of crossover impedance anomolies. This is why some amplifier-speaker combinations depend on their mutual synergy. The single ended triode designer has to choose whether to maximise the swing of the output valve, relying on the end user choosing a single driver speaker without a crossover (like a Lowther) or compromise the voltage available to allow for higher phase-shifting current demands of more conventional 2-way designs. If speakers all behaved as linear resistances there would be no need for high instantaneous current monster amplifiers. Indeed, the main task facing domestic audio amplifiers is driving the crazy impedance curve offered by the crossover.
Finally, the commonest argument for active crossovers is that they operate at lower voltages and currents than passive crossovers. A lot of amplifier energy gets used heating crossover components instead of moving air. Differing driver sensitivities are matched by resistors in passive crossovers; what a waste of expensive amplifier power. Driver impedance curves are often tamed by Zobel networks that include energy wasting resistors in passive crossovers. It is also more difficult and expensive to make a big capacitor with low ESR and low distortion than a little one. It takes a much bigger capacitor to turn over the same frequency with an 8 ohm resistor than with a 20 kohm resistor. And that biggy also has to handle much bigger voltages and currents. It costs more. It distorts more. Its tolerance is harder to achieve.
The driver impedance is a lot less predictable than the input resistor of a gain stage. As the driver voice coil heats and cools its resistance changes dramatically, so the crossover frequency of a passive crossover changes by a similar order-of-magnitude. If the frequency response of a passive crossover speaker was measured at 2 volts it will be different at less than 1 volt and very different at 10 volts. All of these conditions could happen in the same 3 minute song.
Apart from cost and the vested interest of manufacturers to keep persuading enthusiasts to buy a new amplifier then to buy a new pair of speakers, then to buy a new amplifier then to buy a new pair of speakers, then to buy a new amplifier then to buy a new pair of speakers, then to buy a new amplifier then to buy a new pair of speakers, then to buy a new amplifier then to buy a new pair of speakers, ad infinitum, I cannot think of any reason why domestic audio remains committed to this clumsy technology.
Even with modest sized speakers and modest amplifiers my experience has been that the same amount of money is better spent on an active set-up. As soon as a system reaches separate pre-amplifier and power-amplifier, whether it is solid-state, class D, push-pull valve or single-ended triode, the next step ought to be an active crossover to suit that technology and an extra stereo power amplifier.
Go on to Part II
Copyright © 2006 Mark Wheeler - www.tnt-audio.com