Some random thoughts on speaker system design

Too large a topic to do anything other than jump in.

Before every design the first thing to consider is the intended application. What are the environment, bandwidth, minimum/maximum volume and distortion limits? Reverberent or reflective environmental considerations will point to controlled dispersion or room treatments and speaker placement also.

Budget should be beyond obvious, just don’t get too out of proportion, a $15 midrange and a $150 tweeter may sound good together but the less costly item will probably put limits on some aspect of performance. There are exceptions to every rule.

Crossover point selection involves (usually) more than generalized rules like 3 or 3.6 octave spread (for 3-ways) as well as minimizing disturbance of the vocal range – 300 Hz to 3 kHz is the usually accepted definition of the vocal range. Wavelength, directivity, piston/bell modes, resonance, mounting height, spacing, baffle dimensions, edge refraction, sensitivity, impedance and external resistance are some considerations that interact. All crossovers cause phase shifts, higher orders generally cause more phase shift - driver location can compensate at 1 frequency. Floor reflections are often neglected in design consideration (bad interference can result in the lower midrange). Passive equalization and resonance filtering are sometimes required when using "exotic" materials like fiberglass, Kevlar, titanium and even some paper cones. Single suspension domes can have linearity problems at high volumes. Some popular crossover points are; 80 to 120 Hz from subwoofer to midbass, 300/500/700 Hz into the midrange, 3.5 or 5 kHz to the tweeter (8 kHz for a supertweeter) - two ways often cross between 1 and 3 kHz. I like to have an octave response beyond the crossover point to minimize problems in the passband, even then tweeters with a high impedance at resonance can need special treatment like a higher order crossover or a notch filter.

Sealed enclosures are the simplest, they roll off at 12 dB/octave below resonance. Tuned ported are 4^th order and will play louder with lower distortion due to reduced cone excursion in the lowest octave, but below the port tuning they "unload" and can use filtering for protection. 6^th order ported include a filter/equalizer to extend the low end. Woofer "Q" determines which style of enclosure is compatible. The guidelines are pretty fuzzy; .2 to .3 for horn; .2 to .5 for ported; .4 up for sealed; .7 up for open back. Bandpass boxes have become popular for subwoofers, 4^th order has 1 ported and 1 sealed chamber, 6^th order both chambers are ported, they seldom have much over 1 or 2 octave bandwidth.

I try to voice the speaker for a balanced response (no range emphasized over any other), as flat as possible within the intended passband. I let the record producer/engineer set the tonal balance. I will admit to letting the general music preference set some parameters like power handling capacity. Chamber music does not really stress a speaker system where punk or synthesizer does.

I find doppler distortion to be extremely fatiguing for extended listening above 80 dB SPL. This is the basis of my dislike of most Bose products. At moderate listening levels they sound fine, but I can’t take them loud for long. Levels of harmonic distortion totally unacceptable in electronics are common in most direct radiator systems. Horn based systems generally have much lower distortion but are more difficult to integrate into systems with non-horn components.

Jeff Robinson