I think it is becoming clear that we have two approaches to speaker design:
- science & measurements-based
- ideas-based
Contrary to what you would expect, (1) is not cutting-edge, pushing forwards the technology, but is passive (in the normal sense of the word), retrospective, and maintaining of the status quo.
It works like this:
1. is based on existing science, and measurements or observations of existing speakers. It is all about 'mitigation'; "What can we get away with.". And science can help to inform us about what we can get away with.
The science shows that ... (NO! We will not use that phrase. "Science tends to suggest that..." is much better) ... humans can't resolve frequency response differences less than X dB, so we will use this as our target.
Science tends to suggest that humans can't resolve phase differences in many circumstances, so phase will not enter our consideration at all; we have a free hand to use whatever filters we like and will only measure frequency response magnitudes. Passive filters will be fine, here. Active crossovers do not appear to be any different in terms of these measurements.
Being hard-nosed measurements-driven people we would like to see the widest, flattest frequency response. Bass reflex extends the flatness of the bass further. In terms of the measurements (and therefore the sound) this is a free lunch. Result!
We have a science-based bonus: science tends to suggest that people like their music distorted, so we don't need to worry too much about distortion. If the tweeter is stretched over a wider frequency range, or we go two-way and subject the drivers to higher displacements than they would otherwise have in a three-way, it's no biggy. It might even sound better.
Science tends to suggest that people prefer a drooping in-room frequency response, so as long as we can show that our box gives an in-room response that gets close to a plausible target, we have achieved our objectives. If we play with the crossovers and 'voice' the speaker (a suck-out here, a phase inversion there), a fashionable narrow-fronted box can do this, so we will use that.
2. The ideas-based approach is different.
It ignores the science. It starts from the
idea that the perfect speaker simply reproduces the signal accurately in all respects. There is no attempt to 'get away with' anything; there is no assumption that it will duplicate an existing speaker. Scientific literature is not consulted. It is not a passive, retrospective exercise.
What is needed in order to duplicate the signal? It includes:
- linear phase crossovers with additional driver phase correction
- Time alignment between drivers
- Sealed drivers
- Active amplification
- More than two ways - to keep drivers in their 'sweet spots' and to maintain better dispersion with frequency
- Steps to improve the dispersion characteristics with active means.
- Steps to actively improve distortion.
Basically, make the speaker cones move in the way we want them to move.
No scientific literature is consulted. Measurements are merely confirmation, not the driver of the design. The speaker is not 'voiced'. The idea is not to see what can be got away with in the individual parameters, but simply to work logically towards the overall result of duplicating the signal. Implicitly, the speaker's in-room response is a result of what you get with such a speaker, not a target.
It could be argued that the (2) approach is wrong because most music will have been mixed for (1) speakers. Maybe people have learned the sound of (1) and so are attached to it. These are reasons why the science based on listening tests is as it is. But it is also a reason why the (2) approach is exciting: hearing music recordings as they have never been heard before, as fresh as if you were in the studio or at the venue.