The big picture
The audio chain looks like this:
Original performance > Studio recording > CD > DAC > preamp > power amp > speakers > playback room
Of course there will be variations, but by far the three most critical parts of the chain are:
- Room acoustics
You have no control over the recording, so that leaves the speakers and the room.
Synergy is a popular word in audio but it's often used where it matters least. The most critical synergy exists between the speakers and the room. They should be considered together, rather than simply choosing speakers then patching up it's problems with treatment.
Conventional speakers pay little attention to their interaction with the room. There are three types of speakers that stand out as having a better thought out room interaction - omni, dipole and controlled directivity.
Omnis radiate evenly in all directions. This can be desirable for a background music system, in which their large sweet spot is useful. If you don't have a listening chair, and listen well off axis then this type of speaker can be well suited. For critical listening they may not be the best choice.
Dipoles reduce lateral reflections (ceiling and side wall and potentially floor as well) while adding extra ambient delayed reflections. In effect this is like having room treatment without making the room sound dead. You can aim the null towards the first reflection point and cancel out first reflections off the front and rear walls.
Controlled direcitivity speakers seem to have the most compelling room interaction. With a waveguide they are able to achieve a flat response over a wide area. This means the room reflections are less coloured. The higher directivity means less absorption is required and this is very desirable. In a small room there is a trade off between dealing with reflections and making the room dead. Too little treatment will leave the problems. Too much will make it sound dead.
The key factors
1. Controlled directivity
A speaker should aim for a relatively flat response across a 60 - 90 degree window, and this should be achieved over a wide bandwidth as much as possible.
2. High efficiency
A speaker should not be required to use it's full rated power to achieve output targets. Typical drivers have around 5 db of power compression at their full rated power. Efficiency is a key factor here and is also a big part of dynamics.
3. Low compression
A speaker should be designed to avoid dynamic compression of any kind. This can relate to the input power (power compression), mechanical effects as well as thermal dynamics and BL vs excursion issues.
4. Absence of resonances
Through extensive blind testing of speakers, Harman Audio determined that the absense of resonances was a key indicator of perceived quality. Issues such as cone breakup are key offenders.
5. Distortion profile
A speaker should have low overall distortion, especially with tall order harmonics. THD has proven ineffective at indicating perceived accuracy, however distortion is still important. Ideally we should aim for low distortion of all types, with increasing emphasis on the more obnoxious components.
6. Wide bandwidth
For a music only system, 40 - 20k is a worthwhile goal. Some may prefer to go a little lower depending on music preferences. For a home theatre system the extension should at least aim to reach down to 20 Hz, or perhaps as low as 15 Hz for bass nuts. I don't recommend extreme goals here, since a great price is paid for extra extension. For a music only system I would much prefer to achieve 40 Hz with high efficiency than push an inefficient subwoofer to pump out 25 Hz at high levels.
7. Minimised very early reflections
Early reflections cause many problems. The worst example is perhaps a speaker placed right in a corner. It is not only the bass that goes out of control - early reflections create many problems. The sound is highly coloured, with boomy/peaky bass and very poor imaging. Where speakers are placed near any boundary, acoustic treatment is necessary. Ideally it would be better to either avoid this completely, or flush mount.
Another form of very early reflection is baffle diffraction. Sound waves travel along the surface of the baffle until they reach the corner. They diffract at that point, causing a secondary delayed source.
8. Live vs dead acoustics
The room design and treatment should aim to avoid problems without the over-use of absorption. Ideally this would involve room design as well as careful treatment which includes absorbers and diffusion panels. Too little leaves the problem. Too much makes the room lifeless.
9. Smooth bass response & rapid decay time
Every room has modes in the bass range typically from 40 - 200 Hz. These cause the bass to ring out in the time domain. Modes create peaks and dips in the response and decay at a slower rate. This causes the bass to sound unnatural - boomy at one moment, missing the next.
It is possible to flatten the frequency response without fully correcting the problem. It is necessary to resolve the time domain as well.
10. Time alignment
Time alignment is an essential part of maintaining accurate imaging. Crossover filters as well as misalignment of acoustic driver centres cause the sound to arrive at diferent times. This can be corrected and allowed for in the design.
11. Response shaping
A ruler flat response is the ultimate only in a perfect world. In reality, some response shaping is a better goal. Overall a flat direct field response (from the speaker without any room impact) should be the starting point. The tweeter level should then be reduced slightly, typically to around 2db. Then with a farfield measurement, the bass level should run at about 6 - 10 db higher the the midrange. This combination will be perceived as accurate with decent and balanced, although the bass may be too much with movies where the bass level is already increased.
12. User adjustment
Recordings vary a great deal, with various tonal balances. Ideally it should be possible to adjust the bass and treble level quickly and easily, although tone controls will often be inadequate since they provide EQ that will often mess up the response.
Now, let's see how these goals can all be achieved >