The aim of hi-fi is to reproduce sound as accurately as possible. This means that it must reproduce the correct musical note, at the correct time, for the correct length of time, and as near as is practical at the correct volume. Of course, “correct” means that which was on the original recording.
The errors produced by hi-fi loudspeakers are generally far greater than those produced by the other components in the chain. It is often said that deciding which is a better loudspeaker is a subjective judgment. This is not true (assuming that the amplifier and CD player have been competently designed and do not have gross errors).
If a loudspeaker is any good then the bass end of the frequency range will play a recognizable tune and not just a loud, repetitious, dull thud. It should be possible to distinguish whether a bass guitar, double bass or large drum is being played. Few recordings have been made where the bass player can only play one or perhaps two notes on an unknown instrument. The bass notes should not be too long. In comparing speakers the one where the bass notes are of shorter duration is the better loudspeaker. This applies to any note in the audible range but it is generally easier to judge low notes.
The next easy test is the stereo image. If the sound appears to come from the loudspeakers with a gap between them then this is wrong. If the sound comes from a restricted area somewhere in between the speakers then this is also wrong. Most stereo recordings spread the music right across the whole width between the speakers and each instrument should stay in position and not appear to wander about. The more defined the image the better the speaker. There are some recordings where instruments are located to the left of the left hand speaker and to the right of the right hand speaker – e.g. Bob Marley & The Wailers – Jamming – Island Records.
The next consideration of the speakers which you are auditioning is the timing or phase shift across the frequency range. If the music sounds boring, or even worse, it sounds as though the band had a hangover when the recording was made, then this is a sure sign that the timing is wrong. I use a recording of a piece by Benjamin Britten, Diversions for Piano and Orchestra Op. 21 – Finale – Tarantella, to check the timing of the loudspeakers. If the music makes you think “no wonder this is not the world’s most popular composer”, then the timing is wrong in the speakers. When music makes sense and is detailed, then you are listening to a better speaker.
The volume of sound reproduced by a loudspeaker at each frequency across the audible range is called the frequency response. A frequency response that is flat to within 2.5 dB of its average level is very good. A deviation of 3 dB from its average level would mean that at a particular frequency it would be half as loud as it should be and at some other frequency it would be twice as loud as it should be. Fortunately our hearing is not very sensitive to such errors in sound volume and these enormous variations are not so important as are errors in the pitch of the note and its timing and duration.
The performance of a loudspeaker can be predicted by mathematical calculation. This is particularly true at bass frequencies and there is no technical reason why a bass note should not be accurately reproduced. However, it is essential for accurate reproduction that the loudspeaker should be critically damped, but this is more expensive than it is to produce one which is underdamped. “Big fat bass” is typical of under damping of bass frequencies and this type of reproduction seems to be favoured by reviewers in the more popular magazines. Such underdamping causes the bass notes to be louder and longer than they should be and greatly reduces the musical detail reproduced by the loudspeaker. Underdamping also makes it much more difficult to make sense of music and can blur its clarity. In extreme cases it produces “one note bass”. The greater the degree of underdamping the less the production costs and I suppose the greater the profit margin.
It is often alleged by reviewers that a particular loudspeaker lacks bass when in fact it is critically damped and reproducing the recording accurately. The lower bass end of the frequency range should be rolled off at 4 dB per octave below a frequency determined by the diameter of the bass unit to reproduce a flat response in the “average” room. This is because the loudspeaker is not radiating into an unrestricted space as it would were it located up a tower in the garden. The energy from a loudspeaker bass unit at low frequencies is dispersed equally in all directions and some reflects off the floor and the walls before reaching the listener. Since the wavelength of sound at 40Hz is about 27ft. or 9m, the listener if 10ft. away from a loudspeaker receives these reflections more or less in phase with the direct wave, (that is they arrive at the same time). As the frequency rises the bass unit becomes more directional and does not radiate equally in all directions. A frequency is eventually reached where the bass unit becomes so directional that, if it is not pointed towards the listener, a “hole” appears in the frequency response, and such speakers need to be placed close together. The bigger the bass unit, the lower is the frequency at which it begins to “beam”, and in general the closer thogether the speakers need to be placed.
The “crossover” is an electrical circuit that prevents higher frequencies being supplied to the bass unit, and instead feeds these frequencies to the tweeter. Lower frequencies (below about 3O00Hz) are responsible for the great bulk of the power so that the tweeter only needs to handle a small percentage of the total power. This is fortunate because it is difficult to design a tweeter that will handle significant power in a correct manner. This limitation introduces severe problems in designing an accurate loudspeaker.
The difference between one kind of musical instrument and another is entirely due to differences in the quantities of the harmonics and their phase relationship, compared to the fundamental. A violin and a flute sound different when playing the same note because the harmonics produced by these instruments are of wildly differing magnitude compared to the fundamental.
A loudspeaker should reproduce the recorded fundamental and harmonics in the correct proportions and at the correct time (i.e. phase relationship). It is possible (in fact very likely) that the harmonics will not be produced at the correct time with respect to the fundamental. This alters the sound of an instrument; with many loudspeakers it is impossible to tell whether a violin is a Stradivarius or one bought at the local second-hand shop. It should sound obvious which instrument is being played and if it is an orchestral piece then a 90-piece orchestra should not sound like a 15-piece ensemble. If it does then the reason is that the rest are lost in a forest of resonances and phase errors. Resonances in the walls of the cabinet or any other component of a loudspeaker including the bass unit cone and chassis can only detract from the quality of the sound reproduction by reducing the detail and the imaging and in bad cases sounding “boxy”. Resonances in any part of the bass unit or tweeter will also result in non-linearity of the phase response of these units and can contribute to the muddiness of the sound reproduced.
Simple first order crossovers will produce some phase shift. Second order crossovers will produce greater phase shift and introduce an anomaly about the crossover region. Third order crossovers will produce even more phase shift but without an anomaly but on the down side they also produce resonances. A crossover is essential to protect the tweeter from bass frequencies that would destroy it. The higher the order of crossover the more protection is afforded. It would appear that some reviewers do not recognize that each type of crossover has a characteristic sound because of the different phase shifts across the audio range and their inherent resonances. A full third order crossover has a 360° phase shift across the audible frequency range and a fifth order has a 720° shift. This means that the harmonics of a particular musical note are delayed in time by different amounts. Some will be going backwards instead of forwards and all will be starting too late. It also means that they stop too late. If the woofer and tweeter are not linear in their phase response this can make matters worse. A manufacturer should consider the phase characteristics of the woofer and tweeter when deciding which type of crossover to use – this does not always happen. There is no such thing as “euphonic distortion”, which some manufacturers and reviewers think improves the music. It should be left to the artists, conductors and record producers to decide how the music should be interpreted. A “euphonic distortion” may be appropriate for some recording artists but will surely insult the rest.
One way out of this quagmire is to use a simple first order crossover of modified form, which produces less phase shift than a standard first order. The problem with this approach is that the protection afforded the tweeter is not adequate for general-purpose use. You cannot use such a speaker when having a party and playing music extra loud. It will only cope properly with listening at home at normal listening levels. If what you want is disco type sound then this is not for you. The advantages of a first order crossover in a loudspeaker using a bass unit and tweeter with linear phase characteristics and critical damping are immediately obvious to the listener. Bass notes not only have correct pitch but also reveal texture because all the harmonics are pretty well in their correct phase relationship and there is negligible overhang. The stereo imaging is greatly improved, the timing perfect and instruments recognizable. Singers sound natural and the dynamic range is reproduced with accuracy. For enjoying music nothing else comes even close.
Joe Akroyd © 1998, all rights reserved.
Reproduced with permission from original www.roydaudio.com website.