AFTER ALL IS SAID AND DONE, a hi-fi system is finally judged by its owner ( and by his friends ) in terms of the realism or 'accuracy' it imparts to the reproduction. There is no doubt that to many critical ears one system or, indeed, a specific item of hi-fi equipment may well give a different listening experience from another. The problem is to determine conclusively whether the difference is for the better or worst . There are some cases where the difference is obvious and all agree that , say, amplifier A 'sounds' better than amplifier B. With hi-fi at its nth degree, however, we are always dealing with extremely small, subtle differences and unless one is extremely careful judgment can be colored by personal preferences and sometimes imagination! Orchestra Sound No hi-fi system yet evolved is capable of absolute accuracy of reproduction.
Nevertheless , all systems worthy of the hi-fi label are capable of excellent sound quality. The sound produced by an orchestra in a concert hall undergoes an astonishingly large number of complicated processes before it arrives as sound again in the listening room. It is amazing that the accuracy of reproduction is as high as it is.
Not all hi-fi listeners are regular concert goers and most of us receive the majority of our music electronically from loudspeakers. We have established a fair impression of the 'norm' for this kind of reproduction - from small TV loudspeakers and the mini-speakers in transistor sets - so when we experience something in advance of this we are impressed. Hi-fi to most of us, then, is top quality sound reproduction. To the devotee, though, this is by no means the end of the matter.
Subtle Differences
This highly dedicated enthusiast has often trained himself to detect the most subtle of differences between systems and components of systems; but even for this chap to state conclusively that one item renders greater 'accuracy' than another he must have the original live music datum established clearly in his mind. He must, therefore, be a regular concert goer and possibly music lover.
There are not many of this type of person around who are able to recall from memory the sonic impression experienced live at the concert hall. Other subjective factors may also be influencing judgment.
Comparative Audition
There is a current trend for the evaluation of hi-fi systems and parts of systems to be made by comparative audition alone. A number of amplifiers, for example, may be compared one against the other or against a previously established reference and then judged for sonic desirability. This scheme undoubtedly has its merits; but unless handled correctly literally bristles with ambiguities.
Sonic differences there certainly will be. These are noted mostly between loudspeakers and pickup systems, including record decks; but even different amplifiers and FM tuners 'sound' different. Whether one 'sounds' better or worse than another is left to the judgment of the listener . If a solitary listener makes the judgment the accuracy of the results obviously cannot be very high.
The results are, then, the opinion of one person. Another person may come to a different conclusion! Apart from this danger, the expression of the differences noted presents very real problems. Since this mode of hi-fi equipment evaluation has come into being a whole new vocabulary has been evolved in an endeavor to define some of the subtle sonic differences so experienced. Unhappily, some of the expressions coined are extremely hazy in meaning and can be interpreted differently by different people. This will always be so when endeavors are made to express the performance of hi-fi equipment subjectively.
Nevertheless, a full evaluation of a hi-fi system or an item of hi-fi equipment certainly demands listening tests. The tests must be undertaken by a panel of listeners under controlled conditions. Moreover, each member of the listening panel must himself be 'tested' and weighted. Furthermore, these tests must complement - certainly not replace - the more scientific measurements undertaken in the laboratory.
'Specmanship'
A number of factors have led towards subjective attempts of evaluation. One of these is best described as 'specmanship'. That is, in recent years there has been increasing competition between equipment manufacturers and designers to achieve what is believed to be the best set of engineering specifications.
Current technology makes it possible to design hi-fi equipment with specifications far in advance of those of the earlier valve era; but there is reason to believe that not all these advancements lead to improved sonic performance or, indeed, to greater accuracy of reproduction.
Tests undertaken by our own listening panel and also by other people have indicated that in some cases a preference is shown in favor of some of the earlier valve amplifiers when compared with certain very highly specified transistor amplifiers.
In some areas this has led to the conclusion that valve amplifiers 'sound' better than transistor amplifiers. In our judgment this is not intrinsically true. If such a transistor amplifier is re-engineered to yield parameters similar to those of the favored valve amplifier subsequent subjective assessment is then rendered extremely difficult with judgment often going in favor of the transistor amplifier.
To secure this condition it is necessary to 'de-value' some of the parameters of the transistor amplifier. In other words , it appears that developing some of the parameters of transistor amplifiers to their engineering optimum, as now provided by the state of the art, the subjective experience can be impaired rather than enhanced! Opinionated Approach Another factor relates to the very critical nature of objectively measuring parameter differences and relating these to sonic differences . Such measurements call for highly sophisticated and hence expensive test equipment, engineers and technicians with extensive background experience, with highly developed hearing and an appreciation of live music, and a very reliable listening panel. These requirements are tending to shift evaluation emphasis from the scientific towards the personal 'opinionated' approach. The overall cost of the latter is far less than the former, which is another factor.
After spending many thousands of pounds in the development of a new item of hi-fi equipment, it can hardly be regarded as 'fair ' to the designers and manufacturer concerned if the item is given a relatively poor assessment from elementary listening tests alone.
E.S.P.
The much more scientific approach is to endeavor to discover which parameters have the greatest influence on the sound experience and how they should be regulated for the greatest accuracy of reproduction. Although this is a protracted and costly business there are now signs that this is happening in the more scientifically involved areas of hi-fi.
Our lab, too, has been researching into these aspects of hi-fi equipment design. Some of the hi-fi equipment manufacturers themselves are also investing in research along these lines . Hitachi, for example, is one company which gears electronic design to the subjective requirements . This is called 'Emotional Response Sensation and Physical Characteristics' (ESP for short !). The scheme aims at establishing electronic design procedures based on the sonic experience which compensates for the lack of objective/subjective correlation of the more conventional electronic design approaches . The firm has evolved a way of measuring 'emotion' - translated simply to 'pleasure' and 'displeasure' - scaled to a wide range of subjects and taking account of the known pertinent parameters of a reproduced sound field.
A large number of listeners were employed, and as the various parameters were changed under controlled conditions so the reactions of the listeners were recorded and statistically analyzed. The data resulting are then used at the electronic design stage, which seems to us to be a very logical way of hi-fi design.
Specs Don't Tell All
It is certainly true, then, that equipment specifications fail to tell the whole story about the performance. Indeed, the basic engineering parameters tell very little. Nevertheless , they are essential not only from the design aspect but also so that the consumer can compare one item against another in terms of technical quality and value for money.
There are, though, parameters which do have a close relationship to the auditioning of the equipment and it is a great pity that more manufacturers do not include these in their specifications -- in addition to the basic engineering parameters.
Effect of Interfaces
It must never be forgotten that the listening impression can be greatly influenced by interfaces (see Section 7) and by the acoustics of the listening room.
For example, if a listening panel evaluates comparatively, say, ten different amplifiers with a given pair of loudspeakers, and then the same test is undertaken again with a different pair of loudspeakers it is possible that in some cases the second results will differ from the first . This is because some of the amplifiers may be more favorably inclined to the first pair of loudspeakers than to the second pair, and vice versa.
Similarly, a change in pickup cartridge may give a different set of results again. Some cartridges interface more happily with some amplifiers than others (see Section 5). Moreover, the relative characteristics or parameters of the cartridge and loudspeaker pair can influence the sonic results . Certain cartridges are notoriously bad partners for certain loudspeakers.
The interfacing of one make of preamplifier or control amplifier with a different make of power amplifier can quite significantly affect the listening experience owing to the resulting differences in impulsive characteristics (see Section 7). Tuner/amplifier interfacing is less critical, particularly when the tuner is designed with a relatively low output impedance.
However, differences can certainly be detected when the output impedance of the tuner is not all that low and connection is made to the amplifier via fairly long-length screened cables . It is also now being suggested that the nature of the connecting leads themselves can influence the final sound result *.
* Jean Hiraga, Can We Hear Connecting Wires?, Hi-Fi News and Record Review, August 1977
Effect of Acoustics
If a skilled panel evaluates comparatively a number of loudspeakers operating from a common amplifier and high quality master tape signal source one set of results will be obtained in one listening environment, while a different set of results may well be obtained in a different environment. The acoustical characteristics of the listening room can thus influence the results.
Turntable systems and record decks are affected by resonances, as discussed in Section 5. This is now fairly well established (e.g., see Martin Colloms, Hi-Fi Choice - Turntables and Cartridges, published 1977 by Sportscene Publishers Limited). However, how these resonances are likely to influence the sonic results are related to other resonances in the overall system, including those of the pickup cartridge, loudspeakers and, indeed, the listening room; also by how well the record deck is isolated from the sound field and by the intensity of the field (see under Acoustic Feedback, Section 5) . If you are in a position to make a good quality tape recording from your hi-fi and disc record source, try making such a recording (a) with the amplifier volume control turned right down (monitor on headphones if you wish) and (b) the same recording again but this time with the volume control well advanced for high intensity reproduction. If the record deck is subject to 'coloration' as the result of the sound field inciting resonances , etc. you will certainly hear this when tape recording (a) is compared with tape recording (b) . By now you will have gathered that the permutations are 'endless' . You will also begin to realize the danger of assessing on a conclusive basis the performance of an item of hi-fi hardware by listening tests alone.
Our Listening Tests
As already noted, for our assessments - linked to lab results - we employ a test panel of, at least, ten people. During the course of the tests , and unknown to the members of the panel, we deliberately switch several times to sounds with wide measurable differences between them and also to sounds with no difference at all between them (e.g., a repeat of the reference sound) . If the panel is working accurately we expect to achieve consistency of results from each member of the panel on each repeat. This is a sort of 'confusion' matrix which sorts out the panel members most suitable for this kind of assessment. Also, . from the results obtained with different program indices, this detailed processing leads to 'weighting' of each panel member. This, of course, is merely a rough outline of the techniques involved.
Each panel member is given a sheet comprising 50 or more (depending on the range and type of tests in hand) scales graduated over ±5 with zero at the centre, as shown in Fig. 8.1. These avoid the use of cloudy adjectives and hazy descriptions . The plan is for each member to mark the appropriate scale as to whether the sound he (or she!) is hearing is by judgment above or below the quality and 'accuracy' of the sound previously heard as a reference. In this way the panel members are subjected to the least pressure and defining brain work, which is important.
Fig. 8.1: Marking scales used by our listening panel (see
text for details) .
The panel is made comfortable in a room which is removed from the room housing the equipment under test, the test equipment therein and the comparative switching unit . The listening room is optimized for domestic listening reverberation time (circa 0.4 second average) and is acoustically insulated (avoiding spurious sounds getting in and unsociable sounds getting out!).
Sometimes a second, smaller room housing a smaller listening panel - also acoustically insulated and optimized for reverberation time - is used in parallel with the larger listening room and larger panel . This helps with the assessment of the influence of different loudspeakers on the results . A great deal of information is expanded from the 500+ results so obtained, and a computer may be used to analyze it.
Over the years we have obviously detected how variations of certain parameters can affect the listening experience, and some of these will now be looked at .
Parameters Which Influence Listening
Amplifiers with a greatly extended small-signal, upper-frequency response are commonly less favored in a listening test than amplifiers whose -3dB roll-off point occurs round 35kHz. Also an amplifier whose treble roll-off follows a natural 6dB/octave law generally attracts a higher vote than one whose roll-off is 12dB/octave or more. On the other hand, a too--early upper frequency roll-off tends to give low marks to a transient type music index.
These things are shown in Fig. 8.2.
1 Highly voted
2 Generally poor vote (but see text )
3 Low vote
4 Poor vote on transients
Fig. 8.2: Aspects of small-signal upper-frequency response which have
been proved to influence the listening experience.
An amplifier, whose intermodulation distortion entering the audio passband (20Hz -20kHz) exceeds 0.5 % as two equal-amplitude signals displaced by 1kHz are swept up to 100kHz with the amplifier set and signal adjusted for 1 dB below peak clipping on the composite signal at middle frequencies, is commonly given lower marks than one whose intermodulation distortion under these conditions holds at 0.1 or 0.2%. It is noteworthy that an amplifier whose small-signal upper-frequency response reaches the - 3dB point at not higher than 40kHz is less likely to produce relatively high, in-band IMD than one with an extended small-signal upper-frequency response. This is a function of the slewing rate of the output transistors in the power amplifier, the effect being slewing rate induced IMD and transient intermodulation distortion (TID). More information on TID is given in our The Audio Handbook (Newnes-Butterworths), pages 43-44, 102 and 124- 126.
The power amplifier should not be capable of being driven to slewing rate limiting as the frequency of a signal , applied to a normal input at 1 dB below clipping at 1kHz, is increased up to, say, 100kHz.
Fig. 8.3: Harmonic distortion spectrograms at 20Hz driving signal. (i)(a)
one amplifier at 1dB below clipping and (i)(b) same amplifier at 40dB below
clipping . (ii) (a) another amplifier at 1dB below clipping and (i i) (b)
the same amplifier at 40dB below clipping. Amplifier (i) was judged by
a listening panel less highly than amplifier iii . In fact, amplifier (ii)
was judged the most highly from a large number of amplifiers. This can
be understood from the far 'cleaner' spectrograms of amplifier (ii) and
the relative freedom from mains supply harmonics. Scale 200Hz per division
horizontally and 10dB per division vertically.
Pickup Overload Margin
An amplifier whose pickup overload margin is less than 28dB at 1kHz commonly attracts fewer marks than one whose overload is 35dB or more when the program index is from a disc record. It has also been found that slewing rate induced distortion in pickup preamplifiers seriously detracts from the overall vote given to a disc replay index, particularly when this is rich in transient material.
On some amplifiers certain cartridges, particularly those with a relatively high d.c. resistance (see Section 5) , are given a low vote on a program index containing bass information (also see the comments in Section 5 about the upper-frequency effects of pickup cartridges when loaded into certain amplifiers).
Distortion Distribution
An amplifier whose harmonic distortion distribution from a fundamental signal at 200Hz holds constant in accordance with a constant transfer characteristic at all reproducing levels from 1 dB below clipping to 60dB below clipping invariably attracts higher overall marks than an amplifier whose harmonic distribution changes wildly with changing signal level. We undertook a research program in this respect*, and the spectrograms in Fig. 8.3 illustrate some aspects of the results . An amplifier whose mains frequency components (fundamental and harmonics up to 1kHz or more) are at a very low level of -80dB or more ref. the full output of the amplifier when the rated power of the amplifier attracts higher marks for an ambience index than· one whose mains frequency components are at a higher level and extend to relatively high frequencies (see the spectrograms in Fig. 8.3). Distortion produced by contact resistance of speaker leads etc. have a strong bearing on the results, so changeover of circuits must be performed so that contact resistances are eliminated; ordinary switches are unsuitable.
These, then, are some of the more subjectively correlating parameters of amplifiers. There are many others; but, unfortunately, there is just not room left in this guide to detail them all -- perhaps another guide is called for dealing specifically with this subject!
Tuner De-emphasis
A tuner whose de-emphasis is incorrectly engineered and whose treble roll-off starts taking effect round 5kHz is always given fewer marks than a counterpart whose response is 'flat' to 10kHz. The panel can sometimes detect differences between the frequency response of tuners over 10kHz to 15kHz, depending on the type of information transmitted (our lab utilizes a closed circuit FM stereo transmitter of ultimate state-of-art quality taking signal from master tapes for these tests) . It must be remembered, though, that the transmitting authorities terminate the upper-frequency response at 15kHz, so there is hardly any justification for the tuner designer to maintain his response to a higher frequency.
Indeed, this can often detract from, rather than enhance, the sonic results, particularly on stereo. A tuner should really be equipped with a pilot tone and residual sub-channel filter to avoid the 19kHz tone and other higher frequency rubbish from entering the amplifier. On average, our panel has indicated that a tuner with an effective pilot tone filter auditions better than a tuner of otherwise similar characteristics but devoid of such a filter.
[* Gordon J. King, Amplifiers - Measuring What we can Hear, Hi-Fi News and Record Review, July 1977.\
Some members of our panel can actually hear the pilot tone from the loudspeakers when there is no filtering at the tuner. Moreover, the pilot tone, when of relatively large amplitude (circa -34dB below full modulation without filtering), can incite in-band intermodulation products both in the tuner itself and in the partnering amplifier; and these are detectable under certain conditions and with certain music indexes.
Stereo Separation
A tuner whose intrinsic stereo separation averages round 40dB over the entire frequency spectrum is generally given higher marks than one whose separation is less than 25dB and whose distortion on the breakthrough signal 15% or more.
A tuner whose speaking-channel distortion rises above 1 % on stereo at full modulation always achieves fewer marks than one whose stereo distortion does not reach this level until the deviation rises to ±100kHz (e.g.; 33% above full modulation) . Other factors appropriate to tuners include high residual hum level, particularly modulation hum on strong signals (the panel notices this by consistently marking down on ambience indexes) and restricted and non-linear phase i.f passband.
Interface Intermodulation Distortion (IlD)
Another recent finding is IID, researched by Matti Otala, Jorma Lammasniemi and colleagues, which has been mentioned.
Harmonic Distortion
In certain cases it has been found that a small degree of even-order harmonic distortion is sometimes better subjectively than virtually zero distortion, as this can disguise the high odd-order harmonic distortion of some program signal sources, giving a more palatable sound.
= = = =