A visit to any university's electrical or electronic department's library will reveal vast quantities of books about filters, often from floor to ceiling. Given the huge amount of work expended in this area, and by necessity, another mere scratch of the surface is offered here.
Purists require that equalisation, in any form, must never be used, claiming consequential and objectionable detriment to the 'purity' of any given signal. In the real world the opposite proves to be the case. Notwithstanding the variability that naturally occurs in human perception, most forms of transmission and recording use pre- and de-emphasis techniques to improve, say, signal to noise ratios and frequency responses without which there would be very little point in proceeding further with the relevant technology.
Similarly, defficiencies in the equipment used may require modification to increase listening comfort. Some commercial RIAA deviations can amount to ±6dB from 50Hz to 18kHz. One of the original functions for the inclusion of tone controls in an integrated amplifier was to accommodate differing recording curves like the old Columbia, AES and London types or those used for differing tape recording/playback standards/speeds.
Taking the highest quality signal source available to many domestic users as an example, some tuners, like the Technics ST-9600, had, apart from special 'hi-blend' noise cancelling and 'pink-noise' circuitry, a record function whereby a tape recording could be made before the de-emphasis (post-MPX), playback then passing through the de-emphasis thus reducing FM and tape noise (4 channel MPX outputs - horizontal and vertical - were also available). Many FM MPX decoders use output filters that allow mixing of frequencies above 2kHz to subjectively reduce noise without adversely affecting the stereo image. Examples include
Perhaps the most extreme case in domestic terms is that used by vinyl records (RIAA). Here, bass frequencies will be boosted by some ten times, and high frequencies reduced to a tenth, relative to a reference at 1kHz. Some systems, like the RCA Dynagroove, attempted to remove cutting lathe non-linearities. Similar slopes are used for tape replay. Another striking example are the crossovers inside a loudspeaker, notwithstanding the modifications of a frequency response made to cope with the anomalies that arise when using speakers in any room. Other frequency dependent conditioners are commonplace, including noise reduction systems like Dolby, dbx, DNR, etc, and digital systems (anti-aliasing, etc).
Although it might upset haughty audiophiles, it is a fact that in virtually all situations, the first tool a professional will reach for, after setting initial volumes, will be the equalisation. A mere glance at any mixing desk will confirm this.
It is arguable that, for example, with equalisation signal to noise ratios can deteriorate especially with dirty or worn switches or controls. Some will insist that the 'sonic qualities' of capacitors are objectionable. However, most normal humans indicate agreeableness with the ability to vary a system's frequency response, or the slope of same. Contrary views appear to uphold the belief in perfect systems, speakers, listening rooms and ears that all somehow miraculously combine simultaneously, often by sheer will-power, self-belief, unknown laws of physics or because of the amount of money that has been spent. Dispassionate hardware always contradicts these.
At the same time, signal content that is outside the audio band can cause distortion in the later stages of the system's amplification or even contribute to the destruction of speakers. These can arise from infrasonic tone arm defficiencies or moving-coil cartridges that, because of their low inductance, can easily provide an ultrasonic frequency response. Tuners, tape decks and digital sources can all exhibit unwanted high frequency byproducts or a dangerous DC bias. Improvements will then occur if these are then omitted. One audio bandwidth filter is shown below, but other forms and types abound.
Tolerances of ±5% can give satisfactory results, but to reduce channel mismatches ±1% components are recommended. Best driven by a low impedance source, say 100R, these can precede the power amplifier or tone controls.
Although graphic equalisers are often seen (best placed before an active crossover, or power amplifier), preference was given to parametric controls which, in live performance particularly, proved more useful. Whichever approach is used, a light touch is recommended lest system limitations or architectural resonances predominate. Reducing a system's overall output, even by a relatively small amount, will allow greater flexibility, whilst increasing enjoyment. When recording, even experienced engineers will forget that cutting the low frequencies will have a similar subjective effect to boosting the higher ones. Cutting will give more headroom and less likelihood of clipping. So instead of raising the treble, say, cut the bass and boost the overall gain and see if this sounds better than simply raising the treble. Input peak detectors integral with a system's equalisation can be useful for avoiding distortion.
Even in the most rudimentary audio gear some form of 'tone' control can be found. Here, a simple RC network is placed across the volume control.
In this case, an acceptable level of 'treble' would be set, giving detail, the bass performance being determined by the speaker's enclosure. It should be noted that the use of high-value (500k+) potentiometers can introduce problems because of the stray capacitances between the wiper and the track's ends. At low volume levels, a few picofarads across the upper arm can boost the treble response by several dB at 12kHz or so, or introduce a 'spikiness' to square waves.
More sophisticated Baxandall tone control configurations are universal. With variants as diverse as language (Ramon Vargas Patron offers an examination of the James-Baxandall Passive Tone-Control Networks here), a typical arrangement is shown below
Insertion losses will be near that of the gain available referred to 1kHz and log pots are used which may not match as closely as linear types.
Another version intended for the NE5532/4 opamps which 'recovers' the losses and uses linear pots. To cope with high Qs and frequencies, a good gain bandwidth product (20 x Q^2 x Fc) and low-noise is recommended. Regulated and well-smoothed supplies are a prerequisite, include decoupling for each IC.
Experience has shown that often the depth of cut and lift exceeds requirements. A more subtle control can have less of a spread, say ±6dB, compared to the normal commercial range of ±12dB. This can be switched. Other configurations can be used, eg;
Sometimes manufacturers can integrate tone controls and filters into the feedback path of a power amplifier, eg; Technics SA-100/600 tuner amps, Luxman, et al. This is not recommended since unlooked-for complications and instabilities can arise apart from, usually, a drop in performance. This can be partly due to the circuit having to be 'spread out' around the PCB. Personal preference requires a 'clear' feedback path and separate circuits (as in the SA-700), a failure then in one section will not take another with it, making fault-finding easier, unless of course, a compact objective is required, eg;
Slope or tilt controls can also be seen as has step frequency adjustment. John Linsley Hood proposed a 'Clapham Junction' or spot-frequency step-type lift and cut-type tone control in his Modular Preamplifier (Wireless World, Nov, '82) which is worth a look. However, this is reliant on a large number of mechanical switch contacts whose noise performance could deteriorate with use. Some amplifiers rely on large numbers of switches to give flexibility.
These were often accompanied by so-called scratch and rumble filters intended to diminish inadequacies found in records and record decks with varying degrees of success since, obviously, mechanical noise will vary from record to record and deck to deck. Fixed filters' parameters chosen by manufacturers then could not necessarily meet all needs, so switched or variable types became preferred amongst enthusiasts. An interesting commercial example could be the Leak Stereo 30.
For these filters TL07x series opamps were specified.
As with the JLH design mentioned above, another design could suffer from switch noise over time (NE5532/4s). Electronic switching can be used although this can give rise to a complex solution.
In this kind of situation the use of make-before-break switches will avoid open-circuiting a feedback path which can then give rise to interesting, but unwanted, sound effects. To give a continuously variable facility, a customised multiple potentiometer could be considered.
The next design varied the slope of a treble filter, following the pattern used in the Quad 33 and others, giving a more comfortable psychological feel. Any reasonable, low-noise transistor will suit.
Such filters, though often peaking supersonically after the fall-off, found uses with noisy tapes and broadcasts as well as records. Alternatively, the frequency of a steep slope can be switched.
Mains hum and rectification byproducts will invariably have to be removed. To this end 50Hz and 100Hz (UK) filters can be useful. The values then chosen for one circuit can then be easily doubled or halved (paralleled or put in series). When the output of a filter is fed back to a node that is normally grounded the notch width can be varied.
This can be useful when, with temperature drift, f can vary. The absorption of humidity can increase a capacitor's value, sealed types do not suffer this problem. Although there is nothing wrong with ±5% carbon resistors, for best results use low-noise ±1% metal-film resistors and polystyrene, mica, polypropylene or polycarbonate capacitors for filtering and equalisation. A low temperature coefficient is considered vital. With the node tied at close to the same potential as the output, the notch becomes vanishingly small and peaking can occur as can oscillation.
The following variable filter was used to subjectively determine bandwidths before a fixed type was built.
At lower volume levels the perception of lower frequencies falls off (Robinson and Dadson curves). Some means of boosting these is often required and 'loudness' switches appeared. These, most often, reduce the higher frequencies giving the psychological perception that the bass has been boosted. In most cases, one or more taps were taken off the volume control to accommodate differing volume levels, although simpler arrangements can be found, eg;
Again a manufacturer's fixed settings did not necessarily meet personal needs. This arrangement combines volume, balance and a variable loudness (NE5532/4s).
Personal preference would place the volume control on the output. A simpler form is shown below.
A good range of audio filters of a variety of forms have been designed by T Giesberts appearing in the magazine Elektor and these are recommended for study.
Many excellent variable filter designs call for the use of ganged potentiometers. Dual-ganged types are easily obtainable. However, those with four or more can be virtually impossible to obtain. This would be the case if the above design was required to function in stereo with infinite attenuation (additional pot on output). Before embarking on a design that employs, say, quad-ganged pots research the sources for these and their replacements. Digital potentiometers can overcome these issues and others like matching and mechanical wear.
As noted above, a listening area's acoustic characteristics can change with the re-arrangement of the items it contains or, say, when a mobile system is required to function at differing venues. Reflections, frequency peaks and 'holes' will occur that cannot normally be anticipated. For example, the onset of feedback can be a very distinct variable. To achieve the best results, some means of dealing with these is required.
Graphic equalisers are characterised by a large number of sliders each with a fixed frequency that can be either cut or boosted. The greater the number of sliders there are, the greater the number of frequencies that can be varied and, therefore, the tighter the resolution becomes. Smaller domestic units can have as few as five, but will usually consist of one filter per octave and then have ten per channel. Professional designs will have filters spaced at half or third octave intervals (28 or more per channel). Obviously the more filters there are, the tighter the component tolerances (and Q / bandwidth) must become. The appearance of the front panel can give rise to the mistaken belief that the frequency response of a system is then graphically represented, although this is not the case.
Since the noise from the combined filter stages is summed, this tends to cancel itself out. Earlier (and some later) designs used inductors in the filters. The individual resistance of each of these must then be taken into account when calculating each filters' characteristics. Ferrite-encapsulated inductors are recommended to reduce magnetic coupling, and to keep crosstalk at relatively high frequencies down to an acceptable level (<-60dB @ 10kHz). For these reasons, inductors have largely been superseded by gyrator circuits, although, as with all filters, a number of diverse forms exist.
Most commercial designs would employ 4136, 4559, LM833, etc, opamps although the (10MHz) 4560 was a good choice for eq. To increase flexibility, more advanced models can include functions such as input / output switching to accommodate record / playback, a bypass switch that allows comparisons between equalised and flat settings (some designs being able to reverse, or invert, them), switchable filter gains (say ±3dB, ±6dB or ±12dB) and variable outputs. Others can include remote control, noise sources, metering and/or spectrum analyser displays.
If greater care is taken setting up a sound system, for example professionally, a pink noise source can be used. In conjunction with a narrow bandpass filter and meter, equalisation can be applied producing surprising results with perhaps only a little filtering. Graphic equalisers are common, but many prefer, as does the author, parametric types which are felt to offer greater flexibility. This is because a single frequency that falls between the filters of a single, half or third octave graphic equaliser (each being progessively more complex and expensive than the next), can be found faster and precisely tuned with a parametric, not only the gain but also the Q (which determines the bandwidth) and frequency being adjustable. To achieve this the 'state variable filter' is often used. A simpler version for a portastudio.
A more sophisticated system may use filters
and matching tone controls.
With the two circuits shown above, it is ventured that an adequate solution for a hall or living-room would consist of up to three filters and one set of tone controls per channel.
A compact solution that the author favoured for halls, studios and domiciles was the very excellent Technics SH-9010 equaliser, shown below.
To prevent feedback, or 'howlaround' which can damage speakers, in a hall without using delay or frequency shifting techniques (assuming that sensible microphone / speaker placement is observed) set up the system for as flat a response as possible raising the gain so that when empty the slightest echoing just becomes apparent. As the hall fills with people, reflections will be reduced retarding further any onset of feedback. Alternatively, set up as above, then reduce the system's overall gain by some 3-4dB. In disco situations there is sometimes a compulsion to 'crank it to the max'. The inexperienced are then easy to spot as soon as they switch on a microphone, with consequent rushed fumblings and complaints from the audience. True professionals take the care, time and trouble to never have feedback.
Contact me at firstname.lastname@example.org
especially if you want additional content to this page
or if you find any links that don't work. Don't forget
to add the page title or URL. Take care!
Back to index, sound, tips or home.