LED + LDRs
With this method, the control signal drives an LED so that full brightness is defined as either maximum level or full attenuation. An adjacent light-dependent resistor (LDR) acts as the upper or lower arm of a passive attenuator. The intrinsic circuit isolation and physical separation that is possible makes LED/LDRs attractive in systems where isolation (of both grounds and common-mode voltages to 2.5 kV or more) is important for safety or EMC. These parts provide remote control connections analogous to connecting digital feeds via opto-isolators.
Tolerance is an issue and is dependent on the constituent parts, both semiconductors. Because the tolerance of both LDRs and LEDs is rather wide, manufactured combination devices are likewise broadly specified. The performance of both devices also varies widely with temperature. Also, in many circuits, there is no negative feedback loop to keep these variables within limits. Thus LDR/LED combinations are unsuited to system gain control due to inconsistencies of say +/–3 dB. They are fast enough to be used as limiters for bass and even midfrequencies in active crossover systems, and sonic quality is regarded as among the best. However, the above gain variation (in a population) would translate as a spectral imbalance, making overdriven conditions in a large system unsafe and/or uncomfortable, as well as drawing attention to the limiter action.
An LDR may also be partnered with an incandescent lamp. Even if small, the lamp is relatively slow to turn on and off, preventing its use for clean-cut dynamics processing, and lamp life span is more vibration sensitive and so not as certain as solid-state parts in road-going use.
Junction Field-Effect Transistors
JFETs are the lowest cost elements and can be made operative with little support circuitry. They are normally applied in the lower arm of an attenuator network. Without introducing complications of increased noise, noise pick-up, and other sonic degradation caused by introducing high ohmic value series resistors, attenuation is limited in range, and unless added circuitry can be justified, mild attenuation (around –6 dB) produces high (1 to 10% but mainly benign, low order) harmonic distortion.15 Low distortion control can be attained by placing the JFET in a control loop, comprising two or more op-amps and other active parts. However, as most JFETs’ Ron is in the order of a few tens of ohms, attenuation is still typically limited to –20 to –30 dB, enough for limiting, but not as a VCA gain and mute control.
Multiplying Digital-to-Analogue Converters
Multiplying digital-to-analogue converters (M-DACs) involve a resistive ladder, usually binary, with semiconductor switches, usually small-signal MOSFETs. They are the solid-state equivalent of a relay-controlled attenuator ladder (see later). Types suitable for high-performance audio must have dB steps—awkward in binary format—and special MOSFETs for low distortion and absence of “zipper” noise. The latter undesired sonic effect occurs in low-grade M-DACs; it is caused by step changes in DC levels or feed through from the digital control signal. Unlike the previous elements, an M-DAC has discrete resolution—just like a stepped (“detented”) pot. At low attenuations, step size must be no more than l dB for precise control; below –30 dB, larger steps (2 dB) are usually fine enough. To attenuate down to –70 dB in l-dB steps, 12-bit M-DAC is required.
R&R Array
Comprising resistors and relays, this is the mechanical counterpart of the M-DAC, with relays opening and closing paths in a “ladder” or other array of (usually) discrete attenuator resistors. Only high reliability, ATE-grade, sealed reed relays are suited for high-performance audio on grounds of both reliability and sonics. Such relays can act in under l mS and have fast settling, but are still not really suited to dynamics processing! Getting dB steps to act binarily with a resistor array takes some lateral thinking. Although the relays required are relatively expensive, by ingenious network adaptation to increment in binary dB, a mere seven can offer a 60-dB range in I-dB steps. With suitably well- specified resistors, this type can offer the highest transparency of any gain control device.
Summary
Motorized pots, lamp + LDRs, and relay/resistor arrays are good for remote- or machine- controlled gain trim and setting. The latter are the fastest and likely most reliable.
J-FETs and LED + LDRs are good for dynamics processing, but attaining accurate, noninvasive performance takes from the initial simplicity.
VCAs and M-DACs are elements that can do both kinds of jobs well.
Remote Control Considerations
Computers regularly feign precision that is only virtual. Until gain control elements become self-checking, self-calibrating, and self-aligning, they require careful specification.
Temperature
Pots (particularly conductive plastic), JFET, LDR, and particularly VCA elements are quite temperature sensitive. Unless designed with very low tempco, then when used in two or more channel amplifiers, they must be placed isothermally, that is, cosited to be independent of all the major temperature gradients, dependent on drive patterns, siting and even amplifier and rack orientation, as a hot gas usually rises upwards relative to the earth’s surface. This is true even with amplifiers employing forced venting, when small signal parts are not in an air path and are left to cool by microconvection, conduction, and reradiation.
Without such precautions, differences in channel gains of 2 dB have been observed in an amplifier employing VCA-controlled gain when driven up to working temperatures. This is enough to cause howl round or upset spectral balance.
Repeatability
Remote gain settings must not drift or have repeatability errors, which can accumulate to cause more than (say) +/–0.15-dB total error. This may seem stringent, yet on top of an initial tolerance of another +/–0.15 dB, it allows a worst case total difference between speakers of 0.6 dB. Other errors (cable losses, driver mismatches) are of a similar order and add to the differencing toll so there is no room for complacency. Least is best.
Conclusion
M-DACs and relay-resistor-array attenuators have the highest stability against temperature and time. Other types may prove acceptable with ameliorative engineering. Setting precision should not be taken for granted.
Compression and Limiting
Compression and limiting (comp-lim) are gain reduction, alias dynamics processing techniques, that are employed (among other things) to protect speakers, ears, and amplifiers from excess, distorted signal levels. In professional, active crossover-based systems, they are usually embodied within the active crossover. This is the best position for logistics in traditional large systems, with only one comp-limper band to worry about. Positioned within the filter chain can also be the best location for sonics.
Where power amplifiers are driven full range or where active crossover filter sections are integral to the power stage, compression and limiting functions may take place within individual power amplifiers.
Compression must be used sparingly, as average power dissipation in the drivers will be increased, potentially part-defeating the object, as speakers may then suffer burnout.
Paradoxically, the compression threshold (at least for bass frequencies) should be increased if the gain reduction exceeds about 6 dB. Also, attack and release times require careful setting to avoid pumping on strong low bass.
Limiting is a higher ratio, more brute force (many dB-to- l) gain reduction. Its raison d’etre is to catch fast peaks, hence “peak limiting.” Attack times that are useful for protecting most loudspeaker drivers are in the order of 10 μS. Faster rising peaks that “get through” rarely cause damage to hardware, but may be reproduced efficiently by metal- diaphragmed drive units (cf. paper cones) and perceived and found highly unpleasant by the ear. Hence faster-acting peak limiters may enhance sound quality under many real conditions of “operator abuse.”
Clipping (Overload) Considerations
Driving any power amplifier with excessive input results in clipping because the output’s excursion is finite. Amplifiers offering higher power into a given load impedance provide a higher voltage swing into that impedance so clipping for a given sound pressure level is less likely to arise. However, linear increases in power give only underproportionate, logarithmic increases in headroom (in dB) and cost linearly ascending amounts of money.
At some point, whatever more swing could be afforded would make no difference, and a limit is set. Exceeding this is clipping. For short periods it can be benign but else it is unpleasant and potentially damaging to hearing and positively damaging to hf and
bass drive units in particular. Moreover, considerable overdriving, into hard clip, as can happen at any time by accident, even with domestic systems, can heavily saturate and thus vaporize the BJT output stages of inadequately designed power amplifiers.
Clip Prevention
Destructive and antisocial clipping may be prevented with comparatively simple circuits performing like a dedicated, fast limiter. There are as many names as there are makers. Some examples are shown in Table 8.6.
In these and related schemes, clip prevention does not occur until a dB or so of clip. Using the 100-W analogy, the usual low % THD does not rise until the signal passes above about 50 to 70 W. If headroom is adequate, this point should hardly ever be reached with the majority of recorded sound. With live sound, it may be reached quite often, but the fact that the deeply unpleasant point only l dB higher is not crashed through is of far more importance.
Soft Clip
“Soft clip” is a feature that aims to defeat the suddenness of the onset of hard distortion above the clip level in conventional, high NFB power amplifiers. It may be provided as a fixed or switchable option. Unlike compression and limiting, there are no time constants, no settings, and no attempt to avert serious distortion of a sine wave. However, the clipped waveform does not readily square off and retains some curvature (dV/dt) even with heavy overdrive (e.g., at +10 dBvr). This greatly reduces the massed production of unpleasant, high harmonics and intermodulation products of hard clipping. One apparent (but not necessarily actual) snag is that because hard clipping is a real limit, soft clipping has to begin to occur up to –10 dB below full output (–10 dBvr). This is tantamount to saying that distortion (%THD say) with a 100-W amplifier begins rising from above about 10 W, as opposed to rising very abruptly above exactly 100 W, while remaining extremely low up to this point. Here is one difference between low and high global feedback amplifier behavior.
Soft clipping restores the more forgiving behavior of low feedback to a high NFB amplifier. The extent to which it undoes all the high feedback’s other benefits is unqualified. At least the high NFB is in operation for most of the time, for with proper headroom allowance, most of the musical content should lie below the –10-dB threshold or so, whence the soft clip is inactive. Usually soft clipping is arranged to be symmetrical. This may not create the most consonant harmonic structure. Figure 8.12 shows a classic circuit.