Electronic switching

Connectors, switches and other mechanical contacts in controls can introduce anything from a faint crackling to an open circuit, consequently causing problems when worn or dirty, a good contact only being possible when one contact breaks the surface of, and penetrates, the other. At low signal levels, the consequences are obvious.

Most mechanical switches consist of two frictional contacts connected by a shorting element. The commercial standard usually consists of silver-plated brass which tarnishes and wears easily. Each such contact in the signal or feedback paths is in series with every other contact in the signal chain. Thus, the more contacts, the more noise. A DC presence will exacerbate the problem, particularly with volume controls.


In the example above, the mag circuit has 6 friction contacts, the tape has 10, per channel. The DIN connector convention, in essence current-driven into a high impedance source, was particularly troublesome giving rise to external pick-up because of the low signal levels used. Some amplifiers rely on large numbers of switches to give flexibility. Tape and cassette recorders were also vulnerable in this context when, to avoid the cost of separate circuits, extensive changeover record/playback switches would 're-wire' the circuitry for these functions, apart from those used for bias and equalisation (even a low-end version having 10 or more per channel). Crucially, there may be two switches in series between the preamp input and the signal (head or source input) and/or a changeover switch between ground and the head. Similar problems were met with multi-waveband tuners where sensitive RF/IF stages were constructed around invariably noisy switch banks (some with as many as 72 solder tags). Test equipment can include a large number of contacts, one DFM used a 14-pole 6-way switch (84 tags). Higher quality designs might have employed silver-plated contacts which tarnish easily. Even gold-clad contacts, when new, can generate thermal EMFs of 1-40µV.

Some preamp muting arrangements will instead of breaking the output paths with relay contacts, say, will short them to ground.

Diode-switching using a single junction, considered innovative at the time, was useful in 4 channel preamps and organs where multiple channels were controlled and resolved issues with corrosion and mechanical problems with input selector switches. With damped switching, like that shown below, sources would ‘cross fade’ as the diode voltages changed providing a more pleasant and less abrupt change, particularly if the newly selected input was comparatively loud.

Transistors or FETs were often used in semi/professional equipment to fulfill switching functions around a PCB (like the Technics RS-1500US open-reel tape deck which used 9 transistors per channel to select the capacitive elements in the record and eq amp filters), a technique then adopted in later commercial hi-fis when micro-controllers appeared or in cassette decks with multiple noise reduction, bias, eq and servo systems. The Technics RS-M235X (Dolby B/C and dbx) is a good example.


In these situations FETs were often used to block a circuit path whilst transistors were used to short them to ground, a technique first seen by the author in the Lecson AC1 preamp.


A later Technics approach, used to switch signal paths, is shown below.


Although dedicated switching ICs like the Mullard TDA1028/9 appeared, FETs or analogue switches such as the 4016, 4053, 4066, LM1037N, DG211CJ, SSM2412P, etc, could be used instead, eg;


and


but in any case, the complexity, and cost, of the controlling hardware should not be underestimated. For example, the Lecson AC1 used some 11 'signal' devices and 16 switching devices per channel, apart from the controls for these. Common and low cost ICs can be used. The switching systems of two designs are compared below, the first from the Quad 34.

Another, a Giesberts input selector design employs bistable gold-on-gold contact relays.




Some commercial ICs can be used to good effect in less critical compact preamp applications, especially those intended to run from low-voltage single supplies. The Hitachi HA12219NT audio signal processor for cassette decks can offer electronic switching and amplification, for example, as can the Panasonic AN7345K which includes an automatic level control that can be employed to reduce the incidence of clipping. Another, the SIL packaged Rohm BA3126N 2-channel head switch for radio cassette recorders has been used to switch inputs.

Switching elements can not only block signals but, if cross-talk is a major consideration (always separate channels on a PCB with a ground plane, at least), can short sections of the signal path to ground (mute, eq, etc, see TOA application below). For example, each bus to a summing point can have a resistor in series with a FET switch, between these can be another FET switch to ground, the gate of one switch fed by an inverter connected to the gate of the other switch thus switching on when the other is off and vice-versa. At the same time, a design can consider separate record source selection so that recordings can be made of one source whilst monitoring another, eg; tape-to-tape.

If 4000 logic series cmos devices are used to switch analogue signals it can be useful if the supply voltages are asymmetrical, say +7V6 and -8V2 to balance the on resistances of the p and n-channel mosfets. Naturally, no input signal must exceed those of the supply so clamping the input line is wise. Internal capacitive coupling between the control signal and switch terminals can, as with fets, producing switching spikes. The 4051 (8 switches with 3-bit 1 of 8 decoder) is one example. A 1M resistor in series with a 'gate' and a 1n8F cap between the 'gate' and ground can provide a soft switching characteristic. Although the 4016 and 4066 (pin-compatible but lower on-resistance) quad switches are often seen, more critical designs have used the 4007 (dual complementary pair plus inverter). With an on resistance of about 300R and maximum variation of 200R, this equates with the tolerance of a 1% 22k resistor.

Volume controls

Volume controls are most usefully applied at the listening position. A DC presence and the ingress of dust or moisture, particularly with earlier sliders, can present problems. Electronic versions can interest in that the switching must present a continuously variable aspect, subjectively speaking. Analogue approaches used in the swell pedals of upright organs often used a photo-cell (LDR) in series with the signal that was then illuminated by a light bulb. This was then obscured by a vane attached to the pedal to effect volume change. Adequate perhaps, but with it's limitations. For example, to overcome non-linearities evident with some LDR designs, a DC component was sometimes added to the signal. Noise could also be problematic with high attenuations. However, if the current to the bulb could be varied, remote control was possible.

B&O tried this in their (1976) Beomaster 1900 (type 2903) and Beocomparators using a bulb to illuminate four LDRs that then grounded the signal. By it's nature digital switching is stepped. For a smooth transition a relatively large enough number of steps must be employed. So, apart from switched resistive elements, a clock, counter and some means of retaining a default setting (say, for power up) are required. Unfortunately, this early B&O attempt used a 4-bit code resulting in 16 steps (including mute). Complaints from customers in the respect of the perceived coarseness of adjustment resulted in the modification to a 7-bit or 128 step control (Beomaster 1900-2, 2400-2, types 2903/1). Although this proved satisfactory, the bulb's reliability continued to be an issue as did the use of an ultrasonic transmission medium.

A later B&O design, with full remote control, employed TDA1074 ICs in it's preamp. These device's gains were voltage-controlled and although remote operation of the tone and balance controls was feasible with a more complex microprocessor system, only the volume was adjustable electronically, the others being set manually.

Contemporary to these, infra-red TV remote systems (eg; Plessey ML920, ML922, etc) used 5-bit D/A or 32 step arrangements that then drove integrated IF amp / FM demodulator and luminance / chrominance control ICs (TBA120, TBA560, TDA2560, etc). These could offer a 32 command set covering programme selection (up to 20 channels), standby, mute, recall, normalise and stepped programme, colour, volume and brightness selection. There is no recollection of these being applied to an audio system, the reason for this being that without additional coding (pre-RC5) the remotes of both the TV and the hi-fi would have interacted, and in commercial terms TVs had priority.

Some non-conventional configurations can possess disadvantages, such as a reliance on the virtual earth of a shunt-feedback amplifier for absolute attenuation to zero, not possible with conventional carbon-track pots.

With this approach if the pot wiper open circuits, the gain peaks, sometimes disconcertingly.

The next approach achieves that, for a 20dB gain, whilst using a more easily matched linear device to achieve a near audio taper.

With this approach if the pot wiper open circuits, the gain dips.

Operational transconductance amplifiers (OTAs), like the dual 13600 (available as the EXAR XR13600, the National LM13600 and the Philips NE5517), enabled a concise component count to the voltage control of one or more channel's volume. These were useful in synthesisers, organs and disco mixers and could be used in filters, automatic gain amplifiers (AGCs), dynamic noise reduction circuits, amplitude modulators and even timers and tachometers. These could be driven digitally by employing digital to analogue converters (DACs). Although these could be bought fairly cheaply as ICs they could also be built quite easily. Sometimes the range of resistor values and tolerances required could be problematic. For example, a 10-bit circuit would require a resistor range of 1024 to 1 with a tolerance of 0.05%. R-2R ladders overcame this problem.

An 8-bit word gives a resolution of 1 part in 256, ie; a signal-to-noise ratio of approximately 50dB. A 10 bit a-to-d converter will give a s-to-n ratio of about 60dB.


Switches, like the 4053, could feed the signal to the network, but, in general, the range of attenuation would be limited, to say 48dB or even less. Alternatively, the network could be wired into the feedback network of a non-inverting amplifier to vary an output (see below), but again, the step size would be relatively large (3-3.5dB). Resistor tolerances, however, could be less tight, say 1%, and a non-linear range of values chosen. The amplitude response can be limited by the common-mode behaviour of the amplifier and frequency limited by stray capacitance across the resistors. The variable resistance can be provided by a variety of means - FETs, light or temperature dependent resistors but signal levels should be low enough not to be affected by device non-linearities. Constant impedance stepped attenuators with stages of -1, -2, -2, -5, -10 and 3 x -20dB could be cumbersome and noisy, requiring two switches in series with the signal per stage. Without the control interface, four or more ICs per channel were required.


For those interested, a well-considered design for a stereo programmable audio attenuator, by J. M. Didden, appeared in the May/June 1980 Wireless World with ample notes. Achieving a gain of +16.8dB to -46.2dB in 1dB steps, this used a two-wire control interface, 8 opamps, 24 logic ICs and 6 bit code. Balance was executed by toggling each channel respectively and the control word could be converted to a 2-digit bcd output to drive a 7 segment display. Attenuation steps were 1, 2, 3, 4, 8, 16, 32 and 48dB, these being made by 4007 switches.


Standard or mask-programmed logic, although occasionally cumbersome, can prove more reliable than EPROMs or PICs which can 'fade'.

The AD7110 digital attenuator IC (LOGDAC) offered an attenuation of 88.5dB in 1.5dB steps using a 6-bit code with options for loudness correction and this was run with an 8-bit modification of the later B&O circuit, both of which are shown below.