Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L R By Jake Rothman Audio switching Part 4 – real electronic switching Solid-state switching The first solid-state switches used lightdependent resistors (LDRs) enclosed with a light source (originally a filament lamp, and later, LEDs). These were often home made but later produced as complete units called Vactrols (Fig.62). These are still popular in some guitar effects pedals but are rather elusive and expensive. Also, cadmium sulfide* is banned under RoHS legislation. They + 1kΩ Output +5V for mute 2SC2878 Low VSAT RON 1Ω N-type G D In/Out 5V S S Bilateral switch In/Out D P-type For negative operation use PNP type 2SC4213 Fig.63: bipolar transistors are only used for muting. Toshiba makes these transistors especially for muting (RON = 1Ω). Often, two stages are wired in cascade with 100Ω resistors rather than 1kΩ for lower output impedance. Fig.64: a bilateral switch. Two complementary lateral MOSFETs wired in parallel provide a degree of linearisation of the on-resistance. offer very clean but slow (~100ms) switching. The current required to drive the LED is high, typically 10-20mA. Bipolar transistors don’t make good audio switches since they are unidirectional, and their base operating current contaminates the audio. They are only used for output muting, such as stopping a CD player output thumping upon turn-on, as shown in Fig.63. FETs are the most common solidstate audio switch. MOSFETs can be bi-directional if a complementary N and P channel pair are paralleled; this Basic 4016 circuit X1 1 14 VDD+ Ch 0 1 16 VDD+ Y1 2 13 A1 Ch 2 2 15 Ch 2 X2 3 12 A4 Com 3 14 Ch 1 Y2 4 11 X4 Ch 3 4 13 Com A2 5 10 Y4 Ch 1 5 12 Ch 0 A3 6 9 X3 Inhibit 6 11 Ch 3 VSS– 7 8 Y3 VEE 7 VSS– 8 4016B / 74HC4016 4066B / 74HC4066 Binary code input selection 9 B 00 01 10 11 4052B 2 off 4-way ganged switch 16 VDD+ BY 1 16 VDD+ CH 6 2 15 CH 2 BX 2 15 B com Input 3 14 CH 1 CY 3 14 A com CH 7 4 13 CH 0 C com 4 13 AY CH 5 5 12 CH 3 CX 5 12 AX VEE 7 11 A Select 10 B 9 C VSS– 8 4051B 8-way switch Binary code input selection 000 001 010 011 100 101 110 111 Ch 0 Ch 1 Ch 2 Ch 3 Ch 4 Ch 5 Ch 6 Ch 7 Fig.65: the ever-popular CD40xx series of bilateral switches. 10 A Select Ch 4 1 Inhibit 6 46 Control input ON = 1 OFF = 0 100kΩ 1kΩ * The International Union of Pure and Applied Chemistry (IUPAC) made “sulfur” the accepted spelling for element 16 in 1990; see https://pemag.au/link/abyg Fig.62: a Vactrol, an early type of electronic analog switch. Its control characteristic is proportional and linear enough for vibrato and tremolo circuits. Inverters + W e have been looking at the topic of audio switching for a few months now; this article (and the following one) will concentrate on solid-state switching techniques. Switching audio signals with transistors is cheaper than mechanical switching (switches/relays) and increases flexibility. Disadvantages include increased distortion, control breakthrough and a limit near the power rails where signal breakthrough increases significantly. The on-resistance (RON) is usually relatively high, and is modulated by the audio signal, which is one source of distortion. Also, the isolation is worse than a mechanical switch when it is off. These problems have not been eliminated, but they can be minimised with more advanced circuit design and the help of op amps. Inhibit 6 11 A VEE 7 10 B VSS– 8 9 C 0V or V– 4053B 3 off change-over switch Ch 0 Ch 1 Ch 2 Ch 3 Switch control inputs Practical Electronics | September | 2024 CHANNEL ON RESISTANCE (RON) / Ω AMBIENT TEMPERATURE (TA) = 25°C 600 SUPPLY VOLTAGE (VDD – VEE) = 5V 500 400 300 200 10V 15V 100 0 –10 –7.5 –5 –2.5 0 2.5 5 INPUT SIGNAL VOLTAGE (Vis) / V 7.5 10 Fig.66: input voltage vs RON of a bilateral switch. This is the cause of the distortion. The resistance increases with lower supply voltage or higher temperature. ‘bilateral topology’ is shown in Fig.64. It cancels out some of the RON modulation. It’s possible to buy these in multistage ICs of the 4000 or 74HC series, containing complementary MOS or CMOS bilateral switches (introduced by RCA in the late 1970s). This is the cheapest approach, although they have a much higher RON than a discrete approach. Quad popularised these in their 34 preamp. The most common types, the 4016 and the lower on-resistance 4066, give you four switches in a package, costing only around 50p. Other multi-pole devices include the 4051, 4052 and 4053, shown in Fig.65. Unfortunately, these devices have the distinction of being the most unreliable small-signal semiconductors in the audio world. I suspect this is because the antistatic protection is compromised, being different for the linear devices than the logic devices. Always put them in sockets. Another limitation is that the total power rail is 18V maximum, which causes a headroom loss when used with op amps, which have a total rail excursion of 36V. The RON modulation and hence distortion gets less as the power rail is increased, as shown in Fig.69: DG408 switch chips in the Arcam 65+. There were four; they must have got a special deal! Note the input protection diodes. Fig.66. So always use at the highest voltage possible. When using these ICs, I will often use ±8.2V zener-regulated rails with the op amps feeding them running at ±9.5V because their outputs can only swing to within 2V of the supply rails (rail-torail output op amps could use the same ±8.2V supply). Note that, as with op amps, the signal needs to be biased to remain within the supply rails of the bilateral switches. If using a single-rail supply, that means biasing the signals to half-rail. For dual-rail supplies, the signals should be DC-biased to 0V. The control voltage for the CD4016/66 is V+ for on and V− for off. The other chips, the 4051, 4052 & 4053 have internal level shifters, so they are controlled with normal logic levels, meaning logic low can be 0V instead of −7.5V. If an external level shifter is needed, you can use the configuration shown in Fig.67. Voltage protection is usually needed when interfacing with op amps running on normal rails or the outside world. The Quad 34 used two back-to-back 6.8V zener diodes on its CD input. The distortion was often in the region of 0.1% at a few volts peak-to- +VE 15V Input Input from low-impedance source BC547B 5V TTL 0V 1 5.1kΩ Level-shift circuit for gating FETs BC557B 30kΩ +VC –15V 7 0V –15V Fig.67: a level shifter for standard logic to control FET-based analog switches. Practical Electronics | September | 2024 2 13 Pull up to +8V to turn on Pull down to –8V to turn off 0V Output +VSS –8V +VDD +8V 14 1/4 4016/66 RON* peak when used as straight substitutes for mechanical switches, like in the rudimentary circuit shown in Fig.68, much worse than the rest of the audio electronics. However, these devices were still a godsend for early 1980s mixer designers. Some engineers, such as Steve Dove at Alice, developed circuits to make the distortion quite acceptable for broadcast mixers (more on that later). There are 74HC versions of the CD4016/66 chips; they have a lower maximum rail voltage of 10V but perform better at low voltages. All these ICs work well in synthesisers, where I still use these devices. With all these switch chips, the FET gates are driven by internal fast logic, causing clicks from charge injection into the audio channel due to the gate capacitance. Special chips, single ‘sauce’ There are higher-specification MOS switching chips, such as Maxim/ Vishay’s DG408, used in the Arcam A65+ amplifier (see Fig.69). These are rated at 40V and give eight switches each. The catch is that the DG series is designed for fast analog multiplexing, a feature that is wasted for simple audio channel selection. Output 10kΩ RLOAD 0V THD = 0.1% at 2VRMS for 4016B THD = 0.05% at 2VRMS for 4066B *RON = 115Ω to 350Ω max, ±7.5V rails for 4016B *RON = 60Ω to 175Ω max, ±7.5V rails for 4066B Fig.68: a ‘data book’ 4016 switch circuit. It works, but the distortion can be greatly improved with some tweaks. Fig.70: an opto FET. These are useful where high isolation is required between the control and audio signals. I used one in a tone burst generator. 47 Fig.71: the Braun SK4 Phonosuper made interlocking push buttons trendy. They switched many things at the same time, resulting in a lot of electrical noise and mechanical clunks. Another problem is that they cost almost £10 each when I last looked at Mouser. The distortion was not that good either, at around 0.04%. An excellent chip giving 0.003% was the hybrid bipolar/JFET SSM2412, alas now long deleted, a fate always befalling single-sourced audio chips. It even had internal ramping circuitry for clickless operation. There is another dedicated chip from JRC (New Japan Radio Company), the NJM2750, a four-way input/output selector with 0.005% THD at 1V RMS. It’s based on bipolar transconductance cells, and it’s cheaper at £1.50 from Mouser. I await my samples for evaluation, which may be tricky to test since it’s only available in surfacemount packages. Discrete shopping When using discrete FETs, it is possible to ramp the control inputs on the gates to reduce clicks to an inaudible low-level thump, even though there is the same amount of charge transfer. It is even possible to get opto-FETs, a form of linear opto-isolator (Fig.70). They are free of control voltage breakthrough, but the H11F1M types I have exhibit a rather high on-resistance of 300Ω. The price is around £4 for a single switch. If fast switching is needed, say for a tone burst generator, operation at the zero crossing is possible with electronic switching. This is another, more complex way of avoiding clicks, at least in theory. However, subjectively, it doesn’t work, as Studiomaster found when developing muting blocks for their mixer MIDI muting system. The audio signal is ramped up fast and still sounds like a click. The result is that we are back to making discrete circuits using individual FETs for a top-quality Hi-Fi switch. MOSFETs such as the VN10 can be used, but they are asymmetrical, almost always have parasitic diodes (unless they are unusual four-lead types) and need static protection. JFETs are a better choice and are symmetrical at audio frequencies, allowing the drain and source to be interchanged. Luckily, switching type JFETs such as the J112 and J175 are still cheap at around 35p each, unlike lownoise linear audio types. We’ll look at some practical electronic signal switcher circuits shortly, measuring the distortion and giving Fig.72: Practical Wireless’ Winton amplifier with traditional clunky input selector switches. This was first the MOSFET output amplifier in a UK magazine, in 1979. 48 Fig.73: the wiring on the PW Winton has screened cables running from the sockets at the back to the front of the unit. Note the Alps interlocking switch assembly. some useful multi-purpose PCBs to build. Sent to Coventry I visited Retrotech UK in Coventry on May 11th, an exhibition full of historic gear for sale. Here was the whole history of audio switching laid out before me. I was amazed to see a pristine Braun SK4, a true pushbutton pioneer, shown in Fig.71. However, at £450, it was too much, so I bought a £15 1979 Practical Wireless Winton instead, the first UK constructor’s MOSFET amp (Fig.72). This design shows one of the disadvantages of mechanical switching, with eight screened cables having to run to the front from the sockets at the back, as shown in Fig.73. Electronic switching allows the signal circuitry to all remain at the back, with only four unscreened DC control wires to be routed to the front. However, it was still a competent amp. Quad 34 preamp circuit Quad’s design philosophy was to spend more money on the box and pots (shown in another Retrotech photo, Fig.74) while keeping the electronics cost effective. It ran Bi-FET TL071 op amps on low voltage rails of +8.6V and -9.4V for symmetrical clipping. Fig.74: the epitome of middle-class classical elegance, the Quad 34 preamp and 405 power amp. This was the first successful design to use electronic input switching. It still works after 35 years and is priced at £500. Practical Electronics | September | 2024 From interlocking switch logic +7.5V Phono input 1/4 4066 470nF RIAA 39kΩ 0V –7.5V 0V Attenuator 680nF +8.6V 39kΩ + TL071 100µF 330nF Voltage limiter 0V 100kΩ From interlocking switch logic 1/4 4066 6V8 0V 10MΩ 10kΩ –9.4V + TL071 0V Radio input 1/4 4066 – 10kΩ 0V From interlocking switch logic + CD input Fig.75: electronic input switching in the Quad 34 preamp. Some Hi-Fi buffs replace the 4066s with reed relays. 6V8 0V The switching was done with 4000-series logic and 4066 audio switches running at their maximum of ±7.5V. This gave a 6dB headroom penalty compared to circuitry using normal ±15V rails. This was overcome by placing an attenuator on the CD input to reduce the operating level. The 4066 switching elements were wired in standard series mode, feeding an input impedance of 10MΩ, which was only possible because of the use of FET op amps. Clicks were minimised since there was no input bias current to worry about. The circuit of the input path is shown in Fig.75; the delicate 4066s are buffered from the outside world. The subjective Hi-Fi fraternity didn’t like the design, claiming this system had a sonic “MOSFET mist”. I suspect there may have been prejudice against these devices because a DIY audio site states they were used in Plessey’s System X telephone exchanges. I’ve found no primary source to corroborate this. Q u a d ’s m a i n d e s i g n e r a n d f o u n d e r, P e t e r Wa l k e r, a l o n g with their consultant, Peter Baxandall, didn’t design by ear but by maths and physics. This strategy didn’t bother Quad’s primary market, the classicalmusic-loving middle classes, with around 40,000 sold. Quad gear still turns up all the time, commanding high prices. 0V – Output to volume control in old Neve modules, such as the BA714/5 filter modules. Interestingly, they used germanium AA142 diodes, but 1N4148s or Schottky types are fine. Dual-switch circuits All these circuits have quite poor offstate attenuation at high frequencies (10kHz), mainly due to internal capacitance across the series switching element. If the lower resistor is replaced by a shunt switching element with a control pin driven by an inverter, as shown in Fig.78, excellent attenuation is obtained while also solving the breakover problem. Care must be taken that there is no possibility of both switches being on simultaneously, or the noise gain will be so high that a loud ‘splat’ will occur. Sometimes, a small gain-limiting resistor of around 100Ω (R4) is placed in series with the virtual earth pin as mitigation. Complementary outputs are needed for the dual-switch circuit; one solution is the flip-flop circuit in Fig.59 (from last month). The finished Veroboard version is shown in Fig.79. The op amp The distortion caused by the variation in on-resistance (RON) with signal was minimised in the Quad circuit by loading it with a 10MΩ resistor (R6). This minimises the attenuation because of the resulting potential divider action. In turn, this makes the effect of variation of RON negligible. In the RCA data book circuit, the load was 10kΩ, giving quite bad distortion, in the order of 0.4%. The Quad circuit reduced this to 0.004%. Another way of minimising the effect of RON variation is to put the switching element into a virtual earth at the input of an inverting op amp circuit, as shown in Fig.76. This reduces R2 Electronic 10kΩ the voltage across the switch (FET or IC) switching element to very V+ RON low levels. This depends Input R1 10kΩ ~150Ω 1 2 7 2 Output on the value of the input – 6 TL071 resistor, R1. It is, as usual, a 3 13 + compromise between noise 4 and distortion. A typical 0V Control value is 10kΩ. V– A possible disadvantage is Fig.76: using a virtual earth minimises the that the circuit is inverting, voltage across the switch when it is on, although another inverting reducing distortion due to RON modulation. op amp stage can fix that. R2 A problem with the Electronic 10kΩ switch (FET or IC) bilateral gates is that the break-over limit when off V+ R1 RON is ±7.5V. If a resistor is Input 10kΩ ~150Ω 1 7 2 2 Output – dropped to ground (like R3 6 TL071 in Fig.77), an attenuator is 3 13 + R3* 4 * An improved formed, limiting the voltage circuit replaces 10kΩ 0V across the switch to half. Control R3 with inverse V– 0V parallel diodes. There is a catch; the signal input current to the virtual *An improved circuit replaces R3 Fig.77: break-over can be prevented when the earth is halved and the op with parallel back-to-back diodes switch is off by adding attenuation resistor R3, amp has to make up for this, although the noise increases by 6dB. giving a noise gain of +6dB. R3 can be replaced with R2 Electronic 10kΩ two diodes connected in switch (FET or IC) inverse parallel, avoiding V+ R1 R4 RON the noise gain associated Input 10kΩ 100Ω ~150Ω 1 7 2 2 Output with the resistor. When – 6 TL071 the switch is on, the 3 13 3 + voltage on the virtual 4 NC Electronic earth is so low the diodes NC switch 0V 5 do not conduct. When replaces R3 V– 4 the switch is off, the 0V Control diodes limit the signal, preventing breakover. Fig.78: adding a shorting switch element stops breakThis technique was used over and avoids noise gain (unless both are accidentally Practical Electronics | September | 2024 49 Improving CMOS switching circuits switched on simultaneously!). Fig.80: the op amp based flipflop from Fig.60 (last month) can deliver ±15V with ±17V rails, ideal for JFET switching. Fig.79: just to prove I build and test stuff, here’s the over-engineered switch latching circuit from Fig.59. The J111 has a 30Ω RON but it might need -23V to switch off in the worst case, with a 10V RMS signal and VP at the limit. P-channel JFETs also switch on at 0V but need a pos­itive voltage to turn off. This is more convenient, but their RON is higher. Only one type, the J175, is readily available and has an approximate RON of 80Ω (up to 125Ω in the worst case). flip-flop circuit (Fig.60) will also work if the rails are reduced to suit. This is shown in Fig.80. JFETs All the above circuits can use discrete JFETs as the switching elements, with the advantages of no power rails being needed, higher headroom, greater reliability and less clicking noise. However, they often require high negative switching control voltages. Normally, N-channel JFETs are used for switching because they are the cheapest, most readily available and have the lowest RON. Like all depletionmode FETs, they require 0V to turn on and a large negative voltage to turn off, the pinch-off voltage or VP. The low RON varieties require a higher voltage. The most popular type is the J112, with the older TIS73 and U1898 occasionally encountered. These devices have an approximate RON of 50Ω. The VP is typically -5V. A -15V control signal from op amp power rails will switch them off even with high-level audio going through them (which could subtract from the control voltage, causing break-over). Boss effect in/out The rudimentary circuit shown in Fig.81 works well for effects pedals where distortion does not matter much. This uses the 2SK30A JFET, which is popular with guitar pedal builders because it has a relatively low pinch-off voltage (VP) of around -0.4V to -2V. This is needed because guitar pedals use 9V batteries, which drop to 6.5V as they discharge. The on-resistance is around 300Ω. Mechanical ‘stomp’ switches are pretty unreliable, often suffering from latching and contact oxidation after a few years. An expensive way of preventing this is to use an Eaton EAT8943K32 switch from E. Preston Electrical Ltd, which has wiping contacts lasting 20 +4.5V bias + 10µF FX Output 4558 220kΩ – 1N4148 Mix 22kΩ 2SK30 220kΩ 50kΩ 1N4148 0V + That covers the basics of solid-state electronic switching of audio signals, but there are a lot of nuances, so we will have more on this topic next month. PE Low-impedance source – eg, op amp Output J112 100kΩ (Load) 22kΩ 1µF 0V Bootstrap resistor Effect On A neat trick for minimising the RON modulation in JFETs is to bootstrap the gate with the source voltage since the variation of the gate-source voltage (VGS) modulates RON. A simple way of achieving this is to use a blocking diode and bootstrap resistor (22kΩ) joining the source to the gate, as in Fig.82. This way, the audio rides on top of the control voltage. The diode also protects the gate from being forward-biased excessively and effectively holds VGS at 0V for switch-on, even if +15V is applied. If the diode is grounded, the distortion is higher than if it is taken to a positive voltage since it is easier for the diode to become forward-biased. 22kΩ is typically a suitable value; the lower its value, the lower the distortion. It can sometimes be partially bypassed with a capacitor for lower HF distortion. The effect of the 22kΩ bootstrap was to lower the distortion from 0.003% to 0.0005% at 4V RMS when feeding a load impedance of 330kΩ. 4.5V 1µF 10kΩ Bootstrapping Next month + + Buffer 10µF 2SK30 +4.5V bias Input 1MΩ + 1MΩ years. I use these in the most expensive Colorsound pedals. The cheap way to do this is to go electronic with this circuit. Off 0V 0V 1N4148 >37.5V to turn off +4.5V to turn on –37.5dB attenuation at 10kHz Distortion: 0.01%, 4VRMS – no bootstrap and diode Distortion: 0.0015%, 4VRMS – with bootstrap and diode Fig.81: a rudimentary JFET effects in/out switching circuit from Boss guitar pedals. This was coupled to the discrete flip-flop in Fig.61 (again from last month). The 4.5V half-rail biasing enables the JFETs to switch off with their gates at to 0V. Fig.82: bootstrapping the gate voltage with the audio signal reduces distortion. 50 Practical Electronics | September | 2024