Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L R By Jake Rothman All-pass filters, part three – building an all-pass filter We have been looking at the implementation and uses of all-pass filters over the last couple of issues. After a bit more details on their use in loudspeaker crossovers, we’ll shortly get to a useful one to eight stage all-pass filter board you can build from standard parts. acoustic Butterworth slope with correct phase, hence the aB suffix. The T27’s high-Q fundamental resonance became inaudible. The new network improved on the original network so much that you could buy a drop-in replacement crossover to upgrade it. The circuit effectively removed Funny filters the bump from the tweeter’s resoKEF used asymmetrical filters nance and removed its additional with second-order bass and third-­ second-order slope from the roll-off. order treble slopes for many of their The difference was clearly noticeadesigns. This provided some extra ble on massed strings, the most refilter delay for the tweeter, compen- vealing test for crossover problems. sating for its physical offset. They This filter has been used on many then finished the job by moving the systems, such as the KEF 101, the tweeter up the baffle a bit. Wilmslow Audio CS1a and Malcolm This technique was executed per- Jones’s Falcon LS3/5A derivatives. fectly on their 104aB speaker (Photo Also see my article in the Sep17) and no all-pass filter was re- tember 2019 issue of PE on building quired. The tweeter crossover filter LS3/5A style speakers. circuit shown in Fig.30 for the T27 I use the same circuit topology in drive unit gave a proper third-order most of my designs as well. The KEF 104aB has a lovely laid-back sound, but this was really a defect, resulting from the lack of off-axis midrange energy. Using a 200mm bass unit up to 3.5kHz is too high and results in a very directional output. This then crosses over to a wide directivity output as the tweeter comes in. This results in an inaccurate ‘dog-leg’ response curve off-axis. The LS3/5a had a complex phase equalisation arrangement of an asymmetrical wide overlap crossover and a tweeter with polarity inversion. The flanges for the drivers almost overlapped. To accommodate all these variables, they mounted the bass unit on the rear of the front baffle to introduce additional displacement (see Photo 18). The result was almost perfect integration between the two drivers. There’s a little resonance due to the extra cavity introduced in front of the 600nF ‘Bridged’ T capacitor Original R104 network 3.3µF 10µF 3.3µF Input Improved R104aB network 3Ω 2W 10µF Input 10µF – 0V – KEF T27 tweeter 0.3mH 0.55Ω DCR + Polarity reversed relative to woofer KEF T27 tweeter 0.3mH 0.55Ω DCR + 0V Polarity reversed relative to woofer 0dB –6dB Amplitude (dB) Original network Improved network 60 180° Phase (deg) Photo 17: the KEF 104aB speaker is one of my favourites, apart from the auxiliary bass radiator loading (not shown), which gives bad bass group delay. The upgraded crossover is on the small removable panel. I replaced the fuse with a PTC using the hole for a tweeter level control. Fig.30: KEF pioneered the use of computers to design crossover networks that had a total acoustic output following the correct curve. In this case... it’s a third-order Butterworth filter. Too rapid roll-off New network response provides a better approximation to an ideal Butterworth curve. 3kHz Hump due to Frequency (Hz) tweeter 1.2kHz resonance 90° 0° –90° –180° 500Hz 5kHz 20kHz Frequency (Hz) Practical Electronics | June | 2026 200nF Input from buffer 200nF – 2.2kΩ – 2.2kΩ 2.2kΩ Low-pass output 2.2kΩ + + All op amps are 5532 100nF 100nF 0V 500Hz 4th-order Linkwitz-Riley low-pass filter Decoupling/coupling capacitors not shown 0V 100nF 2.2kΩ 1kΩ 4.4kΩ 100nF – 2kΩ 2kΩ 1kΩ – + + 0V 0V 2nd-order all-pass filter 1kΩ 1kΩ – + High-pass output 0V Mixer Practical Electronics | June | 2026 Group delay (µs) Amplitude (dB) Amplitude (dB) Fig.33: a 500Hz fourth-order linear-phase subtractive crossover filter. Notice the second-order all-pass network. subtraction, resulting in a first-order on accurate component values to slope even if the high-pass filter is ensure complete cancellation. Photo 18: the famous BBC LS3/5a of a higher order. A residual bass signal being fed has the woofer mounted on the rear Adding an all-pass filter is the into a tweeter could damage it and of the baffle for extra delay. way to make it work, as shown in cause intermodulation distortion. cone, which I don’t like. When I make Fig.32. This gives you the added It may be possible to trim this out, “LS3/5As”, I instead front-mount the advantage that the overall filter but it would involve not only amplibass driver and use an all-pass filter. responses, when combined, have tude but phase correction as well. I a constant group delay, resulting could just imagine the response of a in what is incorrectly called a lin- production engineer if I presented a Subtractive filters One interesting point about all- ear-phase crossover filter. This has three-way active speaker with eight pass filters is that the phase shift is the advantage of reducing off-axis trim adjustments! Having said that, it is not a probdouble that of a high-pass or low-pass notches in the speaker’s response. It filter of the same order because the also makes the bass-to-mid crossover lem for home construction because phase shift effect of one capacitor is more time-coherent. However, all tweaking to perfection is part of the 90° in a frequency filter and 180° in an subtractive systems are dependent fun. I could specify high-precision ±1% polystyrene capacitors, but all-pass filter. This is useful to know if 4th-order then the accountant would have a you want to make subtractive filters. low-pass filter heart attack. Linn use these for their An example of this would be Low-pass studio monitors, but at £15k a pair, I trying to make a low-pass filter by output would expect the best components! subtracting a high-pass filtered sigInput A bit of residual bass would do no nal from the direct signal. If this is High-pass – output harm on the low/mid crossover, so done as shown in Fig.31, it doesn’t this would be the best place to try a work because the phase shift introSubtractor linear-phase crossover, especially as duced by the filter messes up the it’s the low end, where varying group 3rd-order 2nd-order delay is most audible. Fig.33 shows low-pass filter all-pass filter a 500Hz fourth-order version. I fancy (Phase delay) Low-pass designing a PCB for this. output The advantage of a phase-linear Input High-pass filter crossover is that the on-axis lobe High-pass – output can be wider before the nulls occur. Filter slopes the same This is because there is a less rapid Subtractor phase shift around the edges of the crossover frequency region. These phase-linear crossover filters were Frequency (Hz) developed by Stanley Lipshitz and John Vanderkooy and should not be 6dB/oct slope Constant confused with the Linkwitz-Riley 18dB/oct slope (LR) crossover (I know I have!). The LR crossover provides good Sum of filter outputs Frequency (Hz) Frequency (Hz) phase-related directivity control, but has a peak in group delay around Fig.31: if the output of a low-pass Fig.32: the subtraction process can the crossover region and a dip in filter is subtracted from the direct be corrected by adding a phasethe idealised overall acoustic power signal, a mirror-image high-pass filter compensating all-pass filter. The response. Importantly, that does not should result, but it doesn’t work crossover filter becomes ‘linearnecessarily mean there will be a dip properly because the phase shifts phase’, an audio marketing term for in the perceived acoustic response. upset the subtraction process. constant group delay with frequency. 61 R1 R3 R6 TP1 TP2 C5 + C6 IC2 R4 C13 R18 R17 R15 R16 R22 R21 R20 R19 C9 C12 C8 C23 IC3 C14 + C20 + R14 R13 R12 R11 C10 + C7 R2 C4 C17 C16 R26 R25 R23 R24 C15 R10 1 R40 J2 R39 C32 + R9 C19 R29 R28 R27 C18 C11 TP3 TP4 J6 1 1 J5 R7 IC1 C2 C3 Power 0V 0V – 0V + R8 R5 C1 Output 0V 0V 0V Out J8 C31 TP10 C26 IC4 C22 R31 R30 R32 C21 C25 R35 R34 R33 C29 TP9 R37 R36 R38 IC5 C27 TP7 TP8 Fig.35: the PCB component overlay for the universal all-pass filter PCB version 01, including the switch and potentiometer mounting boards. SW2 12 VR1a-d To expand on that last point, an LR crossover is designed so that each output is 6dB down at the crossover frequency. On-axis, the acoustic output of the loudspeaker at this point is the result of constructive interference between the two drivers, so the pressures simply sum. Since the signals sent to the two drivers are identical at the crossover frequency, the two acoustic outputs add coherently, resulting in a doubling of pressure (+6dB). The -6dB level from each driver therefore sums to give a flat on-axis response. Without this -6dB level in each branch, the two drivers would produce an acoustic peak at the crossover frequency. The off-axis behaviour is more complex. There is usually a large dip in the vertical off-axis response, depending on the driver spacing and directivity. Soldering time Enough theory for now; it’s time to get building. I suggest using the generic op amp boards described in the December 2022, January 2023 and February 2023 issues to build the switched inverter, attenuverter and single or two stage all-pass filters. For the more complex circuits, the universal all-pass filter circuit shown in Fig.34 is needed. This has up to eight cascaded allpass sections or poles available, built around four dual op amps. Note that 16 stages (two units in series) are required for the effect to be clearly 62 heard on sounds such as kick drums, where the thump will appear to lag behind the click. Grindle has designed a board; its overlay diagram is shown in Fig.35. There is a rotary switch to select the number of all-pass sections required (SW2) and a fourgang potentiometer (VR1) to provide tuning. It would be nice to vary all eight sections to get the biggest 3 – J9 adjustment range possi+ Input ble, but I’ve only recently 2 1 found eight-gang pots as stock items (eg, the Bourns Metal work PTD908-1015F-B103 from Mouser). Those could be used in a future design. Both these controls can be omitted to make a fixed delay unit if desired. The provision of the pot connector (J8) allows a modulator board to be plugged in to make a sweepable studio phaser. This could use LM13700 transconductance op amps, VCAs, Vactrols or optoisolated JFETs. Ordinary JFETs won’t work here, since they have to be referred to ground; these are better used for a 9V guitar phaser pedal, which we’ll describe later on. There is a balanced input section to drive the all-pass filter chain, with a polarity switch, SW1. This is built around two differential stages wired in antiparallel to give equal input impedances on both input pins. There seemed no point in employ- ing a dedicated single op amp (an NE5534) here. Surprisingly, the dual version is cheaper than the single version, and every audio engineer has tubes of 5532s. Poka-yoke The connectors on the PCB are all different sizes, in accordance with the Toyota poka-yoke assembly technique, to avoid mistakes. I blew a board up recently because the output and power supply connectors were both three-pin Molex plugs. I’m never going to make that mistake again! The completed PCB is shown in Photo 19. The switch and pot are on little adaptor boards shown in Photos 20 & 21. Both of these C28 C24 TP5 TP6 C30 + J9 Polarity switch + COM – + Balanced input + – C2 47pF R3 4.7kΩ C1 10µF 25V R2 4.7kΩ V+ 8 IC1a 2 NE5532 – 1 3 R1 1MΩ R4 4.7kΩ C33 47pF + C17 4 100nF R5 4.7kΩ Polarity flip + – V– SW1 R41 22kΩ 0V C4 47pF R8 4.7kΩ C3 10µF 25V R7 4.7kΩ 6 5 R6 1MΩ R9 4.7kΩ C34 47pF – IC1b NE5532 7 + R10 4.7kΩ 0V Photo 21: the quad-gang pot also has its own PCB. This could be useful for other filter projects. Practical Electronics | June | 2026 Photo 19: the universal all-pass PCB. There’s no chance of getting the connectors mixed up... Photo 20: the switch is mounted on a sub-board with a handy SIL header connection. ... Note the extra space for mounting precision axial capacitors vertically if needed. TP9 1 + 2 3 4 + 5 J5 *R11, R15, R19, R23 are 150kΩ only if VR1 is fitted. SW2 V+ C30 100µF 25V C31 100µF 25V 0V Fig.34: the universal all-pass filter circuit, enough for most of our studio needs. To IC1-5 V– + C5 10µF R12 1.2kΩ V 3 C10 100pF TP1 R17 1.2kΩ V– 2 R15 1.2kΩ (150kΩ*) + C20 4 100nF C13 100pF TP2 R18 1.2kΩ 8 IC2a NE5532 – 1 C6 100nF 4 5 6 7 1 2 3 4 5 6 7 8 9 10 0V V+ 2 3 J6 R14 1.2kΩ R11 1.2kΩ (150kΩ*) 8 TP10 C7 100pF R13 1.2kΩ 1 2 + C8 10µF 3 1 VR1a 100kΩ Antilog R22 1.2kΩ R21 1.2kΩ IC2b NE5532 – R19 1.2kΩ (150kΩ*) 6 + + C9 100nF R16 1.2kΩ 2 3 4 C16 100pF TP3 C11 10µF 5 6 7 8 VR1b 100kΩ Antilog R20 1.2kΩ R26 1.2kΩ R25 1.2kΩ V+ 2 3 8 IC3a NE5532 – 6 R23 1.2kΩ (150kΩ*) + C23 4 100nF V– C12 100nF TP4 + C14 10µF R24 1.2kΩ 2 3 IC3b NE5532 – 6 + C15 100nF J8 VR1d 100kΩ Antilog VR1c 100kΩ Antilog V C19 100pF C22 100pF TP5 R29 1.2kΩ R28 1.2kΩ V+ 2 3 R27 1.2kΩ 8 IC4a NE5532 – 1 + C26 4 100nF C18 100nF V– C25 100pF TP6 R32 1.2kΩ R31 1.2kΩ 6 5 R30 1.2kΩ Practical Electronics | June | 2026 – R35 1.2kΩ R34 1.2kΩ IC4b NE5532 7 V+ 2 3 + C21 100nF C28 100pF TP7 R33 1.2kΩ 0V 8 IC5a NE5532 – 1 + C29 4 100nF C24 100nF V– TP8 R38 1.2kΩ R39 100Ω R37 1.2kΩ 6 5 R36 1.2kΩ – + IC5b NE5532 7 + C32 10µF J7 Output R40 100kΩ C27 100nF 63 under the board from the centre pin of the switch connector to the ground end of R40. Two other small parts I forgot are capacitors C33 and C34 on the input balancing amps for RF filtering. Simply solder them across R5 and R10. These mods are shown in Photos 24 & 25. I’ll be doing a mounting board for SW1, which will incorporate R41 and a few other resistors and capacitors to give minimum clicking. Photo 22: this photograph shows how the switch and pot boards are connected to the main PCB. boards could be handy in other projects. They put all the wiring in a neat line so that ribbon cables can be used. This is shown in Photo 22; it is much quicker to assemble than the tangle of different lead lengths that wiring such tagged components normally entails. These boards are so useful in their own right that they are available separately. The pot and switch are not mounted on the main board, since this gives uneven heights on the control panel and the components would be subject to more mechanical strain. A 12-way single-pole rotary switch from Alpha or Lorlin is used because of its standard pinout. Pin 1 is marked on the board, and this should enable the other pins and the wiper pin to all line up to Photo 23: setting the switch to 8 positions with the special washer. 64 push the board on with a bit of judicious bending of the tags. Remember to install the rotation limit washer at position number eight, as shown in Photo 23. Fitting the components All the resistors associated with all-pass stages are 1.2kΩ because this is the lowest value that can be accommodated, giving the lowest noise. 1.2kΩ is the limit because the lowest impedance the NE5532 standard lownoise audio op amp can drive without significantly increased distortion is 600Ω. The two input resistors, R11 and R13, are effectively in parallel at high frequencies. If the tuning pot is to be omitted, resistors R11, R15, R19 and R23 have to be selected to give the desired frequency (from 1.2kΩ up to 100kΩ) as well. If the pot is used, the resistors have to be increased to 150kΩ to maintain the DC path. Powering it The board should be run on a dual-rail supply from ±12V up to ±21V (the absolute maximum for the NE5532 is ±22V). The noise does not increase as the voltage increases, so running at the highest voltage possible gives higher headroom and therefore a better signal-to-noise ratio (given sufficient signal). However, the power consumption and dissipation increase proportionally. The quiescent current consumption per chip is 8mA, so with five chips on the board, that’s 40mA. Given a total power rail of 44V, that’s a steady power consumption/ heat dissipation of 1.7W. In the worst case, loading at full power (clipping at 10kHz) would be 20V into 600Ω, giving 33mA per op amp. This could possibly add up to a shocking 330mA or 6.6W. I’ve never known this to happen in practice, since the op amps shut down, but it’s worth bearing in mind when dimensioning the power supply. Using a standard LM317/LM337 regulated supply with decent heatsinks, such as the one I published in the April 2022 issue, will do the job for up to four boards. That’s sufficient detail for the audio alchemists among you to get started. Next time, I’ll provide some specific applications for the all-pass board, such as a speech symmetriser, giving all the component values, wiring details and PE tests. Photo 24: minor modifications; there are always some! Capacitors are added across resistors R5 and R10. Minor modifications To avoid a loud click when polarity switch SW1 is operated, it’s a good idea to fit a 22kΩ pull-down resistor from the input to the first all-pass filter to ground (R41). I inadvertently forgot this component when giving the design brief to Grindle! This can easily be fitted Practical Electronics | June | 2026 Parts List – 8-Way All-Pass Filter Board 1 double-sided filter PCB coded AO0626ALL, 144 × 58.5mm 1 double-sided rotary switch PCB coded AO0626RSW, 27 × 36mm 1 double-sided potentiometer PCB coded AO0626POT, 22 × 34mm 5 NE5532 dual low-noise, low-distortion op amps, DIP-8 (IC1-IC5) 1 100kΩ quad-gang anti-log (C law) 16mm potentiometer (VR1) [Alps, Alpha, Tayda, AO shop or Blore Edwards] 1 small SPDT toggle switch (SW1) 1 single-pole, 12-way PCB-mounting break-before-make rotary switch (SW2) [Rapid or Tayda] 1 3-pin 2.54mm pitch polarised header with matching plugs and pins (J2) 1 5-pin 2.54mm pitch polarised header with matching plugs and pins (J5) 1 10-pin 2.54mm pitch polarised header with matching plugs and pins (J6) 1 4-pin 2.54mm pitch polarised header with matching plugs and pins (J7) 1 8-pin 2.54mm pitch polarised header with matching plugs and pins (J8) 1 2-pin 2.54mm pitch polarised header with matching plugs and pins (J9) 5 8-pin DIL IC sockets (for IC1-IC5) Capacitors (all ±20% tolerance or better) 3 100µF 25V radial electrolytic (C30, C31 & C32) 4 10µF 25V radial electrolytic (C5, C8, C11 & C14) 2 10µF 25V radial non-polarised electrolytic (C1 & C3) 8 100nF* 63V polyester/MKT (C6, C9, C12, C15, C18, C21, C24 & C27) 5 100nF 50V X7R ceramic, 5mm pitch (C17, C20, C23, C26 & C29) 8 100pF 50V C0G/NP0 ceramic, 5mm pitch (C7, C10, C13, C16, C19, C22, C25 & C28) 4 47pF 50V C0G/NP0 ceramic, 2.5mm pitch (C2, C4, C33 & C34) * value and dielectric depend on the application; this is a good starting point Resistors (all ¼W ±1% axial metal film) 2 1MΩ (R1 & R6) 4 150kΩ (R11, R15, R19 & R23) (only required if VR1 is fitted) 1 100kΩ (R40) 1 22kΩ (R41) 8 4.7kΩ (R2-R5 & R7-R10) 28 1.2kΩ (R11-R38; R11, R15, R19 and R23 not required if VR1 is fitted) 1 100Ω (R39) Order direct from Electron Publishing PRICE £8.99 Photo 25: another minor change. An additional DC path resistor, R41, is mounted under the board from the centre terminal of the polarity switch SW1 to the ground end of R40. 5-year collections 2019-2023 £49.95 2018-2022 £49.95 2017-2021 £49.95 2016-2020 £44.95 Purchase and download at: www.electronpublishing.com EE FR -ROM CD ELECTRONICS TEACH-IN 9 FROM THE PUBLISHERS OF GET TESTING! Electronic test equipment and measuring techniques, plus eight projects to build FREE CD-ROM TWO TEACH -INs FOR THE PRICE OF ONE • Multimeters and a multimeter checker • Oscilloscopes plus a scope calibrator • AC Millivoltmeters with a range extender • Digital measurements plus a logic probe • Frequency measurements and a signal generator • Component measurements plus a semiconductor junction tester (includes P&P to UK if ordered direct from us) Teach-In 9 DVD-ROM – Get Testing! £8.99 PIC n’ Mix Including Practical Digital Signal Processing PLUS... This series of articles on DVD-ROM provides a broad-based introduction to choosing and Teach-In 9 using a wide range of test gear, how to get the best out of each item and the pitfalls to avoid. It provides hints and tips on using, and – just as importantly – interpreting the results that you get. The series deals with familiar test gear as well as equipment designed for more specialised applications. The articles have been designed to have the broadest possible appeal and are applicable to all branches of electronics. The series crosses the boundaries of analogue and digital electronics with applications that span the full range of electronics – from a single-stage transistor amplifier to the most sophisticated microcontroller system. There really is something for everyone! Each part includes a simple but useful practical test gear project that will build into a handy gadget that will either extend the features, ranges and usability of an existing item of test equipment or that will serve as a stand-alone instrument. We’ve kept the cost of these projects as low as possible, and most of them can be built for less than £10 (including components, enclosure and circuit board). YOUR GUIDE TO THE BBC MICROBIT A LOW-COST ARM-BASED SINGLE-BOARD COMPUTER Get Testing Three Microchip PICkit 4 Debugger Guides Files for: PIC n’ Mix PLUS Teach-In 2 -Using PIC Microcontrollers. In PDF format © 2018 Wimborne Publishing Ltd. www.epemag.com Teach In 9 Cover.indd 1 01/08/2018 19:56 PLUS! You will receive the software for the PIC n’ Mix series of articles and the full Teach-In 2 book – Using PIC Microcontrollers – A practical introduction – in PDF format. Also included are Microchip’s MPLAB ICD 4 In-Circuit Debugger User’s Guide; MPLAB PICkit 4 In-Circuit Debugger Quick Start Guide; and MPLAB PICkit4 Debugger User’s Guide. ORDER YOUR COPY TODAY from www.electronpublishing.com Practical Electronics | June | 2026 65