Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L R By Jake Rothman All-pass filters, part four – new PCBs & symmetricisers To make full use of the all-pass filter board presented last month, we need a better polarity-inversion switch. Coupling capacitors and pull-down resistors to stop offset voltages building up have been added to a switch mounting PCB, preventing switch clicks. This means that the resistor, formerly R41 in Fig.34, described last month as being mounted under the board is no longer needed. It is now R4 on the switch board. This added switch circuit is shown in Fig.35, the overlay in Fig.36, and the final construction in Photo 26. It plugs into the new main PCB shown in Photo 27. These additional components have their own numbering. This assembly is useful in its own right for adding audio change-over switching functions to any directly coupled circuitry. There’s provision for a dual-­ colour LED. When the switch is Photo 26: the polarity switch assembly using an older-style Alps switch. Photo 27: the new version of the allpass filter PCB for an 8-gang pot. 10 pushed in, it glows red and the polarity is reversed, while green’s for normal. This LED is a two-terminal back-to-back red/green type, such as the Tayda A-1076. This is driven at around 2mA via a 10kΩ 0.6W current-limiting resistor R9, which gives enough brightness with modern panel LEDs. Its ground current is kept separate from the signal ground to prevent clicks on switching. This connection should be taken to the ‘dirty ground’ on the power supply capacitors’ Earth node if possible, although it is fine going to the ground sides of decoupler capacitors C30 and C31 on the main PCB. There is also a spare changeover switch section that can be accessed from the tags on top of the switch (with the older Paxolin top Alps switches). The board also acts as a convenient way of mounting those Alps Alpine (manufacturer’s part number SPUJ193700) style push switches to the front panel with a couple of fixing holes/angle brackets. Another board is available for an alternative switch from E-switch, as shown in Photo 28. This is available for £1.60 from Mouser (PBH6UEENAG1RBLK). These switches can be difficult to find nowadays. I used to get them from Rapid, but after ploughing through their 50 pages of industrial push-buttons, I couldn’t locate them. Luckily, I still have a bag of 100 older types, so I can supply them via AOShop. Farnell and CPC still sell them; CPC’s are the cheapest, while RS is mega-­ expensive. It is a good idea to add a bypass switch as well. At the moment, only the input is balanced, so the circuit in Fig.37 should be used. To make it fully compatible with balanced professional broadcast gear, a balanced output with a hard bypass arrangement would be needed. Practical Electronics | July | 2026 1 2 J1 R1 100kΩ R2 100kΩ R3 100kΩ 1 1 C6 47µF 25V + S1b 2 All-pass board In Bypass switch Unbalanced Output 0V Metal work 2 J2 Output (to all-pass board) + C5 47µF 25V ‘Long’ lead Alps/Schadow version Out Input Output All-pass effects All-pass filtering can be an effect 3 Input – in itself, giving an illusion of exR5 R6 R7 R8 tra bass, a char100kΩ 100kΩ 100kΩ 22kΩ acteristic ‘chesty’ 4 booming sound 0V signal on male speech. J3 1 This effect is also V+ ≈2.2mA Polarity S1c 2 produced by small 0V (dirty) indicator bass-reflex speak3 V– ers with 6th-orTo all-pass Green + Red – der bass roll-offs, board typified by most LED1 Fig.35: this small active studio Dual-colour polarity switch R9 monitors such as 10kΩ circuit is those by Genelec, 0.6W NC useful in many Adams and KRK. S1d S1d unused – available situations, like NC for user applications From micromixer inputs. NC phone, to preamp, to equalisation, to –phase Output –phase Output Power in AC-coupled audio interface, to the 0V 0V +phase + 0V – +phase signal signal amplifier, the sum of all the phase shifts due to the culminating highJ1 J3 J2 pass poles can be quite destructive Note: three poles are to bass coherence. R9 unused and available for user appplications Electronically, a total of around S1 16 all-pass poles with 100nF capacC3 C6 + + itors are needed for this psychoacoustic delay-based ‘bass boost’ C2 C1 C4 C5 + + + + to be audible, but this is testimony R R R R R R R R to how insensitive to phase human 4 3 2 1 8 7 6 5 hearing is. Compare this to our perception of frequency, where just one or two poles of filtering + ‘Long’ are clearly audible. lead Phase shift can E-switch version be modulated by D1 a low-frequency oscillator or LFO, giving an effect J1 J3 J2 similar to vibrato, but it has a strangeNote: the S1d pole is R9 unused and available for user appplications ly nausea-inducing S1 S1d S1c quality. In stereo, it C3 C6 + can almost give rise to seasickness! That’s C2 C1 C4 C5 + + + + because it isn’t pitch but frequency modulaR R R R R R R R 4 3 2 1 8 7 6 5 tion, which demolishes harmonic relationships. S1a S1b Dedicated frequency Input + – Use J2 channel if stereo is needed C4 47µF 25V + J2 3 – Balanced Input 4 0V signal – + R4 22kΩ Fig.37: wiring a bypass switch into the circuit. Polarity switch board + Output (to all-pass board) + 3 Input – S1a C2 47µF 25V C3 47µF 25V + Input + C1 47µF 25V + J1 shifters produce a static version of this effect. Harvey (my son) and I have been messing about with allpass filtering in music software. Although a research rabbit hole, this can lead to insights that can then be made into analog products. New eight-gang pot version Since the Editor pointed out that eight-gang pots were available from Bourns via Mouser, it seemed desirable to have the all-pass board redesigned for them, along with a new pot mounting board. It’s a bit silly that I had never found them, since I’ve just noticed they were used in John Clarke’s active crossover, back in September 2018, and I’ve been needing some for years. The part number is PTD9082015K-C503, where the C503 means 50kΩ reverse/anti log. The new allpass circuit is the same as before, except that four extra 47µF pot ‘anti-scratch’ coupling capacitors, C35 to C38, have been added. Originally these pot coupling capacitors (C6, C8, C11 and C14) were 10µF, but I found some 47μF 16V bi-polar types that fitted fine. It’s always worth going bigger and bipolar if you can. There are also an additional four 150kΩ DC path resistors, R41 to R44. Basically, it’s + D1 Practical Electronics | July | 2026 Fig.36: the polarity switch PCBs. The boards plug into the eight-gang pot board. Photo 28: the PCB to suit the E-switch part with closer pins. 11 a duplication of the original circuitry around IC2 and IC3, and making it the same around IC4 and IC5. The new overlays are shown in Fig.38 and Fig.39, while the completed assembly can be seen in Photo 29. This board has been made with vertically mounted polystyrene capacitors, shown in detail in Photo 30. The eight-gang pot board is also useful for stereo units and gives a wider adjustment range on a single all-pass filter board compared to the four-gang board. The assembly is shown in Photo 31. While redesigning this board, we also took the opportunity to add the missing pair of 47pF capacitors, now numbered C33 and C34. Testing Start by testing it with a sinewave signal from an oscillator or waveform generator. The unit should have no effect unless you rotate the pot rapidly, where there will be a slight frequency modulation due to the varying delay. A pulse wave or uneven markspace ratio square wave is a good way to test that each of the delay stages is working. As the switch is incremented, the number of ‘squiggles’ in the waveform should increase. Photo 32 shows the effect of one all-pass stage on a square wave. Speech waveform symmetriciser This is a popular project among radio amateurs who need to max- Photo 30: axial polystyrene capacitors can be mounted vertically if the PCB holes are plated, but it is not recommended for single-sided PCBs since the pads may tear off with an impact. Photo 31: the Bourns 8-gang pot and its associated PCB. Such pots should be mounted on a pluggable board for easy replacement when they wear out. imise their speech intelligibility in poor reception conditions. This is achieved by scrambling the phase of the speech harmonics relative to the fundamental to make the waveform more symmetrical using an all-pass filter. This reduces the peak-to-peak voltage, enabling higher volume, maximising the transmitter modulation. This circuit is sometimes called a phase shuffler, an ambiguous term first coined by Blumlein for his binaural stereo-widening circuit. Michael Gerzon also used the term shuffler to describe a side-signal +8dB low-frequency booster for his stereo microphone array. The original way to process speech was by crude symmetrical diode clipping, but this causes gross distortion like a fuzzed guitar. This is minimised by telephone-style band-pass limiting using 300Hz high-pass and 3.4kHz low-pass filtering before and after the clipping. A more advanced system is to convert the audio to RF, clip it, then convert it back. Photo 29: this shows how the new 8-gang PCB is wired to the other boards. Photo 32: the effect of a single-pole all-pass filter on a square wave. 12 Practical Electronics | July | 2026 The question is: how to optimise the phase shift? It’s all very well using an oscilloscope to check waveform symmetry on the bench, twiddling the pot while making various vocal sounds. It occurred to Harvey that a meter would be a good idea. I thought that rectifying the upper and lower halves of the waveform separately, averaging them out, then reading the difference with a centre-zero meter might do the job. I’m working on it. Alternatively, there are some cheap LCD oscilloscope modules out there, such as the Kuman DSO 138 kit. Photo 33: Symmetra-Peaks ensured maximum modulation of radio transmitters. Broadcast symmetriciser Now for a bit of history. Thanks to Photo 34: a full view of the Symmetra-Peak, showing the big audio Harvey’s extensive eBay searches, transformers at each end and potted circuitry in between. I’m able to show you a unique bit L1 of gear called the Symmetra-Peak 7H in Photo 33. As most people know, 15kΩ 15kΩ radio in the USA is still very imT1 T2 portant, and there has always been C1 C2 2 3 4 300nF 300nF Input Output huge commercial competition between stations. This means that 600Ω 600Ω ‘standing out’ on the tuning dial, Asymmetric waveform Symmetric waveform such as having the loudest sound, L2 Three more lattice filters was crucial. 7H In 1959, Mr. Leonard. R. Kahn Lattice filter (one stage) designed this secret processor box Fig.40: the Symmetra-Peak circuit. It is fully balanced to avoid hum pickup. that was sufficiently Hi-Fi to symmetricise the signal going to a radio The most popular design was The centre frequency chosen is station transmitter, giving maximum the Datong processor. This hetero- normally 200-400Hz, depending modulation. It was an expensive dyne-based technique is complex, on the speaker’s voice, to delay the bit of audio engineering, but the inbut it pushes the intermodulation fundamental relative to the harmon- creased listenership and consequent products out of the audio band. ics. The number of poles required is increase in advertising revenue All-pass filtering, along with high- at least eight. It is worth adjusting made it a worthwhile investment. speed amplitude compression, gives using one’s own or a client’s voice. Inside was an epoxy-encapsua clearer result, suitable even for Male voices are more asymmetric, lated passive lattice all-pass filter studio work, for which this design lower in frequency, with more har- terminated by input and output is primarily intended. Compressors monics than female voices and thus balancing transformers, as shown and clippers always work more con- respond better to this technique. in Photo 34. Luckily, I didn’t have To polarity switch board sistently when fed with symmetrical to dissolve the epoxy with waveforms, because DC shifts due to toxic solvents to bring you In-phase output Switch board LED the asymmetry are avoided. the circuit, since it was alOutput from switch – 0V + ready in the patent. This is Output Power Balanced input Out of phase output shown in Fig.40. + – + – 0V 0V – 0V + 0V Notice how the transformers have a high turns ratio. R8 R1 R5 J9 C9 C12 + C8 IC3 C14 TP3 TP4 J6 1 To S2 PCB Practical Electronics | July | 2026 C16 C19 R26 R25 R23 R24 C15 R29 R28 R41 R27 C35 C36 1 C25 R35 R34 R43 R33 R31 R30 R32 R42 IC4 C18 J8 C22 C26 C24 C21 TP5 TP6 To 8-gang potentiometer PCB C31 TP10 C29 IC5 C37 C38 J1 VR1a-h J5 C30 J2 + C32 + R9 C11 1 J3 TP9 C28 R37 R36 R38 R44 J1 To J1 all-pass board R22 R21 R20 R19 J7 To J8 all-pass board R18 R17 R15 R16 C23 + TP1 TP2 C13 J2 R40 R39 + + C5 C6 R4 + IC2 R10 + R14 R13 R12 R11 C4 C17 R2 C10 C20 R7 C34 + R6 C7 IC1 C33 + R3 C2 + C1 C3 C27 TP7 TP8 Fig.39: the PCB overlay for the separate 8-gang pot board. Fig.38: the PCB overlay for the 8-gang pot version of the all-pass filter board. 13 This was necessary to match the high impedance (15kΩ) of the lattice filter to the low 600Ω line impedance that was the audio standard of the time. The filter was high impedance because the only passive audio components available in the valve era were high-impedance inductors and relatively low-value capacitors, with 7H and 30nF being used. If you were to attempt this passive approach today, modern pot cores of a few 100s of millihenries and high-value polypropylene capacitors of several µF would be used, with no transformers. A high-current buffer stage would be needed to drive it, though. Audio inductors are expensive, so the more cost-effective way is to use active op amp circuitry to eliminate them altogether (eg, using gyrators). The lattice can be replaced with four op amp all-pass filter stages. Mr. Kahn found that a good compromise between complexity and performance was to make the delay around 5ms below 300Hz and under 3ms above 300Hz. The effect of the Khan Symmetra-Peak on a square wave is shown in Photo 35. This reduced the energy deviation on general audio signals with an energy imbalance of 20-30% down to 1-2%. The phase plot of the Symmetra-­ Peak is shown in Fig.41. Note that Photo 36: the Orban Optimod. Source: www.proaudio.com each vertical line on such plots denotes a polarity reversal of 360°, each representing two poles of 180°. Therefore, the unit has four poles. 360° is often called a phase rotation; all-pass filters are sometimes called phase rotators. This plot was a pig to make. I had to get Harvey to set up the Audio Precision analyser software. There have been a few all-pass speech symmetricisers produced for the ham radio scene. Research by Gary Blau (amateur radio callsign and website W3AM) and by James L Tonne (callsign W4ENE) has resulted in an active version of the Symmetra-Peak with the frequency set to 340Hz (R = 10kΩ and C = 47nF). They doubled the number of poles from Kahn’s original four to eight, which further improved the +180 Symmetra-Peak phase shift plot +150 +120 Phase Shift (°) +90 +60 +30 -60 -90 -150 +180 8-pole all-pass filter (22nF) phase shift plot +150 20 50 100 200 500 1k 2k Frequency (Hz) 5k 10k 20k Fig.41: the Symmetra-Peak phase plot. This indicates that it uses a fourpole all-pass filter. Amplitude (dB) Sound source +90 Phase Shift (°) -180 Barcus Berry Sonic Maximiser Attempts were made to make similar processors more affordable by companies such as Aphex and Circuit Research Labs, but the biggest seller was the BBE Sonic Maximiser. This was a summed second-order output state variable filter, which provides some all-pass action in Distant mic +120 +60 0 +30 –15 Pre-amplifier +0 Phase cancellation notches -30 -60 -90 -120 -150 -180 20 50 100 200 500 1k 2k Frequency (Hz) 5k 10k 20k Fig.42: a phase plot for the 8-pole allpass filter with 22nF filter capacitors. The pot is in the middle position. 14 The Optimod The Optimod started a trend for units to optimise the modulation of transmitters. It culminated in the Orban Optimod, a five-band multi-way limiter. The internal active crossover provides a large amount of phase shuffling, along with a bit of dedicated HF all-pass delay, to provide ultra-fast ‘look ahead’ limiting of peaks. It was so effective that artists said their records sounded better on the radio than in the studio! There were two versions: a mono version optimised for AM transmission, and a larger stereo one for FM, shown in Photo 36. I saw an AM unit at Retrotech last year for £300 and regretfully didn’t buy or photograph it. They were originally very expensive, built to almost military standards, with hermetic metal-can op amps. +0 -30 -120 Photo 35: the effect of the Symmetra-Peak on a square wave. symmetry. With a 50kΩ anti-log Bourns pot and 22nF capacitors, this frequency is obtained with the pot at around one-third rotation, as shown in Fig.42. 50Hz Sh Makes delay paths de ort almost equal to pring lay Near mic ... phase into line. ... if or z is direc ero us t i de ed nje la All-pass filter unit cti y and polarity flip on Pre-amplifier Fig.43: an all-pass filter in the near microphone of a pair can minimise the effect of cancellations. Frequency 10kHz Amplitude (dB) 0 –15 Using an adjustable all-pass filter notches are moved out of critical area Mixer Mono 50Hz Frequency 10kHz Output Practical Electronics | July | 2026 Photo 37: the Radial Engineering Phazer, available from Thomann. the normal way, but it is little more than a bass/treble booster circuit with steeper slopes, like a Hi-Fi Baxandall tone control on steroids. There is a dynamic noise-filtering VCA circuit on the treble output, which avoids the noise-boosting problem that aural exciters typically suffer from. Of course, the marketing hype has the usual exhortations about psychoacoustic phase correction, reinserting lost information etc, shrouding what is, in fact, quite basic audio circuitry in secrecy. The circuit was disclosed in the data sheet for a BBE chip, the NJM2153, made by JRC. The unit worked well on the low-resolution sampled drum beats of the early 1990s. Sadly, it didn’t work on smooth atmospheric or orchestral music. The use and abuse of phase Because of the Optimod, artists wanted that sound in the studio, and units such as the Aphex Dominator were made to achieve this. The result was that the music was subjected to ‘Optimod style’ processing twice if played on the radio. This didn’t sound twice as good; the result was a strange, loud audio ‘mush’. The Red Hot Chili Peppers and their 1999 album Californication took this to extremes; it is a loudness landmark. It was so heavily processed that the waveform con- Phase alignment unit An all-pass filter is possibly the last effects unit a studio would buy; phase alignment units are very expensive, typically £350 due to the low demand, so it’s a box worth making yourself. They are mainly used in the studio to ensure that a sound source recorded with multiple microphones or a directly injected instrument, such as a 8-pole all-pass filter (100nF) phase shift plot +180 +150 +150 +120 +120 +90 +90 +60 Phase Shift (°) Phase Shift (°) +180 tains flat clipped sections within the waveform, not merely at the peaks. This is what can happen if you phase-shuffle audio that has already been clipped! It sounds unnaturally loud, which is great for grabbing attention in adverts, but fatiguing for music listeners, especially over a whole album. Thankfully, the loudness wars seem to have peaked in the mid-2000s, no pun intended. [Editor’s note: the loudness war was so bad that Rush’s 2002 album Vapor Trails was remixed and re-released in 2013 after years of complaints about its harsh, fatiguing sound. Even the band members agreed! Metallica’s 2008 album Death Magnetic was another infamous example; later digital releases used a less punishing master.] Loudness standards such as the European Broadcasting Union’s R 128, along with loudness normalisation on streaming services, finally made this kind of nonsense far less common. +30 +0 -30 -60 +0 -60 -90 -150 -150 5k 10k 20k Fig.44: a phase plot for the 8-pole allpass filter using 100nF caps with the pot fully clockwise for maximum delay. Practical Electronics | July | 2026 All-pass filter minimum delay phase shift plot -30 -120 50 100 200 500 1k 2k Frequency (Hz) An unfortunate news flash We’ll have to interrupt the phase project now, since I’ve been informed by former PE contributor, Ben Duncan, of a looming audio industry chip supply problem. Ben Duncan introduced me and many home constructors to the original Philips TDA1034, later becoming the Signetics NE5534 in September 1978. The dual version, the NE5532, is now ubiquitous. These original devices are shown in Photo 39. However, the analog audio world is now about to suffer a semiconductor price-gouging catastrophe. The full-specification NE5532 op amp is soon to be no more. +60 -120 20 bass guitar, with a mic’d-up amp speaker, sum properly. This trick makes it possible to avoid frequency notches caused by cancellation in critical frequency areas, as shown in Fig.43. Of the few units made, the only ones I’ve used are the Radial Engineering Phazer in Photo 37 and the Little Labs Analogue Phase Alignment Tool (Photo 38) which they have at my local studio, Giant Wafer. For the eight-gang pot all-pass filter, the total amount of available phase shift is much greater than these units. The phase shift at opposite ends of the pot rotation is shown in Figs.44 & 45. +30 -90 -180 Photo 38: Little Labs. Source: eBay. -180 20 50 100 200 500 1k 2k Frequency (Hz) 5k 10k 20k Fig.45: a phase plot for minimum delay, with the pot fully anticlockwise. Photo 39: left-to-right, top-tobottom: an original 1977 TDA1034, Signetics NE5534A, Signetics ceramic NE5532, various TI NE5532s and JRC’s SIL NJM2114. 15 Use multimeter diode test to look for 0.7V voltage drop. Fakes and NEW Texas NE5532s do not have this. V+ 8 3(5) Inputs ±0.7V 5532 – Dr Craig Sawyer, an associate of Ben’s, has questioned Texas Instruments, the main producer, about changes in the specification. TI deliberately engineered the NE5532 and NE5534 down in February 2024, removing 8V from the maximum supply rail voltage, from ±22V to just ±18V! The slew rate has also been reduced from 9/µs to 5/µs and the current consumption has dropped from 8mA to 6mA. The die geometry must have been shrunk as well, because the static voltage maximum has been reduced from 2kV to 1kV. One has to wonder if these are really the same parts, given their wildly different parameters from the original. Was the NE5532 too good? The original NE5532 was just too good for a low-priced generic product, and its wide availability was stopping customers from buying expensive, proprietary audio op amps. It’s the usual accountancy-led dodge of die shrinkage and making legacy analog chips fit in with modern digital processes. TI was originally castigated for making the high-frequency distortion of the NE5532 worse on their first versions compared to those made by Signetics when they took the company over. Also, the lower-noise single version, the NE5534, was temporarily deleted. The same thing is going on everywhere; shareholder value is all that counts now. Quality be damned. Supermarkets have been trying to shift consumers from cheap generic drugs to high-margin brand-name drugs for years. Why sell ibuprofen painkillers for 35p (the typical NE5532 price) when you can sell Nurofen Period Pain tablets for £4.69 (a typical expensive ‘audio’ op amp price)? The example is perhaps even worse in that both contain the same cheap active ingredient; at least the more expensive op amps are a different design. I used to buy my NE5532s from Rapid, since they were the cheapest, but they told me that they are 16 1(7) 2 Output 3 2(6) Input protection diodes 5532 top view V+ 8 1 + Fig.46: checking for the internal diodes in an “NE5532” op amp IC. – + 7 + – 6 4 4 V– 5 V– only selling the new TI version now. Fake op amps I’ve also come across a few fake NE5532s, normally provided with low-cost Chinese Hi-Fi project kits on eBay and the like. Why they think it’s worthwhile faking a cheap (at the time) chip, I don’t know. These are often renumbered 4558s; their low 3.5mA current consumption is the giveaway, as a real and proper NE5532 draws 8-9mA. Another ‘feature’ is that the fakes don’t have the back-to-back protection diodes across the two input pins, making them easier to damage. Interestingly, a YouTube video from Michael Beeny (https://youtu.be/ Jr-AXau7va4) said that the new TI NE5532s omit these diodes as well. Fig.46 shows how to check it. However, the new ones I bought from Rapid did have the diodes when I checked (I later confirmed they are the old version). Photo 40 shows some fake NE5532s from an AliExpress preamp kit. Oddly, they drew 2.6mA, but the current consumption took off rapidly at ±20V. Also strangely, they had a 0.9V drop between the input pins, both ways on diode test. I know of no other op amp that has this. There’s only one way to find out what it is: dissolve the encapsulation. Overrated? Some audio equipment makes use of the NE5532’s high voltage rating; I certainly do in some of my designs, running them at ±21V for years with no failures to date. Using them at ±22V will give a headroom of +24dBu. One company (can anyone out there remember who?) even ran NE5532s at ±24V with an acceptable initial infant mortality rate. Someone told me it was Hill Audio, who advertised their 1980s mixers as using 5532s throughout, but I checked and they ran them at a conservative ±17.5V. Rupert Neve’s ISA110 mixer channel used ±15V, with NE5534s. Its high headroom of +26dBm was achieved by audio transformers, one being a specially Photo 40: fake NE5532s from AliExpress. A bad copy or a completely different chip? made toroidal ‘orange doughnut’ step-up output transformer. The wonderful Clive Green designed Cadac consoles employed in West End theatres definitely used ±20V and NE5532s throughout. One has to be careful with the power supply when running close to the maximum. Overvoltage protection must be in place on the power supply in case a regulator goes short-circuit, letting through the higher unregulated voltage, or all the op amps will die. For this, I sometimes strap a 5W, 43V Zener diode across the power rails on the board. The loss of one rail while the other is still running must also be prevented, since this can blow the protection diodes on the input pins, causing permanently high distortion. The old Kemo filter company trick of putting 560Ω resistors in series with the input pins works to protect them. I recommend it for followers on the inverting pin and on the non-inverting pin for summing amps and inverters. These techniques are shown in Fig.47. 22Ω Fusible ¼W 3 + 2 – 8 Output 43V 5W Zener 1 4 22Ω Fusible ¼W Follower 3 + Input 560Ω 1 2 – Output 560Ω Rf RIN Input 2 – 3 + 1 Output Inverting 560Ω 0V Fig.47: added protection for an op amp to prevent damage from overvoltage. The 560Ω resistors help if one power rail goes missing. Practical Electronics | July | 2026 Back in phase, distortion Looking at the distortion of the all-pass filter using different chips was interesting. The plot is shown in Fig.48. The ‘new’ NE5532s from Rapid were only slightly worse than the original old Texas devices. They initially told me that I had purchased the new (2026) versions, but I later confirmed that they are still selling off stock of the pre-2026 version, which explains why these ones still perform well. This is the problem with releasing a new part with the same model number as the old one! All-pass filter old vs new NE5532 1 0.5 0.5 0.2 0.2 Total Harmonic Distortion (%) Total Harmonic Distortion (%) 1 At least the new TI NE5532s will still work at ±15V, which is the standard analog audio supply. Luckily, the noise spec is still the same at 5.5nV/√Hz. I will have to check the maximum output voltage swing into 600Ω, though, because it is now omitted from the data sheet. Rule 101 of data sheets: If a specification is not there, it’s usually bad. The Nisshinbo (formerly JRC) NL8802 at £1.35 with a ±22V rating looks like a viable NE5532 replacement contender, but it’s only available in SMD packages. Nisshinbo is the only company still making the original device to the ±22V specification, including the 8-pin DIP device, the NJM5532DD, although it carries the worrying text, “not for new designs”, on the data sheet. Profusion Semiconductors, who are the UK distributor for Nisshinbo, only had the SMD version, the NJM5532CG-TE2 in stock recently, but they are getting a load of the 8-pin DIPs (NJM5532DD) in, before they cease production in March next year. Profusion has been very helpful to me, giving me free THAT Corporation chip samples and real humans on the telephone, so I feel a bit of a plug is in order. 0.1 0.05 0.02 0.01 .005 .002 New .001 .0005 .0001 All-pass filter NE5532 vs LM833 THD+N Old Old LM833 LM833 0.1 0.05 0.02 0.01 .005 .002 NE5532 NE5532 .001 50 100 200 500 1k 2k Frequency (Hz) 5k 10k 20k Fig.48: the new NE5532s from Rapid had slightly higher distortion but I don’t think they are the latest version. Practical Electronics | July | 2026 .0001 All-pass filter film capacitor THD+N comparison 0.2 0.1 0.05 0.02 Polyester Polyester 0.01 .005 .002 Polystyrene Polystyrene .001 400V Polyester 400V Polyester .0005 .0002 20 1 0.5 .0005 .0002 I took a risk and tested a Rapid NE5532 (batch no 43A3K3M with no clear date code) at ±25V. There was no rapid rise in current, starting at 7.4mA at ±15V and holding at 8mA at ±25V. This pretty much confirmed they were the older version, especially, given what happened next. Not long after that, I received a batch of TI NE5532s from Mouser. I’ve confirmed they have no diodes across the input pins and they blow up when a ±18V supply is connected. The current consumption is indeed lower, at 5mA with a ±15V supply. The batch number is 59A7CLM and they have a little underlined G4 in the top right corner, which is a useful way of identifying these new parts. The peak distortion on the all-pass filter rises from 0.0007% to 0.0025% at 8V into 600Ω load using ±15V rails – a nearly fourfold increase in distortion from the ‘same’ part! So the new NE5532s are definitely inferior and will blow up in some old circuits. Strangely, the LM833 performed very badly in the all-pass filter at 8V RMS and 4V RMS, as shown in Fig.49. I think they must have been upset by the low impedance and high capacitance loading. Another more subtle aspect of distortion is that generated by the frequency-determining capacitors: C6, C9, C15, C18, C21, C24 and C27. Polystyrene or polypropylene types give no problems. Using cheap polyester capacitors will give an increase, as shown in Fig.50. The distortion is 10 times worse at 8V RMS compared to polystyrene using 22nF 63V AVX 5mm polyester box capacitors. [Editor’s note: polystyrene, polypropylene and NP0/C0G ceramic capacitors can be considered ‘ideal capacitors’ for audio applications. Polyester types are definitely a step Total Harmonic Distortion (%) Back to Texas Still, the idea of someone changing a chip for another with the same part number and it blowing up is totally irresponsible. The prospect of Les Misérables grinding to a halt due to a new Texas NE5532 being inserted as a replacement does not bear thinking about. (Sadly, the theatre desks are digital now because the old analog Cadac J series took the space of 10 seats selling at £200 a night). The onsemi SMD versions of the NE5532 were downrated to ±20V because the dissipation was too much for the small package, and they stopped making them in 2022, although they are still available (for now). It looks like TI’s OPA2891 video op amp is a possible new NE5532 contender for the top audio amp, but it costs £5.00 and is over the top and so fast it will be unstable unless the circuit board is designed impeccably. Craig informs me that there have been noise problems with some of the LM4562 (ex National Semiconductor) op amps, and they’ve gone up from £2.50 to £6.00 recently [we’ve never managed to get them to work well; they result in much higher distortion when swapped into a circuit that’s happy with an NE5532 or LM833, we suspect because they oscillate with less-thanperfect supply bypassing – Editor]. TI is even thinking of deleting these and replacing them with the LME49720, which looks identical, but of course costs £7.00. TI took over National Semiconductor’s wonderful linear range, which I suspect is to be decimated. National’s LM833 (more on this later), which was always recommended as a slightly lower-cost NE5532 replacement where the high voltage rating was not required, also seems to have gone up in price. .0002 20 50 100 200 500 1k 2k Frequency (Hz) 5k 10k 20k Fig.49: distortion with the LM833 is much worse than the NE5532 in this circuit, likely due to output loading. .0001 20 50 100 200 500 1k 2k Frequency (Hz) 5k 10k 20k Fig.50: regular polyester filter caps are much worse than polystyrene but HV polyester caps are even better. 17 down; good enough for some tasks, but they are definitely not as linear, as demonstrated in Jacob’s plot.] With 22nF X7R ceramic capacitors, the distortion was almost 2%, even at 1V RMS, as shown in Fig.51. [Editor’s note: never use X7R, X5R, Y5V, Z5U or similar ceramic capacitors for audio; even for signal coupling. They have horrible non-linearity due to their high voltage coefficient. C0G/NP0 types are good for audio as they lack this fatal flaw. X5R/X7R types are good for supply bypassing, though.] Clearly, this all-pass filter is a stressor of capacitors and op amps, 10 All-pass filter X7R ceramic capacitor THD+N Total Harmonic Distortion (%) 5 2 1 0.5 0.2 0.1 .05 .02 .01 .005 .002 .001 20 50 100 200 500 1k 2k Frequency (Hz) 5k 10k 20k Fig.51: the distortion with X7R ceramic capacitors. What a mess! Only use them for supply bypassing. as is to be expected. After all, it’s a string of eight filters! I made a version using 400V 100nF Suntan dipped polyester capacitors (MPN TS02002G104KSB0D0R) from Rapid (order code 10-4167) and the distortion was fine (see Fig.50). The thicker 400V dielectric was suffering less stress. To safely mount these larger types of dipped capacitors, it’s a good idea to provide strain relief on the leads, as shown in Photo 41. I hate to think how it would sound made with TL062 op amps and Y5V ceramic capacitors. Maybe it could be a new distortion effect; the Phaso-dilator, perhaps? Losing the plot With the horrific real UK inflation rate (only 3.5%? Who are they kidding?), it’s time to do some hedging or ‘prepping’ of desirable components if you have spare cash. The huge price increases in solder and wire prices recently were a warning. Low Value Import (LVI) duty relief below £135 on our Chinese component and PCB parcels to the UK is to be removed in March 2029. “Making Tax Digital (difficult)”, with a possible software subscription, also worries me. I feel like falling on my soldering iron sometimes, but Teach-In 9 CD/DVD-ROM – Get Testing! This series of articles provides a broad-based introduction to choosing and 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 analog 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. Photo 41: preforming capacitor leads using an RS tool. It’s no longer available but Grove Sales PNR5000D (https://grovesales.co.uk/ products/c-shape-forming-pliers) looks even better. I will try it out. that would be a health and safety compliance failure! We’ll continue the phase shuffle dance next month with putting boards in boxes and the phase-linear crossover board. Thanks to my son, Harvey Rothman, for additional research in putting together this article. The PCB designs are by my usual partner in PE crime, Mike Grindle. EE FR -ROM CD ELECTRONICS TEACH-IN 9 £8.99 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 PRIC E 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 PIC n’ Mix Including Practical Digital Signal Processing PLUS... YOUR GUIDE TO THE BBC MICROBIT Teach-In 9 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 The CD/DVD-ROM includes the article series in PDF format, plus the software for the PIC n’ Mix series of articles. It also includes the full Teach-In 2 book – Using PIC Microcontrollers – A practical introduction – in PDF format. JUST £8.99 (includes P&P to UK if ordered direct from us) ORDER YOUR COPY TODAY: www.electronpublishing.com 18 Practical Electronics | July | 2026