Fullscreen HTML5Med Res (??MB)HTML4

AUDIO OUT AUDIO OUT L R By Jake Rothman Transformers in audio – Part 2 performance of real (as opposed to ideal) transformers. CW1 Input CP Primary series resistance Primary leakage inductance RP LP Primary lumped winding capacitance Interwinding capacitance Primary inductance RO Secondary leakage inductance Ideal transformer LP Core loss: hysteresis and eddy currents (non-linear) NP Primary turns Secondary series resistance LS Losses RS Output Secondary lumped winding capacitance CS NS Secondary turns Interwinding capacitance CW2 Fig.9. The equivalent circuit of a transformer with all the parasitic elements added. L ast time (July 2022) we looked at the uses and advantages of one of the oldest components in electronics – the transformer. This month, we will examine the limitations of real, as opposed to ideal transformers, as well as exploring some common audio transformer applications. Transformer equivalent circuit and ‘parasitics’ The theoretically ideal transformer has no losses and an infinitely wide frequency response. However, no component is perfect, although power transformers come close in terms of efficiency. Audio transformers have very real frequency response limitations at each end of the audio spectrum and because of the way induction operates – depending on a rate of change of magnetic flux – transformers cannot pass DC, which is generally a good thing in audio. The limitations of a real-world transformer lead us to the transformer equivalent circuit, which takes account of the non-ideal Fig.10. No-waste E and I shaped laminations. These are normally inserted alternately (interleaved) into the wound bobbin. 58 All components can be represented by an equivalent circuit consisting of the ideal component with associated undesirable losses and ‘parasitics,’ which are represented by additional ‘components’. This is shown in Fig.9 where additional passive components (resistors, capacitors and inductors) have been added to a truly ideal transformer. The most obvious loss is the DC resistance of the windings, which are called ‘copper losses’ and are represented as resistances in series with the windings. Next, we have iron losses (core losses) due to eddy currents and hysteresis. Energy is lost by the emission of stray magnetic fields, which is worse for standard transformers compared to toroidal transformers. These losses are real. The old crude doorbell transformer in my house continuously dissipated a significant 2W of heat. Replacing it with a toroid eliminated this energy loss. Over 22 years I paid £85 for a lot of electricity units (385kWh) to run a doorbell that only rang for around three seconds a day! Fig.11. Toroidal transformer cores are wound from a long strip of steel into a ring. Practical Electronics | September | 2022 currents are induced in it by the coils, causing heating. These can be mostly eliminated by laminating the steel into strips with an insulating coating. The flat strips allow the passage of magnetic flux but break up any eddy currents circulating in the conducting iron by reducing the cross-sectional area. The higher the frequency the thinner the laminations need to be. Westinghouse commercialised the production of transformers by developing a ‘no-waste’ cutting process of making E and I shaped laminations, as shown in Fig.10. These are interleaved into a ready-wound bobbin. Toroidal transformers use a strip of iron wound into a ring – see Fig.11. (It’s only relatively recently that efficient machines have been developed to wind the wires efficiently through the core.) The steel is normally grain-orientated, where the crystals are all lined up the same way by the rolling process, which gives greater permeability in the direction of the flux. B (T), magnetic flux (tesla) (x10,000 for gauss) Silicon iron 1.0 Mumetal 15% Fe 80% Ni 5% Mo 1.0 n Hysteresis region –20 –10 10 20 H(A/m), magnetic field strength –0.5 –1.0 Vout Time Hysteresis ‘kink’ Saturation Fig.12. Magnetisation curve of silicon steel showing hysteresis kink. In power transformers this gives rise to a small heating loss. In signal transformers this gives a ‘soft-sounding’ third-harmonic distortion on low-level signals. All ferromagnetic materials have a maximum level of magnetic flux density they can accommodate, above which little to no additional flux can be carried. At this point the material is said to be ‘saturated’, or ‘in saturation’. It can be considered as a magnetic circuit form of clipping. For silicon steel (which is really magnetically soft iron, with around 4.5% silicon) this occurs at Eddy currents Eddy currents arise because iron, the magnetic core of the transformer, is an electrical conductor and circulating 2.2kΩ 12kΩ 1.0kΩ Fig.13. Butt jointing where all the Is are on one side rather than interleaving the laminations allows the transformer to accept much more DC current through the winding without saturation. This construction is employed for single-ended Class-A amplifiers and smoothing chokes. Toroidal transformers cannot accept any DC since they have no gap at all. +30V + 47µF 10V 4.7kΩ around 19,000 gauss. Mumetal, which is used for magnetic shielding saturates at only 8,500 gauss, and Radiometal at 16,000 gauss. When saturation occurs, the input impedance can suddenly drop, overloading the stage driving it. High-nickel-content materials such a Mumetal (80%) and Radiometal (50%) have higher permeability (ease of magnetisation) than silicon steel, so they 1.2kΩ 0V 820pF DCR 5Ω 2N1711 + 82µF Input 10kΩ 22kΩ CT TR1 BC184 22kΩ 150Ω 2N3055 27Ω 1.0Ω 2.5W 2N1711 + 100kΩ DC trim on driver stage TR2 VA1040 50Ω NTC 10kΩ + 150µF 1.2kΩ I2 82µF I1 + I2 2N3055 100kΩ 220pF 47Ω 1.0kΩ 4.7kΩ 47µF 50kΩ 330Ω 1W VA1040 50Ω NTC 820Ω 1W Negative feedback 1.5nF 150Ω 1500µF 50V 10Ω 8-15Ω 10Ω + + + 10µF 10Ω I1 TR2 BC184 1.0Ω 27Ω 2.5W 100nF 0V –30V Fig.14a. The circuit of the Rogers Ravensbourne amplifier used two emitter followers with their currents flowing in opposite directions to cancel DC magnetisation currents. The sound of this simple amplifier circuit was highly regarded in 1979. See inset photo Fig.14b. Practical Electronics | September | 2022 59 Vout Typical transformer frequency response Drive impedance, Z < 50Ω Load impedance, Z = 1MΩ 0dB –3dB 12dB/oct slope C4 4.7nF 630V C2 + 22µF 250V R2 220kΩ 0V 6dB/oct slope 30Hz 20kHz 80kHz Log f need fewer turns for a given inductance, but the maximum flux density is less. The maximum power and lowest frequency, ie the total magnetic flux of the system, that can be handled is determined by the physical size of the core. This is why big heavy transformers are needed for high powers. For audio output transformers, one old rule of thumb is 0.17lb (77g) of steel per watt. ~80V 8Ω ~5W R9 10kΩ Input Load impedance, typically 1MΩ for tuning with ‘scope R = 560Ω to 12kΩ C = 100pF to 39nF Select R Output C R1 4.7kΩ R3 12kΩ R8 4.7kΩ 3 2N5457 R4 1MΩ PL84 2 ~6V There is a kink in the BH curve of all magnetic materials due to hysteresis. This is where a reverse force is required to get the curve back to zero, as shown Zobel network 7 9 C3 47nF MPSA42 300V rated C1 10nF +170V 17:1 ZP = 2.4kΩ 22mA Hysteresis 1kHz square wave R6 47kΩ 70mA Fig.15. Typical frequency response of an audio transformer. A first-order bass roll-off below 30Hz and a hump at the high frequency second-order rolloff point. Input +250V for EL84 ZP = 5kΩ R12 3.9kΩ Resonance R5 100Ω 4 5 Heater 15V 300mA R10* + 135Ω R7 680kΩ C5 47µF 0V Negative feedback gain control C6 100µF CW + +10dB *91Ω + 43Ω in series 10kΩ A-log R11 1.5kΩ Fig.17. Small guitar practice amplifier using a PL84 valve. The PL84 is the cheap TV low-impedance version of the EL84 with a different 15V 300mA heater rather than the standard 6.3V. in Fig.12. This non-linearity causes the current drawn from the source to be distorted. Silicon steel has a wider hysteresis area than nickel-based alloys, so it produces higher distortion. Distortion can be minimised by driving the transformer with a low source impedance. It can be made lower still by using a negative impedance to cancel out the primary’s series resistance. The lamination material can be identified by its appearance. Silicon steel has a dark grey finish and nickel-based laminations look bright and silvery. DC current No Zobel network Zobel network, C too high One cycle of slight overshoot Optimum Zobel values dB Zobel network, R too high (not enough damping – ringing moves to lower frequency) Resonance (no load) Optimum Zobel network Too much compensation (C too high) f (Hz) Typically 30kHz – 90kHz Fig.16. Adding a Zobel network damps the high frequency resonance. The values have to be determined empirically by feeding a 1kHz square wave into the transformer primary and tuning out the ringing on a ‘scope. 60 Transformers do not like DC current in the windings as it eats up the device’s saturation headroom. Magnetic induction voltage is proportional to rate of change of magnetic flux, so the lower the frequency the less efficient it gets. DC is zero frequency. If DC current has to be accommodated, such as in the output transformer of a single-ended class A amplifier, a ‘resistance’ needs to be inserted in the magnetic circuit to limit the flux which is analogous to electrical current. This is achieved by introducing a gap in the circuit. In the case of EI laminations, this can be achieved by a butt joint between the lumped ‘E’ and ‘I’ laminations, as shown in Fig.13. A plastic spacer can be inserted to increase the size of the gap if necessary. A side effect of this is to reduce the inductance. Clever circuits have been devised to cancel out the magnetisation of DC currents by running equal currents in opposite directions in the windings. This is one of the great benefits of push-pull operation. It was added to the driver transformers of the Bowes transistor amplifier and the Rogers Ravensbourne, as shown in Fig.14. Capacitance and leakage inductance Capacitive parasitics are responsible for causing high-frequency roll-off. At the upper end of the response curve these parameters often form a resonance (see Fig.15) similar to an Practical Electronics | September | 2022 R7 1.5kΩ R1 56kΩ C2 100µF+ R3 10kΩ –14V VR8 2kΩ Lin C4 100µF+ 0V R17 270Ω 10% 0.5W VR13 5kΩ Lin R9 1.0kΩ TR3 OC42 R10 820Ω TR5 OC22 TR2 OC42 R14 3.9kΩ TR1 OC42 + C1 100µF R15 1.2kΩ R18 560Ω R16 1.2kΩ R19 560Ω R11 820Ω Input C3 100µF+ R2 12kΩ 3W TR6 OC22 R4 3.3kΩ R5 22Ω 1.65+1.65:1 TR4 OC42 R6 8.2kΩ R20 1.65Ω 10% 5W R12 1.0kΩ 0V R21 560Ω C5* All resistors 0.25W, ±10% unless otherwise stated Electrolytic capacitors 12V DC working *C5 – add for stability typically 470pF to 3.3nF Fig.18. a) (above) Mullard 5W class-A transistor amplifier circuit from 1961; b) (below) I have the output transformers and transistors for this design and will build one for its historical and teaching value. Apparently, they sounded quite good. under-damped low-pass filter. This effect is usually suppressed with a series RC circuit called a Zobel network, named after the telecoms engineer who invented it – see Fig.16. Output transformers In pre-semiconductor days, valves were the only audio amplifying devices available. They are high-voltage (90 to 800V) low-current (10 to 100mA) devices, which means a high output impedance (Zout) of around 1500 to 10,000Ω. Moving-coil loudspeakers are the opposite, having a low-impedance, typically 3 to 16Ω input impedance (Zin). It is not possible to wind a reliable 5000Ω speaker voice-coil, so the only solution to this dilemma is an output matching transformer which Practical Electronics | September | 2022 transforms the load imposed on its secondary to a suitable one for the driving device – we looked at this last time. A typical transformer ‘matching circuit’ is shown below in Fig.17, a good valve amplifier using the trusted (and still available) EL84 output valve, which needs a 5kΩ load. This calls for a turns ratio of 40:1 to match a 3Ω loudspeaker. The result is an impedance ratio of 1600:1. Note that with a 3Ω loudspeaker this is reflected back to the primary, that is transformed to 1600 × 3 = 4800Ω. OEP do a suitable audio transformer. If an 8Ω speaker were to be used then an impedance ratio of 625:1 would be needed, giving a load of 5kΩ. Working backwards, the square root of the impedance ratio is taken to get the turns ratio, giving 25:1. I once found a box of cheap PL84 valves used for old TVs. These have a relatively lower impedance, so when designing an output circuit around them I dropped the HT voltage and transformer ratio to 17:1 to produce a 2.4kΩ impedance. As an interesting aside, early transistor amplifiers briefly used output transformers (Fig.18) but as their output current capability increased it became possible to drive loudspeakers directly. Output transformers do tend to be big because they have to transmit significant power. A 20W transformer may weigh a couple of pounds (1kg). Cores are normally grain-orientated silicon iron for high power handling. Negative feedback is often applied around output transformers to reduce the distortion and widen the frequency response. However, there is a limit to the amount of feedback that can be applied due to phase shifts associated with the transformers high- and low-frequency roll-offs. For class-A stages, output transformers enable the greatest theoretical efficiency to be achieved of 50%. (Fig.19a) Without a transformer, using a push-pull constant-current load the maximum possible efficiency drops to 25%. With a fixed current source or a simple loudspeaker load it’s 12.5% and 61 Note: PNP transistors use positive earth VC = – 16.8V 5:1 AL21062 Output = 1.3W (3.19W in 40% efficiency) 2. 5mA 6mA Interstage transformer 7.5:1 1W AL21063 . Gapped core ‘ DCR’ : DC resistance OC35 2.7W on heatsink 4.7nF 1 DCR 1.5µF . DCR 100µF 6V . Alternative transformer: e anco as used in car radio amplifiers 190mA VC = 16.5V –630mV 1000µF 3V –2.5V with RC coupling (Fig.19e), it’s a paltry 8.3% with half the audio power going into the RC network resistor. The high efficiency in the transformer circuit occurs because the transformer has a high AC load resistance and a small DC resistance. This means the output device sees the full power-rail voltage and the resulting AC swing is twice that. . 0.5W et + 100µF 15V 3W 5-inch Celestion + OC72 Input Fig.19. For single ended Class-A amplifiers, transformers provide the most efficient way of coupling the loudspeaker. a) Transformer coupled – 1960s car radio amplifier, theoretically 50% efficient. b) No transformer, modulated constantcurrent load. Efficiency 25-30% due to quarter of power rail voltage lost across current source, c) unmodulated constantcurrent load, 12.5% efficiency, d) direct loudspeaker connection, 12.5% efficiency since half the supply voltage lost. e) RC coupled, only 8.3% efficient because audio is also dissipated in the power resistor. All figures just before clipping. At lower volumes it’s even worse. + 0V Negative feedback onitoring unit e lace trans or er ith constant current source odulated Input plus ractical VC VC 25% – 30% efficiency 12.5% efficiency V V = VC Input plus + –1 theoretical e iciency o 0.5VC ias 1000µF ush ull lass si ilar to ohn insley 0.5VC ias c) Constant-current load ood design VC 8.3% efficiency igh current throudh speaker V 12.5% efficiency 0.5VC d 0.67VC Input plus ias irect louds ea er drive ow X L R input 1 o Moving to the ‘other end’ of an audio circuit, a common use for audio transformers is impedance matching on the input. A common example is a preamplifier for moving-coil microphones (which are typically 600Ω source impedance). The classic NE5534 op amp likes to see a source impedance of 4.5kΩ for lowest noise, so a step-up transformer of around 1:7 is required – see the circuits in Fig.20. Input transformers only need to be capable of transmitting a couple of milliwatts before saturation, so they tend to be quite small and employ Mumetal cores. A microphone input transformer is shown in Fig.21. They typically have turns ratios in the region of 4:1 to 12:1. Input transformers are also used to 1000µF VC Input plus Input transformer = VC + ye anguard = VC + a 1000µF ias e esistive loading ca acitor cou ling om Sowter 3195 1:7 o +15V l wo 3 + 8 R1* 1 IC1a 2 –5532 R2* 2 R4 Jensen JT-13K7-A, 1:5 R3 Vigotronix VTX-101-003, 1+1:6.3+6.3 Lundahl LL1530, 1+1:3.5+3.5 C6 22pF C3 22pF R8 C5 CW 220µF R5 R6 + C2 220µF R9 6 + + *Alternative transformers C1* 150pF C4 220µF 1 + 3 IC1b 5 –5532 VR1 Lin R7 7 O utput 4 –15V 0V Fig.20. Transformer-coupled microphone pre-amp. Note how a centre-tap can provide a lossless phantom power connection. Whole circuit gain: +14dB to +70dB. (Based on a Steve Dove design). 62 Practical Electronics | September | 2022 –12V R1 1 DCR C1 33µF + R4 FM424 1. 34:1 1 DCR O utput NKT218 FM424 1. 34:1 OC71 Input C2 100µF + R3 R5 R6 + R2 C3 100µF 0V Positive earth Fig.22. Lan-Elec Ltd audio-frequency No.5, two-stage transformer-coupled pre-amplifier. Fig.21. Typical microphone transformers. Note the round Mumetal can on the Sowter transformer. The Swedish Lundahl, along with American Jensen transformers are possibly the best. create interference-rejecting balanced inputs. These low-level transformers are enclosed in Mumetal cans to shield them from hum. lowest distortion. A common interstage transformer turns ratio range of between 1.5:1 to 5:1 was used for simple transistor pre-amplifiers such as the circuit shown in Fig.22. They are always step down to match the high-impedance collector output with the low-impedance base input. The low DC resistance of the transformer also provides a perfect path for any base leakage currents resulting in stable bias conditions. This circuit (shown built) in Fig.23 was used as an educational board – it makes a fantastic fuzz box! Next month, we’ll build something! – some transformer mounting PCBs. Interstage transformers When amplifying devices were very expensive it was essential to maximise the gain from each stage, so an impedance-matching ‘interstage transformer’ was employed between stages. These disappeared from valve amplifiers around 1940, only to reappear again in the early days of expensive transistors. These transformers were also called driver transformers, where dual secondaries were used to provide the phase splitting for push-pull output stages. A medium level of power is needed, around 20-200mW, so Radiometal is normally employed for Your best bet since MAPLIN Chock-a-Block with Stock Visit: www.cricklewoodelectronics.com O r ph one our friendly kn owledgeable staff on 020 8452 0161 Components • Audio • Video • Connectors • Cables Arduino • Test Equipment etc, etc Visit our Shop, Call or Buy online at: Fig.23. The circuit in Fig.22 was used as a teaching aid in an old Radio and TV servicing college in South London in the 1970s. Now redeployed as a germanium fuzz box. Practical Electronics | September | 2022 www.cricklewoodelectronics.com 020 8452 0161 Visit our shop at: 40-42 Cricklewood Broadway London NW2 3ET 63