Specification Measured Charactersitics Sensitivity 2.5mV @ 1kHz R.I.A.A. +-1dB 20Hz-20kHz Output 250mV Overload margin 20dB @ 1kHz Z in 47k Z out < 500R R.L. Min. 22k T.H.D @ 250mV, 1kHz < 0.01% Unweighted O/P noise < 0.4mV Supply Ripple Rejection > 56dB @ 100Hz without C8 20dB Supply current @ 24V 1mA Design Philosophy Most amplifier designs use what can best be described as a sledgehammer approach to overcoming problems. Many active devices are used and then 100% negative feedback is employed to smother any defects. However, especially in the field of equalisation from transducers such as magnetic cartridges this has two shortcomings. The first is that with so much gain in a closed loop, stability can become rather marginal. Feed that from a very lively inductive source and there can be problems. Marginal stability often doesn't not show up with sine wave testing, and freqently manifests itself as short bursts of very high frequency oscillation on signal peaks, only visible with a fast oscilloscope. The second more insidious problem is that of transient distortion. This is best explained by taking an extreme case. All devices have a finite time delay, now if we have an amplifier with say, 100 devices each with a propagation delay of 100 nanoseconds then we could assume that the total delay is now 10 uS. If we apply a signal to the input of this amplifier, then the output will be delayed by this much, and if significant feedback has been applied the feedback will be 'late' arriving at the input. If we now hit the input with a transient of say, 5uS (quite possible for a mag. cartridge) the pulse will have been and gone before the feedback loop can respond. What is worse, the feedback loop will then provide another pulse corresponding to the reverse of the input! As there is no effective feedback in this situation the amplifier will swing into heavy distortion, and although the results will not be directly audible, being at too high a frequency, there will be all sorts of intermodulation effects with other signals. As I said this is an unrealistically extreme example. However in a real situation the effect acoustically is for the sound to be unaccountably 'muddy'. The equaliser here works in a quite different way. There are only two gain stages and each one has only minimal local feedback in the form of a series emitter resistor. There is no overall feedback, and the equalisation itself is done by a passive network between the gain stages. Stability is excellent and any transient products that may be produced will be very small and localised. Circuit Description The transistors run at unusually low collector currents. This has the effect of increasing their input impedance, and making them look more like resistors than diodes. It also makes for a very significant improvement in noise performance. Another unusual feature is that the input transistor is a PNP type. This is because these have a significantly better noise performance than their NPN counterparts. For example, under exactly the same conditions the BC560 has a noise factor of 1dB while the BC550 is 1.2dB There is of course always a downside. Running transistors at such low currents and with so little feedback tends to increase harmonic distortion. However, the figures speak for themselves. Most people can't detect less than 1% simple harmonic distortion, especially where, as in this case the distortion products are mostly second harmonic, with some third harmonic. The rest of the distortion is effectively buried in the noise floor. The biasing arrangement for the transistors is chosen for simplicity. Good modern silicon transistors have extremely low leakage. Also, temperature tends to affect the the base/emitter pedestal voltage a lot but hardly affects the current gain at all. At DC the base is being current driven with 100% negative feedback. Using such high impedance biasing resistors means that decoupling can be done with quite small dry capacitors. These will generate far less noise than even tantalum ones would in a conventional arrangement. Finally, The use of a PNP stage followed by an NPN one improves the supply noise/ripple rejection. This is because the collector of Q1 approaches the behaviour of a true current source so any variation in supply voltage will not appear across R4 - if the current is constant then so must the voltage be. After going through the equalisation network, all of which is also referenced to the supply rail, the signal is applied to the base of Q2, who's emitter is again referenced to the supply rail. Now the collector of this transistor has the same constant current characteristic, so again, negligible supply variation will be developed across R11. We now have the output referenced to the same line (0V) as the input, and supply noise and ripple is no longer an issue. Component Choices Metal film resistors are recommended and specifically R1,3,5 should be low noise types. The dry capacitors can be mostly polyester types. The closer the tolerance of all the passive components, the better the channel matching and the closer the equaliser will be to the correct RIAA curve. C10 should really be a tantalum type, but you may get away with a good quality conventional electrolytic. The optimum transistor hfe is 600, but anything in the 'C' classification will be OK. Only a single channel (1/2 of a stereo equaliser) has been shown. For optimum crosstalk performance the other channel should have its own supply filter (C8/C9/R13), otherwise halve the value of R13 and double the value of C8. C5/C6/R9 form a rumble filter that rolls off at around 18Hz The only function of R15 is to prevent any leakage currents from C10 disturbing later amplifier stages, or giving high amplitude pulses if the unit is plugged in after being switched on. The circuit was originally characterised to be battery driven (3xPP3) and drive a relatively high impedance amplifier. If higher current demand can be accepted R14 may be as low as 2k2 with corresponding improved driving capability. For the best possible noise and interference performance I recommend building a battery operated version with a simple switch in the supply, and housing the unit in a diecast metal box. W Godfrey 1980-2006