The Le Pacific phono preamplifier was introduced by Jean Hiraga. The circuit is based on two jFets and a passive RIAA equalization in between (fig. 1). Its simplicity, cost effectiveness and safety made the Le Pacific preamplifier a popular DIY project. A number of schematic variants circulate on the internet, either with or without source resistors, with the RIAA network in front or behind the coupling capacitor, drain resistors vary between 2.2kOhm and 3.3kOhm (fig. 2 to 4). 24V to 36V power supplies and low idle currents allow battery operation.Fig. 1 to 4: Variants of the Le Pacific preamplifier (source Internet) Problem #1: Low GainFigure 5 shows a SPICE simulation of the concept. For clarity the RIAA network is omitted from the circuit and only the basic amplifying structure analyzed. The simulation in figure 5 reveals a gain of 62db. This is not impressive for @ 1kHz the passive RIAA network will subtract 20db from the overall gain. This would be 500mV output voltage with 4mV at the input.Fig. 5: Simulation if the original Le Pacific preamplifier Fig. 6: Simulation with two 2SK369 jFets instead of the 2SK170. (RIAA omitted). Gain is 62db. Gain is 72db.To increase gain the degeneration source resistors may be omitted and the jFets selected for idle currents of 5mA @ Vgs=0 as in the design by Walters (fig. 2). Drawback is a positive Vgs with the input signals. The gate draws current and the input impedance may be compromised. A better solution might be jFets with higher transconductance such as the 2Sk369. The 2SK369 require 20Ohm source resistors to settle the idle currents at 5mA (fig. 6). Also the 20Ohm source resistors ensure stability and some local feedback. The SPICE simulation in figure 6 reveals a gain of 72db. With 20db subtracted from the passive RIAA 4mV input voltage would give 1.6V at the output.Problem #2: Limited BandwidthA fundamental problem is the limited bandwidth of the Le Pacific concept due to the Miller effect. The Miller effect is nicely explained on this website. It causes a high input capacitance Cin of the jFets. Cin is derived from the following equation:Cin = Cgs + Cgd x (A+1) (Miller capacitance)Cin = input capacitanceCgs = gate-to-source capacitanceCgd = gate-to-drain capacitanceA = stage gain (first stage in figure 5 or 6)For the simulation in figure 6 (2SK369) the calculated input capacitance Cin isCin = 75pF + 15pF x (70 + 1) = 1140pFAn input capacitance of 1.14nF is far above the recommended capacitive load for MM cartridges. Even with the 2SK170 as input transistor Cin still will be high. According to the data sheet and the simulation in figure 5 Cin of the first 2SK170 isCin = 30pF + 6pF x (36 + 1) = 252pFIncreasing Bandwidth by Cascoding the jFetsCascoding the jFets with two BJTs will eliminate this problem. The BJTs keep the voltage at the jFet drains constant and the variable A in the above equations becomes 1. Figures 7 and 8 show SPICE simulations of the original Le Pacific design and a cascoded design, both with 2SK369 jFets. Power supply voltage is set to 36V, because an additional 9V to 10V are needed for the cascoding transistors. The impedance of the input voltage source is set to 5kOhm to simulate high impedance MM cartridges. Figures 7b and 8b show the frequency responses.Fig. 7a. Le Pacific layout with 2SK369 jFet Fig. 7b: Frequency response of the circuit from figure 7aFig. 8a. cascoded Le Pacific layout with 2SK369 jFet Fig. 8b: Frequency response of the circuit from figure 8a With reference to 1kHz the conventional design shows a -3db roll-off @20kHz (fig. 7b) while the cascoded design looses 0.006db @20kHz. Its -2db roll-off is beyond 350kHz in the simulation (fig. 8b). The modification to a cascoded design will need few more components. It is safe and simple. Costs are minuscule. Modified RIAA NetworkA cascode is an almost perfect current source. This allows the use of a so called transconductance RIAA. The network comprising R1, R3, R10, C3 and C5 in figure 9 resembles a variable load which is fed from the current source Q2. With rising frequency the network impedance decreases and the voltage drop at R3 corresponds with the RIAA curve. Main advantage is the omission of the series resistor in the conventional passive RIAA network. Figure 9 shows the cascoded circuit with the recalculated RIAA network. The input signal is fed through a precise anti-RIAA filter. Thus the frequency response at the output must be flat. As shown in figure 10 and 11 the the diversion from the ideal RIAA is within a range of 30mdb. Further information on the anti-RIAA filter can be found here.Fig. 9: Circuit diagram with transconductance RIAA network and anti-RIAA filter at the input. Fig. 10: Frequency response with anti-RIAA filter at the input. The scale is +/- 1db on the ordinate axis. Fig. 11: Frequency response with anti-RIAA filter at the input. The scale is +/- 30mdb on the ordinate axis.. Enhancing the Gain to MC Level. . . . . to be continued soon.