Memphis MC 1   Discrete Solid State Moving Coil Phono Stage  
The basic Concept „Axiom #8 - keep it simple. …. why use two stages if one can do it as well? Switches, pots, connectors, tube, cable and all other toys we love can only loose us information…“. Allen Wright, The Tube Preamp Cookbook, second edition, vacuum state electronics, 1997. The new BJT based phono stage shall comprise one single gain stage only that amplifies the MC signal above level, followed by a passive RIAA equalization and a subsequent emitter follower at the output (fig. 1). A step-up transformer elevates the MC signal close to MM level. The first stage will be designed as a high gain voltage controlled current source. It injects the amplified audio signal current into a frequency dependent load. The voltage across the load drops with ascending frequency according to the RIAA equalization curve. The Schematic
Fig. 1: Simplified block diagram of the Memphis MC 1 phono stage
At first glance the schematic looks like a PNP overall concept with the negative voltage of -24V on top the positive +24V at the bottom. That results from the Sziklai configuration of both the first and the second stage. However, at a closer look all crucial transistors are of NPN polarity. In the schematic the input sage is set to suit a Benz Micro Gold cartridge. However, with a very few tweaks it can be modified easily to suit any other MC cartrige. First Stage A Lundahl LL1933 step-up transformer multiplies the MC output voltage by the factor 8. R3 at the secondary coils plus the transistor input resistances reflect an impedance around 1000Ohm for the MC cartridge. A differential input is necessary to cancel out the base currents and keep the transformer free from any DC. The two input PNP/NPN compound transistors are connected in a Sziklai configuration. The compound transistor resistors R17, R18, R19 and R20 are set to achieve 0.7mA idle current of the PNP transistors and 5.3mA idle current of the NPN transistors. At these operation points the current gain of both transistors are quite constant (fig 3a and b). Both compound transistors are cascoded to ensure identical voltages of 12.0V and thus similar thermal behavior. Q13 is a conventional cascode, Q7 is a folded cascode. The transistors Q6 and Q8 are a current mirror with a 1:2.3 ratio.  Thus the cascoding transistor Q7 draws a current of 7.8mA. Its current is fed into the frequency depending load R1, R2, R14, C1, C2. The requested high gain of the input stage requires a high load impedance at the collector of Q7. Normally the 33k resistor (R14) would limit the collector current to a value around 0.5mA with respect to its voltage drop. This would severely compromise the overload capability of the input stage. Therefore a 810H choke (Lundahl LL1667, 5mA) is in parallel with the collector resistor of 33k. Its copper resistance is 2400Ohm. Thus the DC load of the cascoding transistor Q7 is 2400Ohm while its AC load is almost 33K at the sub-bass of the audio spectrum then and decreasing with ascending frequency according to the RIAA equalization curve. At the low end of the frequency spectrum the inductivity of 810H in parallel with 33K reveal voltage drops of -1db at 13Hz, -3 db at 7Hz and -16db at 1Hz. Raw gain of the input stage (RIAA network R1, R2, C1, C2 disabled) is 79dB plus the eightfold voltage step-up of the LL 1933 MC transformer. The input stage can easily be adapted to the gain needed by adjusting the degeneration resistors R22 and R23 according to table 1. Furthermore the step-up transformer can be changed to a 1:16 winding ratio if real low output MC cartridges are favored.     Table 1: Raw gain (db @ measuring point 1) and input impedance     (Ohm) of the amplifying first stage depending on the degeneration     resistors R22 and R23 without RIAA network (fig. 4). All data are     taken from LT Spice simulation   degeneration              gain              input impedance    resistors (R22, R23) _____________________________________________________________     5 Ohm  83db                  83k   10 Ohm  79db              140k   20 Ohm  73db                244k   50 Ohm  66db              565k      100 Ohm  60db          1100k        Fig. 4: degeneration resistors of the input stage. Second Stage The second stage is just an impedance converter with unity gain to pass the audio signal as unaltered as possible from the RIAA network to the output and to enable the phono stage to drive even critical loads. The PNP/NPN compound transistors have an impedance of several Megohm at the base of the transistor Q1 and do not alter the RIAA equalization. The compound transistors are cascoded by a p-channel MosFet configured as a bootstrapped source follower. Its source follows the output signal. Thus the voltage across the compound transistors is kept constant to reduce distortion (fig. 5a and b). With 0.5mV @ 1000Hz at the input output voltage is 3.4Vp-p. Maximum output voltage before clipping is 34Vp-p. RIAA Equalization Network Several web sites with active or passive RIAA network calculators can be found on the internet. Mh-audio, KAB, hagtech and others provide calculators for passive RIAA networks. Obviously they all are based on the equations published by Stanley Lipshitz in the 6/1979 Journal of the Audio Engineering Society and at least these calculators are a good point to begin with.  Contrary to a voltage driven network (figure 6) the RIAA network of the Memphis MC 1 phono stage is current driven, i.e. its signal comes from a current source with almost infinite impedance. Thus the circuit has to be modified as shown in figure 7. More information on current driven versus a conventional voltage dividing passive RIAA can be found here. Fig. 6 and 7: The impedance of the voltage source V1 is close to zero while the impedance of the current source I1 is close to infinity. Thus C1, R1 and C2 "see" the same impedance of R3. Click on figures for RIAA equalization curves. However, the Memphis MC1 network is somehow more complex because of the inductor involved. Furthermore stray capacitances and all other high frequency losses in the layout have to be considered to ensure an overall frequency response that suits the RIAA equalization. A convenient way to evaluate the RIAA equalization is to feed the phono signal through an inverse RIAA filter into the circuit. The resulting frequency response should be flat within the tolerances allowed by the designer. Inverse RIAA Network Searching the internet I found an intersting inverse RIAA filter (figure 8) on the hifisonix web site. To verify its accuracy I simulated the filter in figure 8 using LT spice. Table 2 shows the simulated inverse RIAA response in comparison to the ideal RIAA response. Differences were between 0.01db to 0.03db across the audio band which might be considered sufficient. Simulation of RIAA Equalization I started RIAA simulation using the calculator from Mh-audio with R1 = 34.1k (R14 in the schematic). The corresponding values calculated (R1 = 4.96k, C1 = 64.13nF, C2 = 22nF )and the simulated RIAA response using the above inverse RIAA filter are illustrated in figure 9a and 9b. Figure 10a and b show the RIAA response after simulated adjustments. to be continued soon . . . .
Fig 3a and b: Operation points of the PNP (2N3810) and the NPN (BC550C) compound transistors.
Fig. 2: Preliminary full schematic of the new phono stage Memphis MC 1 (click on image to enlarge).  
Fig. 5a and b: Simplified schematic of the impedance converter (a). Voltage swing (b) at measuring points 2 (red) and 3 (light blue) @ 2mV (1000Hz) at the input of the phono stage.
Tab. 2: Ideal RIAA response and simulated inverse RIAA response from LT spice simulation (fig. 8).
Fig. 8: Inverse RIAA filter to evaluate RIAA equalization.
Fig. 9a and b: RIAA network values from online calculator (a) and RIAA response within the overall layout (9b). RIAA accuracy is +/- 330mdb between 20Hz and 20kHz. Click to enlarge.
Fig. 10a and b: RIAA network values after adjustments (a) and RIAA response within the overall layout (10b). RIAA accuracy is +/- 130mdb between 20Hz and 20kHz.  Click to enlarge.