1.1 Amplifying stages
The amplifying stages should be as simple as possible. No use of op-ams or global feedback. There is no "a priori" decision how much amplifying stages will be needed, but the overall signal path should be as simple as possible. The amplifying stages should provide a balanced signal path from the input selector to the output.
1.2 Volume control
The volume control should have a range of about 60-70db and a step size of 1db. Because of the balanced concept the volume control must have 4 channels with very less deviation between the single channels, to keep the advantage of the balanced concept. It must support an IR-remote control and the control of balance between left and right channel.
1.3 Input selector
The input selector should not use a single mechanical switch to guaranty the channel separation. On the other hand it is not acceptable to simply use two different switches, one for each channel. Then you have to handle to switches every time or you must use tons of mechanical components to couple both switches together.
2 Final Design
2.1 Volume control
I have done a lot of experiments with
different types of volume controls. Mechanical switches, potentiometers,
electronic potentiometers and ADC's.
At the end it became clear to me how
the appropriate volume control for such a amplifier concept should look
like. It is shown in picture
1. There is a fixed high quality resistor in the signal path (R1).
By using electronic switches high quality resistors (R21,...,R24) will
tie the one end of R1 to ground, creating a voltage divider with varying
low end. The signal input will be connected to the other end of the R1.
The amplifying stage will be connected to the node between the R1 and the
so called "varying resistor".
The question is how to implement the electronic switches. For example if R1 is a 10k resistor and the range of the volume control should be, 60db the series resistance of R2* and the electronic switch should be 10Ohms. That means we cannot use CMOS switches there because they will have a too high "on resistance". Relays will work but they are slow and the quality will vary over time also if you we use very high quality ones.
From a book I got the idea for the final implementation. It is the German book "Halbleiter-Schaltungstechnik" written U. Tietze and Ch. Schenk. There is also a English version available from Springer. This book is the one of the bibles for circuit design in Germany. There is a section about electronic switches, and this section describes the usage of bipolar transistors for such purpose. The final implementation of the volume control uses bipolar npn transistors as switches (picture 2). For the positive half of the signal they work in a normal conducting state. For the negative half of the signal they operate in a so called inverted mode. This works very well for small voltage drop across the transistor. Further they provide a very low on resistance and a good linearity for the expected milivolts voltage drop. The only problem is that the base current creates a positive voltage appearing at the preamplifier input. If connected devices have problems with this DC bias you should use a coupling capacitor in series with the preamplifier input.
A 4 bit signal is used to control 10 switches. The common input of the 4067 analogue multiplexer is tied to +5V. The power supply pins of the multiplexer are connected to +5V and -5V. If all 4 inputs are at low level (-5V) the 0 output of the multiplexer is at high potential (+5V). The base terminals of all bipolar switches are connected to -5V by high ohmic resistors R4*. Therefor the bipolar transistors are switched off and the signal is connected through R1 to the amplifier input. R4 can be in the range of 10k because the transistors do not need base current if they are switched off. If input A becomes +5V the 1 output of the multiplexer becomes high also. This provides the necessary base voltage and current through R31 to switch T1 to a conducting state. Now the signal sees the voltage divider R1 and R21. The level of the signal seen by the amplifier input can be calculated by (R21 + Rt1)/(R1+R21+Rt1). With this formula you can calculate the necessary resistor values for R2*. The voltage drop at T* is very small which guarantees good linearity. R3* should be in the range of 1k to provide enough base current to the transistors. But be aware of the maximum current provided by the 4067 multiplexer. If you like to use power bipolar transistors as switches you should use additional driving transistors in between to provide the necessary base current.
The final design of the amplifier uses 6 switches at the input stage controlled by a 3 bit signal. The series resistor is a 10K one. The R2* are calculated to get -10db, -20db,...,-60db signal values at the input of the first amplifier stage. Between the first and the second amplifier stage a second volume control stage is implemented. This stage uses 9 switches controlled by a 4 bit signal for -1db, -2db,...,-9db. This creates a volume control with a range of -70db and a step rate of -1db. The 4067 multiplexer guarantees a "make before break" of the single switches. This is important to get a silent working of the volume control without "clicks" between the single steps. It is up to you to implement a control circuitry to generate the coupled 3 and 4 bit signals. I use a couple of CMOS counter and decoder to do this job. Therefor you can use 4 switches "up", "down" , "left", "right" to control the signal level and the balance. You can use also a single micro controller to do this job. But I do not like to have some Megahertz clock signals for a micro controller in my preamplifier. We get some benefits with this type of volume control:
I will not show here all final values and element types I use for my implementation. It depends heavily on the power supply and attenuation rates you would like to have which resistor values you should use. Also the used transistor type will have an impact on the final resistor values. The best way to find a working implementation is to do some simulations (like spice) to check the basic functionality and the attenuation levels. Finally you should listen to the realization to find the best transistor types and resistor values for your ears. In my mind this type of volume control works very good and has an important impact on the great performance of the overall preamplifier.
2.2 Amplifying stages
The amplifier stages are variants of the "balanced amplifier stage" from Nelson Pass. I like this type of simple differential pairs to get a line amplifier stage. It can be driven by a balanced or unbalanced input and produces both a balanced and unbalanced output.
The first stage connected to the input of the preamplifier and the first volume control stage is shown in picture 3. It uses a 2SK389 dual fet transistor to guarantee a good matching between the balanced signals and a low input capacitance of the preamplifier. Instead of resistors as current source at the sources of the fets as in the zen stage I use two active current sources. These current sources should be matched very well. I use LED type diodes for the voltage source in the current source because of the very low noise figures. The drains of the 2SK389 are driven by resistors to get the output signal of the first stage. The sources of the fets are connected by a resistor controlling the gain of the first stage. For the details of such an amplifier stage please read the project description of the "balanced line stage" at http://www.passlabs.com/. The gain of the first stage is close to one, because I will use the power mosfets in the second stage for the gain of the amplifier. The current sources provide 3mA for each branch. You can use higher values but you should be careful with the total power dissipation of the 2SK389. The outputs are coupled through high quality capacitors to get rid of the DC voltage.
The second stage (picture 4) is again a simple differential pair build up from power mosfets (IRF 610). It uses the same type of current source as the input stage but with much higher current. The gain the this stage is about 3. Again the outputs are coupled through capacitors to the preamplifier output. I have done some listening experiments to find the best current values for the amplifier stages. From this experiments I also decided to use only a single current source for the second preamplifier stage instead of two ones as in the input stage. The current source provides 40mA overall current, 20mA for every branch. The power mosfets can handle much higher values but you need larger heat sinks and power supply therefor.
2.3 Power Supply
The principles of the power supply is shown in picture 5. There is no need for strong voltage regulation from the power supply. All amplifier stages and the volume control consume a nearly constant current. Therefor the power supply uses simple voltage follower. Zener diodes (D1) are used to get a constant voltage and power mosfets are used as followers to supply the current. Because of the high ohmic behavior of the mosfet gate I use a RC circuit (R3, C3) to filter the noise of the zener diodes. A pi circuit (C1, R1, C2) is used to filter the original transformed input voltage. Another RC circuit (R4, C4) filters the regulated voltage directly at the amplifier stage. A second similar stage is used to get the +5V/-5V voltages for the volume control. Every channel of the amplifier has its on power supply circuitry to guarantee a high channel separation.
2.4 Chassis
The preamplifier is build up in 3 different parts using 3 separate chassis. One chassis contains the power supply and the control circuitry. The control circuitry controls the input/output selection relays, generates the 3bit and 4bit signals for the volume control of each channel and implements the IR remote control. The two other chassis contain the whole circuitry of the two preamplifier channels. Each channel has its own chassis. I use shielded cables to provide the supply voltage to the amplifier chassis. The control signals are carried by simple computer cables.
2.5 Finish
To find the best values/types for all circuit elements I used a two step approach. First I have done tons of spice simulations. You can use a public domain spice simulator like spice3f4 which can be found on the web. For most of the bipolar and mosfet transistors you will get spice models for free from the semiconductor vendors. I 'm working for a semiconductor company and therefor I used several commercial simulators. Typically they provide better graphical probing tools not better values from the simulation. You can get the transient behavior, the bandwidth and the noise and distortion figure of the circuit from such simulations. The selected values/transistor types can be used as starting point for listening sessions.
As a second step I used an experimental implementation of the preamplifier to find the final values/transistor types by listening tests. I did not focussed on the type of the passive elements like resistor or capacitors (with exception of the coupling capacitors in between the two stages and at the output of the preamplifier). I focussed mainly on different power mosfets and on the current values of the current sources for the differential pairs.
3 How does it sounds?
This is very easy to say. It sounds marvelous. It is absolutely quiet in absence of an input signal. The volume control works fantastic, no "clicks", very small step size and a very good channel symmetry and separation. One major point of the overall performance of the preamplifier is the overall channel separation. It improved the imaging of the complete system dramatically. I own a two way speaker system using a broad band bending wave transducer for the frequency range from 100Hz to 30000Hz. One major advantage of this transducer is the reproduction of human voices. Also this part of the reproduction is further improved by the preamplifier.
One last point is the related to my wife. She is very critical with my audio enthusiasm because it costs a lot of time and she does not like the large speaker cabinets in our living room. But she was very impressed by the improvement of the preamplifier to the overall music reproduction.
That's it. I cannot tell you the "bla bla" of hi-fi magazines. I simply say: try it and listen to it.
4 Gallery
mail me your comments ( maik.herzog@infineon.com)