I am now returning to all-transistor development, after a relatively fruitful Summer of EMS circuit development with various integrated circuits. The work this Summer included some innovations with data conversion and digital approaches. Then, some LM3900 DC circuits were tested, some computing circuits with transistor arrays were built, and various tests were done with the LMC6484A CMOS op amp. Integrated circuits are wonderful, they do amazing things. But I want something really different for the Model III EMS.
A goodly part of this thinking has been the self-discovery that I've spent vastly too much time developing for extreme precision, while I have simultaneously not taken enough advantage of the benefits of approximation. The focus has been too much on extremely precise electronic circuits, with little thought as to how they sound! This needs to change, because the Model III EMS is basically an art project for hearing the "Sound of the Transistor," to experiment with new sound generation and signal processing techniques. It's not at all a high-fidelity project, where adherence or not to extremely demanding specifications has an impact on the results.
What is motivating this shift toward BJT implementation are the following things:
First, I have already developed over time some very useful building block circuit designs, and in particular some specific ones with very good temperature stability. Among the circuit designs are a good op amp, a good comparator, inverted-mode switch, paraphase circuits, very temperature stable current sources and sinks, voltage-controlled current sources, voltage buffers and line-drivers, sweep generators, a recreation of a 1968 antilog amplifier, ... and more. Basically, I've already developed a catalog of practical and tested building blocks for construction of analog systems. Nearly all of these circuit designs offered high-performance, compared with IC op amps. Particularly in realms like switching speed, current drive, and bandwidth. A great deal of the useful discrete development work was done in the early- and mid-2000s, but then 2017 was a particularly prolific year with discrete circuit designs. Doing things in discrete form is challenging in certain aspects, but no longer seems very difficult.
Second, in more recent testing of LM3900 DC circuits for temperature stability, I became a lot more aware of how to do temperature compensation with BJTs, and also much more aware of how much precision is possible with 5% resistors, esp. creating more accurate custom resistor values by joining pairs of low- and high-valued 5% resistors. This is not to say 1% and 0.1% resistors don't have a role. It's more a dawning awareness that some things in analog circuit design are satisfied with "good enough." As well, the ubiquity and utility of the Zener diode is also more apparent in relation to BJT circuit design.
Third, translinear techniques showcase innovative design ideas for analog integrated circuits in computing transcendental functions. The utility of the BJT for these functions using IC processes is rather clear. The unexplored terrain though is implementing transcendental functions with discrete transistor circuit designs. And all of it possible with truly inexpensive components. Even with just two devices, the 2N5088 and 2N5087, a huge variety of circuit designs are possible.
Fourth, the economy and extreme versatility of function with BJTs is very attractive. As just one example, differential outputs are relatively easy to obtain, and this impacts designs and applications in a virtuous manner. The economy and versatility of function are quite evident after studying Moog's 1960s circuit designs in depth. A modest number of what were then expensive silicon devices yielded incredible results. Now that BJTs are inexpensive, and many more of them can be used, more complex system designs are possible now than were afforable 55 years ago. Another aspect of this approach is simpler circuit layouts, whether implemented as prototype breadboards, or as PCBs. The components are all more similar in size with respect to each other, which contrasts with mixing discrete and integrated circuit technologies. This does mean potentially larger circuit board area, but at the same time PCBs are inexpensive. More power is likely required when contrasted with integrated circuit implementation, but the synthesizer circuits are not particularly high-power.
Fifth, and this is key: BJTs allow for very wide experimentation with novel circuit designs to make musical sound. Designing systems using them is unconstrained from numerous limitations: there are more degrees of design freedom compared with integrated circuit implementations. This is also a holistic approach, preserving synergies that are not usually possible with most integrated circuit implementation technologies. For example, high-performance PNP devices are available to offer complementary symmetry with NPN devices. Developing the Model III EMS this way is engaging in electronic music experimentation more like it was in the 1960s, an experimental decade. This is a very different way to New Sound. An experience not an event.
Comments
Post a Comment