After a lot of years of development, a workable antilog generator circuit for the LM3900 operating on a single +15V power supply has finally emerged. This circuit differs from a classic Serge design which uses the internal LM3900 current mirrors directly for antilog waveshaping. This circuit has a topology similar to conventional bipolar opamps and matched dual NPN transistors, but instead use a matched dual PNP transistor in a configuration intended for the Norton opamp and single-supply operation. This involves as well a reversal of current directions, compared with a NPN circuit. The transistor pair is a 2SA798 device, which has worked remarkably well for low-noise and antilog ideality. A developmental schematic is shown below, along with a scope foto of the input/output characteristics.
Antilog Generator Circuit
Antilog Input/Output Characteristics
The design derived from another design done yesterday that had much more thermal stability (because it used a 1KΩ +3300 ppm/°C TEMPCO resistor), but that design lacked adjustment range, had various I/O interoperability problems, and also operating the transistors in a non-ideal way. The present design was developed to address these flaws, but leaves some other circuit shortcomings intact for the moment. So, the design does not yet have a thermal compensation scheme, and the precise 1V/octave relationship is not yet adjustable (it was only approximated, in order to get to a generally working circuit first).
Operating with a single supply is more difficult in some ways for transcendental computation circuits, because a ground reference level is unavailable to represent positive and negative information (as voltages). The LM3900 is unique due to it's current mode of operation, which can invite a number of very different alternative application circuits. Because the inputs of the LM3900 operate in a current-mode, resistors are normally used to convert voltages to currents.
But currents can be injected directly into the inputs as well, which was done in this design. The inputs though are clamped to +Vbe above ground. A short-coming of the previous iteration of this design was the Vcb for the transistor was not 0v, which can cause deviations from ideality for the antilog transfer characteristic. Because the Norton opamp inputs are at +Vbe, for Vcb to be ≅ 0, when collectors are fed to the inputs, the base must also be a +Vbe. And for the transistor to be turned on, the common emitters between the matched pair must work at ≅ +2Vbe. However, at room temperature a 1000:1 dynamic range of collector current from the antilog transistors only requires an input ∆Vbe of 180 mV. So, the problem that was to be solved, was biasing the transistor bases at +Vbe, their emitters are +2Vbe, and arranging 180 mV of signal swing. How this was done will be explained in a more detailed post later.
Comments
Post a Comment