As part of research into multipliers and transcendental waveform converters based on LM3900, over time it became clear that the DC circuit precision could be improved by understanding much better how the +input behaves. This input drives the current mirror for current differencing action into the -input, where otherwise negative feedback is used. The measurement results included this VBE versus input current graph, which clearly obeyed a 60 mV/decade logarithmic slope:
By having this graph, it's now possible to choose +input currents far more accurately, so that the amplifier output for DC bias is more precisely set to a value, mindful of voltage swings that will occur because of relative current changes based on the -input. An alternative situation is using the +input to set a reference voltage for which switching action will occur based on a matching current into the -input. This does depend on the current mirror gain accuracy, which is imperfect, but very close to 1:1. More accuracy is obtainable from the current mirror by using reference currents for which the VBE at both inputs match, which per AN-72 is ~10 μA. However for all other input currents that might be set for an application purpose, much more precise DC output results are obtainable. This is possible by reducing the initial error of the voltage difference from a voltage reference across a current setting resistor versus the true value of VBE seen at the other end of the resistor, at current mirror for the -input.
The measurements were taken nominally at room temperature, and for most circuits, temperature variation has little effect. Reducing the effect of temperature induced changes may be important in some circuits though, e.g. where DC values may be important. The LM3900 shows a lot of potential even here, because with monolithic construction, the ΔVBE for the 4 amplifiers will track extremely well, affording a number of circuit designs where precise ratios are what is most important, not absolute values. Such is often the case with music and sound.
By having this graph, it's now possible to choose +input currents far more accurately, so that the amplifier output for DC bias is more precisely set to a value, mindful of voltage swings that will occur because of relative current changes based on the -input. An alternative situation is using the +input to set a reference voltage for which switching action will occur based on a matching current into the -input. This does depend on the current mirror gain accuracy, which is imperfect, but very close to 1:1. More accuracy is obtainable from the current mirror by using reference currents for which the VBE at both inputs match, which per AN-72 is ~10 μA. However for all other input currents that might be set for an application purpose, much more precise DC output results are obtainable. This is possible by reducing the initial error of the voltage difference from a voltage reference across a current setting resistor versus the true value of VBE seen at the other end of the resistor, at current mirror for the -input.
The measurements were taken nominally at room temperature, and for most circuits, temperature variation has little effect. Reducing the effect of temperature induced changes may be important in some circuits though, e.g. where DC values may be important. The LM3900 shows a lot of potential even here, because with monolithic construction, the ΔVBE for the 4 amplifiers will track extremely well, affording a number of circuit designs where precise ratios are what is most important, not absolute values. Such is often the case with music and sound.
Comments
Post a Comment