AN-72 had dropped hints about non-linear action being available from the diode-connected transistor at the positive-input of the LM3900. This is the input to the integrated current mirror which subtracts current from the negative-input of the opamp -- which after phase-inversion is considered actually forms a non-inverting function, relative to the output.
The LM3900 inputs operate at +Vbe above ground, and are at low-impedance in accepting currents. The input current can vary over wide-range, though 10uA is the nominal current at which delta-Vbe between the transistors at both inputs are said to match quite well. Directly sourcing a signal into the non-inverting input for the express purpose of modifying the Vbe immediately there should have a non-lnear impact on how the voltage input is "interpreted" as a current. This effect will only occur over a small range, but it should be non-linear. And it's a very interesting case of taking an output directly from the input of the opamp, at other side of a current-conversion resistor.
This design concept actually works. The circuit below transforms a triwave attenuated at the non-inverting input into a fairly clean sinusoidal wave:
The circuit is very sensitive to the input DC baseline, but the non-inverting input can be appropriately modulated so as to provide a non-linear output such as for sine-shaping. As shown in the scope photo, the range of effect for the best shape actually extends slightly below ground. The input point was only monitored with a high-impedance pickup, the 10M scope probe, so as to avoid drawing out any additional current. And the feedback resistor from output to inverting-input just completes a circuit where the linear input on the input side of the voltage-division input resistors is merely reflected at the opamp output -- linearly! The circuit is a otherwise a standard-issue LM3900 voltage follower.
An intriguing range of application functions could open up if this non-linear effect can be exploited more fully. First of all, there are four opamps on the same die, and the transistor matching over temperature should be relatively good. A differential mode of operation comes to mind, for balanced effects. Possibly a clever circuit could exploit this non-linearity for log, antilog, and other transcendental computations. In particular, the low-impedance low-capacitance nature of the LM3900 inputs invites wideband operation. While critically dependent on the DC level input, the sinusoidal shaping effect seen was lossless easily up to 100 KHz, and probably extends to MHz.
The LM3900 inputs operate at +Vbe above ground, and are at low-impedance in accepting currents. The input current can vary over wide-range, though 10uA is the nominal current at which delta-Vbe between the transistors at both inputs are said to match quite well. Directly sourcing a signal into the non-inverting input for the express purpose of modifying the Vbe immediately there should have a non-lnear impact on how the voltage input is "interpreted" as a current. This effect will only occur over a small range, but it should be non-linear. And it's a very interesting case of taking an output directly from the input of the opamp, at other side of a current-conversion resistor.
This design concept actually works. The circuit below transforms a triwave attenuated at the non-inverting input into a fairly clean sinusoidal wave:
An intriguing range of application functions could open up if this non-linear effect can be exploited more fully. First of all, there are four opamps on the same die, and the transistor matching over temperature should be relatively good. A differential mode of operation comes to mind, for balanced effects. Possibly a clever circuit could exploit this non-linearity for log, antilog, and other transcendental computations. In particular, the low-impedance low-capacitance nature of the LM3900 inputs invites wideband operation. While critically dependent on the DC level input, the sinusoidal shaping effect seen was lossless easily up to 100 KHz, and probably extends to MHz.
Comments
Post a Comment