Current sources

Discussion in 'General Electronics Chat' started by Veracohr, Nov 5, 2013.

  1. Veracohr

    Thread Starter Distinguished Member

    Jan 3, 2011
    I'm working on a project where I need to control multiple OTA's with a single control signal over about 2.5 decades of control current. First I tried using a voltage source and resistor, but because of the range of currents I need (about 3μA to 1mA), I had trouble. It would have required getting a wide-ranging control signal with a very precise lower limit.

    So I thought I would try a current mirror, but I'm unsure if the way I did it is sound, or if there are practical problems that the simulation doesn't show. I can't really have one output transistor for each OTA control input because one of the circuits will have 8 OTA's that need to be varied together, and it would require a lot of transistor arrays to accomplish. I'm trying to keep component count and power use down.

    So my question is: can I use a single transistor output current source that feeds multiple inputs that all need equal current?


    In the schematic, everything from Q1 to the left (where the "Iout" label and arrow is) is an exponential current source that I got from this source. V3 is the linear control signal and V4 provides a reference. The diodes at the right represent the amplifier bias current (Iabc) inputs of the OTA's. The Iabc inputs are two diode drops above the negative voltage source (from page 4 of the LM13700 datasheet), so I just simulated them as 2 diodes.

    It seems to work just fine in simulation. Each diode branch gets exactly the same current in exactly the range I want. Assuming only minor variations between individual IC's, does this seem like it should work in practice as it does in simulation?
  2. #12


    Nov 30, 2010
    Two difficulties with practical current mirrors. They work in simulation because the transistors are perfectly matched. The work in an IC because the transistors are closely matched. The don't work in real life unless you hand pick your transistors, buy a monolithic chip with matched pairs on it, or add resistance in the emitter circuit to degenerate the differences.

    That's all I have on this one.
  3. Veracohr

    Thread Starter Distinguished Member

    Jan 3, 2011
    Oh, yeah, I would definitely use monolithic pairs. I have a bunch of 4-transistor arrays, although they're not specifically stated to be 'matched'. At least for testing purposes, perhaps I could expect them to be matched enough by virtue of being on the same chip?