0 to 1 Amps Variable Milli Amps Source

Yes, the LM317 three-terminal regulator was designed to be a voltage source, rather than a current source, but nonetheless it's long been a rather popular CS chip, thanks to its low 1.2-volt reference voltage. For example, to get a 1A current source, just add a 1.2-ohm current-sensing resistor to drop 1.2V at 1A, satisfying the ADJ-input's reference voltage (right, from NSC datasheet).

The exact current is Iout = V_ADJ / R1 + I_ADJ where V_ADJ = 1.23 volts and I_ADJ = 50 uA typ, or 100 uA max. Some people recommend a capacitor across the load. The compliance voltage is up to V_IN - V_DROPOUT - 1.2V, where the dropout voltage is 1.7V at 250mA and 25ºC, rising to 2.3V at 1.5A.

If R1 is a fixed plus a variable resistor, the output current can be reduced as R1 is increased, but at currents above say 10mA, i.e., with a 100-ohm pot, this is impractical. But there are many good solutions to this issue.

In the old days at s.e.d. (the usenet group, sci.electronics.design) we used to toss variable LM317 current-source circuits back and forth at each other for hours on end. We also specialized in voltage-controlled CS circuits using LM317. We found the LM317 more attractive to use than a big old BJT output-pass transistor in our circuits, because it featured thermal protection and was easy to drive with its tiny 50uA V_ADJ-pin current. The 3-pin TO-220 package was just like a classic power transistor, and the cost wasn't much higher. Here's one circuit, from a post on 2002-03-22.

For Vin = 0 the output current Iout = 1.25/Rs, but if you raise the control voltage to Vin = 1.25 volts, Iout will be reduced 0 (ignoring the LM317's 3.5mA typ minimum-load requirement). This circuit uses an opamp acting as a difference amplifier to control the LM317's ADJ pin forcing the Rs sense voltage - V_ADJ = vin. The output diodes insure that the op-amp outputs have operating room.

I commented at the time, "if you use an IC difference amplifier such as the Burr-Brown INA105, with its 0.01% resistor matching, you'll have more accurate results at low currents. Don't forget that TI offers free samples! http://focus.ti.com/docs/prod/folders/print/ina105.html"

You'll note the schematic is a famous ASCII drawing, which uses a fixed-font. This was necessary because usenet didn't allow pictures or html. Here I'm forced to use a picture to reproduce the ASCII drawing because CR4 doesn't allow us to select fonts, even Courier, or any other fixed-pitch font.

We quickly improved on the circuit, first adding a 15mA current-sink to handle the LM317's quiescent current, and moving the difference-amplifier sensing across Rs. Then we used a 10:1 amplifier, so the sense resistor could be reduced by 10x, saving power dissipation at 1A. The circuit did require a negative supply.

The file I took these ASCII drawings from has dozens of more interesting and complex examples. Yep, those were the days!

We can't leave this topic without mentioning the new Linear Technology parts meant to replace the LM317 in CS circuits, the LT3080 and the LT3092. Instead of a 1.2-volt V_ADJ they feature a precision 10.0 uA ±0.1 uA current-programming SET pin, and you add a resistor to pick your voltage, e.g. 120k for 1.2 volts, or 12k for 120mV. Nice, huh? The latter part is guaranteed to work without Cout capacitors.