1
\$\begingroup\$

I need a small (somewhat) voltage independent current (about 10 mA), with a voltage range of 3-30V. I came up with the following circuit.

schematic

simulate this circuit – Schematic created using CircuitLab

I drew the resistor U/I characteristic in the Id-Vgs characteristic of the MOSFET:

enter image description here

The idea is the circuit will quiesce at the intersection of both characteristics.

The MOSFET is a depletion mode N channel type. Circuitlab doesn't seem to have distinctive symbols for enhancement and depletion mode.

I choose a type that can handle over 1W of power to keep well away from SOA limits.

I am aware that a similar solution exists for JFET's, but I could not find a JFET type that can handle the voltage requirement. Also I know similar devices with JFET's exist as discrete components (constant current diodes), but these are rather expensive ($5 or more. The aim is to both keep component count and price low. A Widlar source is not an option, because it needs a reference current, which is not available.

Current constancy is less of an issue, it is meant to limit the current of an unknown voltage, supplied to a low resistance (20-50 ohm), 10 - 20 mA is fine. Neither are response times very important.

I don't feel really comfortable working at the lower end of the characteristic. Does anyone see a problem with this setup, something I overlooked, thermal stability issues, and maybe points for improvement? Maybe I am too much reliant on constancy of performance of individual devices?

\$\endgroup\$
  • \$\begingroup\$ RdsOn of Nch shown is <30 to 60 Ohms is too high for circuit shown with 80 Ohms current sensing for accuracy. It's Idss=20mA. but for rough approximation, ok. Using a FET with Idss=10mA is better with R1=0 or a BJT current limiter \$\endgroup\$ – Tony Stewart Sunnyskyguy EE75 Oct 6 '16 at 15:03
  • \$\begingroup\$ The LM317 and a resistor will make a nice ~10mA current source. \$\endgroup\$ – George Herold Oct 6 '16 at 15:22
  • \$\begingroup\$ Like George is indicating LM 317 TO92 can do the trick. Costs amost nothing. \$\endgroup\$ – Decapod Oct 6 '16 at 15:29
  • \$\begingroup\$ @GeorgeHerold 10 mA is right at the bottom of the range for an LM317. I looked for something where 10 mA is within the operating range for a device and found the LT3092, which is handily available in the same package as a BSP135. Could that be a better option? \$\endgroup\$ – Andrew Morton Oct 6 '16 at 18:38
  • 1
    \$\begingroup\$ @mkeith I think the OP is stating that he doesn't have such rails indirectly by writing, "A Widlar source is not an option, because it needs a reference current, which is not available." It there was a rail, then there would be a way to make a reference current. \$\endgroup\$ – jonk Oct 7 '16 at 8:59
1
\$\begingroup\$

The aim is to both keep component count and price low

Given your other limitations, it's hard to beat just a depletion mode MOSFET and a resistor. (I see where you are coming from, setting aside the JFETs.) I gather you are thinking of this:

schematic

simulate this circuit – Schematic created using CircuitLab

I can't tell from your writing just how repeatable you need to make this, though. The value of \$V_{GS\left(off\right)}\$ isn't all that precise. So you may need a potentiometer in there. Which ups the price, physical size, etc. Or you could just select parts, I suppose, and work it that way.

For example, the Supertex DN3525 specs \$V_{GS\left(off\right)}\$ anywhere from \$-1.5\:\textrm{V}\$ to \$-3.5\:\textrm{V}\$, with \$V_{DS} = 15\:\textrm{V}\$ and \$I_D = 1.0\:\textrm{mA}\$. Will that be repeatable enough without either selecting parts or selecting the resistor value?


I am aware that a similar solution exists for JFET's, but I could not find a JFET type that can handle the voltage requirement.

You could use this topology to cover that concern:

schematic

simulate this circuit

However, with the devices shown you may need more headroom than just \$3\:\textrm{V}\$.


I know the load line approach would seem to get you there. But isn't the variability pretty significant? Or do you think it's good enough?

\$\endgroup\$
1
\$\begingroup\$

A bipolar transistor current source can be used for your application. This one still regulates (barely) with Vin2=3v, but wastes nearly 3mA when Vin2=30v:

schematic

simulate this circuit – Schematic created using CircuitLab

This is a low-side current regulator using generic NPN silicon transistors. R3 determines the current-regulation setpoint. About 10mA flowing through R3 generates enough voltage to turn on Q3, which begins to current-regulate. This circuit has a voltage-burden of a bit more than a volt, better than a MOSfet. Q2 dissipates some heat. A generic 2N3904 will get warm with 10mA, a bit too-warm with 20mA. For more than 10mA, substitute for Q2 something beefier, like MJE181 that can dissipate heat more easily.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.