Constant Current Sink

Introduction

A constant current sink is a device which draws a constant current regardless of the supplied voltage. It can be a useful piece of test equipment for load testing DC power supplies, measuring the capacity of energy storage devices or as a current limiter.

A simple constant current sink can be constructed by combining a voltage reference with an emitter-follower transistor amplifier:

Ie=(Vref-Vbe)/R. Ic~=Ie for transistor Beta>>1. If Vref, Vbe and R remain constant, current will also be constant over a wide range of supply voltages (VCC).

This circuit however suffers from drift caused by 3 separate mechanisms being the Early Effect, and temperature effects caused by self-heating of the transistor and emitter resistor. This article will study each mechanism in isolation and propose practical solutions to minimise their effects.

Sources of drift

BJT Early Effect

As Vcb is increased across a bipolar transistor, common-emitter current gain (Beta) rises due to modulation of the base-width. The implication for our emitter-follower amplifier is that a change in supply voltage will cause a small change in collector current, since Ic = Ie*B/(B+1), where Ie is being held constant.

The Early Effect can be reduced dramatically by the addition of a second transistor in a cascode configuration:

The result is that Q1's Vcb is held close to constant with varying supply voltage, reducing the Early Effect in Q1.

Transistor self-heating

Vbe has a negative temperature coefficient, causing an increase in current as the transistor heats up. A complementary feedback pair as follows offers a partial solution to the problem:

The complementary pair configuration has two main benefits - firstly, the effect on the current regulation caused by Q2 self-heating is greatly reduced by negative feedback. Secondly, the collector current of Q1 is reduced by a factor of Beta therefore reducing power dissipated in Q1 and therefore maintaining a more stable Vbe of Q1. The disadvantage is that we have now introduced the Early Effect of Q2, however it is greatly reduced by negative feedback. Q2 could be cascoded as well if required.

Since the current regulation of the circuit depends mostly on the stability of Q1's Vbe, further transistor stages (either emitter-follower or complementary feedback) can be added, each further reducing the power dissipated in Q1.

The purpose of C1 is to limit the slew rate, which may be required to prevent global oscillation when many stages (or slow transistors) are used.

Resistor temperature effects

The drift caused by self-heating of the main emitter (current shunt) resistor can be minimised by choosing a resistor with a better temperature coefficient of resistance, sinking heat away from the resistor by passive/active cooling, or by paralleling two different types of resistors which have complementary temperature coefficients.

Selecting a lower value shunt resistor will reduce the power dissipation in the resistor but can have an adverse effect on stability at low current settings.

Practical Implementation

Voltage reference

An adjustable voltage reference can be as simple as a battery with a potentiometer wired across it, with the wiper providing the output voltage. As long as the current drawn from the output of such a voltage reference is negligible, the reference voltage will only drift as quickly as the battery discharges due to the quiescent current drawn by the pot.

A more novel implementation would be to have a voltage reference which is powered from the circuit under test. This can be accomplished by powering a zener diode from a constant current source as follows:

The current source works by using Q2 as a current limiter. As the voltage across R1 reaches the Vbe of Q2, the drive to the base of Q1 is pulled down by Q2. The current through the load is therefore regulated at Vbe(Q2)/R1 producing a stable reference voltage across the zener. Early/thermal effects of Q1 are mostly compensated for by negative feedback. Since a small change in current through the zener diode will cause an even smaller change in reference voltage, this is unlikely to be a significant source of drift in the overall design of the current sink.

The major limitation of this circuit is that the bias current through Q2 will vary almost proportionally with the supply voltage, varying Vbe(Q2) and therefore also the regulated current output. Choosing a value of R2 to provide an adequate bias current over a wide supply range without being under-biased at one extreme and causing significant self heating of Q2 at the other is not possible. The variation in bias current is also not desirable if the overall current sink design is required to maintain mA accuracy with varying supply voltage.

An improvement can be made by replacing R2 with another current source as below. The bias current for the second current source is regulated from the first, so quiescent current maintains near-constant as well:

Putting it all together

The complement of the above circuits (swap NPN for PNP and vice versa) has been used to allow for the use of NPN power transistors which are generally cheaper and more available than equivalent PNP transistors. D3 removes some of the dead zone caused by the reference voltage being lower than Vbe(Q3) and provides some temperature compensation for Q3. R4 is used to set the upper limit of the current control. The value of R1 could also be chosen instead to limit the current range, but this will cause increased burden voltage for a given current setting and increase power dissipation in R1. Minimum supply voltage is around 4V.

Although not strictly required, a couple uF bypass capacitor from VCC to ground can prevent oscillation, usually due to inductance of the test leads or a powersupply under test with marginal stability.

Obtaining higher current/voltage/power handling

Multiple power transistors can be paralleled for higher current/power handling. Emitter resistors are required to current balance the output devices, due to manufacturing variance in Vbe. There needs to be enough voltage drop across the resistors such that transistors with a higher Vbe enter their forward-active region before the transistors with a lower Vbe become overloaded.

Maximum voltage is only limited by the voltage/power rating of the transistors and physical construction of the unit (isolation, creepage distances). Q6, operating at around 3mA bias current will be approaching the thermal limits of a TO92 package at around 200V. If high voltage operation is required, the bias current can be reduced (by substituting R5 for a higher value, e.g. 680R for ~1mA) at a negligible cost to regulation performance. Alternatively, a higher power transistor can be substituted, although likely a better solution is to parallel multiple smaller transistors which will have higher gain and bandwidth, each with their own emitter resistor to balance currents.

MOSFET output devices

Of course, there is no reason that MOSFET(s) cannot be used instead of BJTs. MOSFETs have several advantages - one being that MOSFETs with high current ratings can be obtained in compact packages and/or affordable unit pricing. MOSFETs also generally don't require a high current driver transistor unless transient performance is a concern and can be paralleled without the need for emitter/source matching resistors due to their positive Rds temperature coefficient and high peak current capabilities.

Cons include generally high Vgs (compared to BJT Vbe) which increases the overall burden voltage and generally MOSFETs offer no better value for money than BJTs if overall power handling is the key design goal.

Use as a current limiter

While the above circuits can be utilised as a current limiter, they perform poorly as the power transistors are unable to enter saturation and a significant voltage drop is seen across the terminals below the set current limit.

By providing a separate open-collector output, and connecting the ground to the powersupply's ground so the rest of the circuit is provided with a constant supply voltage, the power transistors are able to operate in saturation when the output current is below the set limit achieving a much smaller voltage drop between input and output. The downside to this modification is that now not all the current going through the current shunt resistor flows through the output, so a small error can arise between the set current and the effective limit in practice.