New IC yields high performance two-terminal current sources
25 June 2009
Current source design is a more difficult task than designing voltage regulators. Robert Dobkin discusses the ways in which the LT3092 can overcome the problems of two-terminal current sources.
While high quality voltage sources are very common, current sources as components have remained elusive. Two-terminal current sources generate a new set of problems, especially if high accuracy and stability over temperature are desired. The current source must operate over a wide voltage range, exhibit high DC and AC impedance connected in series with unknown reactance, and exhibit good regulation and temperature coefficient. For optimal two-terminal solutions, no power supply bypass capacitor should be used since it degrades AC impedance.
Older solutions using depletion mode FETs have wide variability in current and temperature coefficients. A two-transistor, two-Zener diode, two-terminal current source does a reasonable job of providing a current, but its accuracy is limited to a few percent. The circuit operates open loop and as such cannot provide accuracies of closed loop feedback circuits. Since the temperature coefficients of the Zener and diode do not match the transistor perfectly, the circuit has drift and inaccuracies due to the variability in Zeners and transistor VBEs (Base-Emitter Voltages). Further, the circuit requires a minimum of about 3 V across it in order to operate properly.
A new device, the LT3092, overcomes the problems of earlier two-terminal current sources. Output current can be set from 0.5 mA to 200 mA. Current regulation is typically 10 ppm per Volt. The LT3092 operates down to 1.5 V or up to 40 V. This gives an impedance of 100 MΩ at 1 mA or 1 MΩ at 100 mA.
Figure 1 shows a basic diagram of the LT3092 current regulator where a PNP transistor is used as the output device. Internal circuitry is differential and buffered, with a regulator to isolate it from power supply changes. This isolation allows stable operation without bypass capacitors.
The internal current source and the offset of the amplifier are designed to reject power supply changes by 100 dB or better, so the regulation is very good. Setting the Rset down to zero allows the output to be adjusted down to zero.
A small voltage is impressed upon an external set resistor, 20k in this case, to generate a 200 mV reference. That forces 200 mV across a current-determining resistor R and the total current is then equal to 0.2 V divided by R (plus 10µA). The current regulator works from about 1.5 V across it up to 36 V.
The 200 mV generated reference is chosen to equalise the errors due to changes in the internal current source and changes in the offset of the amplifier with supply voltage. With supply changes, the internal current sources change approximately 50 pico-Amps per Volt. The offset of the internal opamp changes less than 5 µV per Volt. Assuming worst case for both the current source and the offset of the amplifier, using a 200 mV reference gives equal error contribution from both the amplifier and the internal current source. If the 200 mV is increased to 500 mV by using a 50k resistor, the contribution of the internal opamp offset goes down. This improves the regulation of the current source against supply changes.
The set resistor also allows for ease of trimming the total current. Should this device be used for high currents -100 mA - trimming the current would be difficult because of the low value for resistor R. However, the 20k resistor is always easily trimmed to set the current value to the desired level.
Increasing voltage compliance
For higher voltages, current sources can be stacked to operate at a higher total voltage. Two current sources are set up for the same currents and a voltage-limiting Zener is placed across each of the current sources. At low voltage, whichever current source has the incrementally higher current will saturate and the current will be controlled by the other current source. As the voltage increases, at some point the Zener breaks down and starts to conduct. Then the voltage across the saturated current source starts to increase and it regulates the current as the voltage continues to increase. When the current control goes from one current source to the other, there is a small discontinuity in the output current equal to the error between the two current sources. Typically this is less than 1%.
For high set currents and high voltage, there is considerable power dissipation in the LT3092. For example, 30 V and 100 mA equals 3 W of dissipation, which can result in significant temperature rise depending on the thermal resistance of the PC board. An external resistor can shift a portion of the power to the resistor and reduce the power dissipation in the LT3092. Figure 2 shows the basic current source with a resistor RX from the input to the output of the device. As long as the total current is more than the current through RX, regulation is not impaired and the current source impendence does not change.
Current through RX is within the feedback loop and gets compensated as the voltage from input to output changes. The current flows through the internal PNP transistor or the external resistor while the feedback loop keeps the total current constant.
For good regulation and to have reasonable margin, the current through RX should not be any larger then 90% of the desired current for the device at the maximum voltage. The formulas in the illustration show how to choose RX so that the current through RX always leaves at least 10% of the current flowing through the LT3092. This drops the maximum internal power down by shifting some power to the external resistor. This results in a reduction in device dissipation as well as a reduced rise in temperature.
If higher output currents are needed, current sources can be directly paralleled. Two LT3092s can be set up (with or without power shifting resistors) and directly paralleled to get twice the output current.
Figure 3 shows another method of paralleling devices. It requires fewer external components so it is possibly a better method to parallel many devices for high current. In this case, the set pins are tied together, which causes the output pins of the regulators to be within a few mV of each other. Then the outputs of the regulators are summed through 40 mΩ ballast resistors, causing the currents to share. These resistors typically can be made up by a small piece of the PC board that the devices are mounted on. When this is done, we increase the voltage drop from 200 mV to 1 V by using a 100k set resistor. The reason for this is to minimise the temperature coefficient effects of the PC trace ballast resistors.
About 8 mV is dropped across the ballast resistors and these resistors are made of copper. Copper has a temperature coefficient of 0.3% per degree, which will affect the temperature coefficient of the overall current source. Increasing the reference voltage from 200 mV to 1 V makes the percentage of voltage on the ballast resistor smaller and reduces this effect from about 1% to 0.2% for a 100˚C change.
Increasing the output current above 400 mA only requires one additional paralleled device and one ballast resistor, so for high currents this minimises the parts count. Again, we use RX to minimise the power dissipation in the devices by shunting a portion of the current around the ICs at high voltage.
Current sources can be driven into any kind of load. Since this device is actually a complicated integrated circuit, the impedance of the load may react with the internal circuitry and cause some instabilities. While every effort has been made to make the device stable into a wide variety of loads, instabilities can still exist. Stabilising the device is easy. A resistor can be inserted in series with the current regulator or a capacitor (basically a bypass capacitor) or series RC is connected across the device from Vin to Vout. This gives the device a known impedance across it and stabilises it against unknown impedances. Unlike older regulators, the capacitor can be very small.
The capacitor does not affect the impedance of the current source at low frequencies because, like the external resistor RX, it’s within the feedback loop. For AC changes across the device, current flows through the capacitor or through the internal transistor on the LT3092, so the impedance is unchanged. At high frequencies, the LT3092 runs out of bandwidth and the impedance looking into the LT3092 is capacitive.
With loads where there is significant voltage change during operation, the current through the capacitor must be less than the program current. Otherwise the loop becomes broken when more current flows through the capacitor than the programmed current. The slew rate that can be tolerated across the device is shown in Figure 4, and again approximately 90% of the current can flow through the capacitor without disrupting the impedance of the LT3092 current source.
Robert Dobkin is Vice President, Engineering & Chief Technical Officer at Linear Technology Corporation
Contact Details and Archive...
Related Articles...
Most Viewed Articles...