I am trying to build a circuit featuring multi-channel digitally-control current source and voltage meter.

Each current source is required to supply up to 5mA of current to a load that could reach a voltage as high as 10V. I chose a NXP constant current source device which can supply maximum 50mA of current and support a max voltage of 50V. The current source can be adjusted using an external resistor. I use a digital potentiometer to act as this external resistor so that I can control the current source digitally. The current source circuit is illustrated as followed. Current Source Circuit

I have to generate both the 3.3V and -15V power supplies because the Digital Potentiometer requires that the voltage applied to both terminals should not exceed the [0V,3.3V] range in either direction, while my current source load requires a voltage range of 0-10V.

To generate both 3.3V and -15V, I use a power op-amp like TI's LM675 to split a 24V power supply into [0V,+9V] and [-15V,0V] dual ranges. Then I use the +9V to generate two 3.3V supplies for analog and digital circuit respectively. U108 and U109 in the figure below are voltage regulators and generate fixed 3.3V output. Power op-amp

In the last part of my circuit, I use a 24bit ADC to measure the voltage of a signal, which is buffered using an Op-amp that directly connect its output to the negative input (voltage follower). In the figure below, U1 and U3 are two-channel 24-bit ADCs and U2 is a four channel op-amp. The DVDD is only used to communicate with the ADC chip and configure the digital potentiometer. ADC and op-amp

In total, I have 20 channels of those current source and volt meter all sharing the same AVDD, DVDD, AGND and VGND.

Problem: the output of the power op-amp (U114) is extremely unstable. The power supply path is oscillating.

I have tried to remove both the 3.3V voltage regulator, but the same problem remains and now DGND=2.5V while DVDD=2.1V (with reference to the 24V power ground). Can anyone provide a hint?

  • \$\begingroup\$ You probably have too much capacitance on the output of the op-amp. How much total capacitance is connected to the output of the amplifier and small signal AC ground? \$\endgroup\$
    – John D
    Commented Aug 25, 2016 at 1:01
  • \$\begingroup\$ I can't figure out how your U114 rail splitter is supposed to give you +9v/-15v ... Even if it didn't oscillate it would still split your supply evenly into +12v/-12v. And where does the DGND net come from? Your U114's output is driving VGND. \$\endgroup\$
    – brhans
    Commented Aug 25, 2016 at 2:05
  • \$\begingroup\$ Have you tried complementary darlingtons on U114 output with feedback from that? \$\endgroup\$ Commented Aug 25, 2016 at 2:17
  • \$\begingroup\$ @JohnD I am worrying about the output capacitance too. I didn't not have an accurate number. The VGND is connected to 18 IC chips, each draws about 1 or 2 mA. The AGND is also connected to 15 IC chips, each draws about 1mA. \$\endgroup\$
    – Zefu Dai
    Commented Aug 25, 2016 at 14:18
  • \$\begingroup\$ @brhans Sorry I label it wrong, there is no DGND, only VGND and AGND. The output of U114 is set by R59, R60 and R61, which gives about 12*(1+5.5/(2+20)) = 15V. \$\endgroup\$
    – Zefu Dai
    Commented Aug 25, 2016 at 14:21

1 Answer 1


@brhans Thank you for your valuable hint. I separate the input paths of the LM675. When there is no load at the output, U114 acts exactly as expected, giving me a 15V. However, whenever I connect the output with the load, it starts to oscillate.

Then I notice that both input signals are oscillating too. So I tried adding a 1uF cap to each of the input signal and BINGO! the output becomes really stable. So the schematic looks like this now. Voltage split with stable output

I don't have enough knowledge to explain this. The noise may come from the -15V connection to all the current source ICs. But they are all turned off at time of testing.

  • \$\begingroup\$ C66 causes less phase margin with DC load current (lag), while C15 causes more phase margin with Phase leading derivative. and C14 decouples supply noise. I was suggesting output buffers to reduce the DC load current, but if this works ...ok. when phase margin gets near 0 it oscillates. \$\endgroup\$ Commented Aug 25, 2016 at 17:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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