I have a simple circuit that pumps 5A for 100uSec. (Constant current circuit with Opamp).

enter image description here

This circuit is on a different board. 20V is generated locally on this board from the 4V supply (not shown, using a switcher). uP gives a 3.3V pulse and I observe 5A pass through the LED. The 5A pulse is occasional so, I charge the cap and when I need the 5A, the cap provides the current.

Occasionally, I have seen that the 5A return current destroys the CPU and some other components on my mother board. What am I doing wrong?

The GND of this board is connected to the mother board ground via the connector at a single point, I was expecting the current (5A) comes back to the Cap and hardly any current flows back the the motherboard ground. Apparently, something wrong.


Opamp is TLV2465

Mosfet IRLML6344

Cap is 2x 1.5mF (UHE1V152MHD Nichicon)

Schematic above has a mistake, the -/+ are swapped, the CPU goes to the + terminal and feedback from the sense resistor is to the -.

The board receives 4V which is boosted to 20V. Board has only 4 inputs, GND, 3.3V pulse from CPU (hidden behind a buffer), and blank. Opamp is run from the 4V before it is boosted to 20V. I also have a PTC to avoid inrush current between the 20V and the + end of the Cap.

  • \$\begingroup\$ A more complete schematic may help us answer your question. Which power nets and data signals enter the board on which connectors? What does the boost circuit look like. How much capacitance is on the 3.3 V net and at other locations on the 20 V net? How are these things layed out? \$\endgroup\$ – The Photon Aug 14 '12 at 3:52
  • \$\begingroup\$ Is the 3 mF capacitor a single part or a bunch of smaller parts in parallel? What is the ESR and ESL? \$\endgroup\$ – The Photon Aug 14 '12 at 3:53
  • \$\begingroup\$ @ThePhoton 2 caps parallel.. See the updates \$\endgroup\$ – Ktc Aug 14 '12 at 5:55
  • \$\begingroup\$ Was the prototype built on a PCB or on a breadboard? Can you post a picture of it? \$\endgroup\$ – Rocketmagnet Aug 14 '12 at 9:01
  • \$\begingroup\$ @Rocket [OT] - I read in a comment to a deleted answer from you that you can't include pictures in comment. You can: start a new question or answer, and upload the image. Copy the image's location, and cancel the question/answer. Use the image location in a hyperlink. \$\endgroup\$ – stevenvh Aug 14 '12 at 16:08

It is really hard to say what killed your CPU. But here are some things that are either certainly wrong or something to look into.

  1. I assume that you are using an N-Channel MOSFET. Your schematics are not clear about that.

  2. You do not show how you are powering the opamp. If I were doing this, I would power the opamp from +5v and -5v. Alternatively, you need to make sure that the opamp inputs and outputs are not too close to the power rails. The way it is currently drawn it is possible that the opamp will not be able to turn the MOSFET completely on and/or completely off.

  3. If you simulated this circuit using real models for the MOSFET and Opamp then you might find that it is unstable. (Or check this with an o-scope.) A cap between the opamp output and the negative input of the opamp could improve things. It will take a little trial and error to find the correct cap value, but expect it to be in the range of 50 to 500 pF. When I say "unstable", I mean that the output of the opamp or the LED current will oscillate or ring when it shouldn't.

  4. Consider the parasitic capacitance between the pins on the MOSFET. In particular between the drain and gate. As the MOSFET turns on that cap will tend to force the gate voltage down. If the Opamp were not there the gate would go to -20v. With the opamp it won't be so bad. It might not even be an issue. But if the MOSFET is very large then the parasitic capacitance will be large and might overwhelm the opamps ability to drive the gate. You don't need a large MOSFET for this circuit.

  5. Consider the loop area of your circuit. Your current loop is from the cap, through the LED/MOSFET/Resistor and back to the diode. The square inches (cm) of this loop of wire has a direct influence on the amount of EMI that will be emitted. Reduce the loop area to reduce the amount of EMI. It might be possible to emit enough EMI to damage other circuits.

  6. Consider placing a small cap in parallel with R2. This will slow down the LED turn on/off time and thus reduce some of the bad effects of having that 5 amp current spike. Of course don't slow it down enough to make your circuit not useful, but slowing it down even a little can help things a lot.

  7. Place a current limiting resistor in series with your LED. Size this resistor so that if the MOSFET is fully on the the LED is at around 110% of max power. This will help make the circuit more stable and help keep you from accidentally blowing up your LED's. It might be possible to combine this resistor with the current sense resistor (same resistor does both functions), but it depends on the input voltage range of your opamp.

  8. Keep the high current path (Cap-LED-MOSFET-Resistor-Cap) wires short and thick. Especially short. These wires will act like an inductor, and the inductive kickback of 5 amps for 100 uS could be huge. Also, you should put a high-current schottkey diode between the source and drain of the MOSFET. There is already a diode in the MOSFET but it isn't that good. Putting one in parallel with that will improve the protection from inductive kickback by a lot.

  9. Carefully analyze the current paths of your circuit and change it to keep the high current and voltage spikes from the sensitive parts. It is very likely that all of my other points will not identify what is causing the damage. But it is just as likely that whatever is damaging things is happening on every pulse, but only does damage occasionally. Get out the o-scope and methodically go through things.


Are the 4V and 3.3V generated independently? If not, you may be missing a power path.

I'd suggest using a resistor and maybe even an inductor instead of the PTC, to make sure the 20V is not suffering current spikes, since power is provided by the cap. This will eliminate a power path that may be the cause of your troubles.

Not related to your problem, but I suggest a simple high pass filter at the input, in case your cpu locks up, the circuit is not kept on.

In any case, at the cpu board check the power rails and possible ground voltage spikes with a scope when the led is being switched.


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.