First: im new here and this is my first question so if you think that I should ask or behave in a different way, feel free to tell me.

The (My) Problem: We run LEDs on a test stand with ~5V,2.3A and have the Problem of brief (~20µs) disconnects between the LED and the power supply. The disconnect is rapid, so that in some cases the supply will not shut down. In case that happens the voltage at the open contacts increases and there is an amp Peak ~>5A at the Moment of reconnection which destroyes the LED.

I my idea for a circuit to solve that issue was this one:enter image description here

There should be a n-mos with a high power, low resistance Resistor in a series Connection in a parallel configation to the LED. The Gate Voltage should be taken of a voltage divider (someone mentioned to me that instead of R1 i should use a z-Diode, although they did not provide a reason for that) The increasing voltage should open the mosfet, wich creates a way to reduce the energy in the circuit and there is no peak when the LED is connected to the circuit again.

this one is similar, but there is an OP involved to drive the mosfet. enter image description here

My aim is to simulate, build and test a prototype. It is not to create a PCB with the circuit yet. Since Im quite a newbie when it comes to simulations in LTSpice and desinging real circuits I have some difficulties to turn my ideas into a running solution.


-I know i need a resistor before the gate of mosfet, but which one?

-Or how to choose a fitting OP?

-Wich Mosfet shall i use here?

-What are the next steps ?

Thank you for reading, i hope to find help here.

  • 1
    \$\begingroup\$ Please avoid using potentially ambiguous terms terms like "open" or "closed" when referring to transistors. It's sometimes difficult to determine if you're using them like "open valve" or "open circuit" - which mean completely opposite things. "On" or "Off" would be less ambiguous. \$\endgroup\$
    – brhans
    Mar 25 '19 at 14:04
  • 1
    \$\begingroup\$ What power supply are you using? LEDs should normally use constant current supplies which shouldn't behave like this. \$\endgroup\$
    – Finbarr
    Mar 25 '19 at 14:06
  • \$\begingroup\$ Besides which, the current source in your diagrams is pointing the wrong way. \$\endgroup\$
    – Dave Tweed
    Mar 25 '19 at 14:07
  • \$\begingroup\$ I'm pretty sure you mean suppress, not oppress, but I still got your meaning. \$\endgroup\$
    – Hearth
    Mar 25 '19 at 15:12
  • \$\begingroup\$ Sounds like the LED driver is using current sense feedback. During the disconnect, there is zero current, so the control loop goes to max duty cycle and the output voltage spikes. Then upon re-connect, the high voltage stored in the output capacitor blows up the load. \$\endgroup\$
    – mkeith
    Mar 25 '19 at 16:06

To prevent surges from intermittent impedance one must be able to detect impedance rise and shutdown output faster than the part can fail or not change the output voltage fast than it can detect an open conduction.

If you sense current and voltage then differentiate both to produce a protective reaction from either an open or a short, using some capacitance limits the voltage slew rate. BUT one must be aware this creates a line LC resonant condition which creates a different problem.

So slew rate must be managed in a control loop with a Bandwidth defined by your specs.

To systematically find an acceptable solution, one must 1st define measurable acceptance criteria or specs on surge current using Ohm's Law, energy stored with formulae like 1/2CV^2 and Safe Operating Current-time spec limit on your LEDs and drivers.

So always start a design problem with Specs rather than solutions that may become poor band-aids. Such as dV/dt, dI/dt limit, soft-start spec, momentary interupt dV/dt limit.

The essential part of all of this is the current sense Resistor which are typically choosing for 100mV drop max or << 0.25 W . But also remember if you shut off the current at too fast a slew rate the back EMF = V=LdI/dt from cable inductance.

The critical part of learning is how to master these concepts and learn by examples that work if you can find any such circuits. They may not be as simple as you would like.

This is why CC power supplies are full of parts and not simply a CC regulator. Power regulators often have OCP, OTP, UVP and step response times and overshoot to define these characteristics.

This requires some System Control Theory experience, how to read datasheets on every part, then put it all together into a proper Design Spec.

One commercial part for this is called a crowbar clamp or ..

Bourns® LSP Series protectors are electronic shunts that provide a current bypass when an LED element in an LED string fails open circuit. This ensures the remaining string of LEDs will continue to function. There are many cases where high reliability of the LED lighting must be maintained, such as LCD backlighting, transport lighting, avionics, intrinsically safe and low maintenance lighting.

Then you design the trigger and shutdown to release the SCR’s holding current and restart after a suitable delay.

  • 1
    \$\begingroup\$ I didn't read thoroughly, but I think their problem is, in fact, coming from trace inductance. I'm glad to see you catch the point in what you wrote. +1 \$\endgroup\$
    – jonk
    Mar 25 '19 at 18:20
  • \$\begingroup\$ Often the surge current is I=dV/ESR from storage capacitance at over overvoltage but without negative diode clamp protection that can certainly be an issue with long cables \$\endgroup\$ Mar 25 '19 at 18:48
  • \$\begingroup\$ I see some of you asking for the specs. Which Information should be given here? \$\endgroup\$
    – JohannOC
    Mar 29 '19 at 10:04
  • \$\begingroup\$ pls define all input surges by V,I or Z, t and or J with acceptable output criteria or some universal std tests \$\endgroup\$ Mar 29 '19 at 15:21

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.