Hi All;

I have issue at ceramic capacitor in my design.
The capacitor is, 1uF, 50VDC, 0603, X5R Samsung capacitor (CL10A105KB8NNNC)

Let me clarify issue.
- I have a power supply, 24VDC output, max current capacity is 15A.
- I have loads which are LED Drivers that can be controlled by a sytem. Each load is max. 2W, total max. qty of load is 150 qty. Loads are connected as parallel to power supply.
- Loads's can be controlled by system, thus a load can be 2W, 1.7W, 1.2W, 0.8W, 0.3W at any time.
- My capacitor is placed as decoupling cap(MLCC) at VCC-GND terminals at input side of loads. ,
- Distance between LED Drivers and power supply has 50 meters.

When I set load power to 0.5W, everything is fine. Total power is max 75W.
When I set load power to 2W, input capacitor blows and it becomes short-circuited. Total max power is 300W

So, when total power consumption of load is low, everything is fine.
However, when total power consumption becomes higler as drive with 2W, the cap at input side of LED driver very randomly.

What can cause this? And do you have any solution to eliminate of given issue.

Best Regards;

 

What can cause this? And do you have any solution to eliminate of given issue.

The loads combined result in very high current spikes. These result in Joule effect in the capacitor's ESR.
You must increase the number of capacitors.

 

It takes around 500 nano second for a signal to reach the LEDS after it has reached the cap (assuming a traveling speed of c/3 ). Have you tried to add some resistor in series with the cap (and both in parallel with the load) to reduce the initial current seen by the cap? Furthermore, the initial current, C * dV/dt, has to be limited (ref.: ABC of Capacitors, by Würth Elektronik, 1st ed, page 43), unfortunately, that maximum allowable value, Ic, is not often given.

 

High current power distribution can be almost as tricky as resolving ground loops on high-frequency, single-ended low level signal distribution systems. It doesn't HAVE to be that way but sitting here it's not possible to prove that some particular issue IS NOT causing the problem, and you may have to try changing more than one parameter in order to find out what combination of issues will actually resolve the problem. For example almost all modern power supplies are switching and some are operating at hundreds of kilohertz, at these frequencies and the currents you're using a strip of copper sheet a foot long and half an inch wide has a substantial inductance, you very likely need to take the "skin effect" of the conductor into account too. Now we don't know if the power supply has any stability issues with regard to the maximum phase angle of the load that it's "looking into", and we also don't know the nature of the LED drivers, they could be digital PWM, analog current or a mix, so we may have a switching source "looking into" a load switching at a DIFFERENT frequency, this suddenly becomes a situation you're not going to be able to resolve by running a simple SPICE simulation!

Let's just consider the "50 meter" distance between the power supply and the loads. According to a simple table set up for 3% voltage drop at 110 volts AC, for 150 feet the minimum recommended wire size is 8 AWG, and that's just based on resistance not inductance or skin effect or anything. Now maybe you can "get away" with a smaller gauge and higher voltage drop than that and save some money, we're fine with that but what "headroom" can the power supply you're using actually offer, is there even any engineering data provided about that issue? Does the power supply provide "sense terminals" so the 24 volts provided actually is measured at the load, and if it does are you using them?

It's really anybody's guess whether the capacitor is failing based on spikes or a sustained voltage level, based on experience I suspect the latter which means at some point your LED drivers are being subjected to 60-70 volt levels which has the potential to really shorten the life of those drivers and possibly the LEDs themselves, and it COULD even wind up being a higher voltage than the input voltage to the power supply switch! A 15 amp 50 meter power distribution cable is really a rather low impedance analog delay line but at some frequency it's also a resonant tank circuit. Also compared to its characteristic impedance at that frequency a 1 uF capacitor really offers no energy storage at all, BUT simply "bulking up" the size of the capacitor is not likely to yield success. And it's definitely "possible" to see exactly what's going on in time domain at both ends of this cable using a differential probe and a high-frequency storage osciloscope, but from what you've told me you don't have the level of experience to succeed at this - and even I wouldn't try because you'd also need to have design information about the power supply including the software tools that were used to design it, and that information probably isn't available either!

Is there any "natural" packaging interval for these LED driver circuits, or are they all controlled from the same MCU? The reason I'm asking is my "gut feel" tells me it would be easier to get this working by distributing the power across say 5 different cable circuits of 2.4 amps each, and each of those groups would have its own bypass capacitor, they could still all come off the same power supply. If that created a "ground loop" then I would recommend considering breaking up the control circuits using optoisolators. If the cost and complexity of optos is too much then maybe they should each have their own power supply. Or move the single power supply much closer to the load if it's feasible. I'm sorry if I can't provide one of those single "magic bullet" answers but with the combination you've selected you've just "pushed" everything way past the range where simple systems assumptions have any relevance.

 

I agree with jlm1948. If the led drivers employ a "switch-mode constant current source" topology, then the current pulses drawn by the driver from the bypass capacitor are heating it up via the ESR (as well as the supply current replenishing the loss). If the 24V @ 15A power supply is linear (or transformer based), then the ripple voltage AND current on the bypass caps would be even larger. If the supply is transformer based, then a rule of thumb from a Stancor application note indicates: 2000uF / per 1A of current. That would equal a 200uF cap per led driver (I would still keep a low-value ceramic between the larger bypass and the driver). In any solution, you will need to increase the number of caps, perhaps a combination of types, to lower the overall ESR of the driver's bypass capacitor(s).