Dear All;

I have a problem about a ceramic capacitor in my design.


The problem occurs at C6. Note : C2 and C5 is same capacitor but no error occurs at them.
C6 is 1uF, 50VDC, X5R, 0603 MLCC type capacitor (Mfg P/N : Samsung CL10A105KB8NNNC)

Let me define the board and problem.
- This design is low-power RGB LED driver which has LED Driver IC and LDO.
- Design can work at 19~28VDC nom. at 24VDC (VIN power port)
- Design has resettable fuse and TVS diode(SMAJ26A) on LDO power input.
- LDO just fed LED driver IC, not LEDs, LEDs anodes are directly connected to 24VDC, LEDs cathode are connected to LED Driver IC.
- This sheet is called as a one pixel of linear luminarie. One luminarie has 8 pixels(8 pcs of given sheet).
- My power supply can fed up to 30 meters of linear luminarie and it means 240 pcs of pixels.(One pixel is given sheet as specified)
Please note that, 240 pcs of given design is connected as parallel to one power supply.
- 24VDC / GND distributed parallel with help of extra cabling . Data signals (SDI / SDO) are serially connected to each pixels, as in and out fashion.

The problem is;
C6 capacitor blows at a random time and at a random pixel, C6 become short-circuit.
So, LDO stops functioning, F4 fuse protects C6 to fire up. Then, remaining portion of cascaded connected pixels can't work, because data can't go through this defective pixel.
When we remove C6 from blown pixel, system / all pixels resumes to functioning.
Somehow, this capacitor blows and short-circuited, so it malfunctions of this pixel line.

What we try (All our changes and attempts doesn't give any result). Given attempts are done step-by-step and make observations for sufficient time.
- We decrease the power of pixels to reduce total current flowing through VIN line.
- We decrease the voltage of power supply to 21VDC.
- We place bulk capacitors (680uF, 100VDC, Aluminium Electrolytic) at VIN line to reduce overshoots and possible spikes at start, 10th, 20th, 30th meters of pixel line.
- We made observation that issue is happening at power on : Absolutely NOT. Cap may take the knock at power-on, but we oftenly observe total malfunctioning of pixel while system is working.
- We get measurements from start and end of power line(VIN/24VDC) of luminaries with scope. Power seems very very clear at power-on, continous and power-off times, no overshoot/no spikes, voltage rise to 24VDC very soft (5ms)
As seen here, we do most of known modifications and observations.

Pleased your help about what can be the reason of specified blowing cap issue.

Best Regards,

 

Try using a 100v rated cap, or place 2 x 50v ones in series.

 

Hi Mike,

As my measurements, it is not about voltage rating.
Because no spike or overshoot at any time, I got long scope measurements with single trigger mode.
Placing 2 x cap as series can be one of solution, you are right.
But, I need to find the cause of issue.

Best Regards,

 

Just because you can't see it doesn't mean it isn't happening. When the 24 volts is turned on, C6 looks like a dead short. Each turn on cycle contributes to the degradation of the part until failure occurs. An 0603 capacitor is a pretty delicate part, and you need to damp the magnitude of the turn on transient, or find a more robust part.

 

C6 will need a large voltage above its operating limit and / or a high ripple current to heat it and blow it.


Are you certain its the right part, not a polarised capacitor put in back to front , or a counterfeit part with the wrong voltage quoted ?

I hope its something like this as I'm using a bunch of these on a board I'm just finishing.

 

I'll bet that C6 is a tantalum. Try changing to an aluminum electrolytic, or do as some do when they have this problem, omit the capacitor from your design.

 

C6 will need a large voltage above its operating limit and / or a high ripple current to heat it and blow it.


Are you certain its the right part, not a polarised capacitor put in back to front , or a counterfeit part with the wrong voltage quoted ?

I hope its something like this as I'm using a bunch of these on a board I'm just finishing.

C6 is the right part which is defected.
As written in my first post, C6 is ceramic non-pol cap.
We bought this cap from well-known distributor, but counrterfeit can be one of minor possibility.

Best Regards,

 

I'll bet that C6 is a tantalum. Try changing to an aluminum electrolytic, or do as some do when they have this problem, omit the capacitor from your design.

Hi,

C6 is ceramic non-pol cap. One of the best and easiest way is to remove this cap from design, but I want to find what is the cause of given issue to be sure about my design.

For eg; is there any limit or may be the formula that to find maximum parallel connected capacitor in a power line?

Best Regards,

 

I told you what the issue was in Post #4. Is there some reason you don't believe me?

 

Papabravo's diagnosis in post #4 sounds valid. It is a variation of what I was thinking of. If you add some resistance or sufficient inductance in series with the power supply that might solve the problem.

 

I told you what the issue was in Post #4. Is there some reason you don't believe me?

with all due respect @Papabravo
You have put forward a suggestion, as have others, you have not identified "the" problem

 

Papabravo's diagnosis in post #4 sounds valid. It is a variation of what I was thinking of. If you add some resistance or sufficient inductance in series with the power supply that might solve the problem.

I'm really not to certain , given the evidence provided.

" C6 capacitor blows at a random time and at a random pixel, C6 become short-circuit "
" decrease the voltage of power supply to 21VDC "
" place bulk capacitors (680uF, 100VDC, Aluminium Electrolytic) at VIN line to reduce overshoots and possible spikes at start, 10th, 20th, 30th meters of pixel line "
" We made observation that issue is happening at power on : Absolutely NOT. Cap may take the knock at power-on, but we often observe total malfunctioning of pixel while system is working "
" Design has resettable fuse and TVS diode(SMAJ26A) on LDO power input "

So the large 680 uF input capacitor added should have successfully absorbed any start up spike that got through the TVS, and they do not see it die at start up, so kind of rules as low probability the PSU input,

I find this statement interesting " So, LDO stops functioning "

why would LDO stop functioning ? my thought would be over current ,

hence this is a very good question,
obvious "answer" as mentioned above is to put a higher voltage capacitor in,
but that might just be making the problem less likely, and it might catch fire in customers boards.

so I'm personally not to keen on that without some evidence that its a over voltage

a duff batch of capacitors seems the "logical" answer, but as they purchased from an authorised distributor, also seems unlikely.

sorry , I'm no help on a perfect answer,

I'm wondering, you mention about spikes as the strips each cycle through.
what sort of sizes are these as measured at the LDO ?
You say if I understand correct, that the LED driver is powered direct off the Vcc,
so why would the LED power changes induce spikes on the LDO,
it has a feeling of a ring earth as opposed to star point,
but that would have to be terrible to blow a capacitor which is more than 100 % over rated for the 24 V

Can you share a more detailed schematic and a picture of the assembly please.

 

Is there anything in the same circuit that could cause a large ripple current in the capacitor?

Short story: After working on a hand-wired circuit board for a year the printed circuit version arrived. I quickly loaded the board and fired it up. A few checks showed everything working. The board contained some analog circuitry, a boost converter (a couple amps output at least) an a microcontroller. I left the board running while running an errand but when I returned home it was dead. A quick check of the input voltage line showed a short to ground. Then I noticed that the 5V regulator and its ceramic input decoupling capacitor was right next to the 470 uf decoupling capacitor on the input of the boost circuit. I removed the ceramic decoupling capacitor and everything worked fine. The assembly went into production years ago and the problem never came back.

 

I really appriciated about your responses.

We have found one new clue. Let me explain it.
As I clarify in my first post, we decrease total system power consumption of LED strings, then our failure rate decreases a lot but not finished. We made this test more than 15 days, we got 2-3 failures totally.
3 days ago, we again increased total power consumption of system, failures became apparent each days as 4-5 pcs.

I thought that it is a big clue that total current demand of given LED string increases the probability of blowing.
So, in some reason(ripple current etc.), overheating of cap is the main problem I guess.

@Papabravo : I got your advises in Post #4.
@andrewmm : In Post #12, as I mean LDO stops functioning means, PTC became so resistive and there is too much voltage drop, so input voltage of LDO decreases and can't generate relevant voltage for other IC.
Actually about your spike question, I didn't see any spikes after placing 680uF caps.

@DickCappels : As you described, it is nearly same issue, the cap blows at run-time mostly, or get damage at start-up but blows at run-time. As I said, if total current flow increases, it makes more failures at this caps.

I pleased your advises.

 

Question,

How hot does the cap and the LDO get ?

 

Just a thought, That LM317 you have is rated at 100 mA max , its not powering anythign else is it ? Should not kill capacitor, but a separate thought.

The other thign that can kill LM317, is reverse current,
If the input is removed / shorted , and the output has capacitance on it ,then current flows back through the LM317,
if its ever likely that the input can suddenly drop to zero, and output not, then you need a reverse diode across the top of the LM317.
Again should not be reason for the capacitor to die, but.

 

Just a thought, That LM317 you have is rated at 100 mA max , its not powering anythign else is it ? Should not kill capacitor, but a separate thought.

The other thign that can kill LM317, is reverse current,
If the input is removed / shorted , and the output has capacitance on it ,then current flows back through the LM317,
if its ever likely that the input can suddenly drop to zero, and output not, then you need a reverse diode across the top of the LM317.
Again should not be reason for the capacitor to die, but.

Capacitor is at input side, LM317 just fed IC and max current need for that IC is 20mA.
Actually, nothing happens at LM317, it is working if I removed the cap, it is not working if cap still there. As you said, input is short circuited because of cap.

The best and easiest way is to remove this cap, but I'm trying to find solution not to make this modification.
Let me ask you that, connecting serial resistor to main input line(24VDC) of string solves problem?
If high ripple and overheating causes the problem, what are the ways of reducing this effect?

Best Regards,

 

If its heating effect on the capacitor, you should be able to measure it ,
If it is ripple heating , three answers

a) Use a lower ripple PSU
b) Use a filter to lower the LDO
c) Use a higher value capacitor / better ripple current characteristic.

Sorry I have forgotten, Have you seen this ripple ?
If so, is this caused by the current spikes from the led array ?

 

I told you what the issue was in Post #4. Is there some reason you don't believe me?

Papabravo is correct- the cap, assuming it's quality, will only fail due to voltage. Not current. It is a field-effect device. If the PTC1206 can handle a higher voltage than C6, even just a couple of volts, that can kill that cap. What is the specific part number for the PTC?

 

The PTC is 0ZCJ0012FF2E.

https://www.belfuse.com/product/part-details?partn=0ZCJ0012FF2E