Surge Protection for 24V DC Input Circuit

[victorb]

I had a circuit of buck convertor that converting 24V dc to 4V dc.
24V dc input is from a PSU (Meanwell's PLN series), once it go in, it connects to a Polyfuse, then polarity diode (in series), then Common mode choke, then it's my converter's input capacitor (35V 47uF AVX's TAJ series). Occasionally, the input capacitor got burnt when power on the 24V PSU. I'm thinking that the problem maybe:
1. my input capacitor's voltage rating is not enough, some document said I should use at least 48V voltage rating if I'm using tant capacitor.
2. Surge current on my low impedance circuit <http://www.avx.com/docs/techinfo/voltaged.pdf>. But how to solve it? I don't want to add some resistor on the input part to waste the energy. If I add a 2R resistor, it wastes nearly 2W already, and make the board hotter. If I add a NTC, will it be unable to protect the circuit if the user is switching the board off and on fastly after the board is warmed up.
3. Someone told me that a varistor or TVS may help to avoid input side's over-voltage at transient. But when a high voltage cause the TVS/VAR to have low impedance, after sometime, it will be burnt, then what's the general failure mode for a burnt TVS/Varistor? Short circuit? or open circuit? If it's a open circuit, it may make the capacitor to be burnt at consumer side next time.

Anyone have any opinion on my question and 3 methods? What's the general way to avoid those problems in circuit design?
:)

 

[vk6kro]

Hi,

You don't seem to be describing a normal buck converter.
See this:
http://en.wikipedia.org/wiki/Buck_converter
is this what you have?

The switch is a transistor turning on and off, rapidly.
The shunt diode is a special one for high frequency work.
A normal rectifier diode will get very hot and may self destruct in this circuit.
Computer power supplies have one of these diodes on a heatsink and it is worth scavenging it from dud power supplies.

The coil is quite critical and must be designed carefully.

The capacitor should only charge to the output voltage (4 volts) so it won't be voltage that destroys it, but the currents in and out of the capacitor will be very large.
You are producing 4 volts from 24 so the capacitor is being charged only 1/6th of the time.

So, it has to charge at a fierce rate while it can. You may be able to look for a capacitor with good current ratings. A normal Tantalum will possibly not be good enough.
Find the websites of capacitor makers and try to get information on current or on Equivalent Series Resistance ie ESR.

 

[victorb]

Hi vk6kro, thank you very much for your reply. I mean a normal buck converter that same as your link, but I have a capacitor at input side and I have a feedback loop that done by the switching IC. Also I have more components between the input voltage and the buck converter for other purpose.

I'm using LM3102 (a simple buck IC with PWM and internal switch, and it built in the switching diode too).
Those fuse (for overcurrent protection), polarity diode (for polarity protection) and common mode choke (for filtering) are not part of the buck converter, but it's in my circuit placed between the 24Vin and the buck converter.

And my problem is: Occasionally, the input capacitor got burnt when power on the 24V PSU (I'm using 35V capacitor already and tant capacitor). And I can't get to know where is the problem.

 

[berkeman]

Hi vk6kro, thank you very much for your reply. I mean a normal buck converter that same as your link, but I have a capacitor at input side and I have a feedback loop that done by the switching IC. Also I have more components between the input voltage and the buck converter for other purpose.

I'm using LM3102 (a simple buck IC with PWM and internal switch, and it built in the switching diode too).
Those fuse (for overcurrent protection), polarity diode (for polarity protection) and common mode choke (for filtering) are not part of the buck converter, but it's in my circuit placed between the 24Vin and the buck converter.

And my problem is: Occasionally, the input capacitor got burnt when power on the 24V PSU (I'm using 35V capacitor already and tant capacitor). And I can't get to know where is the problem.

Very strange problem. Connect your oscilloscope to the input capacitor, and watch what all goes on for that cap during power-up and normal operation. Re-check the inrush characteristics of the cap and your circuit. Re-check the ripple current specs on the cap and what you expect from your circuit.

And maybe post your full circuit diagram here for us to check out. What "extra" stuff are you connecting?

 

[victorb]

hi berkeman, thanks for your reply.
I also considered will that be too high ripple current for the input capacitor, but I checked the input capacitor's ripple current rating is 0.9A @ 100kHz, 85degC, where my converter's output current is 1.2A. I think it should be fine... And 1 more point is, the burnt capacitor was burnt before my loading start to load the current (More detail information, it's a MCU that control the on/off of the loading, I have to program the MCU after I assembled the PCBA. As I see, the loading is off when the MCU is blank. And the capacitor burnt before I program the MCU).

Normally when I made 100pcs of PCBA, there's about 3 of the PCBA will have a capacitor burnt.

here is my schematic:
24V power in through a connector circuit, then it goes to 4 converter circuit (3 of them are using LM3102, 1 of them is using LM25007

Connector circuit
PhysicsIn.JPG
http://totao.homelinux.com/phpbb/download/file.php?id=3650


LM3102 Circuit
Physics3102.JPG
http://totao.homelinux.com/phpbb/download/file.php?id=3651
3.3V:
CIN: AVX TAJ series, 47uF 35V, and in parallel with some low capacitance (220pF, 100nF) MLCC
RON: 100kR (so switching freq is around 254kHz)
CVCC: 1uF 16V MLCC
CSS: 1nF 50V MLCC
CBST: 33nF 50V MLCC
L: 10uH 5A (for some reason, I have to use such low inductance, but I can allow the abit larger output voltage ripple)
Cout: AVX TPS series, 100uF 10V, in parllel with 47uF 10V (also TPS series) and in parallel with some low capacitance (100pF, 220pF, 1nF, 10nF, 100nF) MLCC
RFB1 and RFB2 are selected for 3.3V output (also for some reason, I didn't put in the capacitor in parallel with RFB1)
Also I connected a diode from SW to ground to futher increase its efficiency, I also have a 100pF MLCC in parllel with it for lesser radiation at EMC (unfortunately it reduced the converter's efficiency)

4.3VA:
CIN: AVX TAJ series, 47uF 35V, and in parallel with some low capacitance (220pF, 100nF) MLCC
RON: 120kR (so switching freq is around 276kHz)
CVCC: 1uF 16V MLCC
CSS: 1nF 50V MLCC
CBST: 33nF 50V MLCC
L: 10uH 5A (for some reason, I have to use such low inductance, but I can allow the abit larger output voltage ripple)
Cout: AVX TPS series, 100uF 10V, in parllel with 47uF 10V (also TPS series) and in parallel with some low capacitance (100pF, 220pF, 1nF, 10nF, 100nF) MLCC
RFB1 and RFB2 are selected for 4.3V output (also for some reason, I didn't put in the capacitor in parallel with RFB1)
Also I connected a diode from SW to ground to futher increase its efficiency, I also have a 100pF MLCC in parllel with it for lesser radiation at EMC (unfortunately it reduced the converter's efficiency)

4.3VB:
Same as 4.3VA

LM25007:
physic2500.JPG
http://totao.homelinux.com/phpbb/download/file.php?id=3653

And I found this from web:
http://www.avx.com/docs/techinfo/voltaged.pdf
Maybe it's the reason, that's why I named surge protection on the topic. if so, then how to avoid?
Or can anyone know what's the reason?

 

[berkeman]

Your diagrams did not appear to come through. Maybe try again? How were you trying to post them?

 

[victorb]

er... maybe the place that stored the pictures are not for all the public, anyway, I put them as attachments now