Ceramic capacitors have this unfortunate property which can cause spikes in voltage when live batteries are connected to boards (hot-plugging), due to the inductance of the wires to the batteries and the very low inductance of the capacitors. It can create spikes of more than double the supply voltage.
On my first board which used entirely ceramic capacitors on the input, I read up on this in detail in the LT1933 datasheet (See page 14) and the recommendation was to use a 1 ohm series resistor, in a 1206 package, because it may dissipate up to 0.2W, and a 0.1µF cap directly on the input to reduce input ripple. However, I am now working on a project which has a 2A buck converter. That poor 1 ohm resistor, at a low input voltage, could be dissipating 4 watts! That is way too much.
The datasheet also recommends using an aluminium electrolytic capacitor, I'm currently using this. However, it is pretty big and ideally, I would like to eliminate it from my board. Also, being an electrolytic makes it sensitive to environmental conditions a lot more than ceramics, and it reduces the lifespan of the device.
So, what are my options for limiting hot-plug spikes?
Answer
You can consider an active hot-plug controller which solves the spike problem by controlling the charge-up of the input caps.
You can also consider a passive NTC inrush limiter which remains in the circuit all the time. It will dissipate power constantly like the 1206 resistor in your example, but if chosen appropriately will dissipate much less power (it will be high resistance during startup, and reduce during steady-state operation.
Another technique used when connectors are involved is a precharge pin. The connector is designed such that the first power pins that make contact have a resistance in series, to "slowly" charge the caps. When the rest of the power pins make contact, the precharge resistance is shorted out and doesn't burn any power.
No comments:
Post a Comment