Wow the decoupler cap IS important!

[Oznog]

I have a project with a buck converter for driving high powered Luxeon LEDs without generating a crapload of heat.

The circuit would fail whenever a load was hooked up- in fact, it was sort of a latchup mode, looks like the pins just tristated. There is no feedback in there except for a voltage feedback pin which has a 2.2k resistor going to the pin and a 5.1V zener to clamp off any excessive voltage, it just shuts off, and removing the load won't make it work again, the power has to be cycled off. This is hard to understand because only the PWM pin goes to the circuit, and it's driving a MOSFET gate so it's electrically insulated. And the MOSFET gets its power from the 12V source, not the 5V reg. There's a tantalum cap on both input and output of the 5V reg. And I've seen this exact same shutdown problem with another PWM that never worked.

After trying various things for this inexplicable problem, I found that I hadn't put in the ceramic decoupler cap on the PIC's power pins themselves, it wasn't far from the reg's tantalum output cap so I didn't think it was a priority to put in there.

Weird- but that looks like the problem. Just thought I'd share that with you, after spending so long trying to figure it out I had to share it with somebody.

 

[Monkeyman87]

Just out of curiosity does it matter were you put the decoupling cap in the circuit? ie. Across the power supply for the whole project or just across the PIC pins? It is the same thing right?

 

[AndyWatson]

I've got a related question: do the caps have to be polarized tantalum caps, or can they be standard non-polarized ceramic caps?

 

[e]

the decoupling caps should be as close to the supply pins of the IC as possible and must be of low esr, not ones like electrolytics

 

[Oznog]

In this case of the PIC18F252, the VDD is only on the right side but board geometry prohibited putting the decoupler there. The location on the other side of the chip was still close enough.

I have to note that this decoupler is not really much closer to the chip than the tantalum at the 5V reg's output terminal, which has much more capacitance. Capacitance was not the issue, distance was not an issue, the tantalum just didn't have the ESR to do the job apparently.

Just buy a crapload of these. They're outrageously cheap, tiny, very effective, and have so many uses.
https://www.allelectronics.com/cgi-bin/category.cgi?category=140240&item=CM-474&type=store

 

[Oznog]

xxx