1
\$\begingroup\$

I have a power supply to power an 8W LED board and a Raspberry Pi.

enter image description here

I have used a 24VAC (12-0-12V) 2A transformer to feed the bridge. I have also used a LM317 to provide a voltage of approximately 15.8V for the LEDs. The output is also used by the LM2576 to provide the 5V for the RPi. The V+ socket is connected to the LED PCB through a switch and the return to ground is an unbroken connection. The Rpi is always ON. The LEDs consume 500mA at 15.8V and the RPi consumes a maximum of less than 1A at 5V (measured using an ammeter.)

When I power ON, with the switch connecting power to the LEDs (full load,) the circuit works without a problem. However, when I switch off the LEDs after power ON, soon after a few seconds the RPi also goes down. The 5V output is now 2.3V and not sufficient for the RPi. The V+ reads 2.8V to 3.3V. Even if I power on without the switch connecting power to the LEDs, the RPi soon shutdowns due to low voltage.

This is really puzzling for me. How can a power supply work fine at full load but fail at lesser loads? Has anyone had such issues before and if so how did you solve it?

\$\endgroup\$
13
  • \$\begingroup\$ I suspect that C2, 1000 µF may be involved. It shouldn't be there as the regulator should handle output fluctuations. A 100 nF would be more appropriate. However, I can't explain the latch-up on sudden reduction of load. \$\endgroup\$
    – Transistor
    Nov 11, 2017 at 22:12
  • \$\begingroup\$ Inductance in the LED load? Switching off an inductive load without a flyback diode could cause a transient. \$\endgroup\$ Nov 11, 2017 at 22:18
  • \$\begingroup\$ You said the LEDs are connected to the output from the linear regulator, but in your schematic you have it on the output from the LM2576. Which one is correct? \$\endgroup\$
    – awjlogan
    Nov 11, 2017 at 23:25
  • \$\begingroup\$ @Transistor C2 is usually advisable, although 1000 uF is probably excessive. The LM317 is stable without an output cap, but will be more stable with one and also will improve transient response. \$\endgroup\$
    – awjlogan
    Nov 11, 2017 at 23:28
  • 2
    \$\begingroup\$ i bet that the LEDs pull down the voltage that's fed to the 5v regulator. When you stop the load, the input voltage climbs, the difference turns to heat, and within a short time overheats the 5v regulator, which probably shuts itself down. is it really hot? \$\endgroup\$
    – dandavis
    Nov 12, 2017 at 6:49

3 Answers 3

1
\$\begingroup\$

Connect the buck converter output to C1, not to the output of LM317. This will reduce current through LM317 and reduce power wasted as head in LM317 too.

Also add ceramic decoupling caps on the input of the buck converter. Something like two 1-10µF MLCCs of appropriate voltage, placed and routed for low inductance. You also need at least one MLCC on the output, like 10µF 6.3V.

enter image description here

\$\endgroup\$
0
\$\begingroup\$

Your LM317 is regulating from 24V down to 16V, that is 8V drop. And you are drawing 1.5A. So regulator has to dissipate 12W of heat. You must use proper heatsink. The same applies to your buck convertor.

\$\endgroup\$
8
  • \$\begingroup\$ Nope, he's paralleling the transformer outputs, so only 17V down to 15.8 V. A bit low for regulation but doesn't account for the observed behaviour. \$\endgroup\$
    – awjlogan
    Nov 12, 2017 at 16:32
  • \$\begingroup\$ @awjlogan where did you take 17V from? You mean 12V * 1.41? Are you sure you are able to draw 1.5A @ 16V DC from 2A @ 12V AC? You want to draw 26W from 24W source? There is a problem in design then. Not mentioning that you need like 2V voltage drop. \$\endgroup\$ Nov 12, 2017 at 16:52
  • \$\begingroup\$ OP has stated that the RPi is "less than 1 A", but there's no exact figure so I don't think we're breaking energy conservation here :) The 17 V is rectified (as you said) but there's also the 1.4 V voltage drop in the bridge. Everything is very close to current/voltage/heat limits, so I think the system is not well designed. \$\endgroup\$
    – awjlogan
    Nov 12, 2017 at 16:57
  • 2
    \$\begingroup\$ @awjlogan "Nope, he's paralleling the transformer outputs ..." - Nowhere do I see the OP stating that fact, so you cannot assume that. It needs verification. \$\endgroup\$
    – marcelm
    Nov 13, 2017 at 10:48
  • \$\begingroup\$ @marcelm Agreed, but if it wasn't parallel it wouldn't work at all, given the power dissipation in the LM317. But yes, does need verification. \$\endgroup\$
    – awjlogan
    Nov 13, 2017 at 11:10
0
\$\begingroup\$

Put the lm317's output on a scope. That chip is known for being a little "ringy" under transient loads, and with that 1000 uF on the output it could be going unstable.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.