DC-DC Switching Regulators and Large Capacitive Loads.

[Bob McCloy]

Hi, I'm designing a control board (PCB) that will be used to power and control strips of Adafruit NeoPixels for various future projects. The board will be powered from a 24V supply and this needs splitting down to three 5V, 1.5 - 2A supplies that will power strips of NeoPixels. In the design I'm currently using the TracoPower TSR 2-2450, DC-DC Switching Regulator to step down the voltage.

https://uk.rs-online.com/web/p/switching-regulators/9068487/

However, on the 'Best Practices' and 'Powering NeoPixels' pages on the Adafruit website...:

https://learn.adafruit.com/adafruit-neopixel-uberguide/best-practices
https://learn.adafruit.com/adafruit-neopixel-uberguide/powering-neopixels

...It suggests using a 1000uF Capacitor near to the start of the NeoPixel strip to "prevent the initial onrush of current from damaging the pixels". However, the datasheet for the TSR 2-2450 has a Max Capacitive Load of 600 μF. I have looked for alternate switching regulators but I'm struggling to find ones that either state a max capacitive load or if they do it is too small. In my mind, it sort of feels odd to me to try to prevent the initial onrush on current into the NeoPixels by creating an initial onrush of current into a large capacitor.

My question is:

Is there a smarter circuit design that could mitigate this issue or another switching regulator that someone knows of that could handle this large capacitance?
Or any other suggestions for a better design to look into?

I have attached an image of the current schematic.

Thanks in advance!

[MagicSmoker]

...
However, on the 'Best Practices' and 'Powering NeoPixels' pages on the Adafruit website...:
...
...It suggests using a 1000uF Capacitor near to the start of the NeoPixel strip to "prevent the initial onrush of current from damaging the pixels".

Yeah, that does seem counterintuitive. If the goal is to reduce inrush current then series inductance would be needed, not shunt capacitance, but that would aggravate voltage overshoot caused by ringing. In fact, I suspect voltage overshoot is the real worry, especially if ceramic bypass capacitors are used in the neopixel thingies, as that would form a high-Q resonant network (this issue has plagued many a laptop, btw)

However, the datasheet for the TSR 2-2450 has a Max Capacitive Load of 600 μF. I have looked for alternate switching regulators but I'm struggling to find ones that either state a max capacitive load or if they do it is too small.

If the switching regulator has pulse-by-pulse current limiting then any amount of capacitance will be fine, it will just take longer for it to come up to nominal voltage when power is applied. However, cheap, lower power (say, <10W) dc/dc converters - especially self-oscillating types - might expire if loaded down with too much output capacitance because they will effectively be overloaded for much of the time it takes to charge the capacitance up. Unfortunately, unless the manufacturer of the dc/dc converter provides the schematic and/or specifies a maximum output capacitance you can't really tell if it will work reliably.

My suggestion is to either stick with the maximum capacitance recommended by Traco, or insert a bit of resistance in series with the capacitor so it turns into a parallel damper (something in the range of 1-10 ohms should work - the exact value depends on the loop inductance of the supply wiring and the amount of bypass capacitance in the neopixels, neither of which might be practical to determine).

[Bob McCloy]

Unfortunately, unless the manufacturer of the dc/dc converter provides the schematic and/or specifies a maximum output capacitance you can't really tell if it will work reliably.

Yeah, that's an issue I was thinking about, because if I ordered the component and it worked initially I couldn't be certain of its reliability.

My suggestion is to either stick with the maximum capacitance recommended by Traco, or insert a bit of resistance in series with the capacitor so it turns into a parallel damper (something in the range of 1-10 ohms should work.

Okay, I'll look at doing this. I had noticed that a few other switching regulators datasheets stated they could take a larger capacitive load depending on the capacitors ESR. So is this essentially just adding an artificially larger ESR to increase the charging time? Would this have any negative effects on the capacitors function with the NeoPixels?

Thanks for the help!

[MagicSmoker]

...I had noticed that a few other switching regulators datasheets stated they could take a larger capacitive load depending on the capacitors ESR. So is this essentially just adding an artificially larger ESR to increase the charging time? Would this have any negative effects on the capacitors function with the NeoPixels?

Yes, adding resistance in series with a capacitor effectively increases its ESR - the only difference is that the ESR of a capacitor is internal to it and therefore causes heating proportional to ripple current, whereas external resistance obviously won't heat up the inside of a capacitor. High ESR reduces the ability of a capacitor to act as a capacitor (e.g., smooth out ripple) so don't go crazy with it. For this case 1 ohm will likely work well. Note, however, that this resistor should NOT have a high stray inductance (otherwise you are just making another RLC network), so don't use a wirewound. Metal oxide through hole and thick or thin film SMD would work best.

[David Hess]

If the goal is to reduce inrush current then series inductance would be needed, not shunt capacitance, but that would aggravate voltage overshoot caused by ringing. In fact, I suspect voltage overshoot is the real worry, especially if ceramic bypass capacitors are used in the neopixel thingies, as that would form a high-Q resonant network (this issue has plagued many a laptop, btw)

Ringing from a long lead length or leakage inductance when the input is switched on hard is exactly the problem they are worried about.  This should never be a problem with the output of a properly designed regulator when the recommended bulk output capacitance is used.  Typically 50 to  100 microfarads of electrolytic capacitance per amp *close to the regulator's output* is sufficient.