power supply smoothing on the cheap

[steveo]

I have a car pc based on a raspberry pi, runs on 5v @1a loaded

I built it around a small switch mode power supply that I poached from a dead gps, rated at 2 amps, its very efficient, low heat, and very clean output on the scope. similar to a cheap cellphone charger.

unfortunately it doesn't survive cranking sometimes. it produces output around 5v but the waveform is madly unstable and the car pc locks up

what would you do with minimal funds? this is for a low budget project that I want to be reproducable..

- build a linear mode supply instead? (don't want to due to heat/inefficient)
- try to put a shitload of caps on input? output? both? how big of caps?
- give up and buy a fancier switch mode supply that'll take a beating?

[dave w]

I have a car pc based on a raspberry pi, runs on 5v @1a loaded

I built it around a small switch mode power supply that I poached from a dead gps, rated at 2 amps, its very efficient, low heat, and very clean output on the scope. similar to a cheap cellphone charger.

unfortunately it doesn't survive cranking sometimes. it produces output around 5v but the waveform is madly unstable and the car pc locks up

what would you do with minimal funds? this is for a low budget project that I want to be reproducable..

- build a linear mode supply instead? (don't want to due to heat/inefficient)
- try to put a shitload of caps on input? output? both? how big of caps?
- give up and buy a fancier switch mode supply that'll take a beating?

I would regulate the input power with a LM317 ... see attached

The input filters caps could be a higher voltage rating to better suppress spikes.

dave w

[Six_Shooter]

I would regulate the input power with a LM317 ... see attached

The input filters caps could be a higher voltage rating to better suppress spikes.

dave w

That's what he wants to avoid. Besides If I were to go that route, I'd use a 7805, even though the rating would be pushing it a little bit, a higher capable transistor could be used to increase the current capability, while using the 7805 to control it.

Could you grab some video/screen shots of what happens before, during and after cranking to see the problem?

My first thought as an easy fix, is to increase the value of the capacitors at the input to help with the cranking transition. You can also use some diodes between the car's 12 feed and the caps to keep them from being drained while cranking, sustaining the storage in the caps, and helping the SMPS to get through the cranking. Off the top of my head a pair of 1N4007s (well pretty much any 1N4001 to 4007) in paralel on the input should work. The .6 volt drop from the diode really isn't a concern here, because you're cutting 12V down to 5V anyway.

Otherwise a different SMPS might be needed.

[RobertISaar]

diodes/caps on the input would work well to keep a somewhat more consistent supply and is the first step i would attempt if a linear supply is out of the question. i don't know how much capacitance could be added to the output before weird things start happening(like any overcurrent protection, if present, kicks in and shuts everything down). the pair of 1N400x in parallel Chris mentioned would handle the 2 amps the power supply can handle. they'll dissipate a combined 1.4 watts or so at 2 amps(assuming i'm thinking about this correctly), so not that much heat/loss.

how much capacitance is needed? no idea. 470uF/25V caps are cheap/easy to find, i'd start stacking them in parallel and add more as necessary. you could probably get away with 16V units, but the price/size difference is pretty minimal.

[Six_Shooter]

It's generally accepted that the WV (Working Voltage) rating of a capactitor should be double the expected voltage in the circuit, or more. The 25V is actually a little under that rating, but is usually more than enough in practice. I have seen problems arise with 16V caps used in some automotive electronics when used on the 12V rails. When they are used on 5V rails it is not an issue.

[steveo]

thanks for the tips guys, i'll give it a shot.. i guess you're right, just throw caps at it until it's good to go, and i never thought of more diodes. that's a good idea. the ones on the input of the supply itself right now are pretty wimpy lookin'.

i wouldn't say a linear supply is out of the question entirely, but my first attempt at a primitive homemade one resulted in enough draw where i might drain my battery overnight.

that's the thing, i dont power it down often. with the switch mode supply, when the processor and wireless adaptor are idle, takes so little input power that i can leave it on for days.

you want pictures? i dont think my tube scope would photograph well, or handle the rain;

so here's some windows paint pseudo...cillosc...whatever of my observations from memory.

kinda looks like the switchmode supply has some kind of input undervoltage protection to prevent an idle device draining your battery, and it's just doing a piss poor job of actually... switching off.