Best way to power 4 x Raspberry Pi

[RazziaDK]

I am thinking about building a small cluster with 4 Raspberry Pi (3 b+), but I am not sure how to power them.
I have looked at a few desktop chargers, but they claim to be able to provide 2.4A on some of the ports, and "up to 2.4A" depending on which device is connected. I am not convinced by this.

Would a 5v 10a power supply like this be better?


Do you have other suggestions? Prefable something that can be bought in Europe or cheap on ebay/aliexpress.

I would also like to hear, how you would connect the power to the Pi. Micro-USB or perhaps GPIO?

[Ernst]

I am thinking about building a small cluster with 4 Raspberry Pi (3 b+), but I am not sure how to power them.
I have looked at a few desktop chargers, but they claim to be able to provide 2.4A on some of the ports, and "up to 2.4A" depending on which device is connected. I am not convinced by this.

Would a 5v 10a power supply like this be better?

No - this power supply, according to the specs, supplies only 5.0V and does not have the recommended cables and usb plugs for the raspberry.

Do you have other suggestions? Prefable something that can be bought in Europe or cheap on ebay/aliexpress.

https://www.raspberrypi.org/documentati ... /README.md
https://www.raspberrypi.org/products/ra ... er-supply/

[rpdom]

Yes, that supply should be fine with suitable cables connected to the 0V and 5V pins of the GPIO headers.

[RazziaDK]

I am thinking about building a small cluster with 4 Raspberry Pi (3 b+), but I am not sure how to power them.
I have looked at a few desktop chargers, but they claim to be able to provide 2.4A on some of the ports, and "up to 2.4A" depending on which device is connected. I am not convinced by this.

Would a 5v 10a power supply like this be better?

No - this power supply, according to the specs, supplies only 5.0V and does not have the recommended cables and usb plugs for the raspberry.

Do you have other suggestions? Prefable something that can be bought in Europe or cheap on ebay/aliexpress.

https://www.raspberrypi.org/documentati ... /README.md
https://www.raspberrypi.org/products/ra ... er-supply/

I was kind of hoping to avoid having four seperate power supplies, which is why I am considering the mentioned 5v 10a supply. I am aware, that it doesn't have the correct cables, but that can be fixed. I am just not sure if is stable enough.

[Heater]

That supply would work fine.

If you are up to snipping the ends off some USB cables and connecting them to the supply. Or perhaps connecting to the 5v pins on the GPIO header.

However, with 10 amps available there is serious risk of damage and/or fire if you ever short something out.

But I use such things all the time.

[LTolledo]

check the data sheet for the minimum voltage at full load, voltage must not dip below 4.9v to be safe, and not go beyond 5.2v on light or no load.
if transient dip fall below 4.8v the undervoltage thunderbolt appears,
if transient rise increase above 5.2499v then you at risk of getting "fried RPis"

power your RPi using the microUSB port, using a cable rated AWG20/0.5mmsq (AWG18/0.75mmsq better) so you can still use RPi's on-board power protection circuit.

[davidcoton]

That supply should be OK, subject to good regulation between minimum load and full load.

As above, I would connect power to the microUSB sockets (or solder to the equivalent test points). Using the GPIO header for power input bypasses the on-board protection circuit, thus removing a safety factor. Whether the polyfuse can react fast enough to stop damage from a 10A supply under fault conditions is a different matter.

if transient dip fall below 4.8v the undervoltage thunderbolt appears,
if transient rise increase above 5.2499v then you at risk of getting "fried RPis"

undervoltage detection, IIRC, is at a nominal 4.63V.
The Pi is specced for power input of up to 5.25V. The protection circuit works at a nominal 6V (again IIRC). I would not expect damage to occur below 5.5V (but anyone can be unlucky). Usually the protection circuit will prevent damage from voltage over 6V even up to 12V, sometimes by sacrificing the TVS diode. What happens with a sustained (not transient) 5.9V is not something I want to find out.

Off topic, remember the Pi0 does not have any protection.

[Imperf3kt]

I'd probably go a bit higher than 10A so you aren't running it at its limit at any point.
12 or 15 amp should do.

[timrowledge]

My main Pi system has been happily running off a unit like that for over year. I thin k it is a 6A version and IIRC it cost C$6 delivered. I hooked it up to a ‘scope and a load to pull 5A and the voltage changed by perhaps as much as 2 hundredths of a volt. Pretty good.
It powers the pi and a small SSD via a geek worm board. Oh and a tiny fan.
Going for a 12 - 15A version might be smart for four pi’s. Remember to allow some cooling airflow for the psu!

[rpdom]

My main Pi system has been happily running off a unit like that for over year. I thin k it is a 6A version and IIRC it cost C$6 delivered. I hooked it up to a ‘scope and a load to pull 5A and the voltage changed by perhaps as much as 2 hundredths of a volt. Pretty good.

I've got a Pi Zero running off an 8A version. It's been running happily for about 2 years now. It also powers a WS281x LED strip, which is very rarely on.

[laurent]

Hi,
I would rather use a 12V power supply with cheap DC-DC converters for each Pi.
Such rating currents at low voltages are never a good idea to deal with.

[Imperf3kt]

Hi,
I would rather use a 12V power supply with cheap DC-DC converters for each Pi.
Such rating currents at low voltages are never a good idea to deal with.

Why though? These devices are exactly the same thing, a 240/110 volt AC to DC regulator.
Why step to 12v and then to 5v when you can just drop to 5v directly and be more efficient?

[Heater]

Because shorting out a supply that can deliver 10 amps is a lot more dangerous than shorting out one that can only deliver 2. A much bigger risk of melting wires and starting fires. Ever tried putting a hook up lead across a car battery?

Ideally the supply would deliver the current your application needs but not much more. That is why we use fuses and other protection circuits.

Arguably using a 12v supply and separate 5v regulators provides that isolation and protection.

Delivering high power at low voltage requires higher current than when using high voltage. That means that voltage drops down cables due to their resistance will be higher. You lose regulation and efficiency. So for long distances a higher voltage is definitely an advantage.

[davidcoton]

Delivering high power at low voltage requires higher current than when using high voltage. That means that voltage drops down cables due to their resistance will be higher. You lose regulation and efficiency. So for long distances a higher voltage is definitely an advantage.

Just to elaborate, losses are due to the square of the current. So double the voltage and the loss reduces by a factor of 4. That's why out mains electricity is distributed at up to 400kV -- the increased complexity of managing and converting higher voltages is offset by the reduced losses.

If you stick to 5V, run thick wires for each Pi back to the PSU. Using 12V, place the 12V to 5V convertors as close to the Pis as possible.
TBH, powering a cluster where all Pis are close to the PSU will work with a 5V PSU just as well as with a 12V/5V system, and will cost less.

[drgeoff]

I'd probably go a bit higher than 10A so you aren't running it at its limit at any point.
12 or 15 amp should do.

Unless every one of the 4 RPis will have power hungry USB peripherals or there is significant additional load directly on the PSU then 10 Amps is perfectly adequate.

[Imperf3kt]

Because shorting out a supply that can deliver 10 amps is a lot more dangerous than shorting out one that can only deliver 2. A much bigger risk of melting wires and starting fires. Ever tried putting a hook up lead across a car battery?

Isn't the power supply still connected to a 13A capable mains voltage supply? The risk is still the same either way.
And yes, I know you can weld spanners and stuff if you connect it directly across a lead acid battery's terminals.

[davidcoton]

Isn't the power supply still connected to a 13A capable mains voltage supply? The risk is still the same either way.

No, the power supply has very little risk of allowing the full mains current on its output. Current (even under fault conditions) is limited by the design of the PSU. Other things being equal (YMMV), shorting a PSU with 10A capacity will do more damage more quickly than shorting a supply with 2A capacity. Don't try this at home.

[RazziaDK]

Hi,
I would rather use a 12V power supply with cheap DC-DC converters for each Pi.
Such rating currents at low voltages are never a good idea to deal with.

How do I know how many amps a 12v PSU need to be, if I convert it to 5v?
I have a 12V 5A somewhere around here that I don't use - would that be enough?

Can you recommend a 12v-5v circuit?
I found a couple of options:
With USB
Another option with USB
LM2596

[achrn]

How do I know how many amps a 12v PSU need to be, if I convert it to 5v?
I have a 12V 5A somewhere around here that I don't use - would that be enough?

Can you recommend a 12v-5v circuit?

Probably 12V 5A will be enough, yes, though it does somewhat depend on what circuit you use to step it down.

Personally, I use Hobbywing 3A UBEC when I need to do that. It has better performance as the official plug-in PSU (up to 3A, anyway) - http://www.astounding.org.uk/ian/raspi- ... _test.html - mainly because you can put it really close to the Pi.

But actually, I wouldn't do that here, I'd just use a suitable 5V supply. I disagree that shorting a 5V 50W power supply is "a lot more dangerous" than shorting a 12V 60W supply (you'll need the extra power for the extra losses). Neither one will be cataclysmic - it's not like one will convert your house to a crater and the other will not. Most modern supplies protect themselves when shorted anyway. If you're sufficiently cac-handed that you think it's a likely outcome, make sure you get a supply that does.

Likewise, assuming the Pis are within a couple of meters of the supply, the hand-wringing about line losses are pretty irrelevant - use decent wires.

[davidcoton]

I disagree that shorting a 5V 50W power supply is "a lot more dangerous" than shorting a 12V 60W supply...

Remember that under short circuit conditions the voltage dissappears, you are left with just the current. In the absece of foldback protection (not necessarily the case) a 12V 60W supply will produce around 5A into a short, while a 5V 50W supply will produce 10A. Clearly the higher current will produce more heating and a greater fire risk. Possibly mitigated by the thicker wire needed for the higher normal current, but that's another assumption that won't always be true.

Unless the Pis are some distance from the supply, there is no point in the extra complication of dual voltage conversion.

[RazziaDK]

Likewise, assuming the Pis are within a couple of meters of the supply, the hand-wringing about line losses are pretty irrelevant - use decent wires.

The Raspberry Pi's will basically be stacked on top of the PSU, in a configuration similar to this one:
https://www.thingiverse.com/thing:1897376

[achrn]

I disagree that shorting a 5V 50W power supply is "a lot more dangerous" than shorting a 12V 60W supply...

Remember that under short circuit conditions the voltage dissappears, you are left with just the current. In the absece of foldback protection (not necessarily the case) a 12V 60W supply will produce around 5A into a short, while a 5V 50W supply will produce 10A. Clearly the higher current will produce more heating and a greater fire risk. Possibly mitigated by the thicker wire needed for the higher normal current, but that's another assumption that won't always be true.

Unless the Pis are some distance from the supply, there is no point in the extra complication of dual voltage conversion.

Yes, somewhat higher current. Possibly. Assuming the supply doesn't have protection that prevents it (as I said). It's not going to trigger a thermonuclear fireball. It will be low tens of watts at most. One will have wires that get really hot and might start a fire. The other will have wires that get really hot and might start a fire. Hardly "much more dangerous".

[Zebu]

as an alternative suggestion for an off the shelf component the cheapest corsair (just for example, any brand would be about the same) ATX computer power supply i could find listed has 20 amps on the 5 volt rail, has connectors on the mains side so you don't have to wire up anything running high voltage and includes by default...

"Over-voltage protection, under-voltage protection, short circuit protection, over power protection, and over temperature protection provide maximum safety to your critical system components"

[ejolson]

as an alternative suggestion for an off the shelf component the cheapest corsair (just for example, any brand would be about the same) ATX computer power supply i could find listed has 20 amps on the 5 volt rail, has connectors on the mains side so you don't have to wire up anything running high voltage and includes by default...

When using 5V from an ATX power supply, do you need to wire up dummy loads to the 12V, 3.3V, -12V and -5V rails to keep it happy?

[Imperf3kt]

as an alternative suggestion for an off the shelf component the cheapest corsair (just for example, any brand would be about the same) ATX computer power supply i could find listed has 20 amps on the 5 volt rail, has connectors on the mains side so you don't have to wire up anything running high voltage and includes by default...

When using 5V from an ATX power supply, do you need to wire up dummy loads to the 12V, 3.3V, -12V and -5V rails to keep it happy?

Possibly.
But there's always the 5vsb rail, which should be good for about 3 to 5 Amps without needing to turn anything else on.