Minimum Load for ATX Power Supply: Powering Hard Disk & Raspberry Pi, Voltage Reduction, Stability

Hello.

I have some doubts about ATX computer power supplies, they are constantly searching for some contradictory or incomplete answers and I hope you will be able to dispel them. Well, I'm doing such a small project and need to power up the hard disk and raspberry pi. There would be some power supplies to do this, but it would be a bit fun to reduce the voltage of them to 12v and 5v required by the hard disk, so I decided that I prefer to use the atx power supply for this because they are cheap as borscht and I really like it fits the use of one.

However, I met with opinions that the 5v or 12v lines should be charged with resistors (depending on which provides more A) so that the power supply is stable and only with such artificial load can it be used. As I used to build a laboratory power supply from an ATX power supply, I followed this advice, and honestly speaking, I see it on average now, because these resistors heat up terribly well, and energy is also wasted, and by default this power supply is supposed to run 24/7 .

Also, hence the question, do you think it is safe for the ATX power supply to run 24/7 with a hard disk load + raspberry pi?

What some folly write on forums. ATX power supplies, especially those of an older generation, have a built-in resistor that loads 5V lines and this is enough for such a power supply to work properly without connected devices. In your case you will have a hard disk connected so I don't see the problem.
I do not know exactly how it is in power supplies of the current generation, but looking at their design and the amount of electronics, such a resistor can not be seen, there are also no reliable descriptions as to the requirements of idle load. And also EU regulations do not exhaust or specify the subject.

Check that the voltages at low load are correct. Check (e.g. with an oscilloscope - but it's better to assemble a system for it, because the assessment on the oscilloscope screen is not sure), what is the ripple / noise / noise at low load - it is specified in the standard, what is the permissible peak-to-peak voltage. If this is a new power supply, it may be information in the manual, but you should not rely on the old one without checking, because it usually has deteriorated parameters and it is not known to what extent.

You can check but I doubt that a beginner will have access to the oscilloscope and be able to properly perform and interpret measurements of noise or current peaks. As for the instructions or the parameters of the manufacturer of a given ATX power supply model, it is currently not even worth reviewing, because the reliability of the parameters with measurements has as much in common as the value of the mains voltage in the center of Warsaw. I would prefer to rely on good old power supplies of the 1st, 2nd or possibly 3rd generation ATX power supplies from the 90s. The only condition is the replacement of all electrolytic capacitors, good quality with low ESR. Personally, I stay away from current generation power supplies. All, unfortunately, were "taken off" in the PRC, but as I saw in the movie how they wind the transformers and what dimensions they are and the quality of the capacitors, I prefer to repair old generation power supplies. I generally want to build a power supply from scratch.

Thanks so much for your reply. As I thought, the topic is not so obvious. Unfortunately, I do not have access to the oscilloscope, and even if I had it, I am almost a complete newbie in the subject of electronics and I would not know how to conduct tests. My power supply was also "taken off" ( ) in the PRC and it's about 10 years old 350w from FSP, during his work in the PC he was quite snappy so I hope that he won't let you down now, I will only check the voltages with a multimeter if there are any circuses to have at least a cursory look at him.

Maybe I could build a power supply from scratch and even work, but with my experience, I wouldn't have the courage to leave it turned on when I'm not at home.

I will leave the topic open in case anyone else would like to speak, I will follow him.

You can do such a simple system - you only need to have a voltage slightly higher than the tested output of the power supply: resistor, diode and capacitor. A capacitor is charged through the resistor, diode anode to the capacitor, cathode to the checked output of the power supply. And another similar, without power supply: diode anode for this output, cathode for the capacitor, and a resistor in parallel with it. Diodes rather fast, e.g. 1N4148. Capacitors and resistors selected so that the charging / discharging of the capacitor is below 1V / s. You connect a standard meter between capacitors, measure the voltage, and then disconnect the diodes from the power supply output (they are still to be connected to each other). The voltage indicated by the meter will increase slightly. If? If a lot, the power supply is not suitable for electronics. As far as I remember, for a good power supply there should be up to 50mV voltage change for the + 5V output, and 120mV for the + 12V output.

Regardless, you can open the power supply and look at the capacitors, if they are not swollen - if you find any swollen, then this and all similar (same company, similar parameters) to be replaced, because as the company made swelling capacitors, it is on a massive scale. There should be Low-ESR capacitors there, others do not work long.

If you're not sure of the power supply, the Raspberry Pi can be powered via the LM7805 - but as Pi draws 1A, 7W will turn into heat in this LM.

_jta_ wrote:

Diodes rather fast, e.g. 1N4148.

And according to a friend this is a "fast" diode? To the switching power supply? It has as much in common with speed as pendolino in our country. Such a measurement using a regular diode will be affected by a large error. The Schottky diode and good capacitors will be better, and even better the fast meter with medium voltage measurement mode.

_jta_ wrote:

As far as I remember, for a good power supply there should be up to 50mV voltage change for the + 5V output, and 120mV for the + 12V output.

And this is interesting because, in principle, there are currently no specific standards adopted and still the PRC manufacturers do not comply with them because the voltage spikes in current computer power supplies are sometimes 160mV and more. I have met those that have even 720mV (record) checked in several new copies. SiC! In 1.2 and 3rd generation power supplies, they did not exceed 10mV / 100ms, i.e. the so-called clock time of the serial bus frame (90s). The current standard for industrial power supplies for the worst case. To sum up, in general, each manufacturer applies its own standards for a given application.

Stable 5V voltage for processors is needed when we use an internal generator, because the frequency of its operation depends strictly on the voltage value. However, when we use quartz, the voltage can flow as much as the manufacturer of the given processor allows.

Added after 15 [minutes]:

_jta_ wrote:

look at the capacitors if they are swollen - if you find one swollen

They do not have to be swollen at all, they just need to be overheated and you can tell by the shirts they have been shed. This is the first symptom that the electrolyte degrades. But to assess the condition of the capacitors, you need to desolder them and measure the capacity and series resistance. I usually replace preventive sensitive capacitors in power supplies without having to watch and measure them.

And according to a friend this is a "fast" diode? To the switching power supply?

Is 2ns, according to a colleague, too slow a diode for waveforms not containing frequencies above a few MHz?

And this is interesting because, in principle, there are currently no specific standards adopted

Is. 120mVpp on + 12V, 50mVpp on the others. See ATX 12V Powed Supply Design Guide, any version from 2.01.

Stable 5V voltage for processors is needed when we use an internal generator

Voltage stability is different and the presence of fast pulses is different - the latter cause data distortions. 50mVpp seems to be quite a stringent requirement, but I saw a power supply, which had a few times more, caused frequent hanging or resetting of the PC.

_jta_ wrote:

but I saw the power supply, which had a few times more, caused frequent hanging or resetting of the PC.

More likely capacitors in the motherboard inverter circuits were futile. It is possible that the pin impulses could affect the watchdog although I do not want to believe it because I met with the power supply which "lost" all capacitor capsules and walked stably. Nevertheless, cribs with graphic performance were happening. I will send you a photo

Replacing the power supply with a new one caused the computer to work stably. I found no damage to the capacitors on the motherboard.

Sorry, I'm cutting in
But I'm making a workshop power supply with atx.
My questions;
1. Can I use a 20W 16Ohm resistor as an artificial load?
2. How to make a short circuit protection, e.g. as a short-circuit to 12v or other voltage, so that it does not burn after a few weeks or days.
Unless atx already has such protection.
Please help.
He writes to moderators here because it's a shame to litter the forum with such a topic. And nowhere did I find the answer.
I am asking for help and thank you in advance.

ATX power supplies can be different - some need a load to operate, others do not; some have some protection (e.g. against short circuit), others do not.

As for the load, you can check - if it does not work without load, or, for example, it gives too much voltage, it means that the load is necessary for it.

Security information is sometimes provided in the marking. But an unreliable producer can give information, and not give security.

And can this resistor use 20W 16Ohm for artificial load?
And how to do it so securely?
And here is my power supply;
Qoltec silentline 400w.