Discussion: Power Supplies & Electricity

[TechyInAZ]

Hello Community!

This thread is dedicated SOLELY to the discussion of power supplies and how electricity works with PSUs.


Rules:

1. DO NOT involve yourself in a flame war.
2. Keep everything you discuss civil. (If you do not, you're at high risk of a temporary ban.)

Thank you!

 

[Inkiad]

Anyone has any suggestion to get oneself educated in the inner workings of how power supplies actually work? I have basic background on electricity.

 

[Onus]

It's getting a little dated due to some newer platforms, but start with http://www.hardwaresecrets.com/anatomy-of-switching-power-supplies/

 

[Inkiad]

^Already read that, including the psu article from techpowerup.

I am more interested in how switching is made, power is regulated, power packets added together again after dc conversion, these kind of inner workings.

 

[turkey3_scratch]

The Hardware secrets article is where I started. Kitguru has a pretty good article. Some of the articles, while they say they are for beginners, really have far too complex topics (like the PSU 101 on Toms). The Hardwaresecrets one is great because Torres is a phenomenal writer, but only scratches the surface.

Switching is done through switchers. Usually comes after the bridge rectifier and they are bolted to heat sinks.

I barely know anything of the inner workings. It's too complex for somebody who hasn't been doing this stuff for decades. Even people who do know the inner workings of a PSU probably don't know the inner workings of the inner workings. Instead of what the parts do, how the parts work. How the parts that make up the parts work. How the electrons and protons work in those. String theory.

You have to decide for yourself how deep to dig into it, because once you try digging as far as me you are left with more questions than answers. But I'd rather have the ability to question things than not know at all.

It's also important to note the difference in a physics and EE's perspective on things. EE is about what stuff does, physics is about how it works.

 

[turkey3_scratch]

You can't go by brand, that's incorrect. Secondly, anything can fail; resistors. Poor soldering can cause a short. Fan. Tons of stuff. There's a lot more to efficiency than MOSFETs and inductors. LLC resonant, shorter wires, relays, tons of stuff. You are also wrong in how long they last is the only thing that sets them apart; their performance sets them apart, too.

 

[TJ Hooker]

^+1

 

[Someone Somewhere]

Interesting doc from Intel on power management. It's based on Sandy Bridge, but certainly applies to other chips.

Most telling is slide 14. Thermal design power is just that; thermal. Electrical can be significantly (25%+) higher for up to perhaps a minute, because TDP is steady-state and a cold heatsink can absorb much more energy until it gets up to operating temperature. PSUs can do the same, but not on the same scale.

TL;DR: Don't use TDP as the maximum power consumption for a chip. It's not. It's the maximum medium-term average consumption.

 

[turkey3_scratch]

Interesting doc from Intel on power management. It's based on Sandy Bridge, but certainly applies to other chips.

Most telling is slide 14. Thermal design power is just that; thermal. Electrical can be significantly (25%+) higher for up to perhaps a minute, because TDP is steady-state and a cold heatsink can absorb much more energy until it gets up to operating temperature. PSUs can do the same, but not on the same scale.

TL;DR: Don't use TDP as the maximum power consumption for a chip. It's not. It's the maximum medium-term average consumption.

Of course, but average is still a very important value. If we are to take into account every little spike then we'll get freaked out since even the GTX 1080 (MSI aftermarket one) will spike to 400W for about 1ms every 50 seconds.

 

[Someone Somewhere]

You're talking about 1ms spikes.

I'm talking about sixty second spikes.

One can easily be fed from capacitors.

One cannot.

 

[turkey3_scratch]

Ya but who is using TDP for these measurements? We have plenty of reviews that show the power of these cards, I surely don't look at TDP vs the reviews. Also, pictorally a 60 second spike would be a cliff

 

[Someone Somewhere]

I'm pretty sure you've used TDP to say that a 450W PSU is fine several times.

Also, pictorally a 60 second spike would be a cliff

??

 

[turkey3_scratch]

I'm pretty sure you've used TDP to say that a 450W PSU is fine several times.

Also, pictorally a 60 second spike would be a cliff

??

I use data from reviews. I've looked at reviews for CPUs and graphics cards from many websites. And to be fair, those times when i say a 450W PSU is fine, I am still correct. It's gotta be a good one though no cheap junk of course. But that's what I've learned on the Jonnyguru forums. Gained a lot of knowledge from there, it's a great forum just like this one.

Spike = mountain. But if it lasts for 60 seconds at the pinnacle, then it's a cliff. No, a mesa. That's the word. A mesa of power.

 

[SexySally]

No one seems to have addressed the issue of 'how does it work', which seems to be the theme of this thread?
Specific designs in PSUs are all different in detail, but similar in essence. The idea is to take -say- 240V AC and change it to -12, +5 and +12V DC with some efficiency as a byproduct. PCs use a technology called 'Switched Mode' -hence SMPSU- to produce these lower voltages. SM runs at a high frequency, so inductors are smaller and lighter than in an analogue PSU, and as the power semiconductor is only switching on and off, the heat dissipated is lower as well. That's the what. Now for the how.
Incoming AC is bridge-rectified and fed to the reservoir capacitor, producing about 380V DC. That's why you don't poke around inside one without knowing what you're doing. This DC is then chopped up by a power semiconductor and fed into the primary winding of a ferrite transformer. Other secondary windings pick off the required output voltages as AC which is then rectified, smoothed and output to the PC. One of these, usually the +5, is monitored by the circuitry controlling the power transistor so as to vary the power applied to get the +5.0 exactly. Depending upon the design, the other output voltages might have their own regulators, or just take their chances that if the PSU is loaded as the spec on the box, they will be correct. You get what you pay for.
There are two other small points: the power control circuit is often fed from the same transformer as the outputs, so it needs a kick to start the process. If the PSU has a fault or is shorted, the kick won't be enough to start the control circuit, so it protects the PSU. If the outputs are shorted in use, the excessive requirement is detected by another small circuit, and it will switch the supply off until the input power is cycled. Modern PSUs also include standby voltages, remote on switching and 3.3V rails - but these are all basically covered in the thumbnail description above.
BTW, if you do try to replace any internal parts, watch out for that 380V - it's very nasty, and ensure that any capacitors are low ESR types for SM PSUs. Have fun!

 

[aloff10]

On a power supply the 700W and the 550W are only the maximum power that the PSU can run at and not the amount of electricity that it uses although the 700W PSU can use more electricity(at times) when the box is under load, but to answer your question it won't save electricity.

Also make sure that the computer will work under a 550W PSU because it may require more than 550W in order to run properly(although I highly doubt that will be an issue with those types of components) and just to be safe they have some websites that calculate the amount of wattage required by the PSU by entering your PC's parts.