I am thinking about building a PSU - some questions

Shining Arcanine · Post by **Shining Arcanine** » Mon Jan 23, 2006 4:49 pm

I am thinking about building a PSU for my PC for fun but I have a few questions...

How on earth do you get greater than 81% efficiency (wikipedia says that 81% efficiency is the limit) with a full wave rectifer? I know it can be done after looking at my Antec Phantom's efficiency ratings, but I do not know how. I asked a physics teacher at school and he suggested using a second full wave rectifer. I am not an electrician and I do not quite understand how such a setup would work on paper. Does anyone know what he meant by using two full wave rectifiers?

Mr Evil · Post by **Mr Evil** » Mon Jan 23, 2006 7:50 pm

You're not thinking of a linear power supply are you? They aren't very efficient, so it will need to be huge and expensive. PSUs for computers are almost always switching power supplies, which can be far, far more efficient and compact.

Unfortunately they are also hard to design and build, and potentially dangerous for mains-powered ones.

pawstar · Post by **pawstar** » Mon Jan 23, 2006 11:29 pm

To do that, you need to use a synchronous rectifier circuitry which relies on mosfets rather than diodes.

BTW, why would you be attempting to make your own PSU with little background knowledge? You should be very careful about working with main s power, not to mention you may be violating (local?) regulations.

darthan · Post by **darthan** » Mon Jan 23, 2006 11:33 pm

Yeah, seriously, building your own multi-hundred watt PSU is something that falls under the heading of final project for a college level electrical engineering tutorial or something, not fun and certainly not easy. Unless you've taken some serious EE classes I really wouldn't try it if I were you.

larrymoencurly · Post by **larrymoencurly** » Tue Jan 24, 2006 12:41 pm

Here's a 250W reference design from On Semi that's at least 80% efficient:

http://www.onsemi.com/PowerSolutions/content.do?id=1331

The documentation is 35 pages, and near the end it warns that the design has a few bugs.

mathias · Post by **mathias** » Tue Jan 24, 2006 4:56 pm

Why not start by modifying a PSU? Maybe replace the capacitors, convert an ATX PSU to AT, or write a guide on how to on how to completely discharge a PSU. Or even replace the parts that affect efficiency.

larrymoencurly · Post by **larrymoencurly** » Wed Jan 25, 2006 5:28 pm

Here's a table I found in a book called Power Supply Cookbook, by Marty Brown:

Development time:

Linear regulator: 1 week
PWM switching regulator: 5-8 person-months*
Resonant transition or quasii-resonant regulator: 8-10 person months*

*author claims the shorter time if you use his book

I've also heard of switching regulators blowing out a lot during the development process.

Lack of talent is why I stick to just modding PSUs, such as by changing fans, adding EMI line filters (only the worst cheapos lack them), beefing them up by adding missing diodes or transistors (be sure any that work in a parallel pair match exactly so they share the load equally, even if it means replacing an existing part) -- lots of those on lower power Enermaxes, and adding overheat shutdown.

scalar · Post by **scalar** » Fri Mar 03, 2006 10:40 pm

Some notes on switching supplies..

They mainly work by slicing up the incoming power into small pieces, but safety is a very big deal to prevent the supply from self-destructing and potentially taking out all your connected devices along with it.

- They have a designed MINIMUM amperage draw, and will not work below that point. A supply without minimum-draw protection can destroy itself because it goes into very high-frequency making tiny tiny power slices, which can actually cause high-voltage spikes that blow up the supply and/or the devices being powered.

- They have a designed MAXIMUM amperage draw, which is essentially full-on power draw from the wall up to the load limit of the device. A supply without designed overload-protection can melt down and destroy itself, when then destroys the powered device if the meltdown shorts something out in the supply.

- They have DIRECT-SHORT detection capability, where they can tell if a load is way beyond the ability of the unit to supply stable power, and will immediately shut down to protect both themselves and whatever is shorted.

- The supply may contain three or more individual switching supplies which are all connected together in the end to supply DC power. This way one can increase or decrease output independent of the others. One supply is for the +12v DC, one supply is for the +5v DC, one supply for the +3.3v DC. The safety mechanisms are tied together across all of these separate switching supplies, so if one detects a short/overload/underload-condition, it shuts them all down together and holds them all off until the main power source is disconnected. You do NOT want to have a condition where the +12v drops out but the +5v and +3.3v continue to operate. (The CPU could continue to run with +5/+3.3... but all the 12v fans would stop...)

You can certainly design a power supply without any of these safety considerations. It might even manage to be more efficient without the protection circuits. However, it leads to a condition where only experts can use the power supply safely, and dumb mistakes on your part can destroy it and anything connected to it.

TomZ · Post by **TomZ** » Sat Mar 04, 2006 6:56 am

I've got a fair amount of experience in electronics, and for myself, I would not think it is worth the risk.

You need to think about the worst-case possible outcomes - electric shock or fire, and to realize that it might not just affect you, but it could injure family and friends (e.g., house fire). And weigh these against the benefits.

Just building the circuit and getting it to work is one thing - and you may find reference circuits, etc. to tell you how to do that. What is harder or impossible to find is information about how to make it safe. It needs to be up to UL/CSA/etc. levels of safety.

Sorry to be a wet blanket.