Do you want to learn how to build your own power supplies from scratch?

[Th3_uN1Qu3]

If anyone's interested, i'll share my knowledge so far. I will address designing a SMPS from zero, taking you thru the basic building blocks and their interactions, and keeping theory to a minimum. It's your business if you want to learn the theory or not, in 90% of cases you won't need it.

This topic will be for aspiring engineers from an aspiring engineer, as there are some things that i do not truly understand at the moment of writing and i learn them as i build them myself. One such example is loop compensation, which is IMO the most complicated aspect in designing a regulated SMPS because it can make or break the whole device, and it is tricky to implement by calculations only because real-world parts have quite high tolerances.

I've put together a spreadsheet which does most of the math for you and also avoids the exponents found in most textbook equations - lose one zero and you mess up by an order of magnitude, or worse - the spreadsheet takes all the data exactly as it's given in manufacturer data sheets. Using it you will be able to make your own power supplies or upgrade existing ones - adding capacitors, rewinding transformers, using bigger inductors or heavier size wire. The spreadsheet also integrates several tables for convenience - a wire size table in both AWG and metric dimensions, with current handling and skin effect calculated for all wire sizes. There's also a frequency to time converter with duty cycle, a voltage divider calculator, a transistor peak current calculator and lots more goodies that make your troubleshooting easier.

Best of all, the design tables are all linked to each other. Modify the operating frequency or change the core, and the number of turns changes accordingly. This is just a small example of what the spreadsheet currently does - and for the most part i've just gotten started! So... what do you think?

[Khron]

Re: Do you want to learn how to build your own power supplies from scratch?

That's... enterprising, to say the least

Count me in!

PS: "Happy" (new year's) hangover

[PCBONEZ]

Re: Do you want to learn how to build your own power supplies from scratch?

At one time or another I learned all that's needed but as I never used it for anything much of it leaked out the ears and evaporated.

I would love such a read.
.

[Agent24]

Re: Do you want to learn how to build your own power supplies from scratch?

Sure, I always like to learn new things (or at least try, anyway)

[goodpsusearch]

Re: Do you want to learn how to build your own power supplies from scratch?

We are waiting.

[Th3_uN1Qu3]

Re: Do you want to learn how to build your own power supplies from scratch?

PS: "Happy" (new year's) hangover

Thanks. I've gotten started already.

Anyway back on topic. I've been trying to wrap my head around magnetic amplifiers, with not much luck. Probably due to the fact that i don't have the possibility to build one right now and observe how it works in real time.

But i see most new power supplies have done away with this approach and use a dc-dc converter approach to the 5v and 3.3v rails. That's most likely what i'm going to do as well.

After all is said and done, the intention is to eventually have my efforts result in a totally custom power supply for my dual-PIII. I've built a couple SMPS supplies until now, but nothing of the complexity that a PC power supply requires. Even with 2 PSUs in parallel the motherboard reports 5v a tad low (4.78v) so i will need an independently regulated design, because i don't think i break 3A of 12v usage on this thing, while the 5v and 3.3v rails are overloaded.

[Th3_uN1Qu3]

Re: Do you want to learn how to build your own power supplies from scratch?

I decided to satisfy my curiosity and bust out the TrueRMS amp meter. At maximum load (IntelBurnTest and Furmark together) she peaked at 1A drawn from the wall. With regular gaming (CS Source) she hovers around 0.93, and web browsing around 0.85 on a mains voltage of 225v.

So it's really not that much power - at 80% efficiency this means that the system itself only asks for 180W at most, and honestly i think that the efficiency of my PSUs hovers around the 75% mark, so that's 170W. And i need 2 PSUs to sustain this... i think it becomes obvious that my worst enemy here is group regulation. On the other hand, independently regulated units seem to be found at high powers only and would set me back too much $$ and leave me with a ton of unused 12v, so a custom job still seems like the best idea.

Now... 90% of 170W divided by 5 gives 30,6A, since the 3.3v rail is derived from the 5v one that's probably correct. Indeed, not something i'd trust my Xilence to handle on its own (and it doesn't). But not that difficult to achieve IMO. I think i'll mod a random power supply rather than build one from scratch. The primary part will stay mostly intact, but all other windings on the transformer will be removed and a fat 12v rail installed instead. Then dc-dc my way down from there. Reasoning is simple - i need a 40A 5v rail to be on the safe side, and that requires very thick wire (or a lot of strands of thin wire), which would be difficult to wind. A 25A 12v rail on the other hand will be easy.

[goodpsusearch]

Re: Do you want to learn how to build your own power supplies from scratch?

An Antec Smartpower SP 450W would do the job in my opinion... (40A @ 5V)

[ss627]

Re: Do you want to learn how to build your own power supplies from scratch?

You can start building http://www.onsemi.com/PowerSolutions/content.do?id=1331 or a 80 plus silver rated unit here http://www.onsemi.com/PowerSolutions...nt.do?id=16599.

I started designing the schematic/pcb a while ago but never finished due to calculating the costs. It would be cheaper if you scavenged some parts but I really doubt many people on here would be interested once they knew the total price of such a project. You can buy a 80 plus gold rated 750 watt seasonic for much cheaper.

[Th3_uN1Qu3]

Re: Do you want to learn how to build your own power supplies from scratch?

You can buy a 80 plus gold rated 750 watt seasonic for much cheaper.

Depending on how many parts you have lying around. For me the only cost will be the PCB, and since i etch that myself, $2 for a copper clad board really isn't that much. Besides that, i'll only have to buy resistors and 2 power transistors at $1 a piece, the standby one i already have.

I'm quite sure the Antec Smartpower wouldn't handle it either - simply because it's still a group regulated design and i'm barely using any 12v. Group regulated power supplies are built with the assumption that the rails will see equal usage (+/-20% in the worst case). While my dualie uses 80-90% 5v and 3.3v. I'm back to the initial plan of designing a PSU from scratch since the dc-dcs will need to be synced to the master (off-line) switching IC, and the typical 33kHz used in cheap halfbridge PSUs is too low for achieving high current in a reasonable size.

Besides, this topic isn't about building computer power supplies, my dual-PIII is an exception. It's about general SMPS for general use. Like a bench supply, or something lightweight for a power amplifier, or maybe even for a solar powered cellphone charger. I still have some quirks to fix in the spreadsheet but once those are done i'll post the first version up.

[Th3_uN1Qu3]

Re: Do you want to learn how to build your own power supplies from scratch?

First version up.

Notes: Please let me know if there are any locale problems (you know, with . and , ). For now it deals with 2 transistor forward only and there are some things to consider:

• The output cap calculator is based on a general formula for low-ESR caps, you cannot enter the ESR of your particular cap. Should be safe but i'll give it more attention later.
• The primary cap ripple calculator uses a cheat. It's pretty accurate though.
• The wire tables aren't there yet. You can grab 'em off the 'net for now.
• There is no provision for rectifier voltage drop in the secondary. This is because the number of turns is rounded up to the nearest integer which in practice will certainly overcome the diode drop.
• The lower the minimum current, the larger the output inductor needed. The dI value is what actually hits your output caps, so the larger the inductor, the less stress on the caps. I'll explain minimum current later on or you can look it up yourself. It's not as critical as one might think.

The feature i'm most proud of is the temperature compensated copper loss calculator. This breaks from cookbook constants, most of which are too conservative today - as frequency rises less turns are needed, and since resistance is a function of both cross-sectional area and length, with less turns, you can get away with thinner wire. It gives the loss in watts instead of simply an estimate of how many strands you need because thermal calculations you must do yourself.

Basic version: Apply 1W of heat to your transformer core and see how much it heats up... this gives temp rise in C/W, now multiply by the copper loss you got, add 50C to it and get the operating temp. Oh and keep in mind there's also core losses to think about... But those are usually given in core datasheets.

Oh yeah and it needs some organizing, i'll split it up in several sheets once i do the halfbridge section as well. And color the input and output values differently... But for now i hope you'll find it useful the way it is. Stay sharp - lots more goodies to come.

[Th3_uN1Qu3]

Re: Do you want to learn how to build your own power supplies from scratch?

Updated table...

• Fixed an embarrassing calculation mistake in the wire calculator (sorry! everyone please use this version. all other equations have been verified good, i'm actually doing a 800W supply using this spreadsheet right now).
• Added wire diameter to area conversion. Next up is a AWG/SWG table.
• Applied color scheme. Blue are input parameters (ie what you want), yellow are calculation results.

Enjoy.

[Th3_uN1Qu3]

Re: Do you want to learn how to build your own power supplies from scratch?

Updated again.

• Organized into two pages
• Added a new color - light gray - for linked values (values that track each other).

[ben7]

Re: Do you want to learn how to build your own power supplies from scratch?

Updated again.

• Organized into two pages
• Added a new color - light gray - for linked values (values that track each other).

Do you have any updates on this?

I know I'll probably get yelled at for bringing up an OLD topic, but I'm still curious about this, and I don't want to end up making a whole bunch of threads on the same topic - better to keep it in one place.

-Ben

[goontron]

Re: Do you want to learn how to build your own power supplies from scratch?

i already do! replacing the PSU section of tube radios.....

[ben7]

Re: Do you want to learn how to build your own power supplies from scratch?

i already do! replacing the PSU section of tube radios.....

That's easier than a switching power supply

[goontron]

Re: Do you want to learn how to build your own power supplies from scratch?

That's easier than a switching power supply

till you kill a tranny. then you need to trade an arm, leg, liver and kidney to get one!

[CapSwapper]

Re: Do you want to learn how to build your own power supplies from scratch?

not very cost effective. you can get a decent modern PSU for $20 to $30 after MIB

[ben7]

Re: Do you want to learn how to build your own power supplies from scratch?

not very cost effective. you can get a decent modern PSU for $20 to $30 after MIB

Ugh I dunno how many times I've had to say this.

The point is to LEARN how to design and build them, to gain more understanding of them, for practical purposes like repairing other power supplies, or making a custom one that requires certain dimensions or output voltages. Also, some people like to learn how to make them just for the fun of it.

-Ben

[everell]

Re: Do you want to learn how to build your own power supplies from scratch?

.......some people like to learn how to make them just for the fun of it.

-Ben

Just for the fun of it?????

I have been getting lots of lectures about that one. Something about not being economical....................