Building your own power supply: A simple example

[Th3_uN1Qu3]

I forgot to post this one didn't i. When i was like 5, my cousin had this computer he'd never let me play on. It was a "HC-91" where HC stands for Home Computer and 91 is the year of production. The machine is a Romanian ZX Spectrum 48k clone. I've always wanted one, and recently i found one for cheap along with the tape deck, some tapes with games and stuff and a b/w monitor (which didn't work), so i bought it.

It's a pretty cool machine, and there are plenty of fun games for it, even though it's worse than a NES in terms of capabilities. It also has a nice manual which teaches you how to program it in BASIC, and even a little Z80 assembler! When did you last see a computer come with a programming manual?

The RF modulator output is crap (suspected bad caps actually), but i made myself a composite video output cable for it and it works great on that. It also has a monitor output, but it's CGA and none of my displays support that. I also made an old Sega gamepad work with it. All caps are GSC... And they have no vents. Not even the big 4700uF one on the power supply (not pictured). So GSC existed in '91!

It had one really annoying aspect though - it had a big brick of a supply. Also the cable was wonky and easily disturbed. I fed it with a little SLA i had, however the linear reg it had built in would eat it in an hour or so. And i had other plans for that heatsink...

So, i built a switching power supply for it! It is a simple buck converter using a p-channel mosfet, and a TL494 as controller. I used the same controller board i had designed for my 800W supply, so a few little mods needed to be made to it, but apart from that the supply basically designed itself. The inductor i think came from a 3.3v output in a cheap PSU, it is ~32uH and did not need to be rewound for my application. Since i had no other use for the second error amp of the TL494, i configured it to shut the supply down if input voltage goes over 25v. This is to avoid excess heating of the inductor - the higher the voltage, the higher the core loss. There is also a little 33v zener to protect the TL494 and the input cap in case i do something really dumb. Which isn't likely, but still.

Schematics and PCB layouts are available for whoever wants them. With bigger/better/more caps and diodes this could probably do close to 10A without needing a redesign - the power mosfet is this one. It was only $1.20, so why the heck not.

For the nitpickers:
Yes i wasted 3 transistors by using emitter output on the TL494, then i had to invert and buffer the signal again so as to drive the P-channel power mosfet, but the board was already made and i was too lazy to change it. Anyway it uses just 12mA when idle. There are also some useless copper areas left by the PCB software that crowd up the controller PCB, they will be removed in the next revision. As for the caps... this thing only uses 820mA at 5v so spare me. Input is 330uF 25v Jamicon SK, output is 470uF 16v G-Luxon LZ. Don't throw eggs at me already. The cap is rated for 730mA ripple with 0.093 ohm ESR and it doesn't even see 300mA in this application. Ripple measured ~90mvpp, which is a bit higher than the ATX spec... but this isn't a PC, it uses all 5v logic so it could care less. Oh and the sharp eyes have probably noticed a fukyoo 1uF cap in there. Along with that zener it forms the gate drive circuit for the mosfet. I chose that cap for a reason - it has around 1-2 ohms ESR. Hey look - built-in gate resistor! Again, it'll last darn near forever in this application.

One last word on the measurements - at this low voltage the amp meter drop causes some serious errors as you can see in the second pic where the amp meter on the 5v output is removed, so the efficiency is a fair bit higher than you see in the pic with all the meters hooked up. On a bench supply the HC uses 0.82A @ 5.0v, which makes the actual efficiency 78.7%. Not stellar, but most of that is wasted on those rather lame diodes (2x 1N5819). With little effort this could be turned into a synchronous converter and achieve over 90% efficiency with ease, but at this power level, why bother.

Here's the machine complete with old sticker from chewing gum. It needs a little makeover but i'll take care of that later.

[Th3_uN1Qu3]

Re: Building your own power supply: A simple example

Here are the PCB layouts. The TL494 one has been cleaned and somewhat revised as opposed to the one on this PCB. The cut trace and resistor on the back would still be needed if you make the same type of power supply as i did, because the PCB was originally made for a different application (my 800W supply) and i was too lazy to change it. This version divides the reference of the TL494 to get 2.5v, the big supply uses 5v reference. The reason for this is that the upper resistor of the voltage divider has its role in compensation, which would be difficult without it. There is room for compensation components for both error amps btw. As opposed to the 5v ref version, no trimpot for the reference on the second error amp however, so if you want to do current limiting you will need other parts.

I'll draw a schematic too if you really need it. I never actually bothered to make one - i just worked with the TL494 datasheet in front of me. The TL494 PCB is made with 1/8W resistors (or like i did, 1/4W ones sitting upright), you will need the custom components, attached as well. Extract to your ExpressPCB folder. And there's an inconsistency in the supply PCB - instead of the resistor paralleled with the zener diode there must be a 100nF-1uF cap.