# Building a linear power supply

Discussion in 'Electronic Basics' started by Andrew Wagstaff, Nov 14, 2011.

1. ### Andrew WagstaffGuest

Hi

I'm looking at building a linear power supply. I'm thinking of using a
12V 300VA torodial and linear regulators. I really want a power supply
that can be used to power both AT and ATX computer motherboards, has got
variable current limiting (3.3V, 5V and 12V) and displays current. I
haven't seen anything in the marketplace, so I was looking at building
my own.

The voltage and current ratings I'm looking at currently are:
3.3V 10A, thinking a 5A variable reg and using a pass transistor to get
up to 10A.
5V 5A-8A, thinking a 5A variable reg and a pass transistor.
12V 5A, thinking a 5A variable reg.

The voltage drop from 12VAC (~17VDC using a bridge rectifier) down to
the lower regulators (3.3V and 5V) was concerning me, as I was looking
at producing around 60W-100W of heat just for the 5V reg.

Was also considering having 4.5V, 6V and 9V. Was going to use another
variable reg - only need around 1A and using a switch to change between
them.

Another problem is how do I get the -ve voltages for the 5V and 12V
(only need ~500mA), do I just use some sort of polarity inverter or do I
need a centre tapped transformer?

I already have most of the components (except the torodial), that's why
I was thinking of going with linear. I do have some components to make a
5V switchmode, but I wanted to keep the power supply as quiet as I could.

Thanks

Andrew

2. ### JamieGuest

why not simply use an existing computer supply and have inline current
limits on the outputs of choice?

The current limit circuits can be designed using power fets to pass the
source voltage with current sensing to drive a comparator that drives
the power mosfet. With a little hysteresis you can keep the fet out of
the linear mode and force it to simply shut off if current limit is reached.

Pots on a panel can set the (I) range of each output..

Jamie

3. ### Andrew WagstaffGuest

Hi Jamie

I have tried before to use a computer power supply as a high current 12V
supply before, but found that unless the 5V line was loaded at around
1.5-2A, the 12V line dropped down to around 11V.
I might not be using the 5V line when I want to use the 12V or 3.3V lines.
The current limiting I was thinking of using, was using a 5A linear
regulator wired as a current regulator.

Thanks

Andrew

4. ### Andrew WagstaffGuest

Just revised my design a bit and are now thinking of using the 9V 300VA
torodial, instead of the 12V one. Thinking I'll use a low dropout reg
with a pass transistor to get the required current at 12V.
This should make the heat production from the 5V and 3.3V a bit less,
but still not sure how to get the -5V and -12V.

Thanks again

Andrew

5. ### ehsjrGuest

It's going to take up a lot of bench space and produce a lot of
waste heat. It'll be heavy. And it will cost you a lot. Your
proposed 12V transformer won't work. You'll need an 18V transformer
for the 12V 5 amp supply**. If you use the 18V transformer for all
load. Better would be 3 transformers, or a single transformer with
3 secondaries. You'll still produce a lot of heat, no matter what
you do, if you go with a linear design. A linear supply to power
a PC is a really *bad* choice.

** Regarding the voltage:
12VAC input to a bridge rectifier and capacitive filter will
produce 16.97 VDC, ignoring the voltage drop in the diodes.
Under a heavy load, you'll get a drop in the diodes totaling
about 2.2 volts, reducing the voltage to 14.77 max. Ripple
will reduce that further. The LM338 (5 Amp Adjustable Regulator)
needs a minimum of 2.7 volts overhead at 5 amps, so it will
drop out. In fact, it will only regulate for the briefest of
moments when the AC is at peak. Say we figure about 3 volts
ripple and about 2.2 volts drop in the diodes. That means we
need 3 + 2.2 + 14.7 or 19.7 volts DC peak across the filter caps.
That implies a 14 volt AC transformer which will produce ~19.79 DC
peak across the caps. However, there will be some voltage drop
in the transformer under load, and the input line voltage can vary,
too. Say the line voltage is 110 instead of 120. With an 18 volt
transformer fed by 110 instead of 120, we get 18*110/120 or
16.5 volts AC. We can drop to 14 volts AC and still work. If the
transformer drops 10% under load, it still yields 14.85
volts AC and will charge the caps to 21 volts. Even if the
ripple was 4 volts, we'd still have 15 volts input to the
LM338 which is plenty of headroom.

The above is in no way intended to encourage you to build the
supply you described. OTOH, I don't want to discourage
experimentation/learning, but the proposed project has too many
downsides to be worthwhile without some compelling reason(s).

Ed

6. ### JamieGuest

That's not a problem, have yourself a front panel indicator lamp
operating from the 5 Volt supply with a little extra load on it, that
will satisfy most computer supplies and give you an idiot light, also
add a 5 volt cooling fan..

Jamie

7. ### P E SchoenGuest

"Andrew Wagstaff" wrote in message
A single 16 VCT transformer can be used for all of the required voltages. A
FWCT rectifier with two diodes and a big capacitor can provide the raw 10
VDC for the 5V linear regulator (and 3.3V).

And another, with diodes and capacitor reversed, can be used for the -5 VDC
supply, which is usually much lower current.

You can also get an adjustable buck supply (1.25-30V, 2-3 amps) on eBay for

For the + and - 12 VDC supplies, you can use a voltage doubler made with two
capacitors and two diodes for each polarity, which generates about 25 VDC
open circuit, and is also inherently current limiting because the
transformer is capacitively coupled to the output.

And perhaps the easiest solution is to add a load to the +5 VDC output of
the ATX supply. It's about 20W but for a test setup efficiency and heat are
not much of an issue. Use a 5 volt fan as part of the load.

Paul