Two regulators connected to a single transformer? (linear PSU build)

[[Chrille]]

I am going to build a custom linear power supply for an old computer that requires 5V (about 1.2 Amps) and 12V (0.1 Amps).

I want to design the PSU with a design like this (but slightly modified):
https://www.build-electronic-circuits.com/wp-content/uploads/2023/06/full_rectifier-1.png

This is a classic linear power supply with a transformer, a full wave bridge rectifier, a couple of filtering capacitors and a voltage regulator. Classic and quite straight forward.

But since I need two different voltage outputs my plan is to use a 230V -> 12V AC transformer (2 Amps or higher) and then use the the same transformed, rectified and filtered 12V output and connect two different regulators like a 7805 and a 7812 to get clean regulated 5V and 12V outputs. Is there a problem with this kind of setup or do I need to use a transformer with two secondaries (or two separate transformers)? In theory it should be OK I guess as long as the transformer used has enough capacity to power the two outputs. Is this bad practice or will it work fine?

If I have to use a transformer with two secondaries I will be wasting a lot of capacity since I need at least 2 x 1.5 Amp outputs and I will not even use 10% of that capacity for the 12V output, and I will need 2 bridge rectifiers and two filtering capacitors etc. It seems like a waste.

Feedback appreciated 

[madires]

You can do that but the drawback of that design would be wasting power for the 5V rail (heat dissipation for the linear regulator). 12 V AC is going to be around 15 V DC. So the 5V regulator would have to dissipate (15 V - 5 V) * 1.2 A = 12 W. And you need a regulator type rated for more than 1.2 A. It would make more sense to use a buck converter for the 5V rail.

[[Chrille]]

You can do that but the drawback of that design would be wasting power for the 5V rail (heat dissipation for the linear regulator). 12 V AC is going to be around 15 V DC. So the 5V regulator would have to dissipate (15 V - 5 V) * 1.2 A = 12 W. And you need a regulator type rated for more than 1.2 A. It would make more sense to use a buck converter for the 5V rail.

Yes, I know that there will be quite a bit of power waste but the computer I am building this for is not getting much use so it's not really a problem. The heat dissipation issue is more of a problem but can be dealt with using a large enough heatsink.

But aside from the power waste is there any other concern with this kind of solution?

The 7805 5V regulator was just an example, I will probably use an LM323K instead which is rated for 3 Amps which is a bit overkill but it looks nice and has a lot of capacity. Some 7805's can be used for up to 1.5 Amps so they should work too if properly cooled.

The good thing with this solution is the simplicity, smaller PCB footprint, less number of components used etc.

The reason I don't want to use a switch mode PSU for the 5V rail is because I want clean power with as little ripple noise as possible to avoid any potential damage to the old IC's used.

[madires]

Maybe adding a diode from the 78xx's output to its input (protection for the 78xx). BTW, 78Sxx is good for up to 2 A.

[bdunham7]

The reason I don't want to use a switch mode PSU for the 5V rail is because I want clean power with as little ripple noise as possible to avoid any potential damage to the old IC's used.

Computers are inherently noisy and have been using SMPS for many decades without issue, so you might not be addressing any real issue.  If you want simplicity and reliability I'd suggest a drop-in 5V buck module for the 5V rail.

https://www.ti.com/lit/ds/symlink/tpsm84203.pdf

If you insist on going linear, you'll need to first realize that you need sufficient headroom, meaning the difference between the input and output of the regulators.  For an LM7812 that means at least 15V at the trough of the ripple and that probably means at least 18VDC average.  You'll likely get that from your 12VAC transformer untill you start loading it down, so you'll have to test for that.  A 2-amp transformer is actually just barely big enough for your output specs because the RMS-to-average ratio can exceed 2.0 in a linear PSU.  This means that at full load your transformer will be stressed.

The next thing to worry about is that this is going to generate a LOT of heat.  At 18VDC input and 1.2A output, that 5V regulaor will dissipate 15.6W.  That's almost impossible to deal with, even with a thick-tab LM7805 and a very large heatsink.  IDK if it can be done without fan cooling.  You can use a series power resistor to drop the voltage and take away some of the heat and that's what I'd recommend  if you insist on staying linear.  If you use a high-current LDO like the LM1084 you can drop the full-load input voltage down a bit further for even less dissipated power.

https://www.ti.com/lit/ds/symlink/lm1084.pdf?ts=1741030348715

[ledtester]

The reason I don't want to use a switch mode PSU for the 5V rail is because I want clean power with as little ripple noise as possible to avoid any potential damage to the old IC's used.

If you can find a suitable transformer for the 5V rail you can use a boost converter to get the 12V rail and keep the 5V rail linear.

[PGPG]

If I have to use a transformer with two secondaries I will be wasting a lot of capacity since I need at least 2 x 1.5 Amp outputs and I will not even use 10% of that capacity for the 12V output,

Why you assume that you have to have two windings made with the same wire (the same current)?
Long time ago when building oscilloscope I needed:
+ 5V for TTL,
+12V / -7V for X,Y preamplifiers,
+175V for X,Y amplifiers last stage (to drive deflections plates),
- 1100V to drive cathode of an oscilloscope tube.

All that I got from one 100W trafo. There are no rule telling that winding for 1100V have to be wound with wire of the same diameter as the winding for 5V.