LDO to supply close to 5V from 5V or 12V input.

[paulca]

I want to create a circuit that has a MCU in it requiring at least 4V, the closer to 5V the better from either a 12V supply or a 5V supply.

Using a parametric search I can find lots of 5V LDOs which claim 0.5V drop out voltage or less at the small <100mA current I need. 

I'm always cautious of datasheet marketing headlines though.

For example:
https://docs-emea.rs-online.com/webdocs/140d/0900766b8140da69.pdf

How guaranteed is this to provide me with the headline 4.5V from 5V?  A quick scan of the datasheet only gives the drop out voltage for a single load current.  The graph Fig.3 on page 8/20 suggests that with a Vs of 5V I should get pretty close to 5V out... at 75mA.  What about at 10mA?

[BrianHG]

Just use a KF50BDT.  <0.2v drop at 100ma.  0.3v drop at 200ma.

[Zero999]

The pass transistor is a MOSFET, which means when it's operating in the drop-out region it behaves like a resistor.

A worst case of drop out of 500mV at 150mA is specified, so the highest resistance is 0.5/0.150 = 3¹/₃Ohms, so use that figure in your calculations and you can't go wrong.

[mikerj]

What is the tolerance on the 5v supply?

[paulca]

It's fairly tolerant, but will be pushing data out to LEDs which have a logic transition level of around 3.8V / 4V.

I think I will go with the KF50BDT idea.

[mikerj]

It's fairly tolerant,

I mean what is the voltage tolerance on the 5v supply i.e how much can it vary across the full temperature, load and supply voltage range?  When you are using an LDO with minimal headroom you need to know this stuff.

[paulca]

It's fairly tolerant,

I mean what is the voltage tolerance on the 5v supply i.e how much can it vary across the full temperature, load and supply voltage range?  When you are using an LDO with minimal headroom you need to know this stuff.

You see that's the thing.  The point is to make it universal.  Running either 5V LEDs with a random 5V wall wart (or USB for short lengths) or running 12V LEDs with some other random wall wart, lipo or car battery!  Though I don't think the LEDs would take too kindly to the later if it was on charge at 14.4V.  To allow that I would need to consider a whole raft of other things to regulate boost/buck the high amp side for the LEDs, which I am avoiding.

A timely example is that I have a 0.5m length of 15 LEDs for the back of my monitor and powering it off 2x USB ports (provided by the monitor).  I will avoid 100% white, but it would pull theoretically 9W, circa 1.8A.  My monitor USB charge ports, which are claimed to be fast charge ports, drop to 4.8V with a 0.8A load (EDIT:  I did not verify how much of that drop was in the cable!).  I can test this setup can supply enough power without dropping too many volts using my 12V board and putting a bit of wire across the pads instead of the Reg.  If I can get a similar rig to work with a LDO in place, that's what I'm after.

I think, if we are talking about an LDO with even a 0.5V dropout, then I should be good down to 4.5V, maybe lower.  If the board happens to get powered by a PSU that can't put out 4.5V at the amps required, then the PSU is inadequate, probably overloaded, the LEDs will behave erratically and I need to find another.

[BrianHG]

Just so you are aware, you can construct a 0v dropout regulator with a P-Channel mosfet, and 2 transistors + zener diode with a few resistors and cap.

When I say 0v, between around 2.5v and 5v input, you output will be equal your input voltage, even at 1 amp draw.  Between 5v and 12v input, you output will stay at 5v.

[sleemanj]

I assume you have good reasons for wanting an LDO?

Otherwise, a SEPIC dc-dc converter sounds like it would be the solution you are looking for, set it to 5v output, input can go above or below 5v and that set 5v output should be maintained.

There are plenty of such SEPIC boards available in the usual places for cheap, LM2577 based ones are most common, just search for LM2577 SEPIC.

[paulca]

Size is important.  So the DC-DC converter solution would probably fill the board space on it's own.  At the moment the board is 29mm x 44mm which was quite spacious with an ATTiny85 on it, but now I want to put an ATMega328p + crystal on it.  So the regulator will have to go on the back side of an already via rich board.

Of course I'm considering adding input sensors to vary the LED patterns based on the environment, such as the smallest condensor mic I can find, a dallas one wire temp sensor and an LDR.  Which might force me up a case size.

[BrianHG]

I wonder if it is possible to further simplify my design below:
I'm using a P-MOSFET SSM6J801R or similar.
This design should pass 1.5v in to out with a series resistance of 88mOhm.
It should pass 2.5v in to out with a series resistance of 40mOhm.

So, with a 100ma draw with a 2.5v in, the output will be 2.496v.
A 100ma draw with a 5v in, the output will be 4.9967v.
A 1 amp draw at 5v in will give you a 4.967v output.
Well, you get the idea...

The output should max out at 5v with the input as high as 20v, though, since it is linear, with 20v in drawing 100ma, the device will dissipate 1.5 watts burning up, with 12v in, the device will radiate 0.7 watt.  You might want to use a mosfet with a slightly beefier package than an TSOP6F if you are drawing the full 100ma.

[paulca]

The 5V rail only powers an ATMega328, so circa max 18mA.  However it will be in a sealed, small case, so i can't really have a lot of heat.  Granted my linear LDO will produce, 0.018*7 = 126mW heat at 12V.

Clamping the LED output to 12V might allow an even more flexible power input, but if someone connects 5V LEDs and a 12V supply it's still going to smoke.

Don't want to over complicate things.   It's not a product, but more of a thing I can give family/friends, mostly I will be providing the PSU or checking they have a correct one for it.

Actually as it turns out I might not need to do this at all.  I determined a few converging design limitations might make two versions easier to make.

The 12V board with ATTiny85 and 5V reg can only run about 60 LEDs due to memory limitations.  However as that amounts to 180 LEDs in a 12V set, it would be above the sensible current limit for a single power rail for the LEDs anyway.  So I can't run more 12V LEDs from a single board.

The 5V LEDs come in 144 per meter and 1 meter of them would be the limits of the single power rail supply.

So an option is to just keep the 12V board as is for running 12V strips and make the ATMega328 version a 5V board with no regulator, just some filter caps to remove the LED current switching noise.