EEVblog Electronics Community Forum
Electronics => Projects, Designs, and Technical Stuff => Topic started by: AndersJ on June 07, 2020, 03:29:15 pm
-
I’m testing the protypes of a new board.
Unfortunately I made a power supply mistake.
I have 3 LT1762 5V LDO regulators.
One of them should have been larger, 500 mA.
Not a big issue for the next revision,
but I would like a not too difficult workaround for the prototypes.
Is it a mistake to parallell the LDO outputs?
-
as far as i know its not the done thing but using resistors on the outputs works by sharing the current,ive not tried it tho.
-
Yes, it practically won't work unless by very good luck. The output voltages typically do not match closely enough, so the highest voltage device does all the work. Adding series resistance after each regulator improves current sharing, but now it stops being a regulated supply. You can kind of adjust the resistances to find some compromise between the current sharing, and regulation. Or you can try to hand-pick regulators to make the current sharing better to begin with. Or you can derate massively, like parallel ten 150mA regulators to get 500mA, hoping that at least three of them happen to be quite close to each other (and maybe some of the rest contributing even slightly).
Note that even if you measure to verify you were lucky, thermal coefficients and aging-related drift make that unreliable.
-
Yes, it practically won't work unless by very good luck. The output voltages typically do not match closely enough, so the highest voltage device does all the work. Adding series resistance after each regulator improves current sharing, but now it stops being a regulated supply. You can kind of adjust the resistances to find some compromise between the current sharing, and regulation. Or you can try to hand-pick regulators to make the current sharing better to begin with. Or you can derate massively, like parallel ten 150mA regulators to get 500mA, hoping that at least three of them happen to be quite close to each other (and maybe some of the rest contributing even slightly).
Note that even if you measure to verify you were lucky, thermal coefficients and aging-related drift make that unreliable.
you can help by putting current sharing resistors before the sense point
-
When you start paralleling them, you also face the risk of oscillation. I've had bodged 7805s with 0.5 ohm series resistors oscillate at 80+ MHz under certain load conditions.
-
The current sharing is fine if the current limiting is smooth and nothing's going into thermal overload. The first few (in order from highest to lowest voltage) will sit in current limiting, dumping max, while whichever one is left does the regulation (delivering 0 < Iout < Imax), and the rest near or at zero.
Now the thing with LDOs is, their current limits tend to be shite. Like a "500mA" regulator might limit at 0.7 to 1.5A. Even for a small voltage drop, that's getting awfully likely to enter thermal limit (say for an SMT version).
Thermal limit still isn't a complete stopper, but...
1. You need enough in parallel that, for whatever duty cycle they go into thermal limit at, there's always enough to take over;
2. Ripple is a lot higher, because of the cycling;
3. Reliability is probably shite because thermal cycling means thermal stress, and all the ones that are cycling are holding themselves damn close to Tj(max). Typically LDOs thermal limit at something stupid like 150-180°C, at or slightly above the Tj(max) limit.
And obviously, even if it's not thermal cycling, running in current limit means those parts will be running way hotter than all the rest, and may suffer short life as a result.
So yeah, better to:
- Just use a bigger regulator in the first place, bodge it in if necessary;
- Use current sharing resistors,
- Use LDOs with accurate current limits (say +/- 20% or better),
- Use extra in parallel
Note that you can boost the regulator, if you have enough overhead to do so. All the old school circuits from 7805 and LM317 and etc. still apply, adjusting for the different voltage ranges as necessary. This will affect the current limit (the boost transistor is typically hFE limited, not a safe mode of current limiting!).
Tim
-
Attached is the schematic for the 18650 battery shield V3. This shield is the subject of Andreas Spiess' Youtube video #250. The shield has three ganged 3.3V regulators, and Andreas demonstrated that collectively they provide 1A, although he didn't mention anything about temperature. So it looks like in this case it works ok.
Also notice that the designer went for two parallel Schottky diodes in the 5V boost converter, which I think officially does not work. It seems he's a big fan of parallelism, so perhaps shouldn't be considered as authoritative.
-
I've always heard to never do that, although when I was a kid I built a power supply with three 7812 regulators in parallel and used it like that for about 10 years, never had a problem with it. YMMV though.
-
Schottky are better to parallel than regulators; they have a much softer V(I) curve at least. The main problem is the tempco isn't very convenient. If they're placed adjacent in the layout, with big copper between them, it can work out okay.
Tim
-
Schottky are better to parallel than regulators; they have a much softer V(I) curve at least. The main problem is the tempco isn't very convenient. If they're placed adjacent in the layout, with big copper between them, it can work out okay.
Tim
Good answer expanding on the important point. A poor regulator with a large load regulation will parallel well. Voltage tolerance in diodes (or regulator references) can be tight enough for the (non ideal) impedance to carry the balancing, or it might not! But it can be quantified.
-
Since it has a separate sense pin, couldn't you hook the output to a pass transistor? I couldn't find an equivalent circuit diagram in the datasheet but based on the text it seems it might work.
-
In the last page of the data-sheet they give an idea of how to use two in parallel!
-
Is it a mistake to parallell the LDO outputs?
if the regulator is designed for paralleling, it is not a mistake. See, for example, the LT3080 (https://www.analog.com/en/products/lt3080.html).
-
The lack of a block diagram or simplified schematic for the device in the datasheet is a little disappointing, so there is no information to deduce if there's a good way to parallel them.
Does your circuit have a guaranteed minimum load? Is it a significant fraction of your total load? If so you could put a carefully chosen resistor in parallel with the regulator, such that at max input voltage it passes say 80% of your minimum load current. The regulator then tops that current up to bring the output into regulation and everything is happy. Get it wrong though and the output voltage will rise to meet the input.
-
Is it a mistake to parallell the LDO outputs?
if the regulator is designed for paralleling, it is not a mistake. See, for example, the LT3080 (https://www.analog.com/en/products/lt3080.html).
Yep - some recommend it to help reduce noise (by sqrt(N)) or increase max current (by N).
-
The few linear regulator ICs that do support parallel operation do brag about it on the front page, obviously because it's a very good feature and worth some money.
Of course you can mill a screwdriver out of hammer...
LT1762 datasheet shows a peculiar paralleling example circuit ("Paralleling of Regulators for Higher Output Current") where input currents are measured using shunt resistors and the ADJ pin of the second regulator is fed back from the current mismatch. If you have an extra opamp lying around and can afford a bunch of extra passives, why not... But I wonder how many designers actually implement such "typical application" circuits.