Paralleling Linear Regulators

What don't you just add a big heat sink and liquid cooling if necessary?

"Bobby Joe" <

** If you only want to reduce the heat dissipation in the LM340 - then add a series resistor on the input side.

The resistance value will depend on the max current draw and the lowest expected input DC voltage minus the regulator dropout voltage and output voltage.

.... Phil

Since the combination will dissipate the same amount of heat as the single, and its cooling will be very similar, you won't get much of a benefit from doing this--unless your 340 has no heat sink at all.

Phil's suggestion of an outboard series resistor is a good one.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal
ElectroOptical Innovations
55 Orchard Rd
Briarcliff Manor NY 10510
845-480-2058
hobbs at electrooptical dot net
http://electrooptical.net

If he can get at the pins of the LM340 to solder another one in parallel, the original LM340 can't have much of a heat-sink. If the original was mounted flat on the board, with a thermopad or some other form of thermally conductive gap filler, the free-standing part will offer at least half as much again heat-dissipating capacity as the original - see the small print (note 2) at the bottom of page 6 of the datasheet.

Not least because the series resistor can be run hotter - to 235C for some parts - than the LM340 (maximum junction temperature 125C). It's a trick I've used myself, though not for many years.

-- Bill Sloman, Nijmegen

First, it's a linear regulator; it's supposed to get hot. It's inherent in the nature of linear regulators that they get hot; they reduce voltages by dumping excess energy as heat. If you don't like that, look into switching regulators. If getting rid of the heat is a problem, get a heat sink.

Second, an LM340 is fully protected against overheating and overcurrent; the National Semiconductor datasheet says it is "essentially indestructible". If it's delivering the rated output voltage, it's happy.

Third, if you need a higher-current regulator, get one. There's probably an available part with the same form factor. 3 amp parts are common.

Fourth, it's not clear whether paralleling linear regulators will work. There are ways to make it work, like putting a resistor in series with each output to assist in load balancing. A very small resistance (like

0.47 ohm) has been suggested. But you lose voltage regulation precision if you do that. Plus you need more components. If you parallel them without a resistor, they may or may not load-balance; it depends on the internal architecture of the regulator and the sign of the temperature coefficient. Power MOSFETS, for example, will load-balance because they have a negative temperature coefficient; ordinary transistors have a positive temperature coefficient and, in parallel, the hottest one ends up doing all the work.

Conclusion: get the right tool for the job, unless you like designing and debugging power supply circuits.

John Nagle

"John Nagle"

** The LM340 may indeed be happy enough to run with a case temp of up to 120C - but the trouble is that other parts packed in around it are not going to be AMUSED !!

Particularly, any small electros that filter the incoming DC or stabilise the reg.

Sooner or later, they will take great umbrage at their offensive next door neighbour and go off in a high ESR huff.

..... Phil

You may have to add a quite big capacitor from the regulator input to ground (much larger than generally suggested by manufacturer) to avoid some stability problems.

A zener or a few forward biased rectifiers may also be a good solution to drop some input voltage.

Paul

"Paul Keinanen Alien Nutter "

** Utter Bullshit.

..... Phil

On Wed, 19 Aug 2009 20:45:30 -0700, John Nagle wrote: . . .

--
You got the end result right, but the cause backwards.

That is, MOSFETs exhibit a positive temperature coefficient of
resistance, while for BJTs it\'s negative.

JF

"John Fields" John Nagle

** Shame you are both wrong.

Power mosfets have a large, positive tempco when used in linear mode PLUS have large device to device variations in gate threshold - means parallel devices must be selected AND use source ballast resistors.

Comparable BJTs have a smaller positive tempco and device to device variation in Vbe for a given Ic - so are easier to operate in parallel.

Lateral power mosfets are the exception, they can be paralleled with no special matching or ballasting.

.... Phil

Worst case must be enormous temperature differences between units to compensate for the unit-to-unit Vgs variation. Or don't audio designs use data sheet limits..

en

Yup, But I do sometimes add a cap anyways. Since the resistor is already there why not get a bit of low pass filtering.

George H.

Sometimes a parallel resistor works, too.

John

In their linear regions, power mosfets will usually have such a difference in transfer curves that the minor tc effects don't help enough to matter.

John

"Spehro Pefhany" "Phil Allison"

** The Vgs variation among lateral power fets is quite small in practice, especially if all from the same batch.

The over-riding characteristic is the way the tempco goes from positive to negative around 100mA Id.

This alone means that above 100mA, parallel devices self balance and below

100mA, devices limit at 100mA.

** Experience is a good teacher.

Over-interpreting the published data is the way of fools.

..... Phil

"John Larkin"

** The tempco of Id is strongly positive, so it interferes with current sharing.

.... Phil