piclist 2002\11\03\014755a >
www.piclist.com/techref/microchip/devices.htm?key=pic
BY : Russell McMahon email (remove spam text)

> > *>LM317 adj current is only 50-60uA for 10mA < Iout < Imax
> >
> > Wherever you get that. LM317 adj current is exactly 120uA no matter what
> > the load. That's how it comes to set the voltage by using just a
resistor.
> > Constant current across a resistor == constant voltage == regulation.
> > There is only the 1.2V constant voltage between ADJ and Out to be added
to
> > this.
>
> Ummm...  he got it straight from National's data sheet, page 4 I would
> guess.  I just happened to be looking at the LM317 data sheet when this
> arrived.  The adj current is specified as 50 to 100uA, and according to
> the data sheet the LM317 develops a reference voltage of 1.2 to 1.3V
> between the adj pin and the output.

The current does vary somewhat in practice but typically not nearly as much
as the spec sheet worst case figures. See spec sheet graphs.

This is why the divider uses lowish value resistors. The designers TRY to
keep this current as low as possible and across usual temperature range it
varies by about 10% typically. If this current formed a large percentage of
the current through the lower resistor in the divider than it would have a
severe affect on regulation. Ideally this current should be zero and the
divider current would then only be the current in the upper resistor (with
1.2V across it) and the divider would be in the exact ratio of voltages
desired. eg if there was 1k2 from Vout to Vref then you would get 1 extra
volt  per kohm of resistance from Vref to ground. This (of course) is not an
ideal world we have the added current. to contend with.

National typically recommend a 240r resistor from Vout to Vref giving
1.2/240 = 5mA "regulation" current. A change of 5uA in the bias current
mentioned above will lead to a 5uA/5mA = 0.1% change in the regulated
voltage. This order of effect is probably fairly acceptable in this class of
regulator.

The next to ideal if you can't have NO current is to have constant current
and this is what the IC designers strive for (while necessarily failing
heroically). It's still far better than if you use a regulator not designed
to be used in this way such as the LM340 / 7805

Russell McMahon

--