On Tue, 15 Apr 2008 01:58:25 -0700, Robert Baer
robertbaer@localnet.com> wrote:
legg wrote:
On Sun, 13 Apr 2008 21:10:27 -0700, Robert Baer
robertbaer@localnet.com> wrote:
The requirement of that beastie was minimum dropout, so i looked at a
number of 12V "LDO" regulators, and they either had 1V drop at almost
all current loads, or the drop was 400-700mV.
The only decent LDO regulator i found was the LM2940CT-12, so i
decided to parallel them using a series resistor of 0.2 ohms at the
output of each one toward the load.
I used a 47uF cap right at the regulator output in conformance to the
datasheet reccomendation.
To reduce dissipation of the regulators at full load and 15V max
input, i used four 2.7 ohm 5W resistors.
For testing, i used a FWB battery charger output (bypassed its
regulator) and a Variac to adhust the peak output voltage.
The only load i had was a 12V 50W automotive lamp, as the actual load
to be used is about 3,000 miles away.
It looks good, and the dropout at low inputs (ie: slightly above 12V
output) is excellent.
BUT....
...here is the problem.
I see spike-type oscillation at the input, apparently when the input
voltage is high enough for those 4 "shunt" or "bypass" resistors to pass
current.
There aer zero spikes when this regulator board is not used: ie: when
the battery charger is driving the load (lamp) directly.
Q: What is causing this spiking, and how can it be fixed?
You haven't indicated whether the non-uniform current drain has an
effect on output regulation. If not, it could just be that you are
seeing the 100mA variation of quiescent current that occurs as a
single unit sees minimum headroom. You might consider ignoring it, if
no functional problems result, until proper test equipment and loads
allow a proper solution to become practical.
** The variable load (due to thermal changes) of the lamp appears to not
make a difference; the output looks reasonably regulated (with some of
the spikes showing - undoubtedly due to the "bypass"/"shunt" resistance.
As output voltages are not identical (and even when they are), there
is likely to be motor-boating around the threshold.
** Gets worse as input voltage increses.
Placing series output resistors on an LDO seems rather ham-fisted.
You've not mentioned your specified performance aims; it's not really
possible to offer constructive alternatives until you do.
** Just followed the same scheme recommended for the 1, 2, and 5A LDOs
made by the same company - which is the same thing built into the IC of
another LDO they make.
** Ideally, input should be able to change from12V to 15V and the load
voltage remain at 12V (actualload current is stillunknown but presumed
to be 5A).
Obviously 220mV of load regulation is not considered to be a problem -
this could qualify a lot of LDOs with nominally higher headroom
requirements, as their spec will have assumed a much tighter
regulation requirement in the spec formulation.
The LD29300 goes up to 3A @400mV, which could reduce your component
count and potential modes of irregular parallel performance.
** If i have too much trouble with the presen design, i may consider
that part - especially if the drop decreases as load current decreases.
A discrete circuit is an obvious choice, if you're already going to
the trouble of hooking up 5 linears, but I assume you're stuck out in
the wilderness with only a soldering iron and a Digikey catalog
delivery service.
RL
** Good call!
** Actually, a buck/boost DC/DC would be the perfect solution.
But it seems they cannot be had off theshelf and i cannot design a
workable one from absolute scratch.
This is off the wall. But increase your shunt resistors, or put a
diode in series, or both.
I do not think those ideas are off the wall.
