NoName said:
[...]
National Semiconductor has an article on LDO ESR:
"Capacitors are key to voltage regulator design"
http://www.national.com/nationaledge/jul02/article2.html
The LP2985 specs the minimum ESR at 5 milliohms. This is
compatible with ceramic output caps:
http://lis.epfl.ch/contest/flying07/docs/resources/LP2985.pdf
Then one sunny day the cap mfg "improves" a cap series, resulting
in even lower ESR, phssst. BANG!
That should be caught in incoming inspection.
On ESR? Manufacturers routinely spec a guaranteed maximum, no
minimums. They will refuse to take back and give credit if you return
a shipment because ESR is too low. And rightfully so. I am not sure
1206 caps would be suitable as tiles for the lobby or pavers for the
patio but you'd have to do something with all the "rejects" ;-)
Besides, the ESR "stable range" is usually a figure and not
guaranteed data. That alone is a turn-off point for me. In medical
electronics where I work a lot you can't rely on typical data.
It is never a good idea to use typical values in a design. From the
datasheet:
"The LP2985 was designed to work with ceramic capacitors on the
output to take advantage of the benefits they offer: for capacitance
values in the 2.2 uF to 4.7 uF range, ceramics are the least
expensive and also have the lowest ESR values (which makes them best
for eliminating high- frequency noise). The ESR of a typical 2.2 uF
ceramic capacitor is in the range of 10 mohm to 20 mohm, which
easily meets the ESR limits required for stability by the LP2985."
It sounds like they intend the part to be used with low ESR caps.
Excellent example. Quote from that very datasheet: "The output
capacitor must maintain its ESR within the stable region over the full
operating temperature range of the application to assure stability."
Or how about "...utilizing circuitry which allows the regulator to be
stable across the entire range of output current with an output
capacitor whose ESR is as low as 5 mohm." a page earlier? So what
exactly does "as low as" mean? Is it guaranteed? I doubt that. And how
do you maintain that value? Put a 5mohm resistor in series?
<goose bumps developing ...>
Then there is the typical graph on top of top of page 15. Where are
those limits guaranteed? Why are they not in the min-max tables? I'll
never use a part like that.
I'm sure you could require a guaranteed performance spec in a
purchase contract if you were willing to pay for it.
IME companies are often not even willing to simulate the thing when it
did go wrong. Maybe because then a dirty little secret could pop out.
Anyway, this was just an example to show it is possible to improve
the ESR performance without requiring an extra pin with external
filter components.
Sorry, I do not see that. Same old, same old, there is a minimum ESR
mentioned and IMHO that's not a good thing.
I also had some non-datasheet-mentioned pathologies. For example
on a LM29xx series chip from National many moons ago. Someone
absolutely wanted to have it in there despite my advice against
LDOs. Turned out it did not "like" too high a source impedance.
Made it oscillate, heat up and then fry. Lengthy calls, a team at
National on the other side. You could hear large sheets being
handled. Suddenly a faint "oh drat" could be heard over the phone
and then I knew we were in deep doodoo.
This is an implementation problem in the design. Obviously it is
something you check before committing a part to production.
Absolutely not. This was a stability problem of a chip that was not
mentioned in the datasheet. IOW not supposed to happen but did happen.
Needless to say the guys where rather apologetic on the phone because
they knew it wasn't supposed to happen.
[...]
Take a look at the far right where the price is listed. The
ADP3331 has a realistic quantity street price around a Dollar, way
above budget for a lot of designs. Especially if you can do it
without an LDO.
I checked the parts in Digikey before posting. Certainly the price
is high - that's what you pay for in a monopoly. However, the part
is not affected by ESR, and it does not require an extra pin with
external filter components. So a clever design might get around the
patent and give the same results.
For reference, here's a bunch of articles from various sources on
LDO ESR. The url's are not shown since they degrade with time.
But google finds them quickly.
AN1148 PDF 205,195 Compensating Low-Dropout Regulators
AN682 PDF 879,026 New Generation of Low Dropout Regulators
AND8028 PDF 38,766 Precision Sub-One Volt 1.7 Ampere Output LDO
SLVA068 PDF 68,595 LDO Fundamental Theory
SLVA072 PDF 291,538 LDO Technical Review
SLVA079 PDF 202,342 LDO Terms and Definitions
SLVA115 PDF 85,754 Regulator ESR Stability
SLYT151 PDF 468,245 Compensation Transient Response
SLYT187 PDF 254,973 Understanding the Stable ESR Range
SLYT194 PDF 246,355 LDO Linear Stability Analysis
SR003AN PDF 73,945 Compensation for Linear Regulators
SR004AN PDF 41,598 Linear Regulator Output Structures
Thanks for the links. I don't generally use LDOs but there may
come a situation where it has to be one.
Your complaints about LDO's are well known. You spoke of rolling
your own when needed. Any chance you could post a schematic so
we could see how these problems are solved?
Then I'd get shot ;-)
But seriously, all you have to do is run thorough simulations and pay
special attention to loop stability. In a discrete design it is not a
problem to add a 4700pF cap somewhere, on a chip that is quite
impossible. But I rarely design discrete LDOs these days because when
you are this close to the input voltage you might as well go straight
to a SEPIC or forward converter, which is what I usually do.