You should not be actively driving the ADJ pin. Only be pulling it toward ground.
But if you use some fast schottky diode (0.1V) instead of those cheap diode in the datasheet, you have reduce the risk by 90%, right?
National Semiconductor tells another story for the LM117, read page 9 protection diode http://www.ee.buffalo.edu/courses/elab/LM117.pdf
They also claim of internal protection. I use both protection diode everytime!!!
"The discharge current depends on the value of the capacitor, the output voltage of the regulator, and
the rate of decrease of VIN. In the LM117, this discharge path is through a large junction that is able to sustain 15A surge
with no problem. This is not true of other types of positive regulators. For output capacitors of 25 ?F or less, there is no
need to use diodes. "
if you put a diode between ADJ and output then you can actively drive the ADJ pin. no problem.
the risk with active driving is that you succeed in injecting current into the ADJ pin ( once you pull ADJ more than 1.25 volts above Vout. ) you will turn on parasitics inside theLM317 and you can fry it.
there is a reason they put a diode there when they do the slow start using the capacitor on vadj ... that cap needs to discharge at powerdown. if it can't it can kill the lm317 at poer-off time
_____
in ---|lm317|----+---+-- out
|_____| | _|_
| R1 /_\ <- this guy protects ADJ at shutdown by discharging
| | | C1 into out at powerdown
-------+---+
| |
r2 === C1
| ---
| |
-+-
|
_|_
gnd ASCIICAD (c) free_electron's hand
fully protect lm317 with opamp control
d1
-----|<|-----
| _____ |
in -+--|lm317|----+--+---------+-------- out
|_____| | | | d2
| R1 R3 --|<|--|gnd
| | |
+------- -----------+--r4--|gnd
| ___ |
E \|P / - |-----+
B|N--[r2]--< |
C /|P \_+_|-- control
|
GND
d1 is the backfeed diode. try to yank OUT higher than In and it will conduct protecting the pass element in th lm317.
D2 is a polarity reversal protection. force ea reverse voltage on the output and it will protect the entire supply ( it will fry itself ... this should be a diode that can handle some 'oompf' like a 10 ampere diode )
the PNP can nly pull ADJ to ground . it cannot send current into ADJ send the emitter -base is a diode.
now . there is still a mode of failure. if , for some reason, the opamp cranks the base more than 6 volts above the emitter of that transistor you will fry the transistor ( VBErmax) so, you can add an additional two diode in parallel to prevent that from happening.
d1
-----|<|-----
| _____ |
in -+--|lm317|----+--+---+---------+-------- out
|_____| | _|_ | | d2
| R1 /_\ R3 --|<|--|gnd
| | |d3 |
+-------+---+ --------------+--r4--|gnd
| _|_d4 ___ |
E \|P /_\ / - |-----+
B|N--------+-[r2]--< |
C /|P \_+_|-- control
|
GND
d4 now prevents more than 0.6 volts in reverse across the transsitor, but it allows current injection again ! so we add D3 to limit the reverse voltage between ADJ and OUT not that the diodes are tapped between transsitor and R2 ... we don;t want to fry the opamp output.
this is the kind of misery you end up in when trying to use a component outside of what it was designed for.... you end up throwing so much external crap to prevent trouble.