Question about LM317 with current limiting option

[illusive]

   Hello,
   I'm in a final stage of finishing my circuit design of a bench power supply, but i come across a problem while simulating it.
The basic idea is: The V2 voltage source will be a negative voltage line comprised of LM337. It's voltage will be adjusted so that when the transistor Q3 is turned on the output voltage of the power supply will be as close to 0 as possible. R4 is a trimmer pot and it's adjusted so when the panel potentiometer is 0 ohms the output voltage of the power supply will go to 0 volts. The current limiting is implemented by sensing a voltage drop across the shunt resistor R5 (about 6-10mohm) then it gets amplified about 160 times by U3 and send to a comparator U2. The input voltage to the comparator (i.e current limit level) is set by the voltage divider R11 and R12. Pretty standard approach.
  Overall the circuit works exactly as expected but the problem comes when i set the voltage divider to 10 milivolts, this will translate to about 10mA current limit. At this point my simulation shows pretty big oscillations. All the other current ranges above that value work fine.I have noticed that the base resistor R14 of Q3 has something to do with it, because when i increase the value to something like 47K the oscillations seems to disappear, but i'm not sure this is the correct way to fix this.
  Any suggestions?

[ZeTeX]

   Hello,
   I'm in a final stage of finishing my circuit design of a bench power supply, but i come across a problem while simulating it.
The basic idea is: The V2 voltage source will be a negative voltage line comprised of LM337. It's voltage will be adjusted so that when the transistor Q3 is turned on the output voltage of the power supply will be as close to 0 as possible. R4 is a trimmer pot and it's adjusted so when the panel potentiometer is 0 ohms the output voltage of the power supply will go to 0 volts. The current limiting is implemented by sensing a voltage drop across the shunt resistor R5 (about 6-10mohm) then it gets amplified about 160 times by U3 and send to a comparator U2. The input voltage to the comparator (i.e current limit level) is set by the voltage divider R11 and R12. Pretty standard approach.
  Overall the circuit works exactly as expected but the problem comes when i set the voltage divider to 10 milivolts, this will translate to about 10mA current limit. At this point my simulation shows pretty big oscillations. All the other current ranges above that value work fine.I have noticed that the base resistor R14 of Q3 has something to do with it, because when i increase the value to something like 47K the oscillations seems to disappear, but i'm not sure this is the correct way to fix this.
  Any suggestions?

maybe try a capacitor across U2 inverting input and U2 output?
Also LM317 is not meant to be current limited by extrenal componets, getting it stable might be tricky, also its going to be slow current limiting..

[illusive]

Wow, instant fix! I tried 330pF, 10nF, 100nF and there was almost no improvement. Then i added 1uF and it worked fine.

[Kleinstein]

The input resistors at U2 are way to low, this is the reason for the large capacitance needed. Depending on the model for U2, it may work with with more realistic values to.

Anyway the LM317 is not really a good choice for a lab supply. Already using the extra transistors to get more current is something that might work, but is not really recommended - especially with changing load impedance. So if you need external transistors the LM723 is the better choice than the LM317, and it already has provisions for current limiting.

The simple standard lab supply circuit also only needs two OPs, a reference and the power transistors. So the circuit above does not even same much by using the LM317.

For simulation I would do the main simulation or the regulator starting from a DC source, so doing the rectifier simulation in a separate step. This allows to use AC simulation (e.g. for output impedance)  and thus better check for instabilities.

[exe]

The simple standard lab supply circuit also only needs two OPs, a reference and the power transistors.

And compensation