-15V linear power supply design issue

Thread Starter


Joined Aug 21, 2011

I am designing a dual linear power supply for my home made oscilloscope project. The circuit is designed with 2 stage ripple suppression, the 1st stage is capacitor multiplier, the second stage is LM317/LM337.

Both side 1st stage capacitor multipliers(Q5 and Q6) are working well and providing a <2mV(@50Hz) output ripple.

The +15V side is working well and the output ripple of LM317 is further reduced to <1mV.

The negative 15V side is designed to track voltage of +15V using an OpAmp(U10). Now the problem comes. I have noticed a 70mV(p-p) high frequency oscillation at around 10kHz. I first thought the oscillation is coming from Q5 and Q6 so I connected pin 4 and pin 7 of U10 to the output side of LM317/LM337. I got buffed the high frequency oscillation has jumped twice to 150mV(p-p).

Could anyone help me to understand why is this oscillation happening and how can I reduce it?

Thanks a bunch!


Ripple at +15V output

Ripple at -15V output


Last edited:


Joined Jul 31, 2011
Emitter-followers driving capacitive loads is asking for trouble, I'm afraid. It rather looks like you are indeed correct - Q5 and Q6 are very probably self-oscillating. It seems to be caused by the fact that, above certain frequencies, the transistor gain is rolling off, which means that there is phase shift, and this adds to the base's "view" of the capacitive load, such that the base looks like a negative resistance (like a tunnel diode).

Often you can counteract this by including a "base-stopper" resistor, say about 56 ohms, in series with and physically close to the base. However, I rather doubt this will be good enough. Personally, I'd get rid of the "capacitor multipliers" and just make sure you've got sufficient reservoir capacitance.

I would imagine that powering the op-amp using the regulator outputs would be preferable to using the unregulated (and unsmoothed) voltages.


Joined Nov 20, 2008
  • The LM317 & LM337 each require a minimum of 10mA in the feedback resistor string. Change R16 to 120Ω and recalculate R18 & R19.
  • Ensure the output filters are located close to the output pins of the regulators.

Thread Starter


Joined Aug 21, 2011
Thanks everyone, especially Hi-Z. It is self-oscillating of Q5 Q6. By short all three pins on Q5 Q6 the ripple dropped to sub 1mV. Problem solve.