How to clean up shunt regulator reference voltage?

[Jim from Chicago]

Hi, recently I've been working on using a TL431B shunt regulator reference to create a tight reference signal for an opamp, but my issue is that the reference signal is not very clean. Originally I had 0.1uF between cathode and anode, and it was oscillating like nobody's business. I tried removing that cap, and it got worse. So I increased >10uF and it cleaned up the large lower frequency oscillation, but there is still some higher frequency visible junk on there around ~670kHz, and it's getting on the 48V line which feeds a lot of stuff including RF.

Summary of things I have tried with KA cap:

Removing that cap entirely (awful result)
100pF (huge oscillation)
1000pF (huge oscillation)
0.1uF (big oscillation)
10uF (improved but still clear oscillation)
22uF (more improvement but still noisy - see attached scope shot)
47uF (similar result as 22uF)
47uF with series 100ohm (large oscillation - resistor made things much worse)

I seek wisdom. (using a different reference altogether is on the table by the way)
Also, I made sure the opamp portion was not impacting by totally removing the opamp. I also made sure the 48V line was clean without the reference in there. So I know the culprit is the shunt reference. Please see attachments if you are intrigued, thank you.

P.S. - I am a bit frustrated because I am finding the claims of the datasheet & app note to be false - my reference is quite unstable with no KA cap, but they say it should be stable without any cap (quote "Unlike many linear regulators, TL43xx is internally compensated to be stable without an output capacitor between the cathode and anode.", and it's also unstable with low value caps less than 1/10th of the stability region low-end boundary per the app note. Anyway, maybe I'm missing something.

[Roehrenonkel]

Hi Jim,
 
loose the 390 pF-cap.
 
Good luck

[Ian.M]

Yes, that Ref to Gnd cap is increasing the lag in the feedback loop and thus reducing stability.