The figure 4 is from Application Note written by Analog Device Engineer. It is a great article, worth for reading.
https://www.analog.com/library/analogDialogue/archives/38-06/capacitive_loading.html
Therefore, I tried on the adding a Cf (15pf) capacitor parallel with the feedback resistor. It serves a great deal. The High freq oscillation is gone. The amplifier can handle 33 Ohms load with 0.7 A current without oscillation.
I also try on 100pF and 250pF across the input, but there are still little oscillation going on. So I think I would stick to the feedback capacitor.
Are you using split supplies or are you generating a virtual earth? If you generate a virtual earth then the problem could be there - probe the virtual earth under high load and see if it is stable.
Split supplies = differential input ? How to generate a virtual earth ? Sorry to ask.
By the way, I also have some other questions. Hope can get some answer.
1.) How to calculate the output impedance of a non-inverting amplifier ? All lot document only mention the gain, but not the output impedance.
2.) What actually happen in the High current Op Amp (in this case LM675)? Is same like CMOS ? With a bigger the MOSFET width, the current capacity is increase ? While for BJT, the larger the cross section of the PN junction, the higher the current capacity ?
2.) Can I make conclusion that the Rf and Ri do not determine the loading capacity of a non-inverting amplifier, but only the current capacity of the Op Amp's output does make the different ? E.g. Rf=10K and Ri=1k / Rf=100k and Ri=10k, both have Gain=10. Thus both can load same minimum load value if same Op Amp is used.
Thanks.
Chang