Any output topology can oscillate, there are various tricks to reduce tendency to oscillate, but make sure the overall compensation scheme and output network are sensible first. Faster output devices can give better performance, but you may have to tame their tendency to oscillate at RF with some base-stopper resistance or ferrites. The frequency of oscillation tells you a lot - high frequency is usually local oscillation within the output stage, and typically is dependent on the load and bias and amplitude.
And don't forget layout - poor layout can provide alternaitive feedback paths you didn't know you had!
The Sziklai composite contains a feedback loop. Two "fast" transistor in this configuration often oscillate. The slow MJL21193/4 and fast BD139/140 was stable in my experience, without any "healing" parts (EB cap of power tr.). MJL 3281 and 1302 and BD-s was instable without EB cap. With 1nF EB cap was stable, but non elegant and non sophisticated (only ready for use - as is)
No zobel was added!
The enitial design was to use +/- 20V. The dissipation was around 180mw max.
I still did not completly understood why the FB cap had a small impact at the problem and the cap at the output completelly solved the issue!
I feel like the inductance of the trace in combination with the cap minimized the HF.
BTW the amp project is on: https://github.com/patsaoglou/BeatNTweet-Asma-l
Does adding a series RC Zobel to ground (just before the LR output Zobel) improve the stability?
i have tested adding rc filter on output terminals earlier before i fixed the oscilation. I did not make a difference
If the lone capacitor load helps, so should the RC Zobel. Just keep the leads short and the inductance low.
If it doesn't help, the VAS Miller capacitor may be too small.
If it doesn't help, the VAS Miller capacitor may be too small.
There is local feedback, inherent in a follower (as it follows, LOL), probably less than 'active' feeback via a GNFB loop, but way faster!
I'm not sure if my preferential/personal dislike of GNFB or the fact it messes up the grounding (assuming I'm looking at it correctly) bothers me the most, but I do have some spare transistors, so after the current project and a bit more time in LTspice I might try the ultimate test - to build it and listen 😀
I did this in an SE tube amp and it added great space to the soundstage and 'unmuddled' the sound a lot, so I'm a bit curious as to how a BGT pseudo darlington would fare. I feel it shoudl be magnificent - but the reality could be anywhere 🙂