Just a few easy additions to Tamas' post & (who knows) to revive this thread:
R13 and R14 provide current limiting for the output-transistors
C2 is a feedforward-cap ('stability related')
C1 and R2 tailor the response of the first diff-pair Q1,2 ('stability related')
C3 tailors the response of the second diff-pair Q3,4 ('stability related')
C4 and C5 provide 'HF' supply bypassing
Biasing (tail-current) for the first diff-pair is simply the input-voltage (assumed to be zero) minis neg-supply minus 0.65 divided by R1:
(0V - -15V)/130k = 115uA
So through R3 and R4 flows around 58uA each (assuming equal currents), resulting in base-voltages for the seconbd diff-pair of around 2.26V below the pos-supply.
As said earlier in this thread, the tail-current for the second diffpair Q3,4 flows through R5,10.
Assume it's equally split between Q3 and Q4. The half of the tail-current flows through R6 (assume high beta for Q5). This biases Q5 for a certain collector-current. In essence this also flows through R10, above it.
This is (as I see it) a DC-feedback-loop: if the current becomes higher the voltage drop across R10 increases and the tail-current for the second diff-pair is decreased (since the voltage at the bases of Q3,Q4 is more or less fixed as we saw above - and it's not the result of this feedback-loop under discussion). This results again in a smaller voltage across R6. Voila feedback, so an equilibrium results.
Also, please correct when wrong, please expand etc
regards,
Peter