Today we changed the 15MHz tuned PD, which is also more sensitive to infrared than the one we used before(Thorlabs PDA36A).
After the readjustment of the beam, we got a better error signal than before, about 100mV peak to peak and also the positive and the negative part are almost the same.
With this error signal, we try to lock the cavity but it seems the signal is wrong. After changing the modulation phase, we locked the cavity and it last longer time.
But even with the new PD,there is still the 'jump' when goes near the resonance, so we get rid of another possible reason.
The next step is try to get the transfer function of the loop to see more details.