Power analyses for repeated measures designs are usually done with Monte Carlo simulations: you generate fake but plausible data sets with a hypothesized effect size, then analyze these data sets and count how often the effect is detected.

I agree that N=34 seems low. The effect would have to be quite large to be detected and there may be a risk of Type M (magnitude overestimated) and Type S (incorrect sign) errors. The results should therefore be interpreted in the light of a power analysis.