which correlation for a PCA?

#1
I have data that has both discreet and continuous variables and, that are highly variable, not entirely independent of each other and some aren't normally distributed.
For a PCA, what would be a better method to calculate the correlation matrix:

1. Transform original variables and use a Pearson? Although I have discreet variables as well. Because: Pearson does the additional work of standardising all my variables that are on completely different scales.

2. Standardise the variables (by calculating Z-scores for each) and calculate a Spearman corr. matrix? Because: because originally my data does not meet the criterion to do a pearson's corr. i.e. linear relationship between x and y, continuous random variables, both variables must be normally distributed (small skewness/kurtosis values), x and y must be independent of each other, homoscedasticity

3. Rank the data and (since i have tied ranks) use a Pearson on the ranks (if that is possible)? Because: Wikipedia said so!...or at least that's what i think. It says, "If tied ranks exist, classic Pearson's correlation coefficient between ranks has to be used instead" (http://en.wikipedia.org/wiki/Spearman's_rank_correlation_coefficient)

:eek::confused::shakehead:(