Single-cell-ATAC normalization and merginig

Question

[C4]

Hi, I have single-cell ATAC data. I was wondering what is the consensus on merging and normalizing scATAC-seq data. Since, all the peaks are called separately for each experiment, first I generate a common peak set for all the subjects by the union of peaks and then merge all the subjects to produce one large cellXpeak matrix. Is it recommended to normalize every subject after creating a new feature matrix with consensus peaks and then merging them or merging followed by normalization. Please see I am using TF-IDF normalization and Signac workflow (Seurat equivalent for ATAC). Thanks!