Thanks, I had been doing some reading on this. I did wonder if, as the 595's are being used to both source and sink the current, whether a row is actually subject to a 20ma limit in total - because a row is limited to 1x 20ma pin for the sink, even though the source can provide 20ma x 8 pins (capped at 70ma).
Would this mean that on a fully light row, the limit is actually 20ma divided by 8 LEDs (sink)?
If this was the case, using transistors on the SINK for each row, would increase the current to the maximum of the output (70ma), meaning it would improve the brightness slightly, but still be subject to the 70ma limit for 8 LEDs you mention?
Using transistors on the SOURCE pins and SINK pins would would alleviate the problem completely.
Is this correct? I may not be using the correct terminology. I am interested understanding the concepts here as I would like to build an LED cube next and will encounter the same issue. The plan here is would be to use transistors on the sink (for a layer) and I believe a single PIN would be driving only one LED, but again the total limit on the chip would be the issue.