Also the first PCB, did worked. There are some issue:

• It could only run on 3.3V because of a mistake with the p-fets
• The footprint of the LEDS were so much undersized, that soldering was quite difficult an unreliable
• The TSSOP package of the shift register looks to big

As Solution a made a new revision:

• Now with smart integrated high-side switches (TPS22912 or similar) and because of these with more BGA parts
• Bigger Solder pads for LEDs. Some new routing was needed, because now only one trace fits under each LED. But OSH-Park's new 10mil Drills helps a lot.
• the shift register is now also QFN.

Here some preview of the boards: