that's ok, since there are serious problems with both layouts, and neither of the people posting them seem to have a clue what the schematic of their board is, or how they work. i'd like to help them out, but if they had the basics of op amp troubleshooting down, they should be able to figure it out themselves. there are a few aspects of an LM3886 that are a little bit different (such as the zobel network at the output, and the need for carefully planned ground and power traces). one of the things that makes the board layouts shown in this thread problematic, is the fact that power and ground traces have been laid out poorly, and can cause distortion or instability through magnetic coupling, and neither board used a true star ground. the second board with the isolated input ground (through a resistor) also has the wrong component values in the zobel network. a rule i have with power amp designs, is don't do anything in the design without knowing why. that includes the schematic design, the types of components used (metal film vs carbon film resistors, etc...) and the board layout. if you don't know the WHY, then when something goes wrong, you won't know WHAT or HOW. i was trying to help alexisk10 troubleshoot an offset problem. offset problems are super simple to figure out, if you know the usual causes. in this case, i think he either got the wrong isolation resistor (too high a value) or forgot the isolation resistor from the input ground to the power ground. so the input stage has no ground reference, and you get it latched to a rail. if it were me building that board, i would replace that resistor with a jumper, first of all because National's reference designs don't include it, and the WHY of that resistor isn't made clear. the WHY of National's reference design is simple and clear. the data sheet, and any application notes for the chip should be read carefully to understand the proper way to design an amp with this chip.
i saw where one poster said you don't need a ground at all.... yeah, good luck with that... let me know when you get the output unstuck from the rail...