I completely agree that it looks like an address line stuck at one value.

However, what I meant by " If so, the simulation using data type int should have revealed why as well, shouldn't it? " was that the simulator should have shown how it compensated for the stuck address line, thereby allowing the O/P to see what was going on. Of course, it's not clear that he simulated the version using int data type.

My point, however, really is that the beginner shouldn't fool with 'C' until he's thoroughly familiar with how to program the MCU in ASM. I feel that it introduces confusion and ambiguities that aren't obvious to the beginner, particularly if he fails to grasp how the compiler handles the different data types.

What's more, it's not obvious to the casual onlooker, like me, for example, what the compiler really does, though I, like most "casual" lookers-on have a fair grasp of what it SHOULD be doing.

When there's question as to whether the hardware works properly, the beginner, not necessarily every user, should stick to ASM, where the MCU behavior in response to his code is more clearly defined.

One solid clue that could readily be generated would be to use IRAM rather than XRAM for the data buffer. If they behave differently, then there's probably a construction problem. I'm rather surprised that hasn't already been tried.

RE