A universal Turing machine can simulate any Turing machine. Right got that. But is it always possible to determine the exact nature of a given Turing machine?

Suppose i hand you some kind of hardwired word processor type thing. One of those weird electronic typewriters, now pretend it wasn't programmed in any language you know (and the machine code is unknown too). All you have is the device but you have available to you all the physical information - so you can watch voltages changing across every chip in the circuit as you use the word processor

Purely with that physical data is it feasible to work out how to simulate the machine on a computer? If it's not feasible how simple do we need to make the machine before it becomes feasible and what kind of scaling in terms of difficulty are we talking about?