The software communicates over a serial port, the instrument's comms are handled by a DLL (the source of which I do not have). The instrument is remote and in a chamber where humans cannot venture.

Spoofing data is easy and that works fine, I create test points within the 'receiver' code, to substitute test data. Using this I can test that the onward processing path is working correctly. The rest of the system will 'think' its getting measurement data.

The challenge is to test the system, not to implement a certain function.

Trying to emulate the instrument for test purposes is not easy and it seems I need the proprietary source code. Only with that can I truly emulate the comms unless I can reverse engineer it. If I create parallel code then all I am doing is testing that parallel code. I know it's a silly example but if I am testing the BFlat tone of a trumpet, I vibrate air into the mouthpiece and press valve one. I can of course create a button marked 'Test BFlat' that plays a wav file of BFlat through a loudspeaker, but the latter is a parallel path that spoofs the required result but does not test the trumpet at all. I don't know whether valve one even moves, or the pipework is completely blocked. My job is to emulate a trumpet player not simulate it making a certain noise.

My real problem is that in this case, I'm trying to retrospectively provide test facilities into a hardware/software system that is grossly deficient.