My gut feeling tells me that this could be a problem related to float math losing resolution as the time increases.
Few points:
- C2000 only uses single precision floating point, while your simulation uses 'double' by default. I suggest removing the use of 'double' in your model as it will map to single in the generated code.
- Can you remove the clock block and replace it with a rolling counter? The clock block will continue to increase towards infinity and eventually lose resolution with the single precision floating point data type. The resolution depends on the other operands and I see a divider by .805e7 which will probably lead to small numbers. This is going to be challenging for the floating point data type. Having a clock that continuously increases towards infinity is not a good idea. For a PLL, I would assume that you should be able to use a rolling counter.
- Make sure that you are not using continuous time in your model. Discrete is better for emmbedded applications. While enabling sample time colors, you shouldn't see any black in your model.
I hope it helps,
Let me know,
Thanks,
Antonin.
