Quite silly to reply to myself.
I finnaly edit a new file and copy my original file step by step.
And it works. Until, I don't know why
UTF-16, also known as "wide characters", or simply but imprecisely as "Unicode", is character encoding where 2 bytes are assigned to every character (as opposed to common, "ASCII" or UTF-8, or other types of encoding, where 1 byte is assigned). With UTF-16, you can get 65536 different letters simultaneously, but at the obvious cost of using twice as much space. Fortran compiler (like many others) doesn't like UTF-16 encoded files, as it said.
Much of the underlying OS infrastructure in Windows is, actually, based on UTF-16. When you copy/paste text between Unicode-aware and unaware applications, lots of conversions take place and strange things occasionally happen.
What probably happened is that you pasted some UTF-16 encoded text from the web page, Visual Studio recognized that your file now has a wider range of characters than 256 and automatically saved it as UTF-16 (or asked, but you didn't pay too much attention).
In the Save As dialog of Visual Studio 200x, there's option "Save with Encoding", hidden below the arrow attached to the "Save" button.
