A trimester exam -- or in your case, probably, a semester exam -- should be an authentic evaluation of what each student has learned or not learned in your course. I think of the exam much like a playoff football game, or a state track meet. It's the culminating experience for the season, showing in black-and-white how well your students have done, and how well you've done teaching them.
My friend the football coach points out to me why his job is more stressful than mine. "Put yourself in my shoes," he asks. "Have someone else, not you, administer your exam. Have everyone in the community -- students, parents, alumni, administrators, EVERYONE -- watching, so they see every right answer and every ridiculous answer. And, put up a real-time scoreboard so all can see how your students are doing compared to our rival schools' students. That's my reality, every game, every season."
Although I would accept his challenge at the drop of a hat, I'm not recommending competitive examinations for all. Don't worry.
I'm putting forth this coach's point of view because for him and his team, every game is a test -- a fair, objective measure of his team's skill and preparation. No one ever complains "that's not fair, the other team threw a pass!" Everyone knows the rules up front. The referee and his crew enforce the rules, but do not give advice on strategy. In football in particular, a bit of scouting by watching film will even let the team know what "topics" will be on their "test" -- will it be "Their quarterback is fast and can throw on the run. How do you defend him?" Or could it be "Their offensive line is huge and mean, but their running back is small. What do you do?"
I encourage introductory physics teachers to make the conditions of their exams clear and consistent. The format should not be fundamentally different from the other tests that have been assigned throughout the year. Just as a football team would be thrown for a loop if they were asked to play the playoffs under Canadian rules with a 120 yard field, your students will be less successful on an exam if they see question types that are brand new. My AP tests and exams are all authentically in AP format -- 1.3-minute, no-calculator multiple choice, followed by free response questions of 10-15 minutes each. In general physics this year, all tests and exams include an equal mix of 2-minute, calculator-okay multiple choice; 4-minute "justify your answer" items; and open response items of 2-4 minutes each. No one will have to read the directions on the exam -- they will be able to dive into the physics without worrying about the format.
How would a football team react if the official were ambiguous with his decisions and instructions? If he didn't tell anyone he'd started the clock? If he called a penalty, but marked off yardage without an announcement? If he didn't bother to tell anyone what down it was? Perhaps he could say, with some justification, that a high school football team ought to know when they break the rules, and they should be able to keep track of downs. But that official still puts the teams in a situation in which they cannot show off their football prowess. The game becomes an argument about nebulous rules rather than a contest of skill.
Similarly, exam questions must be crystal clear. It's not acceptable to ask, "In an ideal gas, what happens to temperature when P goes up?" Sure, perhaps every problem you did in class involved a closed, rigid container, so perhaps you expected the class to assume a constant volume and to understand that P means pressure. But why not write the question clearly? Why make the students interpret?
"That's the problem, Greg," you might say. "I wrote what I thought was a perfectly clear and fair question, but still the students had questions about the question. How am I to know what's clear and what's not? How am I supposed to get my students to develop a sense for the level and difficulty of exam items, and then ensure that the exam matches their expectations?"
Here I strongly suggest using some sort of EXTERNAL evaluation. Aim your class from day one toward some sort of test available from someone else. The most common example of such a test is associated with the Advanced Placement program -- the AP exam's topics and difficulty are carefully controlled from year to year. I have hundreds upon hundreds of authentic test items from which I can populate my tests and exams. These questions have already been vetted for clarity, difficulty, and correctness. You don't have to be teaching an AP course to use AP test items! For years I taught a general physics course that covered only about 1/3 of the AP curriculum, but still used authentic AP items on tests and exams.
If you want a lower level than AP physics, there are other vetted test banks available. The New York Regents test has been given since the 1930s, and pretty much all published exams are available online. The SAT II physics likewise has released exams, prep books, and topic outlines available publicly. Some colleges release their freshman exams year after year; pick one and follow their lead.
No matter which publicly available exam you choose to follow, your testing becomes much more authentic through using that external source. You become less the "bad guy" for writing questions that are unclear, or too hard. Rather, you become like the beloved coach, the one who carefully prepares his team to meet the challenge of the playoffs. "That was a hard exam," they might say. "But it was exactly the kind of exam you had prepared us for. I think I did well." That's what I want to hear.