People have different goals when they start on a project. Lately, mine has been getting in the air while spending the absolute minimum cash.
That's lead me to trim costs in a lot of ways. Experimenting with MegapirateNG, playing with cheap props, and flashing my own ESC's.
I've had a lot of success, but one problem has continued to plague me: I could not make Arducopter (or MegapirateNG) fly with cheap props. It would fly like a dream using DJI's excellent 1038 prop (still my favorite for F450-style quads), but with any other prop, my model would bounce up and down in any altitude-hold mode.
Most frighteningly, these oscillations would become fatal if I changed modes while flying at even moderate speeds. I wrote about that particular problem here: http://diydrones.com/forum/topics/violent-over-correction-on-mode-c...
Over the weekend, I finally discovered the source of the problem: I was using simonk-flashed ESC's.
I went back to using stock F-20A ESC's from HobbyKing, and now my quad flies great with any prop in my toolbox. I tried super-cheap 1045's, 9047's, 9443's, so-called "carbon mixed" 1038's, and of course my favorite DJI 1038's. The quad still flies best with DJI 1038's, but all the other props fly quite well in all modes.
Try dropping the value of THR_ACCEL_P. I had a similar problem with a big hexacopter, the motors would keep surging in loiter. I had to drop the value to 0.25.
I think this is an oscillation in the z-direction, similar to an oscillation in roll or pitch when the P value is too high.
I thought my copters flew well on Hobbyking F-30As on stock firmware until I got them dialed in with SimonK (same ESCs reflashed), much crisper with SimonK. Also everything seems to run much cooler, even after giving the motors a hard time they are still at ambient temperature, ESCs or only a little warm.
I tried putting THR_ACCEL_P down to 0.5, which I thought was the lowest it could go. It's OK to go below 0.5?
OK, I tried lowering THR_ACCEL_P to 0.25 and THR_ACCEL_I to 0.5. It did smooth out the altitude oscillations in loiter, but it seemed to exacerbate the violent over-correction I wrote about in the other thread. I have nearly identical F450's with and without simonk, so I'm good to experiment.. but the search continues.
I do use simonk on my multiwii setup, and it's great.