HontaiTec Quadcopters (HT F801, HT F803,...)

Durete wrote: Just tested the emulator (very interesting thing )
I can't see any strange behaviour. According the emulator is sending the right packet always, but in real world is not right

hexfet wrote: Thanks for testing We could stop here if we fix the trims to zero. If y'all are up for more investigation here's what I'm thinking. I'm fine with either choice.

Summarizing:
1) Setting channel and trim data in packet to fixed values sends "20 20 20 00 00 00" to receiver and results in no drift.
2) Setting channel data based on centered sticks and fixing trim values to zero sends "20 20 20 00 00 00" to receiver and results in no drift.
3) Setting channel and trim data based on centered sticks to "20 20 20 00 00 00" results in yaw drift.

If there's a problem it seems likely to be in the trim values being sent. Not sure what I'm missing because the trims look correct, at least with my emulator build. Durete, can you build the emulator (with the driven trims) and verify that #3 is true for your build? Greenfly you might want to try the same and make sure the result is the same. The emulator will print the packet data.
Next step would be back to a receiver capture to verify the correct data at the receiver.

greenfly wrote: What are the drawbacks with going with what we've got?

greenfly wrote: Is this what we are looking for?

4.895263   > RX_PAYLOAD          61 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 => 40 0b 00 00 4a 20 20 20 00 ff 00 0b c5 00 0a c5 91 2b 80

Because it does not look like 20 20 20 0 0 0....

hexfet wrote: The rounding logic is still not correct. I'll take a little time to get it right. It's working as intended for the channels, which is why they're all at 20, but in the trims if the mixer output is just barely less than zero it gets rounded the wrong way. Missed something obvious.

        s32 scale_channel(u8 value, s8 rangeA, s8 rangeB)
        {
            s32 result = 0;

            float offset = (float)(value - CHAN_MIN_VALUE) / (float)CHAN_RANGE;
            if (rangeA > rangeB)                
                result = (s32)(rangeA - (offset * (rangeA - rangeB)));
            else
                result = (s32)((offset * (rangeB - rangeA)) + rangeA);

            return result;
        }

greenfly wrote: The latest build looks good. No drift. Smooth flying. I think we have got it.

Great job Hexfet... thanks for all of your time and help!