Dev 9-3-2011

[spagoziak]

Alex, I tried this dev version today, with the 1.7 stab code enabled. It's the best that MWC has flown to date--the stab code of 1.7 is predictable and understandible. It doesn't drift, and doesn't seem to have tuning characteristics different from the YawPitchRoll PID set. More P makes it wobble too much. Too little makes it lazy. Just like on the 3 gyro axes. The 1.8 stab code seems nervous and a bit hyperactive, like a naughty kid...

I don't know why the 1.8 stab code is so volatile--user experience has varied widely, so such random results seem very odd--but the 1.7 code, while a bit numbing to fly with, is still a more comfortable setup for me.

Incidentally, I tried out 1.8 patch 2 and didn't find much better results than with patch 1 or 0. The level function to be indecisive on where level actually was from moment to moment. Better stick movements of course, but "level" was questionable. I wish I could be more help in puzzling out how to make 1.8 perfect.. but the stuff you do is so far beyond my understanding, I can only say THANK YOU for giving this to us, it's a marvel

spag

[Alexinparis]

Hi,

For all: I released this dev with a special mode allowing the use of the former 1.7 angle calculation. It was to compare the behavior of the 2 codes at the time there was still in some case an ACC drift problem in 1.8.

Now, the drift problem in static situation should be definitively solved in 1.8 patch 2.
What is strange in your observation is:
- the code difference applies only for angle calculation
- the PID level code is exactly the same, so I expected exactly the same feedback feeling.

Maybe you could precise a little bit more where the differences are the most noticeable: FF situations ? static situations ? when you move the sticks ? ...

[spagoziak]

I was pretty tired when I posted this last night, I probably didn't make much sense

The quality of the PID changes in 1.8 (vs 1.7) is amazing; definitely the best yet. I found the change in 1.8p2 to be ineffective, surprisingly. I didn't do much forward flight, though, I only had a few minutes to fly before work. But I did spend the time hovering in a weak wind... I know my descriptions are imprecise, I'm sorry! If I could ever get a day off of work I'd post a video

My impression was that the copter still didn't flatten out when forced to do so. For example, in 1.7, after moving around, the copter will flatten out quickly, but stiffly. That's where tuning is important.. if it's too stiff, oscillations will appear. With 1.8p2, the stick response was much better of course, but the flattening process was more tentative, or looser. Perhaps it's a tuning issue for me? Do I need to increase P substantially for auto level to assertively snap to level?

I'm not sure there was much level drifting in this case (wind prevented me from getting a better sense of it) but it didn't seem that the MWC made it to true flat level; instead it chose slightly different angles every time I recovered from flying in a particular direction (so auto level trim wasn't a factor here since its tendency to lean wasn't toward a consistent direction).

Are there different things to consider when tuning autolevel in 1.8 vs. 1.7? It could make sense, seeing as they're different styles..?

spag

[cardboard]

it didn't seem that the MWC made it to true flat level; instead it chose slightly different angles every time I recovered from flying in a particular direction (so auto level trim wasn't a factor here since its tendency to lean wasn't toward a consistent direction).
spag

agreed, i also experience this.

[rocketray]

Hi,

Love this DEV, using the v1.7 level option i can fly with level on again, i found v1.8 with level on very twitchy and oversensitive, and i did spend a long time trying to find a sweet spot to no avail. i do think my flying style doesnt help though. i normally fly with level on most of the time and only turn it of to do flips etc.
Dont remove it Alex or if you do give us a patch that we can choose to use or not.

[spagoziak]

Alex, while I agree that leaving the 1.7 chunk in would be lovely, I fully understand your motivation to figure out how the 1.8 stab code can be improved. Tomorrow night I hope to do comparison hovering (and a tiny bit of forward flight) in a store after it's been closed for the night. There will be no wind so we can see pure results. I'll load 1.8, 1.8p2 and dev 9/3/11 and video each one. I'll post the links to those vids once they're up. That way you can see what I'm seeing

spag

[cardboard]

@spagoziak
I have just been out flying and have managed to get rid of the issue concerning not coming back to true level every time.
I was using an adc accel (ADXL335) and the nunchuck clone with the ITG3200 gyo (modded with direct connection such as the following, http://imageshack.us/photo/my-images/845/itg3205.png/ from the thread http://www.multiwii.com/forum/viewtopic.php?f=6&t=27&hilit=itg3200+direct

That setup was not fun to fly, I was only using it in level mode so I cant talk for just the gyro but as a combo caused me issues.

Now there is an original nunchuck and WMP in my copter and she flies perfectly again.

[spagoziak]

I'm using an original wmp and a bma020 on 1.8 dev 9-13 with 1.7 stab enabled. It flies perfectly

[rocketray]

I'm using an original wmp and a bma020 on 1.8 dev 9-13 with 1.7 stab enabled. It flies perfectly

Do you mean Dev 9-13 because i can only find dev 9-03 or was that a typo? if not where did you get it from!!!

[spagoziak]

Oops typo!

So I flew both codes tonight and video'd them. However, I'm not seeing much of a difference between the two now, so I need to make sure I didn't make a mistake and upload the same code twice, or something equally as silly. For what it's worth, there's a chance that my observations happened in moving air. There's something to consider in circumstances where wind is present... more testing, results reported as they come

[ziss_dm]

Hi spagoziak,

You probably could notice difference, if you try banking turns.

regards,
ziss_dm

[kataventos]

My impression was that the copter still didn't flatten out when forced to do so. For example, in 1.7, after moving around, the copter will flatten out quickly, but stiffly. That's where tuning is important.. if it's too stiff, oscillations will appear. With 1.8p2, the stick response was much better of course, but the flattening process was more tentative, or looser. Perhaps it's a tuning issue for me? Do I need to increase P substantially for auto level to assertively snap to level?

I'm not sure there was much level drifting in this case (wind prevented me from getting a better sense of it) but it didn't seem that the MWC made it to true flat level; instead it chose slightly different angles every time I recovered from flying in a particular direction (so auto level trim wasn't a factor here since its tendency to lean wasn't toward a consistent direction).

Are there different things to consider when tuning autolevel in 1.8 vs. 1.7? It could make sense, seeing as they're different styles..?

spag

Hi to all, this post describes my problem! When I post in "accumulative drift" I did not describe with so much words, but this is what I feel in stable mode with the 1.8 patch2.

I am sorry for not be able to help more then read, fly and report, C is like a bomb that blows my head!

Cheers,
Henrique