| T O P I C R E V I E W |
| eric.carden |
Posted - 25/06/2012 : 22:30:59
Hi,
I've noticed that LK enters climb mode (goes to circling zoom, shows the thermal locator, saves the thermal position/data, etc.) upon normal/moderate turns while aerotowing. Is this desired for some reason? (I can't think of one and would prefer that it not do this.) I found in the code a comment and a removed line of code that look like someone tried to fix this already but got caught in a "Catch 22" situation - since circling may be needed to detect the start of free flight. I found a simple 2-lines-of-code way to fix this, though, and have tested it successfully via many Condor aerotows. If you'd like the fix, Paolo, let me know, and I'll push it to you.
Thanks,
Eric |
| 10 L A T E S T R E P L I E S (Newest First) |
| eric.carden |
Posted - 29/06/2012 : 16:15:25
Sorry for the long post, but you gave me a lot to answer. :-)
quote:
Originally posted by Coolwind
...with a patch that we know since beginning is NOT repsecting the previously behaviour...
My patch doesn't respect the previous behavior? I don't know what you mean.
quote:
Originally posted by Coolwind
Do you understand Eric that we have consolidated in 3.0 the way this function works?
Yes
quote:
Originally posted by Coolwind
You want to change it moving the 12 degrees per [second] check inside the turning engine, out of freeflight.
So you may spoil both FF engine and turning.
The only change in the turning engine is to add one check just before switching to climb mode (which includes setting Circling to true). The check is to see whether we're on tow and turning at less than 12 deg/sec. If so, then the switch to climb mode isn't made.
All I did in the FF engine was changed the condition under which circling can trigger FF. I just removed the condition that we either must be winch launching or the turn rate must be at least 12 deg/sec. Because Circling will now only be set to true if turning at least 12 deg/sec when not in FF, this removed condition will (almost) always be true (and is thus unnecessary/useless). I say "almost", because it's possible that a momentary spike in turn rate during a long shallow-ish turn on tow could make Circling get set to true with turn rate dropping before FF could be detected. It's only to eliminate this remote possibility of undesirable behavior that I removed this condition from the FF engine. It wouldn't matter much to undo/drop the change I made in the FF engine - it would just leave the remote possibility open for climb mode engaging while on tow.
So in a nutshell, all I've done is made it so that if on tow, a 12 deg/sec turn rate (in addition to the existing requirement to have been turning at least 4 deg/sec for at least 15s) is required in order to engage climb mode. This is the same turn rate required to cause FF detection, so this just keeps climb mode from engaging on tow while keeping the FF detection rules the same (except for the rare 4-criteria winch launching scenario I explained earlier - in this topic).
quote:
Originally posted by Coolwind
I will merge this patch but I will not enable it, but you may do it and do the tests!
As you wish. As I mentioned in my original post in this topic, I've tested this on many flights using my PNA with Condor (simulator). My standard procedure for all changes/patches I make is to do considerable testing (and Condor is a great tool for testing - a close second to real flight testing) before ever pushing the change for consideration. If you have any particular scenarios you'd like tested, let me know, and I'll test them. I'll continue flying (real and simulator) with this change in my version of LK and will periodically report here as long as this patch is still disabled.
Thanks,
Eric
P.S. - LK is awesome! Thanks, Paolo. |
| Coolwind |
Posted - 29/06/2012 : 11:58:10
The question is easy: why do we want to spoil a perfectly working FreeFlight detection engine, costed many hours of work and hundreds hours of testing, and consolidated in 3.0, with a patch that we know since beginning is NOT repsecting the previously behaviour, attempting to fix the turning-before-freeflight glitch?
Do you understand Eric that we have consolidated in 3.0 the way this function works?
You want to change it moving the 12 degrees per minute check inside the turning engine, out of freeflight.
So you may spoil both FF engine and turning.
I will merge this patch but I will not enable it, but you may do it and do the tests! |
| eric.carden |
Posted - 27/06/2012 : 23:06:32
Thanks. Neither of those is the scenario I described. (The 1st wouldn't meet criterion 3, and the 2nd wouldn't meet criterion 4.) So my change will handle these scenarios the same as LK handles them now. So I'm still thinking that the scenario I described looks highly unlikely. :-) Thanks for the thoughts, though. I'm glad someone's getting to fly, by the way. It's been about a month for me... :-(
Eric |
| davesalmon |
Posted - 27/06/2012 : 22:25:22
In the case of a winch launch into hill lift, then the most likely scenario is a brisk turn through about 80/90 deg, which will not take anything like 15 secs. However if thermal lift is found at thr top of the launch, you could be straight into a thermalling turn. I can't remember the FF detection, I'm personally not interested in it for my flying.
Dave |
| eric.carden |
Posted - 27/06/2012 : 20:35:42
Interesting... Thanks for the education. That satisfies parts 1 and 2 (and probably 4) of the above scenario, but what about part 3? If any one of these criteria weren't met in your scenario, then your scenario doesn't present a problem for my changes. Were you flying with LK, and if so, did it detect FF reasonably close to the time of release?
Eric |
| davesalmon |
Posted - 27/06/2012 : 19:41:19
Not unlikely at all Eric. My last flight, straight off the wire into wave enhanced hill lift of 4/6 kts. Also winches are known to produce hot spots, so it's certainly not uncommon to be in lift when you come off the wire.
Dave |
| eric.carden |
Posted - 27/06/2012 : 19:15:20
Before I share it... I found one (highly unlikely?) scenario wherein the current logic would detect FF but my changed logic wouldn't (but perhaps should). This is the scenario where all the following are true:
1. The pilot is winch launching.
2. The pilot releases into lift (i.e., doesn’t lose even 10m after release or doesn’t get a 3-sec avg negative vertical speed).
3. He has turned continuously at a turn rate of at least 4 deg/sec for at least 15s.
4. His current turn rate is less than 12 deg/sec.
I don't have much winch launching experience, but my guess is that this is such an unlikely scenario that it isn't worth worrying about. What do you think? If you think it's important, I'll do a little extra work to clean this up. It would just mean making the "winchdetected" variable global so that it can be used in Turning.cpp, not just in FreeFlight.cpp.
Thanks,
Eric
P.S. - And I changed a 3rd line of code to eliminate a remote possibility of climb mode still engaging without FF detection (due to a brief spike in turn rate at just the right/wrong time). :-) |
| Coolwind |
Posted - 26/06/2012 : 16:13:30
Yes of course! |
| eric.carden |
Posted - 26/06/2012 : 16:07:37
I guess I wasn't clear. :-) I made it so that if you aren't in free flight and you aren't circling with a high enough turn rate to cause the detection of FF, then climb mode (circling) isn't engaged. Essentially, with my patch (which is different from the commented-out line of code), climb mode engagement and FF detection occur simultaneously. The start of FF (including FF detection from circling) does work correctly (in all of my testing, anyway) with this patch. Wanna take a look?
Eric |
| Coolwind |
Posted - 26/06/2012 : 03:29:41
I did that comment, we cannot change it for that reason. We need circling to be detected also during takeoff phase.
If we apply your patch, start of freeflight will not be working correctly anymore..
|
|
|
