Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
On 08/10/2007, Jon S. Berndt [EMAIL PROTECTED] wrote: There are two gains that come into play. One is from FlightGear (0.0 to 1.0, as Dave M. pointed out), and the one eventually sent to JSBSim, which is in ft/sec. It looks like the one set in JSBSim can vary from 0.0 to 100.0 ft/sec. That is the maximum value expected. That seems high to me. The maximum value should be high enough to destroy an airframe. It would be worth investigating research into cumulonimbus clouds, downbursts, etc., but I don't think 100 ft/sec is out of line. Unless you want to make the 0-1 turbulence value non-linear, anything about about .1 should be very unpleasant to fly in. I did stumble into a small, developing storm cloud once in my Warrior. Fortunately, the up- and downdrafts had smooth enough transitions not to cause damage, but they pegged my VSI in both directions, threw me back and forth into uncommanded 60-degree banks, and we gained and lost a couple of thousand feet in seconds. It's almost impossible to believe the power stored in even a small storm cloud until you've seen it. All the best, David - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
This is a report I'm going to be taking a look at to help in evaluating this reported problem with the SenecaII: Evaluation of the Effect of a Yaw-Rate Damper on the Flying Qualities of a Light Twin-Engine Airplane. Technical Report NASA Technical Note TN D-5890, Research Engineering, NASA Dryden Flight Research Center http://dtrs.dfrc.nasa.gov/archive/0797/01/D5890.pdf Jon - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
is as solid as a rock with the default turbulence all the way down the glide slope. With the turbulence zero, both behave the same. The SenecaII wing rock with light turbulence appears to result from a very exaggerated adverse aileron yaw. So I did the same experiment with the c172p and pa28-140 which both use the kap140. With the default turbulence, the c172p oscillates so bad that you cannot complete the approach with the LOC needle going full stop to full stop near the runway. The pa28-161 (also yasim) is as solid as a rock all the way down the glide slope with light to moderate turbulence. Can someone tell me what turning turbulence on in FlightGear does (apart from the obvious)? Which properties are set to model turbulence, which are then presumably passed to the FDM? Jon - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
On 07/10/2007, Jon S. Berndt [EMAIL PROTECTED] wrote: I looked at the configuration file for the Seneca II in flightgear cvs. It appears to me (at least given the quick glance I took) that adverse aileron yaw (Cnda) is turned off - the data is all zeros. I'm not sure about the exact derivatives, but the real PA-28 and the 172 experience negligible adverse yaw in level cruise at bank angles under 20-30 degrees -- you can fly them with your feet flat on the floor and the slip-skid ball barely moves. In both, I think, it's the use of differential aileron deflections that does the trick. All the best, David - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
On Monday 08 October 2007 02:17, dave perry wrote: While optimizing the aitopilot config files for the Century IIB and III autopilots for the pa24 and the Altimatic IIIc for the SenecaII, a significant difference between the values of parameters (gains in particular) that give non oscillatory behavior in yasim and jsbsim became very apparent. I had to completely turn off turbulence to get stability without significant overshoot in SenecaII/Systems/ALTIMATICIII.xml. With the values submitted to cvs, the Seneca still has a wing rock in LOC REV and LOC modes with the default weather that has some turbulence. The pa24 with the same values is as solid as a rock with the default turbulence all the way down the glide slope. With the turbulence zero, both behave the same. The SenecaII wing rock with light turbulence appears to result from a very exaggerated adverse aileron yaw. So I did the same experiment with the c172p and pa28-140 which both use the kap140. With the default turbulence, the c172p oscillates so bad that you cannot complete the approach with the LOC needle going full stop to full stop near the runway. The pa28-161 (also yasim) is as solid as a rock all the way down the glide slope with light to moderate turbulence. If you watch the oscillation for either jsbsim model, you should note that when the yoke is rotating counter clockwise, the nose is yawing right and then finally swings back left, as would be expected with extreme adverse aileron yaw. Most high performance AC show very little AAY except in significant slow flight. I would not expect that small aileron deflections should move the nose counter to the roll in a SenecaII or pa24. Two questions: 1. Have others noticed this difference between jsbsim and yasim? 2. Can this adverse aileron yaw be toned down in jsbsim? Regards, Dave Perry Is auto-coordination enabled? I don't think this is effective for YASim aircraft but it may be complicating things on JSBSim aircraft. Also, are you getting the same frame-rates with both aircraft? Last time I ran FG I found that the autopilot PID controllers ran at the frame rate and not at the Ts rate specified in the controller definitions, which could make them unstable outside a fairly narrow range of fps. LeeE - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
Is auto-coordination enabled? I don't think this is effective for YASim aircraft but it may be complicating things on JSBSim aircraft. Also, are you getting the same frame-rates with both aircraft? Last time I ran FG I found that the autopilot PID controllers ran at the frame rate and not at the Ts rate specified in the controller definitions, which could make them unstable outside a fairly narrow range of fps. LeeE I'm also taking a look at turbulence modeling in JSBSim. That's one area we have not paid a lot of attention and testing to. Jon - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
leee wrote: On Monday 08 October 2007 02:17, dave perry wrote: While optimizing the aitopilot config files for the Century IIB and III autopilots for the pa24 and the Altimatic IIIc for the SenecaII, a significant difference between the values of parameters (gains in particular) that give non oscillatory behavior in yasim and jsbsim became very apparent. I had to completely turn off turbulence to get stability without significant overshoot in SenecaII/Systems/ALTIMATICIII.xml. With the values submitted to cvs, the Seneca still has a wing rock in LOC REV and LOC modes with the default weather that has some turbulence. The pa24 with the same values is as solid as a rock with the default turbulence all the way down the glide slope. With the turbulence zero, both behave the same. The SenecaII wing rock with light turbulence appears to result from a very exaggerated adverse aileron yaw. So I did the same experiment with the c172p and pa28-140 which both use the kap140. With the default turbulence, the c172p oscillates so bad that you cannot complete the approach with the LOC needle going full stop to full stop near the runway. The pa28-161 (also yasim) is as solid as a rock all the way down the glide slope with light to moderate turbulence. If you watch the oscillation for either jsbsim model, you should note that when the yoke is rotating counter clockwise, the nose is yawing right and then finally swings back left, as would be expected with extreme adverse aileron yaw. Most high performance AC show very little AAY except in significant slow flight. I would not expect that small aileron deflections should move the nose counter to the roll in a SenecaII or pa24. Two questions: 1. Have others noticed this difference between jsbsim and yasim? 2. Can this adverse aileron yaw be toned down in jsbsim? Regards, Dave Perry Is auto-coordination enabled? I don't think this is effective for YASim aircraft but it may be complicating things on JSBSim aircraft. Also, are you getting the same frame-rates with both aircraft? Last time I ran FG I found that the autopilot PID controllers ran at the frame rate and not at the Ts rate specified in the controller definitions, which could make them unstable outside a fairly narrow range of fps. I have the frame rate throttled to 25 hz as that is achievable with my setup and both AC. I have tried turning on auto coordination. This helps a little. Also, I included a yaw damper in the autopilot config file for the SenecaII. This helps most of the time but can also add to the problem. Toggle for the yaw damper is a SenecaII menu item in the patches I sent to Andy. Here are the switches from my last test from fgrun. /usr/local/FlightGear-plib/data/bin/fgfs --fg-root=/usr/local/FlightGear-0.9/data --fg-scenery=/usr/local/FlightGear-0.9/data/Scenery:/usr/local/FlightGear-0.9/Scenery-0.9.10 --airport-id=KSFO --aircraft=SenecaII-jsbsim --control=joystick --disable-random-objects --disable-ai-models [EMAIL PROTECTED] --turbulence=0.49 --geometry=1680x1050 --visibility-miles=15 --bpp=24 --fov=65 --timeofday=dusk --nmea=socket,out,5,localhost,5500,udp --prop:/sim/frame-rate-throttle-hz=25 In this test, I turned the turbulence up to 0.49. With this value, the pa24 bounces around a lot and you see the impact of what seems to be thermals and wind shear, but the Century III autopilot flies right down the LOC/GS past the inner marker for RW28R at KSFO. The turbulence means I am constantly adjusting the throttle, but the AP does a good job for all else. LOC and GS stay very nearly centered. With the SenecaII and 0.49 turbulence, it is hardly controllable without the AP, and definitely not controllable with the AP. I think Jon B. is onto something by asking how turbulence is implemented in the various fdms. Thanks for the ideas, Dave - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
I think Jon B. is onto something by asking how turbulence is implemented in the various fdms. Thanks for the ideas, Dave Does anyone know what typical values are for these two properties: /environment/turbulence/magnitude-norm /environment/turbulence/rate-hz The fact that the first property is named magnitude-norm (emphasis on the *norm*) makes me suspect that the turbulence goes from _1 to +1. But, that wouldn't do much for turbulence. And, is that 1 ft/sec? Or is it the value of the turbulence in ft/sec? Jon - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
The SenecaII wing rock with light turbulence appears to result from a very exaggerated adverse aileron yaw. So I did the same experiment with the c172p and pa28-140 which both use the kap140. With the default turbulence, the c172p oscillates so bad that you cannot complete the approach with the LOC needle going full stop to full stop near the runway. The pa28-161 (also yasim) is as solid as a rock all the way down the glide slope with light to moderate turbulence. Another thing I'd like to see is a test from steady state that reproduces the worst of the problem with no turbulence. That way I can set up a similar test in JSBSim by itself and script a set of maneuvers that duplicates the problem. Jon - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
On Monday 08 October 2007 02:17, dave perry wrote: The SenecaII wing rock with light turbulence appears to result from a very exaggerated adverse aileron yaw. So I did the same experiment with the c172p and pa28-140 I agree that they both exhibit unrealistically bad handing characteristics. The 182 and 182rg were also quite nasty, until I brutally hacked the configuration to reduce it. One slight quibble: I'm not sure I would characterize all of the problem as adverse yaw. In addition to whatever adverse yaw problems there were, I noticed an excessive amount of slip-roll coupling. That is, any slip (due to ailerons or rudder or otherwise) produced a tremendous amount of rolling moment. This is a recipe for some bad Dutch Roll behavior, which is pretty much what I observed. We agree that aileron deflection was an easy way to set off this bad behavior, but I'm not sure that adverse yaw is the whole story. Somebody needs to look at all the parameters from top to bottom. which both use the kap140. I don't think that's the primary issue ... although there might be a /secondary/ issue with the kap140 being more vulnerable to bad handling characteristics than some other autopilots are; I don't know. I recommend fixing the flight dynamics first, and only then looking to see what secondary issues might exist. Other oddity in the flight dynamics is: much too much rolling moment due to changes in engine power setting. I would have tried to fix this, but I didn't see any parameter to control this in the configuration file. I know there is some irreducible rotational drag from the propeller, and this rightly belongs in the engine/prop configuration ... but there are other things such as asymmetric wing-root incidence that are used to counteract it ... I didn't do an exhaustive search, but I didn't see that anywhere. There are many other oddities, such as fuel never being consumed from fuel tanks, effective mixture not being sensitive to altitude, EGT reading high and insensitive to mixture, etc. etc. - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
On 08/10/2007, Jon S. Berndt [EMAIL PROTECTED] wrote: Does anyone know what typical values are for these two properties: /environment/turbulence/magnitude-norm /environment/turbulence/rate-hz The fact that the first property is named magnitude-norm (emphasis on the *norm*) makes me suspect that the turbulence goes from _1 to +1. But, that wouldn't do much for turbulence. And, is that 1 ft/sec? Or is it the value of the turbulence in ft/sec? I helped implement this with Tony a few years ago. As I recall, the value was from 0 to 1, where 0 represents no turbulence, and 1 represents the most severe turbulence that we model. You couldn't represent turbulence simply in feet/second, because it consists of both movements and rotations. Note also that the current system has problems for multiplayer mode. For example, if the turbulence were set to .5 (which should be pretty bad), a J3-cub and a Boeing 747 flying into the same airport will experience the same turbulence, while in reality turbulence that would tear a wing off a J3 cub might not do much more than jiggle the drinks in the 747. As soon as you consider a world with more than one plane flying at once, we have to think about modelling the turbulence not by its effects on the plane, but by its source air motion. All the best, David - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
I helped implement this with Tony a few years ago. As I recall, the value was from 0 to 1, where 0 represents no turbulence, and 1 represents the most severe turbulence that we model. You couldn't represent turbulence simply in feet/second, because it consists of both movements and rotations. Note also that the current system has problems for multiplayer mode. For example, if the turbulence were set to .5 (which should be pretty bad), a J3-cub and a Boeing 747 flying into the same airport will experience the same turbulence, while in reality turbulence that would tear a wing off a J3 cub might not do much more than jiggle the drinks in the 747. As soon as you consider a world with more than one plane flying at once, we have to think about modelling the turbulence not by its effects on the plane, but by its source air motion. Dave Yes, the rotational aspects of turbulence will be dependent on the specific aircraft characteristic length in the relevant axis - we've got that covered, so different sized aircraft will see rotational effects of turbulence correctly (theoretically ;-). And we have a direction unit vector and a magnitude vector, which should take care of it all. What I haven't seen yet is how the magnitude of the turbulence is controlled in JSBSim from a setting that is passed through in JSBSim.cxx (the interface). I'd expect to see a maximum value of some kind, because a normalized gain doesn't tell me anything. A value from 0 to 1 is fine, if a total expected turbulence magnitude is passed in, because I want to convert it to a wind velocity. A total range of from 0.0 to 1.0 ft/sec isn't representative - and I'm sure that's not what is intended - but that seems to be what is happening with turbulence. Jon - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
I don't think that's the primary issue ... although there might be a /secondary/ issue with the kap140 being more vulnerable to bad handling characteristics than some other autopilots are; I don't know. I recommend fixing the flight dynamics first, and only then looking to see what secondary issues might exist. Other oddity in the flight dynamics is: much too much rolling moment due to changes in engine power setting. I would have tried to fix this, but I didn't see any parameter to control this in the configuration file. I know there is some irreducible rotational drag from the propeller, and this rightly belongs in the engine/prop configuration ... but there are other things such as asymmetric wing-root incidence that are used to counteract it ... I didn't do an exhaustive search, but I didn't see that anywhere. There are many other oddities, such as fuel never being consumed from fuel tanks, effective mixture not being sensitive to altitude, EGT reading high and insensitive to mixture, etc. etc. This would be interesting - I've not yet seen fuel not being consumed. Which aircraft? I wonder which version of JSBSim you are using in FlightGear? As for engine characteristics being off, unfortunately our piston model author has been very busy (even unresponsive). If I have to fix that, it's going to take some extra time, which I have very little of. If someone else wants to take a look at that, it would be really great. I'd help out as much as I could. I've been wondering if it would be a good idea to turn off some of the dynamic aero derivatives and see what happens to the performance and handling. It could also be that the DATCOM file used to create the Seneca II derivatives needs to be tweaked a little bit. If we can determine some real values for some of the derivatives, the DATCOM file can be seeded with those to help produce better results. Jon - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
Re: [Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
More on turbulence. At least in the copy of JSBSim.cxx that I have there is this code in Copy_to_JSBSim: tmp = turbulence_gain-getDoubleValue(); Atmosphere-SetTurbGain(tmp * tmp * 100.0); tmp = turbulence_rate-getDoubleValue(); Atmosphere-SetTurbRate(tmp); There are two gains that come into play. One is from FlightGear (0.0 to 1.0, as Dave M. pointed out), and the one eventually sent to JSBSim, which is in ft/sec. It looks like the one set in JSBSim can vary from 0.0 to 100.0 ft/sec. That is the maximum value expected. That seems high to me. Today I plotted out some turbulence values in a scripted run using the C172. That was instructive, and I am making some adjustments. Jon - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel
[Flightgear-devel] autopilots, adverse aileron yaw and jsbsim questions
While optimizing the aitopilot config files for the Century IIB and III autopilots for the pa24 and the Altimatic IIIc for the SenecaII, a significant difference between the values of parameters (gains in particular) that give non oscillatory behavior in yasim and jsbsim became very apparent. I had to completely turn off turbulence to get stability without significant overshoot in SenecaII/Systems/ALTIMATICIII.xml. With the values submitted to cvs, the Seneca still has a wing rock in LOC REV and LOC modes with the default weather that has some turbulence. The pa24 with the same values is as solid as a rock with the default turbulence all the way down the glide slope. With the turbulence zero, both behave the same. The SenecaII wing rock with light turbulence appears to result from a very exaggerated adverse aileron yaw. So I did the same experiment with the c172p and pa28-140 which both use the kap140. With the default turbulence, the c172p oscillates so bad that you cannot complete the approach with the LOC needle going full stop to full stop near the runway. The pa28-161 (also yasim) is as solid as a rock all the way down the glide slope with light to moderate turbulence. If you watch the oscillation for either jsbsim model, you should note that when the yoke is rotating counter clockwise, the nose is yawing right and then finally swings back left, as would be expected with extreme adverse aileron yaw. Most high performance AC show very little AAY except in significant slow flight. I would not expect that small aileron deflections should move the nose counter to the roll in a SenecaII or pa24. Two questions: 1. Have others noticed this difference between jsbsim and yasim? 2. Can this adverse aileron yaw be toned down in jsbsim? Regards, Dave Perry - This SF.net email is sponsored by: Splunk Inc. Still grepping through log files to find problems? Stop. Now Search log events and configuration files using AJAX and a browser. Download your FREE copy of Splunk now http://get.splunk.com/ ___ Flightgear-devel mailing list Flightgear-devel@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/flightgear-devel