Atto/Conduit tuning

[Tom Harper]

Had some good flights today. Conduit is behaving well - flies hands off.

On the first flight it wallowed around in roll and pitch - so, I cut the gains in half (from 20 to 10). That cured the problem and the model flew a complete pattern.

The goal here was to position the camera for a known target. The camera is side looking at a 30 degree angle. So the distance to the target should be 2*altitude*.866. For an altitude of 92 meters that's 159 meters distance to target. The flight was planned for the model to fly west of the road at a distance of 159 meters and take photos every 8 seconds. Photos 1 & 2 show the road positioned in the foreground. Not very exciting but it calibrates the images.

The model then returned to the field and flew a loiter circle at 92 m altitude and 159 m radius. The center of the circle is approximately where we are standing on one of the flight pads. So, the photographs should all have the flight pad in the center foreground. Which they essentially do.

It's interesting to note the variation due to changing sun position. The photo with the sun behind the camera is definitely best.

The landing was fine until that little bush ran out on the field and stood in the path of the airplane. There goes the pitot tube again.

The last image is the flight path. The model completed the flight but was not crisp on the path.

So, Atto experts, do I increase the gains? At 20 the model went out of control. At 10 it's sluggish. Seems like a pretty narrow range. What say you?

Stats:

Take off speed ~50 Kmh
Min Set = 55 Kmh
Speed Set = 62 Kmh
Max Set = 80 Kmh

All Gains at 10

Camera Olympus C50 5mpx set at best resolution
ISO 80
Shutter 1/1000


Tom

[ios]

Hi Tom :)

Your flight and photos look great. Considering that your holding loiter circle is around 150m radius, thats a pretty small track to fly. While beta testing, one of the tricks I used to prevent huge deviations from the desired flight path was to fly slower!! It all depends with what your comfortable with because as you said, takeoff is ~50kph, but I would suggest leave the gains as is, and set min speed=55kph, max speed=65kph. This will definitely make an improvement to your turning circle.

[Tom Harper]

ios,

Thanks - the first flights I made were at 60 Kph and I had better tracking. But, if you get close to 50 Kph the thing gives up the idea of flying and falls out of the air. Fifty-five is not much margin.

OK, so we'll slow it down a bit.

Thanks again,

Tom

[ios]

Tom,

If you have aggressive stall characteristics then I can sympathize in your desire to increase your safety margin. I would certainly want to avoid falling out of the sky !! ;)

Have you considered a slightly longer wing or maybe just increase the distances between waypoints. I

I've found to have the best turning performance on flying wings. One of the reasons for this has been because when they bank, they also tend to decend, and this causes some elevator up input from the autopilot which tightens the turn.

Just a thought; Perhaps someone should investigate the coupling effect of bank angle to yaw, against CG location an an airplane. Maybe careful placement of CG on a conventional configuration can have the same effect as what I described above for my flying wings (i.e. increasing bank-yaw coupling instability on conventional planes to improve autopilot based turn rates!!!)

Nick

[Tom Harper]

Nick,

It would be fairly simple to add some wing area. The stab and rudder are large enough to handle it. That would also increase the aspect ratio. Wouldn't add much weight. I'll look into it.

I have a flying wing on the bench - it will be a while.

Tom

[Tom Harper]

Checking the flight log I noticed that the waypoints were hit very accurately. So, I ran the plot on another program with the flight path clamped to ground. That gives a better representation.

Looks like the navigation was very good. I do need some practice in setting up the turns.

[dmgoedde]

Tom - I don't use attitude gains to control navigation quality. I set the attitude gains only so that the roll and pitch stay on their targets smoothly without oscillation but also no lag or slop. Once that is good, then navigation can be tuned up with Dhead, Max roll angle, and proportional band width for heading error. If sloppy navigation is what you got, then you can:
1) Decrease window width of the heading proportional band
2) Increase max allowed roll angle
3) decrease DHead.

Did you get the PID playback of LOG files working on your PC?

[Tom Harper]

Dean,

PID playback of LOG files is working fine. I spend a lot of time with it.

With the plot clamped to ground the navigation looks much better. My path layout is unrealistic in places - learning curve.

I did increase max roll angle and roll rate. I'll narrow the heading window for the flights this morning.

Thanks,

Tom

[Airboatflyingshp]

Suggestion, Cross link your threads Tom it helps when one slides away and someone wants to get the whole picture later.

[Tom Harper]

Dean got us back on track. The gains were too low

Now, stability and tracking are good, but the Lipos are not giving enough duration. I have two new 8000mah packs to try tomorrow.

[Tom Harper]

The new battery pack (8000mah) makes a huge difference. Flew at a solid 68 kph for 10 1/2 minutes and landed with plenty of power. At 10 minutes you could hear the motor begin to sag. I'd say 10 minutes is the safe limit for this motor, Lipo pack and wing loading (21 ounces/ sq ft).

Stuck to the tuning pattern and one flight. The plot of the flight is oriented with North to the top. Wind aloft was gusting from N to NNE at about 20Kph. There was some roll oscillation in the NW quadrant.

Will get back to photography this weekend.

[dmgoedde]

Hi Tom - there is indeed a way I can fix the Nav code to adapt for wind heading versus intended flight path, but it requires I round out and implement the wind model. In general, a given heading error input creates a target roll as output from a nav equation. This relationship is proportional from across the "Heading Proportional Band" in degrees (heading error) and maps that to the entire +/- allowed roll angle. What I would do is depending on aircraft heading relative to the wind is compress or decompress the amount of roll applied based on heading error. I think I've seen some BYU papers on the topic, but think I got it all figured out. Just need some time to implement.

[Tom Harper]

Dean,

I was puzzled by the roll oscillation in the NE quadrant. That is the only place I noticed it. Why there?

Tom

[dmgoedde]

I'd have to analyze the LOG file, but based on experience I'd say it is NOT merely oscillation of the attitude roll, but rather a navigation thing. Is the photo with North up??? The "NE" quadrant should be top right, but that part of circle loiter is very smooth.

Also - your flying field has that TREMENDOUS large mountain right North of it... any non-circularness of the loiter could be the code trimming out the non-flat terrain from thermopile raw attitude. Without seeing a LOG I can't offer much more advice.

[Tom Harper]

Ooops - NW quadrant.

That would make sense - it 'sees' the mountain there.

Tom

[Tom Harper]

1. Purchased some good Lipos from Heads-Up RC - trashed the HK 4000mah units. Photo 2 shows the voltage for the third flight of the day (180 seconds previously used). The current line is Amps/3 to make it fit on the chart.

2. Put time on the airplane. It's something of a floater, even at 7 pounds AUW. I made the grevious error of leaving the ballast box off for one flight. You couldn't tell any difference except it wanted to climb all of the time. I believe the thick section and drooped TE gives it a Cm that will carry a lot of tail weight.

3. The 1/8" piano wire LG axle wasn't enough. Replaced it with 1/4" piano wire.

4. Still had a problem with intermittent spins in AUTO. Carl came up with the cure. There was too much coupling between the rudder and ailerons. Set the rudder gain to 4 and now it is tame and predictable. The aileron and pitch gains are about right or a little high. We noticed some roll oscillation on the due East heading.

5. On today's flight it found and tracked the circle after a bit of indecision (photo 1). The flight was at 56 KMH with an average current draw of 9.7 Amps. That would give 40 minutes duration on the present cells.

6. The only observed problem was that the hold speed was set for 65 KMH and the minimum for 55 KMH. Looks like it was holding the minimum. If Dean reads this maybe he'll comment.

So, it's a reliable platform with a reasonable flight duration. Tomorrow we'll work on navigation.

Tom

[dmgoedde]

If airspeed persists off target, then increase gain. For throttle there is only P (you set this) and I (set automatically as a ratio of P, with fixed Ki reset time).

The loiter circle is about as good as it gets. Unfortunately without an IMU the thermopile autopilot is VERY dependent on good GPS data to calculate an accurate dHead/dt. For accurate navigation the D term requires clean dHead/dt data. In a loiter circle the craft is chasing a constantly moving target of heading. Also, if this is on your flying field, those mountains to the North will make the loiter a bit lumpy. Atto is constantly working to trim the roll bias as it cycles through the terrain view.

Be careful to not simply reduce P gain on roll (if that is what you did) to fix a coupling issue with yaw. I'd prefer to increase roll D gain instead. A low P gain for roll increases chance that a disturbance will be able to push the craft over to tip stall.

[airmcn_3]

1. Purchased some good Lipos from Heads-Up RC - trashed the HK 4000mah units. Photo 2 shows the voltage for the third flight of the day (180 seconds previously used). The current line is Amps/3 to make it fit on the chart.

2. Put time on the airplane. It's something of a floater, even at 7 pounds AUW. I made the grevious error of leaving the ballast box off for one flight. You couldn't tell any difference except it wanted to climb all of the time. I believe the thick section and drooped TE gives it a Cm that will carry a lot of tail weight.

3. The 1/8" piano wire LG axle wasn't enough. Replaced it with 1/4" piano wire.

4. Still had a problem with intermittent spins in AUTO. Carl came up with the cure. There was too much coupling between the rudder and ailerons. Set the rudder gain to 4 and now it is tame and predictable. The aileron and pitch gains are about right or a little high. We noticed some roll oscillation on the due East heading.

5. On today's flight it found and tracked the circle after a bit of indecision (photo 1). The flight was at 56 KMH with an average current draw of 9.7 Amps. That would give 40 minutes duration on the present cells.

6. The only observed problem was that the hold speed was set for 65 KMH and the minimum for 55 KMH. Looks like it was holding the minimum. If Dean reads this maybe he'll comment.

So, it's a reliable platform with a reasonable flight duration. Tomorrow we'll work on navigation.

Tom


Tom that circle looks mighty good. What altitude was that pic taken from?

[Tom Harper]

That's a KML plot on Google Earth - enhanced a bit and cropped in Photoshop.

[airmcn_3]

That's a KML plot on Google Earth - enhanced a bit and cropped in Photoshop.


Sorry Tom,

I am a knuckle head...... :(

[Tom Harper]

Dean,

I reduced only the P gain on rudder. This thing has a fin and rudder the size of a barn door. So, I just reduced the rudder P Gain.

Is there a rule of thumb for the ratio of D to P gain? Starting point?

Tom

[dmgoedde]

Dean,

I reduced only the P gain on rudder. This thing has a fin and rudder the size of a barn door. So, I just reduced the rudder P Gain.

Is there a rule of thumb for the ratio of D to P gain? Starting point?

TomAh - very good. Yes, reducing rudder to reduce that coupling was the right thing to do.

Rule of thumb for P/D ratio... for most planes I fly between 20 and 50 ounces (including a customer's 8' Telemaster, which was about 180 ounces) we set D to be about 1/4 to 1/3 of P. If P were 20, then D of about 5 to 8 works well.

In the case of the 8 ounce MAV plank wing of 18" span, the D gains had to be as big as the P (1:1 ratio), which really suprised me that chaos didn't ensue. In my 20 ounce 36" span blue foam ugly wing (that was destroyed by AltitudeAP's Taylorcraft) I used about 20 for roll P gain and 7 for D gain on roll. On that plane if D was too large then oscillations ensued, presumably because estimating roll rate via differentiation of thermopile attitude has a bit of lag (even at 50Hz) and what starts to happen is the phase delay between estimation of dRoll/dt and control output gets the system chasing its own tail - so to speak. In this case a tiny bit of D gain helps to dampen the system but you also have to rely on the natural dampening afforded by the airframe.

Your Condiut plane sounds like it has a large roll moment due to large wingspan (relative to the small planes I fly). I would bet the 1/3 D:P ratio would work well for you. Now that I think about it more... of course the 18" plank MAV required (and tolerated) a high D gain... the roll moment is very small so the airframe is just not self-dampening much at all. I can tell you that plane flew like CRAP until Jack Crossfire suggested that I rely more heavily on Kd, and I got the C.G. more forward.

[Tom Harper]

AttoPilot has been working great (Photo 0 shows the last test path), but I had some reservations about the airplane. So went back to basics.

The airplane exhibited lateral instability. Just as it broke ground on take off it would swing to the right. And, at low airspeeds, it was prone to fall off into a spin.

The takeoff swing to the right, I attribute to circular prop wash. This would be felt at take off but not at flying speed. That seemed to be the case since there was no trim problem in the air.

The spin tendency I attribute to the large amount of lateral area forward of the CG. Have seen this problem on other profile designs.

Simply adding fin area would move the CLA back but would put more area in the prop wash which would make the takeoff problem worse . The solution, then, is to add fins to the stabilizer tips. A better solution would be to eliminate the center rudder and move everything out to the stab tips. Nice idea, but I'm not up for the linkage issues! The solution is shown in photo 1. Just added passive tip fins.

We tested these mods today. The tip fins seem to be effective. The takeoff was straight and we did not have any problem with spins.

As an experiment we decided to start low with AttoPilot and work our way up. These flights are planned to loiter around a single WP. The loiter circle radius is 80 m. The SET files are:

Aileron, Elevator, Throttle Pgains=25, Dgains=8.

Photo 2 - Max Roll angle = 15 degrees, Rudder gains = 0.
Photo 3 - Max Roll angle = 15 degrees, Rudder Pgain=15, Dgain=5.
Photo 4 - Max Roll angle = 20 degrees, Rudder Pgain=15, Dgain=5.


The wind was from the SouthEast at 20KMH.

In Photo 2 the model is navigating with only 15 degrees max roll and no rudder. It enters the pattern almost downwind, makes a long downwind turn then steers to within 7M of the waypoint. It is able to circle the WP but each circle migrates downwind.

In Photo 3 rudder has been added. The model enters the pattern headed directly into the wind and downwind of the WP. It steers toward the waypoint (it now has some rudder control) and touches the edge of the waypoint limit(50M). It turns downwind and comes around almost onto the circle, but falls off into a fish hook ( I'll have to think about that one - sure looks like a downwind turn - definitely a lack of system gain). It makes another fish hook then makes a long transit to a point upwind of the target and finally turns back onto the path and circles the WP.

In Photo 4 the max Roll angle has been increased to 20 degrees and the rudder Pgain is 15. Now there is just enough system gain to follow the circle (with some wind distortion) and to stay on station.

This exercise gave us a better feel for the effect of the parameters.

Tomorrow we'll increase the roll angle and add more rudder. Then fly the test pattern again.

[Tom Harper]

Been out flying almost everyday. It's getting predictable (or the crew is getting predictable).

Specs:

Speed 60 Kmh - 55 FPS
Max bank angle = 40 degrees
Turn radius ~80 meters
2 Hyperion 3S 4000mah lipos ~30 minutes
PGains at 25; DGains at 7
Rudder gain at Zero!
AUW 7 pounds

Camera - Olympus C50
Shutter 1/2000 sec
ISO 80

When the bank angle was raised to 70 degrees the model hangs in a constant spiral that is too fast for the system to correct. At 50 degrees the system gain is so high that it over corrects and cannot track a loiter circle. 40 degrees is about right.

The 80 m turning radius presents an interesting problem. If the WP satisfy radius is set at 50 m, and the model approaches the WP on a path that is tangent to the WP circle then it will turn toward the WP and never get there. It just circles the WP. We had this problem yesterday. A bit of wind probably would have jostled it into the WP circle, but there wasn't any.

A wider WP satisfy circle doesn't harm navigation since AttoPilot uses departure recognition.

The racetrack pattern in photo 0 is intentionally small to test limits. The camera is set to trigger at a WP indicated by the red line. The lens is run out to full telephoto to pin point the trigger position. The altitude setting is a little high (100 ft) so I got a lot of pictures of the parking lot instead of the pilot platform.

I'm getting confident enough to invest my Canon A650.

[airmcn_3]

Been out flying almost everyday. It's getting predictable (or the crew is getting predictable).

Specs:

Speed 60 Kmh - 55 FPS
Max bank angle = 40 degrees
Turn radius ~80 meters
2 Hyperion 3S 4000mah lipos ~30 minutes
PGains at 25; DGains at 7
Rudder gain at Zero!
AUW 7 pounds

Camera - Olympus C50
Shutter 1/2000 sec
ISO 80

When the bank angle was raised to 70 degrees the model hangs in a constant spiral that is too fast for the system to correct. At 50 degrees the system gain is so high that it over corrects and cannot track a loiter circle. 40 degrees is about right.

The 80 m turning radius presents an interesting problem. If the WP satisfy radius is set at 50 m, and the model approaches the WP on a path that is tangent to the WP circle then it will turn toward the WP and never get there. It just circles the WP. We had this problem yesterday. A bit of wind probably would have jostled it into the WP circle, but there wasn't any.

A wider WP satisfy circle doesn't harm navigation since AttoPilot uses departure recognition.

The racetrack pattern in photo 0 is intentionally small to test limits. The camera is set to trigger at a WP indicated by the red line. The lens is run out to full telephoto to pin point the trigger position. The altitude setting is a little high (100 ft) so I got a lot of pictures of the parking lot instead of the pilot platform.

I'm getting confident enough to invest my Canon A650.


Tom,

It looks like you are getting it all worked out. One suggestion. I have over 250 flights on Atto, I can tell from your pics that the WP are way too close to each other to have a smooth flight that hits all points well within the satisfy distance. Atto can hit a WP field that small but the aircraft cant at that speed. If you increase your turn radius or roll angle in the SET file you will find that you will start causing new problems. I.E. Overshoot of the line, (poor line merge), missed WP which will cause the bird to start hunting for WP, increased risk of tip stall or death spiral. As you may know the less you make Atto work the more precise the flight gets.

Anyhow how am I to tell you, Good luck! You seem to be having good success.


Chris

[Tom Harper]

Chris,

Thanks - We know the course is too tight (perhaps we have learned that the course is too tight).

Will let it rest for a while. Need to tighten things up and glue some loose joints. Then I want to finish the mount for the Canon A650.

Next week we'll do some demos for the BLM and Fish and Wildlife, then go to the East mesa for some runs at known targets.

What is your experience with transmitter range? How far is safe?

Tom

[airmcn_3]

Chris,

Thanks - We know the course is too tight (perhaps we have learned that the course is too tight).

Will let it rest for a while. Need to tighten things up and glue some loose joints. Then I want to finish the mount for the Canon A650.

Next week we'll do some demos for the BLM and Fish and Wildlife, then go to the East mesa for some runs at known targets.

What is your experience with transmitter range? How far is safe?

Tom


Tom,

Looking forward to seeing your A650 mounted!

Transmitter range..

I have flown 2.6km with a futaba fasst system; this was before Atto had the transmitter fail safe feature. 3.2km with fail safe feature but its basically limitless once it has locked out the TX; at this point all communication to the UAS is done via the ground station, as you well know you can all but manually fly the aircraft from the GCS. With this fail safe your tx range does not matter. The TX will not become active until Atto has come back within the fail safe distance you have inputted in the SET file.

Disclaimer: This was done on a private farm; this was also done before the regulations were known as well as they are now. I would not attempt long distance flights like this again unless given permission by the proper authorities.

Chris

[Tom Harper]

Chris,

I fly on a 600 sq mile ranch, with permission. I intend to use the lock out feature, but I was concerned about reasonable transmitter range. Sounds like 0ne mile is a safe assumption.

Tom

[airmcn_3]

Chris,

I fly on a 600 sq mile ranch, with permission. I intend to use the lock out feature, but I was concerned about reasonable transmitter range. Sounds like 0ne mile is a safe assumption.

Tom


Tom,

1 mile is not a problem at all. If you consider the UAS to be safe with lockout at 3/4 of a mile then 10 miles would be no different. I will tell from experience that on a dead flat to the eye area there were spots 2.5km away that increased 200' in altitude. If you were set at home with a 300' flying altitude then all it would take is a 100' tree to jump up and grab the bird when you are out over that point.

Altitude is your friend.

There are also issues such as GPS loss, if you loose GPS and the aircraft goes into navigation fail safe, this is one of the down sides of thermopiles. When this happens the bird is at the mercy of the wind pattern. The aircraft will continue to fly in its holding circle but will be pushed by the wind due to the fact that it has no way to reference where it is in the world. This again can cause you to fly into a hill or tree....... don’t ask how I know ;) :D

It’s nice to have areas such as yours to fly in.

Chris

[Tom Harper]

Thanks for the warnings!

What would cause total GPS loss?

Tom

[Gary Mortimer]

Tom it might be interesting to use Picasa to run all those targeted images as a time lapse, then you could see the red roof jumping around!

Would make a cool bit of video.

In Picasa that function is in create--video--transition style--timelapse then drop all the images in and create the time lapse.

Just a thought, maybe off topic!

Im very impressed if your getting your camera to hit the same area all the time, thats just what we all need.

G

[Irv]

Thanks for the warnings!

What would cause total GPS loss?

Tom


Remember Murphy??

[airmcn_3]

Thanks for the warnings!

What would cause total GPS loss?

Tom

No problem.

Interference between onboard systems, turns that consist of more then 40deg bank angel for extended period of time, poor sat positioning at the time oh ya and sun spots....

It does not require total GPS loss in order to go into navigation fail safe; all it has to do is loose lock which happens at 3-4 sats. If it sees no lock for an extended period of time it then continues to hold until it has received lock again. One problem here is if you have your max roll angle set to above 40deg then that is the angle your aircraft is going to fail safe into, at this angle you have little to no chance at re-gaining GPS lock again. When you get into the 4-5 sat range your navigation goes to complete crap, when you are in a turn at 5 sats you will see that the aircraft has a very hard time determining where the line is, this will cause the aircraft to begin to "hunt" for the line or WP, if your max roll angle or max rate is set too high you will have caused the bird to get lost and have a very hard time to re-gain good satellite lock.

There has not been any sat lock issues until the past 2.5 months, for some reason things have gone to crap. I have heard roomers of satellites detraining and causing poor accuracy in sat information to sat 5 becoming dysfunctional.....

Chris

[Tom Harper]

Chris,

Good stuff. Sounds like it needs some kind of 'limp home' back-up.

Gary,

Neat idea.

Atto put the trigger right on the spot every time. To get the same shot you need to stop any roll oscillation. It may make sense to use the stop motor/stop navigation feature, but maybe for only .6 to 1 sec or so.

Tom

[Tom Harper]

Chris,

I scanned the log from yesterday and I see what you mean. I does not happen all of the time but some high roll angles drop out sats.

That's an argument for keeping the roll angle down - say 30 degrees and keep the turns wide.

Tom

[airmcn_3]

Chris,

I scanned the log from yesterday and I see what you mean. I does not happen all of the time but some high roll angles drop out sats.

That's an argument for keeping the roll angle down - say 30 degrees and keep the turns wide.

Tom


Tom,

Dean and I try to keep it at 30 deg; I actually set mine at 25 deg and keep things mild. One think to keep in mind is Max rate; if this is too high the aircraft will roll too fast and also cause the aircraft to hunt and loose sats.

Chris

[Tom Harper]

Chris,

What are you using for max rate?

Have you varied the max pitch angle setting?

Tom

[airmcn_3]

Chris,

What are you using for max rate?

Have you varied the max pitch angle setting?

Tom


Tom,

My max rate is set at 70

My pitch is set at 15deg and yes I have played around with it a lot. You will find if you pitch is set too high and you are above target altitude when you set the aircraft to Auto it will pitch down rather violently. Now if you are off your target path and above target altitude and you have your roll and pitch angle settings too high the aircraft will violently pitch down to max angle, roll to max angle and proceed to target altitude and target path. For some aircraft this can cause it to go into a death spiral. Max rate will determine how fast it goes to the target angles.

Big turns are the way to go. If you have to, set the flight path over areas you are not interested in so that you can line yourself up with the area you are interested in. Don’t set your loiter circle too small, if the diameter is not matched to the max roll angle and aircraft speed you will have a poor looking circle and will not hit your intended target every time.

Once you have the max rates and max angles set you can work your way over to altitude and line merge (Navigation).

If I am talking too much let me know I don’t mind just trying to help.

Chris

[Tom Harper]

Chris,

Your comments are very welcome. Whether you confirm my observations or reveal new cautions, it's all valuable!

We have been cautious about letting the model range too far. Now that some basic problems have been solved it is time to open up the pattern.

Tom

[Tom Harper]

Flew a much larger pattern today. Used 40 degree bank angle. It held 9-10 sats in the turns.

Airspeed = 60Kph
Altitude = 133 M
40 degree max roll angle
20 degree max roll rate
20 minutes of automatic flight
24 minutes 10 seconds total
Used 6941 mah of 8000 mah available

There is still some roll oscillation. I'll lower the aileron gain for tomorrow.

This pattern is at the extreme of visual contact. The three legs of interest are at the bottom of the image. The first pass was exact. The second was OK and the third was questionable.

I believe the approach to the photo passes is wrong. Anybody know a better way to set up the pass? It might be better to use a straight grid. That would give the model maximum time to null out oscillations.

[airmcn_3]

That would give the model maximum time to null out oscillations.


Tom,

You should not have to do this if the bird is tuned properly. What is your Roll and pitch expo set at? $99, $100 in the SET file?

If you would like you can send me your LOG file from the last flight and your SET file and I can see if there are any recommendations I can make.

It’s defiantly starting to look better but I do see a few things that can be solved.

Chris

[Tom Harper]

Here's the two files. I tried a PM but couldn't figure out how to do attachments.

Tom

[airmcn_3]

Here's the two files. I tried a PM but couldn't figure out how to do attachments.

Tom


Tom,

My first suggestion is going to be $99 and $100.
$99- 15
$100- 20

With those two adjusted you can now go and lower $4 and $5
$4- 16 although I would bet you will get down to 12 or so before you are really happy.
$5- 22

Finally the Max rate is way too low. Line $40

$40- 75

Play around with Steer D Line #38. You may find you need more.

If you can send me your LOG file in its normal text form, the one you have here is for a loiter circle and tells me nothing about navigation.

I will send you a PM with my e-mail address.

To give you a little comfort about these suggestions. I just spoke to Dean and he agrees with the recommendations. (I talk with him more then my wife....... :D )

Good luck and I will give you more detailed suggestions after the LOG file is sent.

Cheers,

Chris

[Tom Harper]

It's in the email.

Tom

[Tom Harper]

I made the changes to the SET file. Will test it on circles and then fly the same pattern as yesterday.

[airmcn_3]

Tom,

Here are my recommendations. You may want to do these in steps and only change one to two things at a time. I understand it is time consuming but that is what is required to tune in an airframe properly.

$4 - 16-12
$5 - 22-18
$6 - 18-12

These settings are still too high but rather be safe then sorry, the charts that I plotted show elevator and throttle as way too high, very fast oscillations. Aileron is better then elevator and throttle but not by much.

No rudder????

Your setting here are too conservative, increase the merge angle and spread it out over a longer distance, this will help keep you from overshooting the line.....

$32 - 40
$33 - 40

I still think line $34 is too much at 40 deg. I would not go above 30 deg. This is up to you but keep in mind UAV not aerobatic airplane.

$38- 8-12 Hard to tell.

$40 - 75

$59 - 0 No need for "D" on throttle, I found this out the hard way as well. I have an almost perfect smooth line when plotting it out with a line graph; this is when the winds are at moderate conditions.

$69 - 75-100 you’re trying to make some very tight circles, as the diameter increases the aircraft will make nicer loiter circles given its not working as hard. Keep in mind thermopile limitations, they do not like too high of bank angles as its starts to get confused.....

Expo will make a significant difference.
$99 - 15
$100 - 20

$101 - You may want to set this as it’s the fail safe feature, flying at 1.5km is not very far but you sure have a chance to run into frequency loss at that range, I have mine set at 600m. At 600m you can barley see the aircraft well enough to fly it in manual so there is your distance to set it at, if your eyes are not so great, fly it out until you are starting to get uncomfortable. Look at the LOG file to find your distance so you can set your fail safe.

I noticed you have not set your servo limits; this can cause problems as well you may want to set them. I set mine with Dual rate setting on low rate; this keeps me from overdriving the servos during test.

Tom,

Take these recommendations with a grain of salt; I am not there so I can only suggest what I see on the LOG file. Keep the changes within your comfort zone....

Hope this helps and remember to not change too many things at once every bird is different.

Good luck!

Chris

[Tom Harper]

Ooops! Too, late - I made the changes and went to the field before I got your post.

Worked out OK though. Flew a couple of circles to make sure there were no death spirals. Those looked good so I put in the same path we flew yesterday. As you can see from the image it is much improved. Wind aloft was 5.3 Kmh. Airspeed was 60 kmh average.

Settings:

I'll continue to back off on the Servo gain settings. There is still some roll oscillation and a bit of throttle hunting.

In the past any rudder has been disastrous. It produces spins. Will try adding it in but I wouldn't want to have a spin at the limit of visibility. Scary!

I too have thought of increasing the merge angle. Will open that up, but shouldn't $33 to be as narrow as possible?

Previously, navigation has been sloppy at bank angles <40 degrees. That may not be the case with the changes you have suggested.

Good idea to set a limit. I'll add that.

I did measure and set the servo limits. These are really old, full size Futaba and Airtronics servos. They will take a full span of signal.

I just noticed that I can upload txt files. It took the SET file but gagged on the Log.

Thanks for the guidance and discussion. It really helps.

[airmcn_3]

Ooops! Too, late - I made the changes and went to the field before I got your post.

Worked out OK though. Flew a couple of circles to make sure there were no death spirals. Those looked good so I put in the same path we flew yesterday. As you can see from the image it is much improved. Wind aloft was 5.3 Kmh. Airspeed was 60 kmh average.

Settings:

I'll continue to back off on the Servo gain settings. There is still some roll oscillation and a bit of throttle hunting.

In the past any rudder has been disastrous. It produces spins. Will try adding it in but I wouldn't want to have a spin at the limit of visibility. Scary!

I too have thought of increasing the merge angle. Will open that up, but shouldn't $33 to be as narrow as possible?

Previously, navigation has been sloppy at bank angles <40 degrees. That may not be the case with the changes you have suggested.

Good idea to set a limit. I'll add that.

I did measure and set the servo limits. These are really old, full size Futaba and Airtronics servos. They will take a full span of signal.

I just noticed that I can upload txt files. It took the SET file but gagged on the Log.

Thanks for the guidance and discussion. It really helps.


Tom,

No problem glad I can help. It looks like a few symptoms are chilling out but as you can see it will cause others. We spent over 100 flights getting the aircraft tuned; this was primarily due to us learning the cause and effect of changing the tuning parameters. Being on the early part of the development team there were a lot of changes being made in the code to get the aircraft to perform like one would expect it to. As you have learned you can get the aircraft to fly reasonable flight paths with the adjustment of only 9 of the settings. The fine tuning comes from slowly adjusting the "Extra" features.

The Rudder...... With the setting at 0 does it even move?

I have a feeling there are a couple of issues going on here. A couple that come to mind when trying to use rudder would be. Rudder sign backwards in atto? Lots of dihedral?

If the servo sign is reversed obviously you would see some very unwanted characteristics.

Dihedral, if your bird has lots of dihedral chances are it would rather turn with rudder then aileron, this would tell me to turn the max bank angle down, and turn rudder on. You should have your auto set to a switch so you can recover the aircraft if it goes into an unwanted spiral. Turn your rudder throws down, set the limit in Atto so that it can not deflect to a point where it will cause the aircraft to spin.


$33, Path Merge Line distance = XXX ' 5 to infinity allowed in meters. This is the half width of "highway" that we merge onto the line over.

You want to spread this distance out as far as you can, you do not want to exceed the distance of your shortest leg between two Waypoints. This will stretch the merge out and will help keep from overshooting your line.

I will try and continue to help if you need.

Chris

[Tom Harper]

Chris,

Wrong rudder sign was an initial thought. Discussed that with Dean some time back. It checked out OK.

I believe that Atto rudder is simply a percent of Aileron. There is no separate calculation. At zero gain it's out of the system.

Perhaps it's time to reduce roll and carefully add rudder. Dihedral is just under 3 degrees on each wing.

It seems that $33 should be as narrow as possible in order to hold a straight line. I may have a wrong view of it's function.

Tom

[airmcn_3]

Chris,

Wrong rudder sign was an initial thought. Discussed that with Dean some time back. It checked out OK.

I believe that Atto rudder is simply a percent of Aileron. There is no separate calculation. At zero gain it's out of the system.

Perhaps it's time to reduce roll and carefully add rudder. Dihedral is just under 3 degrees on each wing.

It seems that $33 should be as narrow as possible in order to hold a straight line. I may have a wrong view of it's function.

Tom


Tom,

Hmm, well that throws out the rudder sign, and your dihedral is not too bad at all..... Man that must be one really effective rudder!

If it were my I would try what you mentioned. Reduce roll max angle, make sure the aileron gain is high enough to hold the bird in the turn but will also keep from oscillating at level flight. Increase rudder gain and decrease rudder throw. I would fly in manual and find out just how much rudder deflection I can give it before I cause it to spin, I would set the max deflection just short of that.......

$33 is measured in meters not angle, $32 is the cone angle.

See attached

[Jack Crossfire]

Interesting to see the reality & not a sales pitch. It can probably nail & bingo the paths at 1.5km scale but it takes very precise PID calibration for the particular airframe & a mental desire to get it perfect. Wonder what happens if you get the calibration perfect & then double the battery weight or add a large payload, throwing a moment of inertia way off.

[Tom Harper]

Perhaps, but the major tuning, here, is to the operator.

Take a look at the SET file above, item $51=Plane mass in grams. For similar airframes (say from LT40 to Telemaster) that may be the only change needed.

Since AttoPilot does more than just aim the aircraft, it will always require adjusting to the actual model. The good part is that the parameters are available for adjustment.

[airmcn_3]

Wonder what happens if you get the calibration perfect & then double the battery weight or add a large payload, throwing a moment of inertia way off.


Actually it’s not all that far off. We have done tests up to 2lb of payload and had no real problems. What we have had to do is change max and min airspeed and widen the distance between WP's.

But, All in All you are right its not perfect........

[Mecha]

Take a look at the SET file above, item $51=Plane mass in grams.

Tom is line $51 the plane mass, actually the mass?? or the weight?

[Tom Harper]

For my purposes, I use weight as a representation of mass, local to Socorro NM, planet earth.

[airmcn_3]

For my purposes, I use weight as a representation of mass, local to Socorro NM, planet earth.


Tom,

Statement from Dean.

"Line $51 plane mass in grams is unused at the moment. It is intended to be used for a scheme that is able to calculate how much motor power is required to lug a UAV of known mass against gravity if you know the climb rate (climb rate is known by Atto).
For example, in level flight a certain amount of motor power is required to counter the air drag. If I want to maintain that airspeed but now climb (or descend) I can know how much extra power to add (or subtract) to maintain the airspeed and now also maintain the climb rate.
A certain climb rate for a certain mass of UAV requires a certain amount of power above and beyond the power already being applied to counter the air drag".


This scheme would be for airspeed hold if the person didn't want to install a Pitot tube for whatever reason

[dmgoedde]

Chris,

Wrong rudder sign was an initial thought. Discussed that with Dean some time back. It checked out OK.

I believe that Atto rudder is simply a percent of Aileron. There is no separate calculation. At zero gain it's out of the system.Rudder is not simply a % of the aileron. That would be no different than a sort of "Y" connection. I started out trying that but it doesn't work well. What happens now in the Atto flight laws is this:
1) Aileron: control is based on roll error (for P), persistent roll error (for I) and roll rate (for D)
2) Rudder: control is based on roll target (for P only). No I or D.

So you see, the ailerons will deflect to roll the plane to target roll, then back off to maintain that roll angle. The rudder simply goes to some delfection based on roll target, thus it stays delfected, and helps the yaw moment. If the rudder gain is set to 0, then the rudder does not respond and just stays where you last left it when switching from manual to any non-manual mode.


Perhaps it's time to reduce roll and carefully add rudder. Dihedral is just under 3 degrees on each wing.

It seems that $33 should be as narrow as possible in order to hold a straight line. I may have a wrong view of it's function.

TomTom and I talked on the phone 1 hour today about various things. He is going to do a 2-step tuning process I described to get normal P gain setup for level flight, then tune the gain for max roll angle using the expo gains.

[dmgoedde]

Interesting to see the reality & not a sales pitch. It can probably nail & bingo the paths at 1.5km scale but it takes very precise PID calibration for the particular airframe & a mental desire to get it perfect. Wonder what happens if you get the calibration perfect & then double the battery weight or add a large payload, throwing a moment of inertia way off.Well, when I advertise of course I describe what it can do if well tuned, based on having flown it across things from 8 ounce MAVs to 15 Lb 8' Telemasters. For me of course I am the key expert on Atto, so it has become easy for me. Mileage may vary for other depending on their desire and location in the learning curve. This is why I brought Bill Nailor on board to help me out. I just don't have time to perfect the manuals and continue development work. We will get this to a higher level for all people starting out. The PID chart playback in the GCS from LOG off the SD card is a great way to tell quickly how well tuned the PID gains are.

[Tom Harper]

Yep, I got my yaw info out of an old manual. Thanks for the clarification.

We spent the morning observing the aircraft in ARC mode. We're getting getting the feel of it.

Actually we have a satisfactory solution. Will be photographing sites for the next two weeks. Now we're trying to get beyond just 'satisfactory'.

[Tom Harper]

Flew some more circles with reduced gains then I made a few judgements on the SET values and came up with a combination I like. The result is shown in the photo. The primary problem is the set up to enter the pattern. It was so skewed that the model missed one WP. I had enough information so I terminated the flight at about 1/3 of the battery capacity.

The targeted photo area is shown in blue/green. The model was perfectly on path through that rectangle. With 30 degree max roll angle it held 10 sats for the entire flight except for a brief period, of 9, where it appeared to be jostled by a gust.

Tomorrow I'll work on a better scheme for entering the pattern. May further reduce the P gains.

Will probably photograph some Prairie Dog towns for the Wildlife folks next week. I believe we are ready.

[dmgoedde]

Tom,

The line-hold navigation is mostly intended for really long paths so you don't end up burning up or using too many waypoints just to keep the UAV on a straight path. I would recommend that for these smaller courses you are flying to reduce the "Path Angle Max" in the SET file to 10 or even 0 degrees. This will cause your Atto to dirctly target just the waypoint and disregard the path, per se. With a courses these small the performance will still be quite good, and you will eliminate the over-prioritization of getting to the path line.

Maybe a compromise of 10 degrees could be tested. If you do set this param to 0, the navigation will default to how the original Atto beta used to operate.