Torque/Sensitivity values

[shersha]

What values have beta testers/pilots been found to be satisfactory for the RV-8? I did my initial flight test on my RV-8 autopilot today and so far haven't found the 'sweet spot'. I have a FltDek180, 2 servos, and an AP74. My heading wanders about 10 degrees left/right and altitude wanders 50-100 feet either way. I've tried torque values from 50 to 100 and sensitivity values of 4 to 25. This was over 2 1/2 hours of flight testing. The closest I've gotten to was with the roll servo set at 55 for torque and 8 for sensitivity, but it still wanders as I said. If I set sensitivity at 10 or higher, it will occassionally quickly twitch back and forth in roll. The pitch servo set at 70 for torque and 7 for sensitivity is the best I could come up with, but it still hunts altitude about 50 feet either way. If it gets more than 50 feet off it seems to use too much pitch for correction, varying pitch as much as 5 degrees above and below the horizon, overshooting the altitude, and then becoming unstable because of greater and greater atlitude excursions. I'm thinking it might have something to do with the AHRS accuracy. When hand flying the only way to hold altitude accurately by reference to the instrument is by using the pink vertical speed trend bar. The horizon line on the instrument isn't always referencing zero vertical speed. If the pitch servo referenced vertical speed trend, it should work very well, but if it's using zero pitch on the artificial horizon, it's not going to work very well. I don't plan on using this instrument or autopilot for IFR operation, but I hope I can fine tune it a little better than it is now. Anybody out there that has numbers that work on the RV-8, please pass them along...

Thanks,
Scott

 

[agrantz]

Scott,
My RV8 flies very well with the following settings: pitch torque = 100%, pitch sensitivity = 20, roll torque = 100%. For roll sensitivity, start with 8, then increase in increments of 1 until the roll response starts to become a little twitchy. I currently have my roll sensitivity set at 15. Your airplane should be very stable in both roll and pitch, and should hold altitude within +/- 20 ft. If you see a yellow highlight of AP, NAV, HDG or TRK, it means the servos are slipping. AP highlited is an indication of pitch slip, while any of the others are indications of roll servo slip. Pay attention to the pitch trim messages, you may see some slippage in a turn as the nose dips. You may have to add up trim in this case, but you will get a message. I have used the AP for IFR flying, and it worked very well.
-Alan

 

[dynonsupport]

Scott,

Alan has it right.

It's important to set the torque values first, and the sensitivity second.

Set the torque to the maximum value that you feel comfortable overriding with the stick. There's no harm in setting it to 100% if you are OK with pushing through it.

The reason to set the torque as high as practicable is that if the servo slips, the autopilot cannot control the airplane. An occasional slip is ok, but if the AP:HDG-ALT is showing a yellow background a lot, you are experiencing too much slip.

That said, try these settings for an RV-8:

Pitch:
Torque: 100%
Sensitivity: 20

Roll:
Torque: 100%
Sensitivity: 5

In either case if the airplane begins to twitch (wing waggle greater than once evey 4 seconds, for example) turn it down one notch at a time.

If the airplane 'wallows' - slowly wanders with many seconds between oscillations, turn it up one notch at at time.

 

[shersha]

I tried the settings you recommended today. This is what I found. With the roll servo setting at 5, the heading wanders +-10 degrees. If I increase sensitivity, the deviations lessen. At 10 it becomes twitchy - rocking the wings back and forth rapidly while trying to maintain a heading. At 9 the twitchy goes away, but heading still constantly wanders +-5 degrees. When HDG is initially engaged, the autopilot almost immediately starts to wander back and forth. In TRK or NAV the couse/trak deviations are more dampened, but they still are there. When I try to intercept a course, it overshoots and takes about 3 zigzags across the course to capture - then it hunts back and forth as it sniffs it's way down the pink line - kind of like a bird dog. As far as Altitude hold goes, it got a little scary at first. Pitch was varying about 4 degrees down to 4-5 degrees up with vertical speeds as much as 1000 fpm - just to hold altitude in smooth air. I turned down the max vertical speed from 1500 to 300 and the altitude deviations became less abrupt. Sensitivity set at 20 to begin with as you suggest was a little harsh too, so i worked it down to 12. It's less abrupt at this setting, but setting lower results in deviations in altitude that are around 100 feet either way. At 12 the altitude deviates about 50 either way and does so continuously once it starts. It will usually stay put at altitude for a minute or so, then something disturbs it and away it goes. If I got into bumpy air the girations got out of hand. Torque settings are at 100, as you recommended here. Yellow torque slip light comes on at the ends of the excursions, HDG & ALT, as the autopilot it switching directions to go back the other way - then the light goes out until it overshoots the target (which it always does), reaches it's deviation limit and starts back the other way, etc, etc, etc.
My settings are: Roll - turn rate 2 degrees/min; bank angle 20 degrees; torque 100; sensitivity 9. Pitch - torque 100; sensitivity 12; min speed 75; max speed 190; max vertical speed 300.
Any ideas???
Scott Hersha

 

[dynonsupport]

Scott,

There are several possibilities for each problem, which may interact. The simplest thing to check right away is for play in the servo linkages.

Put the AP in to servo test mode (which will hold the servos in position) and measure how many degrees the elevator and aileron can move before the servo slips.

Give us a call at 425-402-0433 or email support@dynonavionics.com if you have any questions.

 

[shersha]

Just to keep others in the loop, I talked to Lawrence at Dynon today with the info on control movement prior to torque slippage. On my RV-8, ailerons and elevator both move 4.5 to 5.0 degrees prior to release - that's total movement up and down, so about 2 1/2 degrees each way. Lawrence says that is too much for an RV-8. He gave me some other things to check and send to him before he can make a determination on what may be happening with the servos. There should be no movement at all between the output shaft and the lever arm where the castle nut/cotter pin is. I don't think there is, but I'll check and report back. I think the play is internal in the servo., but I'll be rechecking my control linkage too.

Scott

 

[rvmills]

Scott, et al:

I did some more playing with the sensitivity on my AP on a X-C in my RV-6 today. Torque is at 100% (pitch and roll), max VSI is 500 fpm, turn rate is at 2 deg/sec, max AOB is 20 deg.

Started out with the sensitivity I had set after the last test hop (7 roll, 8 pitch). In ALT, I saw more deviations than on my last hop (smooth air on both flights). I saw +/- 60-80 feet, with VSI as much as 650 fpm in each direction. Just for the heck of it, I went to 25, and started working my way down. 25 was way too aggressive, though it held alt more closely (+/- 10-20', but really twitchy). Dropped to 20, then down in steps of 2 (18-16-14). Turns out 18 felt and performed the best today. Little twitching, and altitude +/- 10'...pretty smooth, though it did pitch up and down a little visually...but without any noticible g feel. At 16 and below, the altitude chase began, though less wallowing than at the 8 starting point (+/- 350 fpm, +/- 50-60').

In a commanded descent (rolled ALT window from 11.5 to 8.5, and rolled in a little nose down trim). At  sens 18, it chased the VSI just a bit, but settled into 500 fpm +/- 50fpm. It also captured 8500' right on, then chased it +/- 20 feet, then settled down pretty well. 18 seems like the nicest spot on my plane for pitch so far.

In roll, at 7 then 6, heading held +/- 2 degrees, and was pretty smooth. Changing the heading bug for a turn caused a relatively quick roll in to 20 deg AOB, and a bit of pitch chasing...need to work that piece a bit, but overall, it was pretty satisfactory. NAV tracking was overall pretty smooth at those settings as well.

Just some numbers for your quest!  ;D

Scott, a question on your control surface tests: How much travel in inches (or fractions) does your 5 degrees (+/- 2.5 degrees) equal. Mine move about 1/8th inch in each direction with light pressure on the corner of an aileron or elevator, and 1/4 to 1/2 inch if I press a bit harder on the center of the surface (with the servos locked in test). Any more pressure and the servo(s) slip. How does this compare to your movement?

Thanks much!

Cheers,
Bob

 

[shersha]

Bob,
I used a protractor with a dial on it to get the angular deflection. I would say that the amount of movement with the controls locked in test is at least 1/2 inch each way. I'll probably go out to the airport today (new years day) - if my wife doesn't mind - and get the data that Lawrence asked for. I'll measure that too and report back. When checking this with the controls locked in test, if it's quiet in your hangar, you can hear your servo making a small noise as it approaches the torque slip limit. When you reach the limit it will release - as you know - making a much louder noise. The deflection you need to measure should be the max deflection just prior to slippage - that's why hearing that strain noise helps to find max play just before slippage. Lawrence says that there is a finite number of steps that it slips. He wants me to measure the amount of deflection in each step too. I'll report here what I find. It may be Friday if I don't get out there today. I'll also be recording servo position values with the controls at their limt of deflection - from the AP>STATUS page.
Enjoy your new years day. My wife says 'things will be finer in two thousand niner'......
Scott

 

[DBRV10]

Hi Folks

just a thought from a "Dumb mechanical engineer " and the issue of of reactions and wandering v osscilations etc can in some instances be attributed to the mechanical and physical installation.

The PID loops in the software will have a hell of a time chasing a target if there if slop in the linkages and the sneekiest thing to find..... Flexing in the servo mounts or control rods/cable mountings. At any point in the system flexing of structure will be a problem, and when normally hand flying your plane the HUMAN PID loop in your brain smooths it out without you knowing it.

Just something to consider, and something nobody seems to have discussed as yet.

What are your thoughts so far on this issue in the installs you have seen Dynon Support?

Cheers
David

 

[shersha]

I measured the values requested by Dynon support today(Friday).

Here are the status values I got from the servos and other settings I measured.

Pitch servo full forward: +00138
Pitch servo full aft: -00016
Roll servo full left: +00097
Roll servo full right: -00095

Elevator travel within Van's spec of 27 TE up; 24 TE down.

Aileron travel within Van's spec of 30 up; 17 down.

There is no play in the aircraft control system with the aileron and elevator surfaces clamped in neutral.

I'm using the top hole in both servo arms.

There is about 3 - 4 degrees aileron movement up and 2 - 3 degrees movement down after each servo torque release before the next step. Similar movement is measured after each torque limit release in the elevator servo.

On both servos, the servo arm moves under the castle nut - there is motion between the arm and the shaft when the servo is locked in test. Moving the control surface - aileron or elevator results in this motion. After removing the servos, the pitch servo has much more movement noticeable in my hand. Holding the thick spacer that the shear pin threads into, it's easy to manually move the arm and notice movement. Doing the same thing with the roll servo, it's not as noticeable, but it's there and more obvious mounted in the plane. They both resulted in 5 degrees of play in the control surfaces. The different geometry of the aileron control system multiplied a small movement at the servo into a large movement at the aileron. The elevator bellcrank is a simple 1 to 1 ratio and the larger amount of play at the servo results in just a little more movement at the elevator.


The status numbers are pretty meaningless to me, but I remeasured the aileron and elevator movement, or play with the servos locked in test with a digital smart level. Total play in the aileron was 4.7 degrees. Elevator movement was 6.1. Dynon is coming up with a fix, I think something you can do in the field to eliminate this movement if you have any. All the servos don't have this problem and I'm betting none of the ones going out from now on will. This 'play' makes it very difficult for a servo to function properly in a responsive airplane like an RV-8. I'll let you know how the performance is when I get the corrected servos in the airplane. Monday I go to the paint shop, so it will be about 3 weeks before I get my airplane home, mount the servos, and test them.

Scott

 

[shersha]

To David,
I really don't think that is the issue here. The competitors servos do not behave this way using the same brackets and the same control system in the same airplane. I've seen this myself on numerous airplanes that are basically identical in terms of control system and servo mounts. I think Dynon's fix to eliminate movement between the servo arm and output shaft will correct the problems of pitch and roll oscillations.

Scott

 

[rvmills]

Scott,

If I'm interpreting your post correctly, the movement you are seeing in your control surfaces when the servo is locked in test (which sounds pretty similar to what I see) is not due to aircraft control "slop", but rather to the movement between the servo arm, and the servo arm shaft...am I correct in that interpretation?

If so, I'll be very interested in the fix that Dynon comes up with, as that would seem to be just what the doctor ordered to eliminate the last of the oscillations I'm currently seeing.

For reference, with pitch sensitivity set to 17-18, and roll sensitivity set to 6-7, I'm seeing gentle pitch oscillations (that can be slightly felt in the seat of the pants), with altitude settling at +/- 10-30', and VSI showing +/- 250 fpm. Sometimes its performs better than that, and once in a while it will start to chase altitude a little more aggressively. In climbs and descents, the AP will chase the 500 fpm target a little, with some slightly noticable pitch oscillations. In roll, it holds heading +/- 1-2 degrees, and tracks GPS Nav pretty well. When intercepting a direct-to from slightly off course (say 10-30 degrees off course), the AP will overshoot (about 10 degrees past) and chase the new course a little, then lock on after 2-3 "bracketing" turns.

Not too bad overall, but I'd like to dial it in further...and if a tightening method at the servo arms becomes available, that would be great. Looking forward to your results with new servos, and for word from DS on what the fix will be.

Thanks much for all your hard work on this!!

Cheers,
Bob

 

[shersha]

Bob,
I doubt that there is any control slop in your RV-6. Yesterday I clamped the ailerons and elevators to hold them absolutely still and then tried to move the push-pull tubes where the servos attach to them. There isn't any movement at all. When you move the stick, you may notice a very small amount of lateral movement of the stick, which is the stick riding on the bushing, but this motion is not transmitted to the controls. The only place you could see slop is if you don't have the rod end bearings adjusted right (very easy to fix) or the rod end bearings are worn - unlikely in an RV with less than a few thousand hours. The movement in my pitch servo between the arm and output shaft is much worse than it is in the roll servo. In fact, if I would have had the roll servo where the pitch servo is, I would probably be seeing what you are seeing in pitch oscillations, rather than what I experienced. It's possible for this kind of autopilot to hold the wings absolutely level while tracking a course or heading and for the altitude not to oscillate at all. That's the way it was in my previous airplane (RV-6). I'm not sure what the fix is, but my control arm was wobbling a little on the output shaft and allowing rotational movement around the output shaft and you can't have that. Somehow you have to allow the arm to freely move on the output shaft if something inside the servo locks and you shear the little brass threaded allen screw. I don't know why it has to be that easy to move the controls in this unlikely event though. If the arm was held on with the nylon bushing like it is, but more firmly so it can't wobble and would allow control movement - even stiff control movement, that would be OK. You just need to get the airplane on the ground at this point, so stiff control is acceptable. I don't know how you determine what torque applied to the castle nut would hold the arm firmly, yet not be too stiff to allow acceptable but restricted manual control movement. That's what engineers are for. I found something interesting while checking my system yesterday. I don't have the overcenter clamps on my servos because the mounting brackets interfere. When I remove the brass shear pin which allows the arm to move freely, independent of the servo motor as in the pin shearing, it's impossible to go overcenter as long as the arm is still riding on the shaft and connected, through the control rod, to the flight control. If I use the overcenter guards, I'll have to cut them in two or cut them down in order to fit, and in the case of the roll servo, modify the servo attach bracket. The only way you can go overcenter in my installation is for the servo control rod to grow longer or get shorter - impossible.

I'll let you know what I find in a few weeks.

Scott

 

[rvmills]

Thanks Scott...all makes sense. I was thinking how tight the feel of RV control surfaces are, and how responsive to control inputs they are, and your clamp test shows that well.

The point on an internal servo freeze and subsequent shearing of the shear screw is food for thought, leading to a question for DS:

DS, could you clarify this for me...In the event of such an internal freeze, or an autopilot hard-over, the first measure of protection is being able to break through the servo motor steps (what we check in the recommended servo test steps). If the servo motor is actually frozen internally, I'm assuming the second level of protection is being able to shear the brass shear screw through brute force, thus allowing the servo arm to rotate freely on it's shaft. Is that correct? In that case, the design would seem to require the right combination of tightness/security in normal ops to prevente unwanted slop at the servo (such as what it appears we may be seeing) and just enough mobility to allow effective control if the shear screw shears (inadvertently through a failure, or intentionally after a lock up). That's not a blast at Dynon at all, as I'm just trying to better understand the system, and I imagine that's what you guys are working hard on right now!

As a follow on question, in checking servos at the factory, does it appear that servo-to-arm movement is the result of the torque of the castle-nut, or is there some play at the shear screw, or is it something else?

Scott, your overcenter tests are also interesting. I have the overcenter bracket on my pitch servo, but not on the roll servo, as it just wont fit in the floor install (the overcenter bracket contacts the rib lightening hole, and I'd have to remove more rib material than I want to to get it in place). Howver, at full throw, it doesn't come close to overcenter, and your test provides more warm-fuzzy on that dynamic!

Thanks again for your efforts and your posts Scott. Good luck in the paint barn, and looking forward to pix of your plane when it's all gussied up!

Cheers,
Bob

 

[dynonsupport]

Bob, you are correct that the servo safeties are:

1) An electrical and software disconnect (the button) which disable the motor,
2) The ability to override the servo by simply forcing it to go where you want it to go,

and

3) The frangible joint formed by the shear screw, which can be broken if the servo mechanically jams inside its casing.

In the last case, the castle nut and washers keep the arm captured so that it can't flop around and jam other hardware, but allow it to move so that the controls are not restricted.

 

[jthocker]

This is a repost from my post on Vansairforce.net

I was having the same issues as Scott was with my servo's. Unfortunately my plane was stuck at the paint shop so I could only watch and assist Scott with his. Since Scott is now taking his plane to the paint shop, he was kind enough to let me have his newly modified(by Dynon) servo's.
We determined that both of our original servo's(4 total) had excessive amounts of "play" between the servo arm and the Shaft/capstan/disc.
This play was translating to about 4.5 to 5 degree's at the elevator(1/2 in. to 3/4 in. at the trailing edge. In my case with an identical RV8 with a tru trak installed that flys rock solid, a comparable test showed less than 1/2 the "play" of the Dynon servo's.
Dynon recognized that this was a problem and has been diligently working on a solution.
The problem lies in the shear screw. A minute amount of "play" in the screw threads in the capstan/disc allows the servo arm to move very slightly, independent of the capstan/disc/shaft. This is BAD!
The fix is to reinstall the shear screw with "Loctite 271 (red)", torque the screw to 1 in./lb, and let set for 24 hrs before proceeding with the re-installation of the servo. Dynon has extensive instructions on how to do all this if you should determine that you might be having a similar problem. Or you can send your servo's back to them and have them do it for you.
Scott volunteered his new servo's so we could carry on as "gamma" testers.
So, as the snow and cold swirled around us yesterday, precluding us from dropping Scott's plane at the paint shop, it was "swap the servo's day" at my hangar. Today, with 10 inches of unplowed snow on the runway at the paint shop airport preventing the delivery yet again of Scott's plane, we launched in my plane for autopilot testing.
After testing with a digital level and seeing elevator/aileron play similar to the trutrak plane sitting beside it, I was hopeful.
In our brief flight we were able to test most modes of the AP74, and I can finally say that this autopilot performs as well as the Tru Trak.
In pitch, it held altitude to +/- 10-20 in light turbulence.
Climbs and descents to a new altitude were at the selected(500 fpm) vertical speed, within 100-200 feet of target altitude the vertical speed rate decreased to about 100 fpm, then a perfect capture of the target altitude.
Heading mode showed +/- 1-2 degree deviations probably due to the yawing in the light turbulence.
Track mode, it kept the pink cdi needle centered.
Nav mode was excellent also. Picking a new "direct to" waypoint 90 degrees off heading resulted in no more than 20 degrees of intercept of the original "direct to" bearing. It then settled in nicely.

I AM EXTREMELY PLEASED WITH THIS AUTOPILOT NOW. Dynon's support has been very good.
__________________
Jon

 

[gianmarko]

is this a problem that affects all servos or an isolated problem? should i check my servos for play?

 

[dynonsupport]

The conservative cutoff date for affected servos is Jan 18. If you bought after that, you're definitely OK. If you bought before there's a POSSIBILITY that your servos will be affected, but it's not a certainty. Generically, if you're wondering, and you bought before that date, give them a look.

 

[gianmarko]

ok. my servos were shipped very early, oct 2008 i think, so ill take a look, are there instructions available on how to eventually correct the problem?

 

[DBRV10]

The screw instructions are here..........
http://www.dynonavionics.com/downloads/Install_Guides/Servo%20Shear%20Screw%20Replacement%20Instructions.pdf

And some discussion on this page here.....
http://dynonavionics.com/cgi-bin/yabb2/YaBB.pl?num=1235026530

That should help!

DB