What's new
Van's Air Force

Don't miss anything! Register now for full access to the definitive RV support community.

Running the Autopilot Gauntlet

Jamie

Well Known Member
First of all, a disclaimer: The purpose of this writing is not to bash any autopilot manufacturer, but rather to share my personal experiences with my fellow builders and dreamers. Each vendor I have dealt with has been unquestionably committed to helping me with my issues and have gone above and beyond to make me a satisfied customer.

When I first flew my airplane on 7/28/2007 it had a Trio Ez-Pilot installed. I had flown in a couple of other airplanes with the EzPilot installed (Mike Stewart's old -6A and his new Super 8) and I was very impressed with the performance. I was generally impressed with the smoothness and the PLED display which showed data from the NMEA stream (very handy).

However, I could never get the Ez Pilot to work right in my -7A, no matter what I tried. [Thread is here] I isolated everything electrically. The EzPilot would rock the wings. I tried every setting. I talked to the Trio guys quite a bit and although helpful, we were never able to come to a conclusion to the issue. I ended up hand-flying the airplane to and from OSH, Pennsylvania, VA, etc. Don't get me wrong, the airplane trims beautifully and I had no problems hand flying it, but an autopilot would have been nice to give the PIC a little break from maintaining heading and especially altitude.

Next I saw an advertisement for Dynon's new autopilot on vansairforce.net. Ok, that was the ticket for me! Only $1500.00 for a two-axis autopilot (I already had a D10-A). The autopilot install was fairly straight-forward, it just required a lot of work (mostly pulling wires).

I was very disappointed that I was never able to get the Dynon to work quite right. [Thread on this issue here] For me it would have been an ideal solution. The AP would hold altitude, but not smoothly. You could always "feel" the autopilot agressively trying to maintain altitude and even then it would deviate by sometimes 30 feet, then hunt sporadically around the target several times until it finally got close again.

The AP would hold a course or heading, but the slightest attitude disturbance in turbulance would rock the wings two or three times. To this day I have not flown in another airplane with the Dynon AP. I did talk to three or four other Dynon AP customers that reported that their APs worked fine, so I am still at a loss to understand exactly what was happening here. Maybe I just have a sensitive rear?

Let me make this abundantly clear. Dynon stands behind their products. They refunded my money, even after asking me if I wanted to take a full refund, keep the servos and become a member of their beta program! That was indeed a tempting offer, but with a 16 month old son at home I simply didn't have the time to do it. My flying is sometimes rare and is therefore very cherished. Consequently, I just want everything in the airplane to work with no squawks whatsoever. If there is a squawk it will just annoy me to no end.

This basically left me with just one other AP vendor that caters to experimentals -- Trutrak.

I spoke with Lucas at Trutrak and explained to him my situation with trying other APs and having issues. He made me an incredible offer. He said they guaranteed their AP to work in my airplane and they would ship me any of their units at no charge to try it (I think they were taking pity on me at this point). If I liked their unit, I could pay them for it. If not, I could just return it.

Ok, so how do you pass up a deal like that? I actually hesitated on the offer for a couple of weeks but then decided with the upcoming summer trips my wife and I have planned (NY, etc) it would be a good time to try it. I called Lucas back up and he asked me which one I wanted. I asked for the Digiflight II, because basically that's the functionality that I wanted. I don't really care for VS select or GPSS. I may one day, and upgrading later will be rather easy.

I received the unit and calibrated it per the installation instructions. My first impression with the Trutrak was that there are a *lot* of configuration settings to allow you to tweak the AP performance to your likes and to your particular airplane. I thought to myself that if the Trutrak wouldn't fly my airplane well that I would just have to give up on the search for an AP.

I made the first flight with the AP, climbed to 10,500 and started testing and doing in-flight adjustments. First I simply tried the roll axis. Perfect. Next, pitch. It was oscillating a little bit. I tried combinations of tweaking the "activity level", "microactivity" and "static lag" settings. I ended up using a static lag setting of 1, a microactivity setting of 1 and leaving the activity level to a setting of 2 (what is called for in the calibration settings). It seemed like I had it dialed in but the true test would only come with a long trip.

Saturday, April 4th we had planned a trip to North Carolina to see some friends. This trip would take us from our homebase (LZU) near Atlanta to just east of Raleigh, NC (LHZ - Franklin County). Going around the Charlotte class B made this trip 317nm. I took off and at 1000ft AGL I engaged the roll axis and trimmed for my desired climb. After leveling off and trimming at 9500 in the smooth air that is so typical of an 0630 launch, I engaged the altitude hold.

Performance? Absolutely Perfect. The flight up was 1.8 hours. The AP never once deviated from altitude. Even when moving my legs around, putting my right arm around my wife's shoulders, etc it never moved. My GX-50 is driving the Trutrak and will display course error to the thousandths place. The most deviation I ever saw was 0.002nm. That's about 12 feet!

This is what I saw on most of the trip. The 0.000 is the distance from the course centerline.
2009-04-04.2233.jpeg


On the way back weathermeister.com was telling me to expect some rather fierce headwinds. I climbed up and the winds aloft forcasts were actually pretty good. My EFIS was showing a 50kt wind, 5 degrees off the nose. Great. It was very smooth up there. Ok, so we stay there at near-Skyhawk speeds or we descend down into the bumps and accelerate to Skylane speeds. We descended to 6500 and put up with the bumps (mostly light chop). The flight back was 2.2 hours and I had the AP on all the way. It handled turbulance perfectly! It corrects the way any pilot would correct -- better even.

The entire flight was made much more relaxing with the autopilot. I never once felt like it was going to do something unexpected, even in turbulence.

I obviously called Lucas and paid for the Trutrak.

I am very happy with this AP. Why did I have so much trouble with the others? Perhaps it will always be a lingering question in the back of my mind. I can only speculate, but I think it may have something to do with the abundance of settings available on the Trutrak. The Dynon really only has two settings; Torque and Sensitivity (which is akin to the Activity Level on the Trutrak or "Gain" on other APs). The trick to *really* smoothing out my AP's performance was the static lag. When I increased that setting from 0 to 1 the pitch performance noticeably smoothed out. Increasing the microactivity setting was just icing on the cake for a near-perfect ride.

And when I say near-perfect...I mean that. The Digiflight II flies the airplane very, very well.

Some more pics from our flight up and back:

Sunrise after 0630 launch.
2009-04-04.2230.jpeg


A blurry shot (taken in low light) of the panel and the DII over on the right (replaced a whiskey compass).
2009-04-04.2236.jpeg
 
Last edited:
Mine too

About the only time the AP when whacko was when it sheared the brass shear screw in pretty heavy turbulence..Oh well not a big deal to replace.

Frank
 
Nice report Jaime! Boy, sure glad that is behind ya!! :) And like ya said, what great customer service from all sides! No doubt due to your extraordinary patience and great attitude!! Happy Flying!

Cheers,
Bob
 
TruTrak all the way!!

I think you'll find that very few people have trouble with the TruTrak. When something does occur you'll get quick, professional service.

I have the Digi Flight II VSGV (or whatever the alphabet is) and have had ZERO performance issues. The only problem I've had is a line running through the display. The control head was sent in and promptly returned with no charge.

I'll be putting the same unit in my new project tying it all in with the AFS Flight Deck!!!

Glad you got your situation figured out. I can't imagine flying any real long legs without it. Heck I use it flying to work and it is only 15 minutes each way.
 
Approaches with the GX50?

Jaime,

Wondering if you have done an IFR approach with the GX50 coupled to the TruTrak. When the GX50 changes its scaling going from enroute to approach enabled to approach active, how does the TruTrak handle that? Or, does the TruTrak use the GX50's GPSS output, instead of its MovingMap output, for guidance?

--Paul
 
Jaime,

Wondering if you have done an IFR approach with the GX50 coupled to the TruTrak. When the GX50 changes its scaling going from enroute to approach enabled to approach active, how does the TruTrak handle that? Or, does the TruTrak use the GX50's GPSS output, instead of its MovingMap output, for guidance?

--Paul

TruTrak uses the RS232 output only.
 
Good write up Jamie. I have delt with Lucas & several others at TruTrac, and they are the best.
 
Jamie
Thanks for the update. It has been my feeling since I installed the Dynon autopilot that it was initially no where near as good as the Tru Trak.
My initial concern when I first flew it, was that the beta testers had never flown a Tru Trak and had nothing to compare it to, ( I know that wasn't the case) and felt the AP was ready for release. When Scott Hersha and I tracked down the loose shear screw,we thought we had found the culprit.
I have not flown any long cross countries to really test the autopilot, and really haven't turned on the AP much since we resolved the shear screw issue.
My gut feeling is the Dynon AP at this point is, it's adequate for the money. It's alot of bang for the buck. I like the features and integration with the EFIS. It suits my mission of VFR cross country workload reducer.
The Tru Trak autopilot reflects their many years of experience in the autopilot business, and if I had an IFR mission, I would probably choose them also.
I hope Dynon continues to tweak their AP( static lag???)and I wish Tru Trak would get their website updated so you could actually compare what an EFIS/AP/engine monitor might cost!
Best regards
 
I chose Dynon for both the price and continuity with the other Dynon stuff. I have everything but the servos and don't intend on getting those until I'm close to flying. Hopefully by then things can get worked out.

Friday I was a safety pilot for a friend on an IFR flight who has an RV-7 with a Digiflight II. We went into the clouds around 2000 and remained IMC at our 4000 assigned altitude. Using altitude hold and heading mode, the Digiflight was rock solid with little or no ocillations after turbulance. After a couple of ILS approaches we did a coupled GPS approach (GPSS Horizontal only) and it was absolutely flawless. Very, very nice.

Dynon is a great company and I'm praying that they can get close to the TruTrak performance.
 
RS-232 AND ARINC both!!

TruTrak uses the RS232 output only.

The Digiflight II series indeed does use RS-232 but also uses ARINC for GPS steering - both vertical and lateral for fully coupled approaches to units that output the appropriate data. That being said, as others have mentioned the older units like the GX series only ouput RS-232, so it's a limitation of the GPS, not the autopilot. I'm not 100% sure, but fairly certain that you will be unable to fly a coupled approach with the GXx no matter what kind of autopilot you have, because as far as I know the GX's don't output vertical guidance on any channels.

For lateral steering this shouldn't be much of an issue, but without ARINC outputs to feed the Autopilot, you're probably not going to get a phenomenal result.

Anyway, just my 2 cents to clear up any confusion. The single axis TruTrak AP's only accept RS-232, but from the Digiflight II series and up, they all will accept an ARINC input.

Cheers,
Stein
 
Last edited:
I'm not 100% sure, but fairly certain that you will be unable to fly a coupled approach with the GXx no matter what kind of autopilot you have, because as far as I know the GX's don't output vertical guidance on any channels.
The GX's don't have vertical guidance on their digital channels, but they can be configured to output GPSS roll commands over RS232, and this works very well for coupled LNAV approaches (my experience here is using this with the Trio Pro Pilot).

On the other hand, there is a potential problem if an autopilot instead uses the XTK (crosstrack error) field in the MovingMap stream from the GX for guidance during an instrument approach, because the GX changes the error scaling by a factor of 16 without the autopilot being told about it.

So I was wondering if Jamie had tried his GX50 with the TruTrak to do an approach. I'd be interested to know how well that works. Or what should have basically the same effect, go into the SYS -> Navigation Info pages and set CDI scaling manually 5.0 to 0.3nm while tracking an enroute flight plan.

We had a thread going on this issue with GX's here:
http://www.vansairforce.com/community/showthread.php?t=36859&page=3
Looking back at that, I see that Jamie said his GX50 didn't have the GPSS firmware, so unless he upgraded, I guess his TruTrak isn't using that.

--Paul
 
I agree with Jon Thocker about the bang for the buck aspect of the Dynon AP and with it's integration with other Dynon hardware. It's adequate for a VFR workload reducer. But having flown both the TruTrak and the Dynon, I would pick the TruTrak if planning an IFR panel. My Dynon just doesn't hold altitude as well or as smoothly, no matter what you program and the lateral modes aren't as perfect either. It'll get you from point A to point B, but if you're trusting it to hold altitude IFR, you may get busted. I have the new modified servos, but they are no where near IFR capable. I think Dynon will eventually make it better. For now it's adequate for me for what I need.

Scott
 
Just one more to weigh in on the Tru Trak - it's very good at holding the course info from the GPS and the altitude. I'm very pleased with it.

Dan
 
The Digiflight II series indeed does use RS-232 but also uses ARINC for GPS steering - both vertical and lateral for fully coupled approaches to units that output the appropriate data. That being said, as others have mentioned the older units like the GX series only ouput RS-232, so it's a limitation of the GPS, not the autopilot. I'm not 100% sure, but fairly certain that you will be unable to fly a coupled approach with the GXx no matter what kind of autopilot you have, because as far as I know the GX's don't output vertical guidance on any channels.

For lateral steering this shouldn't be much of an issue, but without ARINC outputs to feed the Autopilot, you're probably not going to get a phenomenal result.

Anyway, just my 2 cents to clear up any confusion. The single axis TruTrak AP's only accept RS-232, but from the Digiflight II series and up, they all will accept an ARINC input.

Cheers,
Stein

Stein is right - sort of. It IS however possible to simulate a precision approach mode with most TruTrak autopilots even if the device supplying course info does not also supply vertical steering.

My prior plane was a Lancair 235 with fixed pitch prop. When shooting an approach in strong thermal activity, the coupled mode (ILS) on the approach resulted in large pitch changes which in turn caused large airspeed deviations (and could quickly result in being "behind the power curve").

Because of this I created my own "precision approach mode." Couple to the course, and at Glide Path Intercept dial in a descent appropriate to your approach speed and the glide path angle. Thus, if you are 1000' 3 miles from the touchdown zone with a ground speed of 90mph (1.5 miles per minute, or 2 minutes from touchdown), a 500fpm descent will put you on the ground at the touchdown point.

I use this technique even on non-precision approaches, because it almost always places me at the FAF exactly as I hit minimum approach altitude (rather than driving down a lot early and being tempted to zig-zag down low to maintain VFR).

I now have a more powerful plane with C/S prop and a Digiflight II VSGV coupled to a GNS480. I'm looking forward to trying the ARINC GPSS precision approaches - but if I get a lot of pitch over-correction I still intend to go back to my old trick.
 
Analog ILS signals are WAYYY different animals than a WAAS GPS approach. The autopilot gets better information that is more accurate and doesn't deviate throughout the approach as you near the ground like you do with an ILS.

When you get a chance to fly a WAAS GPS approach (LPV or otherwise) the plane will feel like it's on rails compared to an ILS....you'll love it!

Cheers,
Stein
 
So who has had really good results with the Dynon A/P? I assume many have but I am really keen to know as I am launching a new plane with the full suite soon.

Perhaps the next round of software thats in beta test now will iron out some of the bugs.
 
The Digiflight II series indeed does use RS-232 but also uses ARINC for GPS steering - both vertical and lateral for fully coupled approaches to units that output the appropriate data. That being said, as others have mentioned the older units like the GX series only ouput RS-232, so it's a limitation of the GPS, not the autopilot. I'm not 100% sure, but fairly certain that you will be unable to fly a coupled approach with the GXx no matter what kind of autopilot you have, because as far as I know the GX's don't output vertical guidance on any channels.

For lateral steering this shouldn't be much of an issue, but without ARINC outputs to feed the Autopilot, you're probably not going to get a phenomenal result.

Anyway, just my 2 cents to clear up any confusion. The single axis TruTrak AP's only accept RS-232, but from the Digiflight II series and up, they all will accept an ARINC input.

Cheers,
Stein

This is correct. Jamie's particular AP does not have GPSS capability. You need a G in the name to have that. If you do, then it will fly the lateral roll commands from the GX series, although they are not a very good source. The update rate of the GX series GPSS commands is too slow to give reliable performance. That is the info I received from our software engineer who writes all the software for the APs.
 
The update rate of the GX series GPSS commands is too slow to give reliable performance.

Rate is one per second from the GX, which has worked well in our 165kt RV-6 with Trio A/P.

Still wondering how well the TruTrak works when using the GX's MovingMap stream for guidance and the GX changes the scaling on its crosstrack error "G" sentence, either during an approach or manually. Lucas or Jamie, if you could comment on that, I'd be interested to know.

--Paul
 
A Lesson Learned

When I initially installed my Digiflight II in my Rocket, I has some problems with the AP holding the heading correctly. Just be aware that minor problems with your rigging can make it so that the AP is fighting against where the airplane wants to go. I had a few rigging issues to resolve during my flight testing and after I did, the AP did a much better job of holding the heading smoothly, rather than rocking back and forth a little.
 
Thanks Jamie,

Did you have to use the microactivity settings or the static lag settings to get your trutrak AP to fly accurately or did it fly without oscillations with the default settings? My question is did you have the same problems as with the dynon until you used these settings to fix them?

I would say that the pitch behavior was "similar", but not the same. The best I could get the Dynon dialed in, it wasn't oscillating so much as it just wasn't holding altitude very well. Once it deviated from altitude, it would oscillate around the target with decreasing amplitude until it finally got it locked in again.

Without the static lag adjustment on the Trutrak, it was more or less oscillating very smoothly and with a constant frequency.

The micoactivity on the Trutrak was just the "cat's meow", as they say. The airplane wasn't really oscillating very much with this setting on 0, it was just hunting a very little bit. You really had to be paying attention to it to notice it. The microactivity adjustment just smoothed it out.
 
Wow, not one satisfied Trio guy out there?

Jamie, sorry you had such a rough time with the APs. I started with a navaid and didn't like the mechanical gyro, so when I decided to upgrade, I looked at TruTrack and would have to install the whole system, AP head and servo. Well, being cheap, I kept looking and the Trio guys sold me the A/P head only and it worked because they use the Navaid servo. They even made up a wiring harness so I could unplug the Navaid and plug in the EZ-Pilot A/P. It worked from the begining and saved me about $1K. It needed some tweeking so when I went to OSH a couple of months later, one of the trio guys walked with me the 2 miles to my plane and it took him about 1 minute to adjust the settings on my Apolo SL-60 GPS and adjusted the contrast and the EZ-Pilot has worked flawlessly ever since (2 years). I love it. I may go with TruTrack when I pop for the Alt Hold since it stands alone. Actually, I will probably stick with the Trio guys again.
 
Wow, not one satisfied Trio guy out there?
Jamie, sorry you had such a rough time with the APs. I started with a navaid and didn't like the mechanical gyro, so when I decided to upgrade, I looked at TruTrack and would have to install the whole system, AP head and servo. Well, being cheap, I kept looking and the Trio guys sold me the A/P head only and it worked because they use the Navaid servo. They even made up a wiring harness so I could unplug the Navaid and plug in the EZ-Pilot A/P. It worked from the begining and saved me about $1K. It needed some tweeking so when I went to OSH a couple of months later, one of the trio guys walked with me the 2 miles to my plane and it took him about 1 minute to adjust the settings on my Apolo SL-60 GPS and adjusted the contrast and the EZ-Pilot has worked flawlessly ever since (2 years). I love it. I may go with TruTrack when I pop for the Alt Hold since it stands alone. Actually, I will probably stick with the Trio guys again.

There are a bunch of us satisfied Trio flyers out here, but this thread is about Jamie's autopilot travails, so no need to jump in with Trio cheerleading. Lots of threads on the forum about the fine Trio devices. :)
 
Hi Jamie,
What does the "static lag" setting do on the AP?

In our systems we call it "Pitch servo retard".
Tends to be the same thing (perhaps I should rename ours ?).
Effectively it enters a delay value between adjusting the pitch angle and measuring the result. This can be of advantage when the aircraft is very close to the desired altitude (or close to the desired vertical speed) and small changes in pitch take a few momements to register as a change in vertical speed. It prevents over correction. It is usefull to have this setting as different types of aircraft may respond differently.

As a very recent example, we have come across an installation of one of our EFIS's where a pull up increases pressure at the pitot port (so pulling up initially shows a negative VSI). As you can imagine, if your aircrafts static installation does this, even the very best autopilot is going to have to scratch it's head a little as the result of a pitch change is different to what it would expect. Adding "static lag" or "Pitch servo retard" can help smooth this issue over.

I am pretty sure Dynon is trying their very best to make the autopilot as easy to adjust as possible (hence as little things to fiddle with as possible). I tried this too but have now ended up with three settings to influence pitch behaviour (I don't like this but have to admit that it helps as long as you know what you are doing...).

So, similar to some other systems our AP (at this point in time) allows you to tweak the following:

a) Pitch servo magnitude. This setting sets the amount of servo movement desired in a given time to hold a particular desired pitch angle (this does not change the pitch or attempt to hold altitude, just keeps the pitch where the AP wants it). Too much and the aircraft over corrects and starts to oscillate - just like a rookie pilot tends to do. Too little and it takes too long to try and keep a given pitch angle.

b) Pitch control magnitude. This settings is effectively a "gain" setting. It decides by how much to change the pitch angle (over a certain time period) to achieve a given change in vertical speed.

c) Pitch servo retard (static lag ?). Explained above.

Our system is new and there are not a lot of aircraft out there flying it so our experience is limited to a few types at this stage (we know of one RV-7 in the U.S. and it seems fine). We will change/amend/add/remove as we gain more experience in different aircraft.

One aircraft fitted with our AP is a GP-4 that is going to challenge the Henshaw record for a single engined aircraft to fly from Cape Town to London. This one has me worried (to be honest) - this aircraft tends to be somewhat pitch unstable which is going to test our autopilot to the limit. I will report once we have first tests on this aircraft flown. Should be interesting.

Rainier
CEO MGL Avionics
 
Thanks for the detailed response Rainier! Do you have the same retard feature available for the roll servo? I am thinking some planes have adverse yaw where the same feature could help them.
 
Thanks for the detailed response Rainier! Do you have the same retard feature available for the roll servo? I am thinking some planes have adverse yaw where the same feature could help them.

No, for bank we ended up with a completely different system. Here we effectively have two settings:

a) bank servo magnitude
This is the control gain that will attempt to hold a given bank angle. The autopilot tells this part what angle to fly and this will attempt to hold it.
Again, similar to the pitch, too much control will result in oscillation, too little will result in unsatisfactory response.

b) bank control magnitude
This is the actual control gain that will dictate how much bank to throw on to correct a given track/heading error. This is not a direct setting as time comes in as well plus the actual aircraft response.

Apart from this there are other settings like target rate of turn or just how tightly to follow a given track. Sometimes allowing the autopilot a bit of leeway can result in the servos having to work a little less - compare it to "herding the aircraft to a destination" rather than forcing it to fly a narrow track. Much like you would fly the aircraft...

Adverse yaw does not affect our autopilot so the lag was not required.

You can test our autopilot even without spending a single cent (except for a bit of electricity). You need a reasonably fast PC, Microsoft FlightSimulator (2004 is recommended), download the Enigma or Odyssey simulator (Enigma is a bit better as it leaves more resources for FS).
In flight simulator choose a good aircraft (The Mooney works well) and select everything to maximum possible realism (as far as aircraft controls go), also deselect "auto rudder". You might need to set the display detail to somewhat sparse - your FS must not use up too much resources or else the EFIS simulator cannot run at the required speed and autopilot calculations will be too slow.
Start the MGL simulator after you start FS and it will connect. It uses simulated servos that approximate the control response of Trio's Gold Standard servos. Run FS in a window rather than full screen so you can see the EFIS panel.
You have to configure the autopilot just like you would do in a real aircraft.

Here are typical settings that work well for the Mooney:

Bank settings:
Target rate of turn: 2.0
Bank servo magnitude: 10
Heading change magnitude: 5
Track intercept force: medium

Pitch settings:
Low ASI for climb: 80mph
High ASI for descent: 200mph
Target VSI for ascent: 600fpm
Target VSI for descent: 400fpm
Pitch control magnitude: 30
Pitch servo magnitude: 3
PItch servo retard: 4

Select any type of servo (it will use simulated Trio servos regardless of selection). Ignore the engage position setups - that is only relevant for PWM servos.

Once you have things going, engage the autopilot SHIFT+Menu (click on the EFIS buttons). This brings up a menu where you can engage, disengage and select navigation sources. You can also set heading and altimeter bugs.
You will need to dig a bit into the setup menu to enable some navigation sources for use by the AP (look at HSI/GSI setups). Disable the internal GPS simulator - it will use the GPS from FS.

Now, all you need is the various maps, terrain data and a navidata file for your area - I think the navidata for the U.S. can be had from www.MGLAvionics.com (the U.S.) website. The rest you can download from our data server at www.MGLAvionics.co.za.

All that costs you nothing and will keep you entertained for hours so wait for a rainly day. Once you have everthing setup (yes, it is a bit of a learing curve) and your AP is flying nicely, why not try and get it to fly the Extra300 ? Now that is a real twitchy, fast reacting aircraft...

Oh yes, you can try and engage the autopilot in any attitude (even inverted) - try it and see what it will do.
Also, enable some heavy turbulence - no point in flying the easy way.

In case you are wondering about the accuracy of simulation - the simulated aircraft in Flightsim do NOT correspond to their real counterparts 100% - but it gets quite close...

Enjoy...

Rainier
CEO MGL Avionics
 
Not quite ready for prime time in an RV :(

.

Our system is new and there are not a lot of aircraft out there flying it so our experience is limited to a few types at this stage (we know of one RV-7 in the U.S. and it seems fine). We will change/amend/add/remove as we gain more experience in different aircraft.


Rainier
CEO MGL Avionics

I have spoken with the "other RV7" guy with the MGL vertical autopilot setup. His quote was "it will normally hold no longer than 30 seconds or so, then it starts a building oscillation and has to be disconnected". He and I have conversed long on the phone.

I also have an RV7 with the Trio gold servo and MGL Odyssey. I am tearing out my hair. I have two choices of mode:

  • Aircraft bobs viciously up and down while wandering drunkenly across the prescribed altitude (50-200ft off, never settling), then the autopilot servo disconnects from slipping (usually with the nose pointed 15-20 degrees down)
  • Wandering slowly across the prescribed altitude, between 200-250 each leg, airspeed varying between 120kts and 170kts. Sometimes as much as 300 feet high or low. The system is hopelessly "behind the airplane".

These two choices are no more than 1 or two "clicks" away from each other in the settings.

Before you ask, yes, I have iterated EVERY POSSIBLE combination of the three settings.

Breathlessly waiting to hear back from Rainier. I sent a private message describing my experience some days ago. No response yet.

JP
 
OK, I did get your PM but my lengthy response did not get to you.
But it is probably a good thing that I respond in public as this may help others as well.

OK, we know your bank works fine and it is holding course.
So we have to contend with the pitch.

I am assuming that you have the latest BIN file installed (this is important as some of what I say here depends on that).

The pitch system works in a dual loop system. The inner loop needs to be adjusted first. The inner loop compares the actual pitch angle reported by the AHRS with the "desired" pitch angle and creates a correction signal that is sent to the servo.
In order to do this, we need to switch the outer loop off. This is done by setting "pitch control magitude" to a setting of one.

You now engage the AP and start incrementing the "pitch servo magnitude" starting from 1 until you notice the build up of pitch oscillations - then you back off a notch until they stop. The aircraft should now hold pitch angle perfectly - you can see what is happening by going to the "autopilot status menu". Look at the values for desired vs. actual pitch angle - the desired pitch angle should not change while the actual pitch angle should stay close to the desired value.
You need a setting of around 3 or 4 (or even higher). If you need a setting of 1 or 2 here - it means that you have way too much mechanical gain (small change in servo arm position results in large control deflection). Fix this first, there is no point in continueing unless this works 100% up to here.
Don't just punch in a low number - find the point where oscillation starts and then back off. You need to get the gain right or else the system WILL be behind the aircraft.

Right, assuming the above works and the aircraft holds pitch angle with a pitch magnitude setting of 3 or higher (Chalkie in his pitch sensitive GP-4 needed a solid 4 here) - we can continue dailing in some actual altitude control:

Set the pitch control magnitude to a value of (pitch servo magnitude * 5). For example, with a pitch magnitude of 4, select the value 20.
This should be a good starting value for the gain of the outer control loop.

The outer control loop dictates the desired pitch angle based on the desired vertical speed (which would be zero if you are right on altitude). The system does not want a particular pitch angle (it does not care about the actual pitch angle) but will attempt to increase it to gain altitude and then set it at whatever angle it needs to maintain the desired climb or sink rate. These rates you set yourself and they are applied if you are quite far away from the target altitude and become progressively smaller as you get closer to the target altitude.

So, given the above - the pitch control magnitude selects the amount of pitch angle change over time required to chase the target vertical speed. If you enter a setting too low, the aircraft will take a long time to change pitch angle - in this case it will probably never be able to intercept an altitude as it will overshoot badly every time. If the setting is very high then it will overcontrol (like a rookie pilot) - between these extremes there is a good band of numbers that should work well.
The danger here is thinking that slow oscillations need a lower number - they usualy mean that you need the give more, not less authority to the AP.

The "pitch retard" setting applies to the control magnitude only (not the pitch servo magnitude). It will prevent the autopilot from changing the desired pitch angle rapidly when you are close to target rate of climb/descent, instead waiting a while to see how the last adjustment is going - this caters for "static lag". The number works in 1/4 of a second. I find a setting between 4 and 8 does the trick.

The only other things I can think of are as follows:

Take note that the minimum/maximum speeds that you enter are limit speeds. Don't use them to try and set climb or descent speeds. These speeds should be set somewhat higher than stall and lower than VNE respectively.

Make sure that you don't have any mechanical slack in your controls. Use the servo tests and checks functions to engage the pitch servo, then go to the elevator and try and move it up and down by hand. If you can move it more than a few mm then you might have too much play in your linkage.

To check your mechanical gain, use the "servo test and checks menu" and select the servo to full CCW and CW - the resultant position on the elevator should not be the same as the maximum position you can get by using the stick - it should be as little as you can get it without having the servo limit the amount of movement. Try and get the maximum servo positions to be as close to 50% of the total travel as you can.

Hope this helps...

Rainier
CEO MGL Avionics

I have spoken with the "other RV7" guy with the MGL vertical autopilot setup. His quote was "it will normally hold no longer than 30 seconds or so, then it starts a building oscillation and has to be disconnected". He and I have conversed long on the phone.

I also have an RV7 with the Trio gold servo and MGL Odyssey. I am tearing out my hair. I have two choices of mode:

  • Aircraft bobs viciously up and down while wandering drunkenly across the prescribed altitude (50-200ft off, never settling), then the autopilot servo disconnects from slipping (usually with the nose pointed 15-20 degrees down)
  • Wandering slowly across the prescribed altitude, between 200-250 each leg, airspeed varying between 120kts and 170kts. Sometimes as much as 300 feet high or low. The system is hopelessly "behind the airplane".

These two choices are no more than 1 or two "clicks" away from each other in the settings.

Before you ask, yes, I have iterated EVERY POSSIBLE combination of the three settings.

Breathlessly waiting to hear back from Rainier. I sent a private message describing my experience some days ago. No response yet.

JP
 
not sure why I am replying to this thread apart from the fact its a great thing to see.

I have a full Dynon suite, so no bias here, but it is great to see such enthusiastic input from a few manufacturers. While they compete, the services they offer are of importance, and to see both MGL and Dynon and anyone else be open and up front about looking after customer interests is something not seen in many industrial circles.

Well done to all of you! :)
 
The pitch system works in a dual loop system. The inner loop needs to be adjusted first. The inner loop compares the actual pitch angle reported by the AHRS with the "desired" pitch angle and creates a correction signal that is sent to the servo.
In order to do this, we need to switch the outer loop off. This is done by setting "pitch control magitude" to a setting of one.
I've done a fair bit of autopilot testing on transport category aircraft. One thing that installers should keep in mind is that aircraft CG may have a significant impact on the stability and control with autopilot engaged. At aft CG, the aircraft pitch response is greater for a given amount of control input, so there is a greater tendency for pitch oscillations. Thus, some inner loop testing should be done at aft CG, to ensure that there would be no surprises when aft CG is seen in service (e.g, with baggage).

At forward CG, larger control inputs are needed to get a given aircraft response. If the inner loop gain is set low enough to avoid pitch oscillations at aft CG, we need to be sure there the autopilot is commanding large enough control inputs for acceptable control at forward CG.

Bottom line - autopilot testing should be done at the most forward and aft CGs that will be seen in service, to ensure that the final gain settings will provide acceptable performance through the whole range of conditions.
 
Okey Dokey, here we go.

OK, I did get your PM but my lengthy response did not get to you.
But it is probably a good thing that I respond in public as this may help others as well.

OK, we know your bank works fine and it is holding course.
So we have to contend with the pitch.

Yes, but I am using another autopilot for bank. :) I am not fully decided if I want all my eggs in one basket (yet).

I am assuming that you have the latest BIN file installed (this is important as some of what I say here depends on that).

Yes. 1.1.0.7

The pitch system works in a dual loop system. The inner loop needs to be adjusted first. The inner loop compares the actual pitch angle reported by the AHRS with the "desired" pitch angle and creates a correction signal that is sent to the servo.
In order to do this, we need to switch the outer loop off. This is done by setting "pitch control magitude" to a setting of one.

You now engage the AP and start incrementing the "pitch servo magnitude" starting from 1 until you notice the build up of pitch oscillations - then you back off a notch until they stop. The aircraft should now hold pitch angle perfectly - you can see what is happening by going to the "autopilot status menu". Look at the values for desired vs. actual pitch angle - the desired pitch angle should not change while the actual pitch angle should stay close to the desired value.
Ok. With you so far (and this is how I initially tried it).

You need a setting of around 3 or 4 (or even higher). If you need a setting of 1 or 2 here - it means that you have way too much mechanical gain (small change in servo arm position results in large control deflection). Fix this first, there is no point in continueing unless this works 100% up to here.

Mechanical gain appears to be fine.

Don't just punch in a low number - find the point where oscillation starts and then back off. You need to get the gain right or else the system WILL be behind the aircraft.

Ok. servo setting of 3, too slow. 4 bobs the airplane up and down then disconnects. 5 or higher, forget it. Won't stay engaged long enough.

Right, assuming the above works and the aircraft holds pitch angle with a pitch magnitude setting of 3 or higher (Chalkie in his pitch sensitive GP-4 needed a solid 4 here) - we can continue dailing in some actual altitude control:

Phase 1 does not succeed. The servo magnitude controls do not appear to be fine-grained enough. :)

Set the pitch control magnitude to a value of (pitch servo magnitude * 5). For example, with a pitch magnitude of 4, select the value 20.
This should be a good starting value for the gain of the outer control loop.

The outer control loop dictates the desired pitch angle based on the desired vertical speed (which would be zero if you are right on altitude). The system does not want a particular pitch angle (it does not care about the actual pitch angle) but will attempt to increase it to gain altitude and then set it at whatever angle it needs to maintain the desired climb or sink rate. These rates you set yourself and they are applied if you are quite far away from the target altitude and become progressively smaller as you get closer to the target altitude.

Settings of 4/20, 3/15, 5/25 etc have all been tried. (and every possible variation in between) I literally flew every possible setting for servo magnitude of 3,4 or 5 and every possible setting for control magnitude (from 1 to 31). For a while, I thought 3/31 was going to work. But only for a few seconds and then the divergence begins. For servo magnitudes 4 and 5, the ride gets pretty violent sometimes, then the servo disconnects (after slipping).

So, given the above - the pitch control magnitude selects the amount of pitch angle change over time required to chase the target vertical speed. If you enter a setting too low, the aircraft will take a long time to change pitch angle - in this case it will probably never be able to intercept an altitude as it will overshoot badly every time. If the setting is very high then it will overcontrol (like a rookie pilot) - between these extremes there is a good band of numbers that should work well.
The danger here is thinking that slow oscillations need a lower number - they usualy mean that you need the give more, not less authority to the AP.

Fully grasped.

The "pitch retard" setting applies to the control magnitude only (not the pitch servo magnitude). It will prevent the autopilot from changing the desired pitch angle rapidly when you are close to target rate of climb/descent, instead waiting a while to see how the last adjustment is going - this caters for "static lag". The number works in 1/4 of a second. I find a setting between 4 and 8 does the trick.

As far as I can tell, there is no appreciable "static lag" in my system. Observing the vsi while moving the stick shows an almost instant response in the correct direction.

Interestingly, while observing the autopilot status screen, the vsi numbers on the status screen do not directly correlate with the vsi numbers on the display.

The only other things I can think of are as follows:

Take note that the minimum/maximum speeds that you enter are limit speeds. Don't use them to try and set climb or descent speeds. These speeds should be set somewhat higher than stall and lower than VNE respectively.

They seem to have little effect, no matter where I set them. Currently set at 70KTS for low and 170KTS for high. Is this indicated or true airspeed?

Make sure that you don't have any mechanical slack in your controls. Use the servo tests and checks functions to engage the pitch servo, then go to the elevator and try and move it up and down by hand. If you can move it more than a few mm then you might have too much play in your linkage.

No discernible slack in the elevator linkage. While on the ground I engage the servo. Touching the elevator with very light pressure produces almost no movement. The very slight movement is clearly against the servo, which whines at the touch.

To check your mechanical gain, use the "servo test and checks menu" and select the servo to full CCW and CW - the resultant position on the elevator should not be the same as the maximum position you can get by using the stick - it should be as little as you can get it without having the servo limit the amount of movement. Try and get the maximum servo positions to be as close to 50% of the total travel as you can.

Interesting. Trio tech support informed me the powered range of the servo was approximately 45 degrees. They stated there are 1024 steps across the powered range. If the powered range is 45 degrees, that leaves 22.5 degrees for each side of neutral. Each "step" of the servo equates to .0439 degrees (roughly 23 steps per degree of motion). If the tech misspoke and it's a total of 90 degrees powered motion, then it's half that (11.5 steps per degree)

Total available swing of the servo arm is around 170 degrees (not quite 90 each way). Ideally then, you want the servo arm to hit the stop at the same time the elevator hits it's stop, true? That gives the most resolution, the maximum powered travel of the servo is right in the middle of the elevator travel range. You want the servo to move as much as possible and move the elevator as little as possible without restricting travel. This is how my servo is setup. I am using the middle hole on the trio servo arm.

One last note. Why does the autopilot "jerk" the aircraft when engaged? I have the aircraft trimmed for level flight. I engage the autopilot and the stick jumps in my hand as the AP jerks the airplane up or down for no reason. From there, we never recover to level flight.
Hope this helps...

Rainier
CEO MGL Avionics
 
Last edited:
Ok. servo setting of 3, too slow. 4 bobs the airplane up and down then disconnects. 5 or higher, forget it. Won't stay engaged long enough.

Hmmm, this really sounds like something is not coupled right. I can make you a version (for testing) with a finer adjustment but that should not be required, but let's see if that would help. Give me a few days.

As far as I can tell, there is no appreciable "static lag" in my system. Observing the vsi while moving the stick shows an almost instant response in the correct direction.

Interestingly, while observing the autopilot status screen, the vsi numbers on the status screen do not directly correlate with the vsi numbers on the display.

Yes, your observation is entirely correct (not many have noticed). The reason for this is that the AP uses a VSI that is derived from altitude changes via a mathematical route and ignores the VSI that is central to the rest of the system The reason for this is that the normal VSI is actually better described as a variometer. It is sensitive and fast - you don't want that as input to an AP in particular during turbulence. For the AP you need a trend more than an instantaneous value.

They seem to have little effect, no matter where I set them. Currently set at 70KTS for low and 170KTS for high. Is this indicated or true airspeed?

Indicated.
Those limits seem fine for an RV.

No discernible slack in the elevator linkage. While on the ground I engage the servo. Touching the elevator with very light pressure produces almost no movement. The very slight movement is clearly against the servo, which whines at the touch.

OK, that sounds good.

Interesting. Trio tech support informed me the powered range of the servo was approximately 45 degrees. They stated there are 1024 steps across the powered range. If the powered range is 45 degrees, that leaves 22.5 degrees for each side of neutral. Each "step" of the servo equates to .0439 degrees (roughly 23 steps per degree of motion). If the tech misspoke and it's a total of 90 degrees powered motion, then it's half that (11.5 steps per degree)

It is about +/-40 degrees to each side.
Unfortunately there is a significant amount of play in the gears and to the mechanicaly coupled potentiometer that provides the position feedback. So the above figures are rather theoretical.
For applications requiring exact and fast positioning we are currently in the prototype stage of our own servo development - but we will continue to support the Trio servos as mainstream as our own servo is going to be a bit more expensive (it's kind of an "industrial strength" type of thing).

Total available swing of the servo arm is around 170 degrees (not quite 90 each way). Ideally then, you want the servo arm to hit the stop at the same time the elevator hits it's stop, true?

Could not have put it better myself...

One last note. Why does the autopilot "jerk" the aircraft when engaged? I have the aircraft trimmed for level flight. I engage the autopilot and the stick jumps in my hand as the AP jerks the airplane up or down for no reason. From there, we never recover to level flight.

[/QUOTE]

I think we have the "aha" moment here.

In one word (OK, two): Voltage dip.

The system is designed to engage the servo at it's current position. For this the servo sends (at all times) the currently measured position. When you engage the AP sends its first active position comands with this position. As it does this the solenoid clicks in and this takes a bit of current. The servos are somewhat voltage intollerant in our tests and this results in an immediate change of measured position with a resultant change of arm position. The AP does not know this of course.
Can you make sure of the following:
You need a good diameter power and ground cable from a low impedance power source, we do not want the voltage right at the servo to change much when it engages (we don't mind slow changes in voltage - that gets fixed up by the AP). I also found it usefull to place a big 10.000uF capacitor right at the servos to take up any power issue at the point of engaging. The servos themselves have almost no internal capacitors on the power input so to have a nice fat capacitor external is not a bad idea.

The correct behaviour when you engage is: No appreciable change if the stick position when you engage. You can try this on the ground as well.

We today ran our algorithm rough matlab and this did not show any obvious issues. However, we also investigated the possibility of including an alternatively selectable mode for pitch control that mimicks simpler systems. We are thinking of including just a simple method that effectively ignores the pitch angles and just uses barometric inputs to govern the servo arm position - this is the way many current autopilots work with altitude. The disadvantage of this is that is takes longer to intercept an altitude or to fix up disturbences but it does have the advantage that it will probably work better with the trio servo (much less work for the servo). In effect this method uses the natural pitch stability of the aircraft to sort out the pitch angle.

Are you game to try it out ? (I'll test it on our own aircraft first of course)

Rainier
CEO MGL Avionics
 
Some followup

With the linkage connected as described below and running the autopilot test checks for full CW and CCW the stick is moved to between 70 and 80% of full travel. There is about an inch and a half to two inches of stick left in each direction. There is no feasible way to lessen this without restricting elevator travel (a big no no).

The "trend" vsi the autopilot uses is often wrong in domain and certainly VERY slow in magnitude. IE, it often shows climb while we are diving and the converse. We'll be diving at 1200 fpm and the AP vsi shows a 100fpm climb. :(

While sitting on the ground, calm wind in the hangar the horizon "jitters" up and down a slight amount, as if the SP4 cannot decide which degree to land on. The compass used to do this until damping was engaged. This might be a factor. I never noticed it before.

As for the wiring, I am using the harness sold by SportFlying to power the servo. The power wires are not beefy (look like 22 or 20 gauge). They are connected to a dedicated circuit for servo power. The servo is grounded to the airframe as laid out in the installation instructions.

JP
 
OK, I am just finalizing a next update of the Osyssey/Voyager software - this one mainly focuses on the addition of a serial data output stream in Chelton EIS format (similar to the GRT format with some of the AUX fields filled in).

Anyway, we did some interesting tweaks on the AP pitch algorithms that (I am hoping) will help the RV-7 (I just wish I had one to try - but luckily we have you...).

We tweaked the PID for the inner loop a bit to make it a bit less agressive and I changed the "pitch servo magnitude" setting to much finer steps (now you enter a value between 1 and 40 - my guess is that you will need a 10).
Unfortunately local weather is preventing me from test flying for at least a few days but I am uploading this software as "beta" so you can give it a try tomorrow.

Rainier
CEO MGL Avionics
 
Autopilot thoughts

Anyway, we did some interesting tweaks on the AP pitch algorithms that (I am hoping) will help the RV-7 (I just wish I had one to try - but luckily we have you...).

Ahh, If I just got test-pilot pay. :)

We tweaked the PID for the inner loop a bit to make it a bit less agressive and I changed the "pitch servo magnitude" setting to much finer steps (now you enter a value between 1 and 40 - my guess is that you will need a 10).
Unfortunately local weather is preventing me from test flying for at least a few days but I am uploading this software as "beta" so you can give it a try tomorrow.

Rainier
CEO MGL Avionics

Thanks for getting on top of this. Really. Thanks.

Thoughts about the AP:
  • I gritted my teeth and tried it again this morning on the way to work.
  • Pitch servo mag 3 produced long climbs and descents (+- 250 feet) across the target altitude. Pitch control magnitude had little effect on it.
  • Pitch servo mag 4 produced "divergent" weaving. IE, started out close, then moved off a bit, the back through a bit more, then more, then MORE, then violently MORE!!! before disconnecting due to servo slip. This is what is known as a classic undamped feedback loop. Perhaps a "smoothing" function or "damper" could be applied to the outputs (servo positions).

Anyhow, it feels as if "George" has no concept of inertia or time? IE, he moves the servo, looks at the pitch reading and says "hmm, no effect, crank it up". Then about .5 to 1 seconds later(as we are pitching hard) he says "TOO MUCH!!!" and back the other way, where the cycle repeats.

I don't know how fast a GP4 or Jora respond to control inputs. If you have a "measurement lag' built into your formula or not. The RV is probably a bit heavier than a Jora, thus slower to respond to pitch inputs. Maybe a GP4 is lighter as well. Don't know. Heck, an RV might be faster to respond to pitch inputs. Again, I don't know.

Watching the autopilot status screen, it appears the servo is being moved 5 to 10 times PER SECOND constantly. It never "lights" on a setting and waits for a result (as a human does). Maybe there should be a couple of tunable parameters:

  1. Maximum amount of travel for the servo.
  2. Wait period between movements. Give the airplane time to respond to the input before making further adjustments.

Just a thought,
JP

Anyhow, thanks again for the response. I'll give it a try first chance I get.
 
OK, try the Odyssey.bin file you can download from the Odyssey beta page at www.MGLAvionics.co.za

Let me know how this one does. It should do the trick without reverting to a simpler control system.
There was a possible issue with a filter related to the differential part of the PID which makes the system behave well with slightly unstable aircraft (like the GP4) and also like our Jora which is likely to react slower than an RV. But if our model is correct, it can lead to instability depending on the phugoid time of the aircraft vs the filter time constant.
The filter was a left over from earier code that was still able to work with our older AHRS systems that would report attitude to 1 degree resolution (All AHRS delivered for the last approximately 1.5 years can do 0.1 degrees). The 0.1 degree AHRS does not require the filter and the autopilot will refuse to engage if you have an older AHRS. I left the filter in as it seemed a good idea (I was wrong) and I have removed it now.
The finer grained adjustment should also help.

Good luck ! I am eager to hear your results.

Rainier
CEO MGL Avionics
 
Status of MGL Autopilot

I've beem checking this thread every day for a while now so see nothing new. I'm very interested to see how this pans out. Joe, how are things going? Rainier, have you heard from Joe?

I wish I were further along on my 7A to playing with the MGL autopilot. I'm in mid fuselage but plan to go MGL all the way, EFIS, Autopilot, Radio. MGL fits my mission in many ways.

Ed
 
Hi Ed,
Although I have nothing to report, I'm hoping to have my MGL autopilot ready to go by next weekend. I believe Rainier has a new version of code almost ready for release that improves the autopilot in RV's, but I don't know exactly when that will be ready.

As soon as I can report something, I will.
 
I've beem checking this thread every day for a while now so see nothing new. I'm very interested to see how this pans out. Joe, how are things going? Rainier, have you heard from Joe?

I wish I were further along on my 7A to playing with the MGL autopilot. I'm in mid fuselage but plan to go MGL all the way, EFIS, Autopilot, Radio. MGL fits my mission in many ways.

Ed

Yes, a lot has happened since those last posts.
Joe's RV seems incredibly pitch sensitive and needs the tiniest of servo arm movements (he says his RV cruises at over 200mph). Nevertheless, Joe is a good guy and has been testing a dozen or so software versions and we are getting somewhere. The last version he tested used the dual loop system to about +/-10ft tracking but the climb and descends need work (which is what he is testing now). The AP has to treat his aircraft quite litteraly with "kids gloves" but we also now have to make provision to handle his large speed range (i.e. control deflections must become larger and faster with slower speeds).
It's been quite a challenge and sometimes I just want to go and buy a plane ticket to try out his RV first hand - but due to the "power of the internet" it looks like things are moving well.
I am going to wait for his latest tests (should be due any day/moment) and if that is a "go" then I will release it as next version.

There has been no change to the bank/heading/track part of the AP since first release - it seems that works pretty well. The challenge is now to get the pitch working in such a way that it handles a really wide range of aircraft - this has turned out to be more difficult than expected.
I must confess that I learned more about aerodynamics designing the AP than I ever knew as a pilot. I also have new respect for the human brain when it comes to flying an aircraft...

Rainier
CEO MGL Avionics
 
Could a moderator kindly move the MGL support conversation to a different thread? Thanks in advance.

BTW: I'm still really, REALLY liking my Trutrak! It flew us about 6 hours last week out to the coast and back and worked flawlessly.
 
Back
Top