Menu Sign In Contact FAQ
Banner
Welcome to our forums

DFC90 - Sudden "Autopilot Disconnect" in Cruise

The AP has zero tendency to oscillate or over-react.

I am not suprised. If you drive the pitch trim as the primary pitch control, the pitch control loop is almost wholly integral, with very little proportional gain. It isn't going to oscillate. But it will have a lousy transient response.

The SR20 does not have a pitch servo although all of them have a 2-axis AP (S-TEC 55x, now Garmin GFC700 and DFC90 retrofits).

I flew in an SR20 which had the 55X and even in very light turbulence it was all over the place, like a snake. In comparison, the King AI-based APs have solid control (unless faulty, etc).

With the GFC700, are they still driving just the pitch trim?

Administrator
Shoreham EGKA, United Kingdom

The 55X has that problem often when the planes get a little older and when the rigging is not perfect anymore. I know three people now who dumped it for the DFC90. And COPA is full of those oscillations storys

It turns out that I have the generic STEC "flight line" service manual.

Looking at their pitch and roll servo connections, it's clear these are devoid of any electronics. All that is inside is the bare motor (in the roll servo) and the motor and some torque sensing switches (in the pitch servo). Plus the clutch solenoids.

So a minimum 2V to make it turn (no idea what the maximum is) is probably a reasonable test.

But this also tells us that the servo does not return any feedback - either velocity feedback or any kind of "valid" output.

So there is very little scope for e.g. the roll servo to make the autopilot malfunction - until the servo motor is more or less shagged. Especially an intermittent malfunction.

It also means the stability of the autopilot is going to be very sensitive to the motor and gearbox condition. If say the motor runs slowly, the gain of the system will be that much lower, so you will get poor response. Whereas the King servos use velocity (RPM) feedback so the motor gain (RPM per volt) is compensated for. It doesn't matter how clever a computer you hang on the end of the STEC servo...

BTW I totally stand by my "frame of reference" position We will have to sort this over a beer (Czech beer?) with some paper and a pencil.

Administrator
Shoreham EGKA, United Kingdom

Peter, which servos are used is implementation dependant. In the PA-46 it drives both servos. SOme other aircraft don't have a trim servo. The DFC-90 is meant to just slot in and not require new servos to be fitted. So I guess they just install and calibrate with what you have.

EGTK Oxford

which servos are used is implementation dependant. In the PA-46 it drives both servos

Yes, this is correct although in reality, even in a PA46, the DFC90 only operates the trim servo most of the time, it is the primary means of controlling pitch. Only when a fast and strong pitch change is required, it would operate the pitch servo in addition to the pitch trim servo.

It isn't going to oscillate. But it will have a lousy transient response.

This is why the DFC90 can operate both the trim servo and the pitch servo at the same time.

You are right of course about the difficulties with the S-TEC servos that do not provide feedback. This requires the cable tension and the servo behavior depending on voltage to be very close to the specs. The only feedback the DFC90 gets -- apart from change in attitude -- is current sensing. The DFC90 constantly drives the pitch trim even in smooth air level flight to keep the cable under tension.

With the GFC700, are they still driving just the pitch trim?

For a SR20, it would have to be this way given the absence of a pitch servo. Maybe the GFC700 SR20 actually do have a pitch servo?

However, the DFC90 doesn't just drive the pitch trim in all aircraft but the SR20. It uses a combination of both servos with the pitch trim being the primary means of controlling.

However, the DFC90 doesn't just drive the pitch trim in all aircraft but the SR20. It uses a combination of both servos with the pitch trim being the primary means of controlling.

OK that makes a bit more sense. That's not what you were saying before!

But...

The only feedback the DFC90 gets -- apart from change in attitude -- is current sensing.

That isn't going to work because the gear-down from the motor shaft to the servo output gear is of the order of 10000:1 so the motor current bears almost no relation to the aerodynamic load on the control surface. Nearly all the current drawn is to overcome friction and magnetic losses. I found this when doing some research on a related project which used the same kind of motors.

The DFC90 constantly drives the pitch trim even in smooth air level flight to keep the cable under tension.

I wonder how it does that, when it has no way of knowing other than aerodynamic feedback i.e. a pitch accelerometer or similar. I can see it might drive the pitch trim until it just sees a pitch change and then it knows the pitch trim cable is tight. But what if the next gust needs a pitch change in the opposite direction? It then has to drive the pitch trim all the way the other way, firstly to take up the slack and then "just the right amount" to compensate for the gust.

I imagine that the system really needs pretty tight control linkages.

Administrator
Shoreham EGKA, United Kingdom

Peter, Achim,

After discussing my autopilot failure with two Cirrus Service Centers, Avidyne and a Cirrus Certified Instructor Pilot i have not yet heard back from Avidyne about how they would want to proceed in this matter. I am still waiting for an answer from Avidyne to my question who will pay for the new roll servo if i replace it and the failure shows up again ...

Among all the inputs i got about what could be wrong, these are my favourites: - Roll Servo bad (after all the Service Center measure 3.5 volts when it should be only 2) - Autopilot Tray bad (for the DFC90 installation the old sTEC 55 tray is used)

After speaking with Avidyne these are unlikely: - Trim tab bent the wrong way (should also lead to bad tracking, but the tracking is perfect!) Also nothing was changed and it worked perfectly for 45 hours - Autopilot computer bad (unlikely for same reason, but maybe it fails when it gets hotter?)

I will do the following test tomorrow - is the ball centered in cruise (yes, if i rem. correctly) - is the roll trim tab bent the right way (does a roll to 30 deg left/right take the same time?) - Fly without ALT but with GPSS and see if it happens again. and if it does not then check if the pitch servo might be the reason)

One more question: The FLIGHT DIRECTOR mode works perfectly when i fly manually, even with GPSS/ALT. Isn't that a very strong hint that the Computer itself should be okay. Is that a correct assumption. If i understand it correctly then the ONLY difference is that I am the servo in this scenario.... but then it MUST be the (roll) servo.

Alexis,

replacing components in random order when there are ways to diagnose the problem is not a good approach. It is very common in aviation though. The DFC90 has an error log and before doing anything else, it should be read out. Avidyne should be told that they have to put their dealers into a position where they can obtain the logs.

However, in your case you have an issue with the servo. The start voltage of 3.5V is outside the spec and if I were Avidyne I would refuse to even discuss this issue with you until you fix your servo. Go to Straubing, they might be able to repair it for little money and if not probably have on in stock.

The S-TEC servos are cheap pieces of hardware (manufacture/quality not retail price). They have a lot of wear. Even after only 50 hours, there was coal dust in mine.

Achima, that's actually exactly what i thought i should do. What confused me was that the Cirrus Service Center said "3.5 Volts is okay, they are all like that"

(no comment, ....)

Thank you for speaking it out clearly.

The FLIGHT DIRECTOR mode works perfectly when i fly manually, even with GPSS/ALT. Isn't that a very strong hint that the Computer itself should be okay. Is that a correct assumption. If i understand it correctly then the ONLY difference is that I am the servo in this scenario.... but then it MUST be the (roll) servo.

Yes that sounds reasonable, though it could still be within the AP computer, somewhere after the FD signals come off.

All that stuff about trim tabs doesn't make sense. The AP should just work...

I also don't understand where the 3.5V is being measured. On the ground, there should be nothing going into the servo. And in the air, with the AP enabled, the servo control voltage will be according to the control requirements.

It would be different if the servo was removed, put on a bench, and a control voltage was applied to it, and it needed 3.5V to start turning when it is supposed to start turning at 2V.

A servo full of black dust after 50hrs suggests crap AP software which is driving it back/forth all the time, and probably very fast so you get constant commutator arcing.

Administrator
Shoreham EGKA, United Kingdom
Sign in to add your message

Back to Top