For an already installed DFC90, Avidyne show the following test,
In that case, make up a cable with the right connector, get a variable voltage power supply, and see what happens...
The question is what could make a servo sticky intermittently (as reported).
Maybe some mechanical / gearbox problems, e.g. a broken gear tooth, or an open-circuit armature winding on the motor. These scenarios would result in the motor occasionally stopping in a position from which it will not easily (or ever) restart. One way to detect these is to run the motor as slowly as it will go and watch the movement.
Or the issue could be downstream of the servo.
As I say, somebody with deep knowledge of the system needs to look at it, though if the above simple test finds the servo is seized up or burnt out, then just throw it away and stick another one in.
of all the GA autopilots the DFC90 and the GFC700 are by far the ones that are most precise and most fun to fly with. Actually the precisness of the system is almost scary.
That would be a function of the AP computer, not the servos.
using brushless digital servos.
No such thing as a "digital servo" 
Once you get away from 1s and 0s, the whole world is analog. In fact a purist would say the whole world is analog anyway...........
Brushless motors (which I know well, both in the 3 phase world and the stepper world) are nice because there are no brushes or a commutator to wear out, but actually what causes something like 99% of the brush wear in the current servo designs is crap AP software which drives the motors back/forth pointlessly, and poor quality motors. A good quality motor is about €300. The motors used in GA servos are €30 models. See my KFC225 writeup... plenty of photos in there for amusement.
Jason, tell me, what are these additional features? Anything i would envy you for? ;-) What I really love is the IAS mode in the climb. Of course the GFC700 is also a little bit more integrated, but a Perspective SR22 was too expensive for me.
Brushless motors! They exist? Wow!
The question is what could make a servo sticky intermittently (as reported).
We don't know what happened, all Alexis gets is a disconnect with an error message. The whole truth is probably in the AP's error log which only Avidyne in Boston can access.
However, if during the troubleshooting you discover that your servo is outside the spec, then this should be fixed. The voltage threshold is there for a reason. Same holds true for the pitch trim cable tension, the DFC90 is very sensitive to it, more so than the much simpler analog S-TEC. Initially my cable tension was not correct and the DFC90 was oscillating as a result.
That would be a function of the AP computer, not the servos.
not quite. The servos are brushless and digital so operate a little more precisely.
The GFC700 has VNAV modes which can be very good. So I can say I want to be level at FL240, 20 miles before Lambourne. It will fly a 3 degree glideslope to make that happen. It has FLC mode which is the same as IAS mode.
It is a bit better at mode transitions. So I can fly a VS or IAS descent and then automatically capture a 3 degree glideslope.
Finally it can fly outbound on a procedural ILS, fly the VNAV steps and turn inbound to capture the ILS glideslope with me only working the throttles and arming approach while outbound.
And Achim is correct, you need the right cable tensions.
Avionik Straubing is my next stop and the first thing I'll do is check the error log. if anything is wrong with the servo we'll change it. What i don't like so much is that the left aileron has to be removed for that. i never liked it when they remove flight controls...
Avionik Straubing is my next stop and the first thing I'll do is check the error log.
Unfortunately Straubing can't do that. Only Avidyne in Boston can read the error log. When there was a problem installing mine, it had to be sent to the US for error log readout and then back. Very annoying, considering that everybody with an RS232 adapter can read the error log of the KFC autopilots...
What i don't like so much is that the left aileron has to be removed for that. i never liked it when they remove flight controls...
One of the preflight tests I was taught at the FTO is deflect the aileron and check if the right one is going down. Apparently there were cases where the pushrod was installed incorrectly after maintenance. For some reason, I still perform that very test in my own aircraft, just in case somebody entered the hangar at night and changed the aileron orientation 
Jason, i see, sounds great. Actually the descent part can be done nicely with the VNav feature of the GNS430... I just wait until it shows the requiered desc rate i like and press VS-ALt. Not quite as elegant but good enough.
What plane do you have?
They should get you a loan unit if you have to send to them.
They should get you a loan unit if you have to send to them.
Yep, and S&H both ways was covered, too. Still you lose a lot of time for something as trivial as reading a log file. Also very few if any avionics companies have a DFC90 loaner for C182 in stock. The DFC90 is pre-programmed for the airframe with a different part number.
Only Avidyne in Boston can read the error log. When there was a problem installing mine, it had to be sent to the US for error log readout and then back. Very annoying, considering that everybody with an RS232 adapter can read the error log of the KFC autopilots...
Of all the barrels in GA which one can end up bent over and shafted, that is a particularly unattractive barrel...
I am definitely sticking with my KFC225.
I wonder why Avidyne are running such an obviously stupid policy? Keep problems away from the media, perhaps?
Also I can't easily see how a sticky servo motor would be noted by the AP and thus logged. There is no feedback from the servo. The only time when the AP will become aware is if it has driven the servo control voltage all the way up and still the roll axis is not responding.
One of the preflight tests I was taught at the FTO is deflect the aileron and check if the right one is going down. Apparently there were cases where the pushrod was installed incorrectly after maintenance
This should be impossible on a TB20, which is all-rods except for the elevator trim which is a cable, but it has often been done on cabled aileron linkages on which it is usually very easily done.
