STEC55X advice - what can go wrong?

It was that accident that led to the requirement for a type rating for the Extra, and subsequently other single engine aircraft.

AxelF wrote:

I would be delighted to learn anything about a fix. Could you elaborate a bit, please.

Looking back at the records the fix was done in November 2011 by Smets Aviation / Genk Zwartberg. They would know exactly the proceedure for the modification.

Autopilot Computer was sent back to S TEC and converted (whatever that means) at a cost of 1358 Euro
Also the Autopilot Servo Pitch was converted at a cost of 508 Euro
Shipping/ duties 225 Euro
Total 2091Euro

This transformed the STEC55X into something that was useable.

It would be most useful to know how to tell a “usable” STEX55X from a “borked by design” one. Is there any way to tell that the “conversion” was done – preferably by looking at logbooks or at the device, not by going to FL200? A software version change, a hardware revision, anything?

I went to FL190 with an STEC55X (installed in 2005) today and did not notice any pitch oscillations. There is no record in my avionics logbooks concerning the A/P.

What I do notice however is that when I disconnect the A/P it sometimes seems to be out-of-trim in pitch.

I have however also flown a Cirrus which goes into oscillations at any level.

Aviathor wrote:

I went to FL190 with an STEC55X (installed in 2005) today and did not notice any pitch oscillations

I dont necessarily think that all STEC55X are affected….you either have one that porpoises or you dont, luckily there is a fix if you do have one.

Assuming the connections (bridle cables etc) are tightened up to spec, the usual cause of oscillation is excessive open loop gain in the control loop (pitch or roll).

In standard control systems theory, you have PID – proportional, integral, derivative. This was originally developed c. WW2, maybe before, and the whole of industrial control runs on it. With various add-on tweaks, especially where time lags are involved (e.g. temperature controlled ovens). Nowadays you get self-learning ones, though AFAIK these never made it into aviation.

AFAICT, autopilot loops use mostly just PI – proportional and integral. The P is implemented by amplifying the error in degrees (pitch or roll error) and driving a given control surface deflection (in degrees) with that. Servos which drive only the trim are an extra complication and need to achieve the derivative instead, because driving a trim acts as the integral which then cancels out the derivative. The I (integral) term is implemented by sensing the altitude or heading/NAV error (for pitch and roll loops respectively) and feeding in a small piece of that.

I have not seen evidence of the D term in King systems, but there may be something. It is a dodgy one to use, in any system where there is a time lag (or mechanical “slop”) and easily makes the loop unstable. It is used for a better transient response. It seems that transient response improvements have been achieved by sensing something upstream and feeding that in directly e.g. the KC225 computer has a pitch accelerometer which gives it a big advantage over sensing the absolute pitch value, and doesn’t mess up the stability.

In King systems, the classic oscillation is when the tachometer inside the servo is shagged, or a wire has come off it – see here. I don’t know how the STEC servos work internally… have some MMs somewhere I think.

But the general technology used in this stuff is really 1970s, low grade stuff. Like tachometers which are £35 from Faulhaber and due to a finite brush/commutator life are inappropriate for the application.

I would think the computer tweaks described above are something STEC should have done in the first place but didn’t because they got one STC on the back of another one on the back of another one…. without proper flight testing. You can’t do that today, apparently, which is prob99 why the cash-strapped Avidyne are not doing the DFC90 except for the most low hanging fruit. In fact one of their guys told me at EDNY they only really want the pre-G1000 SR22 market (a few k planes, and not growing anymore).

quatrelle wrote:

Looking back at the records the fix was done in November 2011 by ….

Thank you, quatrelle, that is valuable information. Guess I will first check the matter with that German “temple of avionics” firm since they maintain the avionic stuff of the aircraft concerned. Now I know what to specifically ask for which is always a big plus. And if that should not work out then the firm you mentioned will have a go.

Back to topic, there is one quirk in terms of failure modes which may be worth to know:

Although the STEC 55X is a rate based autopilot (i.e. turn coordinator is its primaty source) it is sensitive to failure of heading source, too. If the HDG information fails then all roll modes will not work except one: GPSS roll steering. I discovered that the hard way while vectored on an IFR flight plan but fortunately VMC conditions prevailed. ATC was a bit surprised when I informed them of being unable to follow vectors but DCTs to any navigation fix (waypoints, airfields, …) would be fine.

I consider it a bit paradox that simple “wing leveler” functionality would not be available but the technologically most advanced one, GPSS roll steering, is.