Recently I had the chance to investigate a Piper Autocontrol III (which is a rebadged Century II) 2-axis autopilot. The unit was installed in a Piper Cherokee and had been placarded “Autopilot inop”. The owner reported that avionics shops considered the unit not worth repairing and offered it to me to look at. I enjoy getting old electronics working (I had just finished restoring a 1947 radio), so I accepted the challenge!
Preliminaries and disclaimers
The regulations on installing and repairing avionics are quite strict in most countries (somewhat less strict in the UK if your aircraft is operated on a Permit To Fly), so don’t treat what follows as an invitation to do-it-yourself. If you have a Century autopilot and you live in the USA then Autopilots Central may be able to perform any necessary repairs.
These old autopilots are nowhere near as sophisticated as modern ones and you should expect them to fail at any time. They are unaware of airspeed. At least one fatal accident in the UK occurred when an old autopilot was used to maintain climb angle after take-off into IMC. The engine suffered a partial power failure, which the pilot didn’t notice in time, and the autopilot trimmed the aircraft into a stall.
So in my opinion these old autopilots should only be used in the en-route phase of flying, and auto-trim (if provided) should be turned off.
Although the following relate to the Century II autopilot and the 1C338 amplifier, the Century III with 1D395 or 1C515 amplifiers works on similar principles. The attitude channel in these amplifiers is similar to the roll channel. The oscillator is common to both channels.
Initial tests in the aircraft
The most basic ground test is to ensure that the AI is aligned at least approximately, engage the autopilot in roll mode, listen for a click as the roll servo engages, check that the yoke becomes stiff to turn, then operate the roll knob and see if the yoke turns as the autopilot attempts to level the wings. For this particular unit, the roll servo did engage but the yoke didn’t turn.
Faults in Century II and III autopilots are typically caused by amplifier faults or by bad connections in the circular connectors in the wiring loom. See my earlier post for more. As the autopilot hadn’t been used for an unknown number of years, I decided to test the amplifier.
This Piper Autocontrol III system has the amplifier built into the control console. It was easily removed and taken to the electronics workbench for investigation.
Visual inspection
The console was labelled as a 1C338. The schematic, parts list and some other details of this unit are included in document 68S54 which is available on the Internet. There are two different schematics shown for the 1C338. The differences between them include the following:
- The first one uses two non-polar electrolytic capacitors, where the second one uses pairs of back-to-back polarised electrolytic capacitors
- The PNP transistors in the oscillator circuit driving transformer T1 are MPS6519 transistors in the first but 2N4403 in the second
- The NPN transistor that drives transformer T2 is MPS6515 in the first one but 2N4410 in the second
This amplifier corresponded to the first schematic. These differences were correlated to the faults that I found, as will be seen later.
Inspecting the amplifier revealed that it used a single-sided epoxy glass PCB with rivets inserted where wires were attached. The underside (but not the top side) was coated with a clear lacquer. The parts list mentions this is “anti-fungal” and it should therefore be assumed to be potentially hazardous to health. It can be removed where necessary for component replacement using acetone and cotton buds.
There were no signs of scorching on the board. The components looked OK except that one black cylindrical capacitor had split. This turned out to be a 50uF 5V non-polar electrolytic. More about this later.
The resistors all appeared to be of the old carbon composition type but none showed signs of overheating or cracking.
Electrical inspection
The servo motor output transistors are known to be a common failure point in these old autopilot amplifiers. Using the diode check function on my multimeter, I tested the base-emitter and base-collector junctions of the four 2N3055 transistors in-situ. Three of them measured OK but the fourth measured an abnormally high base-emitter voltage (greater than 1V) and a slightly high base-collector forward voltage. I replaced all four of them. Each transistor had two mica washers between it and the heatsink, with no silicone grease or other thermal compound used.
The cracked capacitor (C19 on the schematic) was obviously faulty. The parts list just specified “50uF” and it was indeed a 50uF 5V non-polar capacitor. Upon removal it read 0.7nF on my multimeter. I dislike non-polar electrolytic capacitors (see my post on the KI525A, which uses an unreliable non-polar electrolytic capacitor that is better replaced by a metallised film capacitor) but I was unable to find a metallised film capacitor that would fit in the available space; so I replaced it by a 47uF 35V 85C non-polar electrolytic capacitor (I couldn’t find any 105C rated non-polar electrolytic capacitors).
Old electrolytic capacitors are always suspect, so I investigated the others:
- C12 and C20 are listed as “47uF 20V tantalum” in the parts list but 35uF 15V capacitors were fitted in this unit. After removal, both read less than 1uF on my multimeter. These are decoupling capacitors so the value used is not critical. I replaced them by 47uF 63V 105C capacitors.
- C8 is listed as 200uF in the parts list (no voltage rating specified). The actual capacitor fitted was marked 200uF 5V non polar. As a precaution I replaced it by a 220uF 16V 85C non-polar After I removed it, as far as I could tell from my multimeter it was still OK.
- The remaining three electrolytic capacitors are marked in the schematic as 5.6uF polar and in the parts list as 5.6uF 20V tantalum. In fact all three capacitors fitted were 5uF 15V non-polar. As the schematic indicates that these can all be polar, I replaced them by 5.6uF 105C 160V polar capacitors. I tested the original ones after removal and they appeared to be OK.
Two of the 2N3055 transistors have 68 ohm 10% 2W (according to the schematic) base series resistors. The resistance of these tends to increase over time. In this unit they measured 75 and 76 ohms so they were slightly out of tolerance. They could perhaps be replaced by modern 68 ohm 2W or greater metal oxide resistors, however care should be take because modern resistors are likely to be smaller and may therefore have higher surface temperatures.
Making electrical connections for bench testing
The amplifier connects has a PCB edge connector to make connections to the DG, AI and roll servo, and to receive power. This edge connector was 15 way single-sided with 3.96mm (0.15″) pitch. I didn’t find a suitable 15-way edge connector, so I used this 18-way connector with the amplifier PCB inserted at one end.
I supplied 14V power to the amplifier from a bench power supply with adjustable current limit. Much of the amplifier circuitry is referenced to the motor common terminal. So the power supply output needs to be floating with respect to ground, to allow you to connect the ground wire of an oscilloscope probe to motor common when measuring signals.
Getting the oscillator working
The Century II and Century III autopilots embody an oscillator that relies on a series LC tuned circuit in the directional gyro. In wiring configurations that use later versions of the 1C388 radio coupler, this is replaced by a series LC circuit in the radio coupler. The component values of this circuit are unfortunately not specified. When the autopilot is used in conjunction with a KI525A HSI and an older 1C388 radio coupler, the KA57 autopilot adapter supplies this circuit.
To bench test the amplifier without the official test equipment, it’s necessary to connect a substitute LCR circuit between the DG EX connections of the edge connector. After some calculations and experimentation, I determined that an inductance of around 4.7mH in series with 150nF and 47 ohms was an acceptable substitute for the LC circuit in the DG. The oscillation frequency was 5.6kHz which is a little higher than the target 5kHz. With the capacitor increased to 220nF the frequency was 4.5kHz, which is too low. I’ll try 3.3mH and 220nF when I get the chance.
Finding the correct values was complicated by the fact that the oscillator would not work initially because one of the two oscillator transistors Q15 and Q16 turned out to have an open base-emitter junction. When I discovered this, I replaced both transistors. The original MPS6519 transistors are obsolete, so I replaced them with BC327 which is a more modern bipolar PNP transistor with higher current capability, suited to this role. Note, the BC327 has a different pin-out order than the MPS6519. Another suitable transistor with a higher rated current would be the 2N4403 as used in the second version of the 1C388 schematic.
Further fault-finding
With the oscillator now working, I measured a 28V p-p between the two ROLL EX pins signal on the edge connector, as indicated in the manual. However, the output transistors were still not being driven when the roll knob was moved.
Tracing the signal from the roll knob through the amplifier revealed that there was a signal on Q7 base but not on Q7 collector, and probing with a multimeter indicated that Q7 base was open circuit. I replaced the original MPS6515 part by a BC337 transistor (again, the pin order is different). The second version of the 1C388 schematic uses 2N4410 in this position, which again has a higher rated current than the MPS6515; but that part too is obsolete.
At this point, I was able to control the motor output voltage by turning the roll knob on the console. The only issue was that when crossing the zero position, sometimes the current draw increased substantially and the oscilloscope revealed high frequency oscillations on Q7 collector. It occurred to me that the lack of a load to represent the motor would make the loop gain abnormally high, so I added a load between the MOTOR and MOT/COM connections. A 47 ohm 3W resistive load didn’t fix the problem by itself, then I added a 10nF capacitor in parallel with the resistor and the problem went away.
Here is the schematic of the amplifier test rig that I ended up using. The letters labelling the pins match the ones on the amplifier schematic. The single-sided edge connector I purchased has the connection spills labelled with the same letters but the pin order is unfortunately reversed.
Testing the roll signal input
To test the roll signal input, I connected the ends of a 10K linear potentiometer between the two ROLL EX pins on the edge connector, and connected the pot wiper to ROLL SIG. This worked as expected: adjusting the potentiometer affected the motor output in much the same way as the roll knob.
Testing the heading hold function
The heading hold function takes an input from the directional gyro to indicate the deviation between the current course and the heading bug. This signal is generated from the directional gyro using the excitation provided by the series LC circuit used by the oscillator. Although the DG signal input is capacitively coupled in the amplifier, most of the amplifier circuitry is referenced to the motor common connection and the capacitor value is quite high (220nF). So I deemed it wise to use a transformer to provide DC isolation between the inductor that substitutes for part of the DG tuned circuit and the DG excitation signal.
However, from the manual it appears that when the 1C388C radio coupler is used, the roll excitation is fed to the DG instead. So it should be OK to use another potentiometer across the roll excitation signal pair to create a fake DG signal instead, obviating the need for a transformer.
Tests using the heading hold mode with the fake DG input suggested indicated the heading hold function was also working.
Summary
This particular unit had three faulty transistors and three faulty electrolytic capacitors. It’s expected that old electrolytic capacitors will fail, and no surprise that one of the motor driver transistors failed. The other two transistors were low-current transistors in positions where the later 1C388 schematic uses higher-current ones. So it seems likely that these are known failure points that were addressed in the later version.
David Crocker's Solutions blog 