I've got a bunch of comp time saved up, and my schedule opened up when good weather occurred this week, so I went flying three times during the week. This time of the year you need to take the flying opportunities when you can get them.
Monday I did some more autopilot testing (flt 101), trying to find a combination of settings that gave acceptable altitude hold performance in turns at forward CG. It works marginally well at shallow bank angles, but I prefer to have the autopilot use 25 to 30 degrees of bank. There is typically a 150 ft altitude loss in a turn with that bank angle.
I also did a test to confirm that the autopilot automatically disengages during the takeoff roll if it happened to be engaged. It disconnected at about 40 kt, just as advertised. I set up the minimum and maximum speed protection at 70 kt and 190 kt respectively. I did a low power climb and a high power descent, both in vertical speed mode, and found that the minimum and maximum speed protection worked as expected.
I called Trio Avionics late Monday afternoon to discuss the altitude hold performance with Chuck. He said that there were several other RV-8s flying with their autopilot, and no one else is complaining about the performance. He also said that they design the system to use about 15 degrees of bank during turns, and accept up to 30 ft of altitude loss. I'm seeing 40 to 50 ft of altitude loss at 15 degrees of bank. We discussed the possibility of Trio changing the upper limit on altitude hold gain - I had already tried the current maximum value of 60. Chuck also reported that one user had found poor performance which was eventually blamed on turbulence form the static port. On that aircraft the interconnection between the two static ports was plumbed close to one of the static ports, and the performance improved when the interconnection was moved halfway between the two ports. Chuck hypothesized that different pressures on the two sides of the fuselage caused flow between the two static ports, and this caused turbulence in the line which affected autopilot performance. He suggested taping one port over for a quick test, as this would prevent any cross flow between the two ports. The other option is to move the pushrod pivot on the elevator bellcrank closer to the bellcrank's pivot, which would increase the amount of elevator movement for a given amount of autopilot servo movement.
Thursday I spent the whole day at the airport. First I did a short flight (flt 102) with one static port taped over, and found no appreciable difference in autopilot performance. Next I removed the aft baggage compartment to provide access to the pitch servo and drilled a new hole in the elevator bellcrank. The new hole gives about 50% more elevator movement for the same servo movement. I did another flight (flt 103) after this change, and didn't see much of an improvement.
Thursday I also gathered some cruise performance data with the wheel pants off. The wheel pants have a tendency to fill up with snow when operating from snow covered runways. The heat from the brakes may cause some of the snow to melt, and the water can freeze on the brakes during flight, causing the wheels to be locked on landing. It is a pain to put the wheel pants on and off as the conditions change, so most RVers just leave them off all winter.
Saturday I flew again (flt 104), experimenting to determine the best way to use the autopilot when manoeuvring, given its performance. I'll set up the autopilot to do shallow turns. If I want to make a sharper turn I'll press and hold the autopilot disconnect switch - if I hold it for more than five seconds the autopilot goes into what Trio calls Pilot Controlled Steering mode. It will reengage when I release the switch, and will then hold the current ground track and altitude (assuming the vertical speed is less than 200 ft/mn when I release the switch). This allows me to easily intervene manually to fly the turns, but have the autopilot do the straight and level bits. This allows me to get most of the workload reduction I hoped for when manoeuvring in instrument conditions.
I also gathered some more wheel pants off cruise performance data. I've got some more work to do to finish analyzing this data, but the initial look seems to show the aircraft is 8 kt slower at the same power than it was with wheel pants on.