Testing the performance of the autopilot typically involves engaging each mode and assessing how accurately it is able to carry out its function. As autopilots rely on external signals for many functions such as VOR tracking or automatic ILS approaches, this type of testing has to be conducted within an assessment of the accuracy of the signals received. Only when it has been established that the external signals are being received accurately can the ability of the autopilot to manoeuvre the aircraft in response to these signals be evaluated. Testing of the autopilot is conducted initially in ideal conditions of low turbulence to establish the baseline performance and finally in operationally representative conditions such as high turbulence and large crosswinds.
Although there are too many autopilot modes to discuss in detail the tests used for the most common functions are detailed below:
• Altitude holds. These are tested by engaging the hold and measuring the accuracy with which it maintains the datum in still and turbulent conditions. As an example a recent UK military helicopter specification called for the barometric height hold to maintain the aircraft within 25 feet of the engagement height. Deliberate use of controls in the other axes is also made to determine if the hold is able to cope with the off-axis interference (for example, the angle of bank is increased, airspeed changes are made, or with a three-axis system which maintains altitude through the pitch channel, power changes are made). Large disturbances can cause the series actuators to saturate and the slower-acting parallel actuators may be unable to respond with sufficient speed thereby causing a large disturbance from the datum. Such testing may lead to restrictions being placed on the flight envelope with the hold engaged. The actuator activity is always examined closely to ensure that saturation has not occurred. For 3-channel systems that also employ trim follow-up the cyclic activity demanded to maintain the series actuators within authority is an important issue. The capture of the hold datum is also explored by, for example, engaging the hold with a rate of climb or descent. The ease with which the hold datum can be changed is also examined; for example, it should not be necessary to completely disengage the hold and then re-engage it when at a new altitude.  vertical speed and its behaviour when the direction of the commanded speed is reversed are also investigated. In addition the ease of changing from the vertical speed mode to an altitude hold mode is assessed.
• Airspeed holds. Testing of an airspeed hold is approached in the same way as a height hold by noting the deviations from the datum in still and turbulent conditions. Power changes are also made to check the hold accuracy as the power is changed and when steady in climbs or descents. As some autopilot systems provide airspeed hold but rely on the pilot to select the datum speed the accuracy and workload associated with selecting a new airspeed can be of great interest. In addition the effect of pressure errors on the system are investigated by inducing deliberate sideslip.
• Turn co-ordination. The effectiveness of turn co-ordination features is tested by conducting turns at various rates without use of the yaw inceptor. Manoeuvres such as roll reversals are also used. A particular area of interest is the manner in which the turn co-ordination is mechanized and integrated with the heading hold. On some systems turn co-ordination will disengage and the heading hold will engage below set yaw rates, bank angles, or airspeeds. For example, in forward flight the SFIM 85T31 autopilot fitted to the AS 355 changes from heading hold to turn co-ordination as the angle of bank exceeds 5°. The heading hold will then automatically re-engage on completion of a turn as the heading rate falls below 1.5 degrees per second. The Boeing Chinook system on the other hand uses lateral cyclic displacement from trim to engage turn co-ordination. Clearly the implications of this automatic changeover requires investigation in a variety of flight conditions and role manoeuvres. This is particularly important if the logic changes above a certain airspeed to account for the need to turn the helicopter in forward flight as opposed to performing a lateral sidestep in the low speed regime.
• Navigation modes. Autopilots can offer a number of navigation features. In each case the testing examines the accuracy with which the autopilot can track the navigation feature being used as well as the performance of the system when capturing the datum. For example, a VOR mode is tested by intercepting the required radial at a variety of intercept angles and airspeeds. The feedback gain should be high enough to correct errors in tracking but not so high that it leads to uncomfortable and frequent roll inputs; this becomes particularly important as the aircraft approaches the navigational aid or waypoint. This type of assessment may highlight problems in the relationship between the roll and yaw channels in turning the aircraft and maintaining the heading. The action of the autopilot at turning points is often an area of great interest as the system logic has to decide when to start the turn onto the next leg and then has to capture this new track. A consideration is the matching of the demands of the autopilot with the capabilities of the aircraft. It should not be possible to programme a search pattern which the aircraft is physically incapable of following.
• Automatic ILS approach mode. Testing of this mode is conducted in a variety of wind strengths and directions to assess how well the localizer and glideslope indications are maintained. The full range of permitted airspeeds is also used. As with navigation mode testing the capture of the localizer is assessed at different approach angles. Three-axis systems are sensitive sometimes to power variations during the approach therefore collective inputs are made once established on the localizer and glideslope and tracking accuracy is noted.
• Automatic transitions. A common feature on naval and SAR helicopters are automatic transitions to and from the hover. As this autopilot mode operates close to the surface, often in conditions of degraded visual cues, it is particularly important to test it thoroughly. Because these autopilot modes are required to control the aircraft from one extreme of speed to the other and climb and descend the aircraft, they are usually given considerable authority to be able to perform their task. Tests are conducted at a variety of weights, sea states, centres of gravity, wind speeds, turbulence levels and entry conditions. If automatic series actuator trim is not provided tests are made with different initial actuator positions to determine at what point saturation occurs. Considerations also include the ease of deselecting parts of the system, the monitoring requirement on the crew, and the safety of the flight path of the aircraft. The actions that the pilot would have to take in the event of an engine failure during a transition are an important part of the assessment.