In order to obtain usable recordings and data, it is important to establish a good recording plan and coordinate this with the end customer. It is also important that the end customer recognize the significance of the recording process in achieving a reasonable sound simulation model. The plan should identify what sounds need recording, under what state conditions (e.g. engine idle, 0 knots ground speed, canopy closed, blowers off), and the sequence of recording events.
Since the recordings will be used both as a general base-line for tuning and potentially as incorporated sound effects in the trainer sound model, the recording plan should account for both situations.
The best process is to first obtain a series of general recordings covering all sounds and significant platform states. These should coincide first with trainer requirements (remember - there is a specification!) and then include any significant conditions identified by the customer. For example the customer may identify an engine rumble at 20% RPM which tells the pilot he has proper engine light off as significant to training.
All equipment on the reference platform that produce audio cues should be identified along with the range of states to be recorded. For example, Engines on an aircraft should be recorded from off through start-up sequence through idle, taxi, take-off, cruise, landing, taxi, and shutdown. A hydraulic pump that is audible (and significant to training) should be recorded during start-up, steady-state, loaded (if applicable), and shut down. Again, the record plan should identify these states and requirements.
Differences in platform configuration, which affect the audio, should also be identified. For example, is there a need to record the range of sounds under different conditions such as canopy open/canopy closed, turret door open/turret door closed, or different runway conditions? If the trainer requires these effects there can be significant differences in audio content for each of these states.
The recording plan should also identify a second and separate series of recordings for sound data that will potentially be used as part of the trainer sound simulation model. These recordings are more specific and need to be recorded in close proximity to the source equipment where possible to achieve the highest quality and usefulness. For example a helicopter hydraulic pump may provide a useful and distinct sound but may be partly masked by other ambient sounds in the cockpit. In these cases the plan should identify recording at the general "earpoint" (e.g. the pilot's headset level or center of the cockpit) and then at a location or locations that provide a better recording of the sound of interest. Again the equipment being recorded should be exercised in all of its performance states (start-up, steady-state, under load, etc.).
Note that some sounds may not be obtainable due to the nature of the sound, safety, cost, or other restrictions. For example, most helicopter pilots will not agree to throw some nuts and bolts into the main transmission just to get that "main gearbox failure" malfunction sound. Likewise, firing an AIM-9 missile a few times to capture the exact "Whoooosshh" may be cost prohibitive. None-the-less these sounds should also be identified in the recording plan and the reasons why the recording is not feasible.
In the cases where sounds cannot be recorded consider:
Having the customer describe the characteristics of the sound. Judicious use of "It Sounds Like" (such as it sounds like a bowling ball striking a metal trash can at 50mph) can actually be useful.
If the "It Sounds Like" model is used, enlist the customer to perform a little "Hollywood Magic" and record something that "sounds like" their description. In the above example, drop something heavy into a metal trash can and record it. Again, this should be identified in the recording plan.
Identify the desired sound cue as a data hole and that a best effort substitute sound will be used unless further data can be obtained.