The problem of how background noise affects a simulator is one that is not often considered, but there are a number of ways this can impact the successful certification of a Level D simulator. At the simplest end of the problem spectrum, the installation background noise can exceed the measured threshold for one or more aircraft/simulator test cases (particularly the quieter ones), and this may require exemption by the certification authorities, since there is often very little that the simulator manufacturer can do about such a situation. More problematic is a situation where a simulator is installed and tuned to meet the plot tolerances, and then is exposed to some external change in the background noise. As seen in the hypothetical situation below, the impact of such a change can very easily be test case dependent, and require that certain test cases require re–tuning, while others may not. Unfortunately at this time technology does not have a viable, simple solution for this issue.
Based on current capabilities, if the background noise changes, then some, or possibly all test case conditions will require re–tuning.
The current state of the art in sound simulation is almost certainly represented by the Level D full–flight simulator. In this case not only is the sound simulation required to satisfy the subjective assessment of experienced flight crew, but additionally the sound system is measured using a tool that represents the sound levels across the frequency range in 1/3rd octave bands known as spectral analysis. Industry has homed in on a requirement to match the spectral “signature” of the simulator, with the capture on an example of the real aircraft, within +/-5dB for each 1/3rd octave band. In practice the requirement is not quite so rigid with individual band exceptions not being uncommon with justification.
Typically simulators are installed in flight simulation centers, where there may be any number of other simulators installed. The construction of these centers varies, and includes custom facilities where there is one simulator per simulation bay with a concrete or block wall separating one device from the next, to large hanger—like halls with any number of devices installed side—by—side.
And it is now we get to the crux of the matter under discussion. The Level D testing methodology includes tests case conditions for all phases of the aircraft operation, spanning eight defined cases including "ready for engine start", through to “final approach”. Somewhat obviously, the sound levels experienced over these conditions varies with the aircraft type, and if the sound system is doing its job, the sound levels should match exactly those found in the original aircraft. Lets discuss very first test case – Ready For Engine Start. The normal condition for this state is that the APU is running, cockpit electrical power is in the normal configuration, and generally the air–conditioning system is in the normal configuration. At this point some thought should be given to the likely environment that the original test aircraft was in during the audio recording for this test. Almost certainly the aircraft was out on the taxi-way, quite some distance from any buildings, and/or other equipment, and it is a requirement that consideration is given to avoid any other external noise affecting the recordings (like other aircraft operations).
For a typical aircraft (lets use a 737NG) the recorded cockpit levels for the Ready For Engine Start condition vary from 66dB at 50Hz, through 52dB at 1kHz, and drop to 37dB at 16kHz. Without any comparative reference those numbers are difficult to assess, so here are some numbers recorded in the engineering area of the ASTi offices where there are something like 30 computers running; at 50Hz measures 43dB, 1kHz 49dB, and 16kHz 16.5dB. Next, here are the numbers for the ASTi server room; 50Hz at 45dB, 1kHz at 55dB, and 16kHz at 20dB.
Well let's look at the second set of numbers recorded in the ASTi server room, and in particular the value for 1kHz. This was recorded as being 55dB, but looking at the levels from the aircraft it was recorded as being 52dB. The bottom line is no matter what any sound system did, the 1kHz band would always show a +3dB. Now this type of problem for a simulator installation is such that might warrant exempting the 1kHz frequency band from the +/-5dB requirement, because it would be quite likely that some of the required aircraft sounds would introduce some additional noise in the 1kHz band in addition to the base–background noise, and push the level over the limit. Generally the certification authorities understand this type of issue, and will accommodate problems such as these, particularly where the root cause can be explained, and indeed measured.
However there is another problem related to this discussion that needs consideration, what happens when the background noise changes? Unfortunately the answer is “it depends”. On the surface that doesn't seem very helpful, but first we need to understand some basic math related to sound levels.
The first thing to recognize is that sound is measured using Decibels (dB) which is a logarithmic unit, which corresponds to how our ears happen to work. Therefore we need to recognize that sound levels do not add according to simple math, but according to log math. To take a real example: one sound that measures 20dB SPL, when added to another sound measuring 20dB SPL, does not produce a sound of 40dB SPL, but in fact would measure 23dB SPL. (Final SPL = 10Log (inverse LOG (SPLa/10) + inverse LOG (SPLb/10)), all values in dB)
Now let's return to the simulator and consider a real problem. First, let's assume that we are installing a simulator in the ASTi engineering area, where the background noise at 1kHz measures 49dB. From the aircraft data we need the signal to measure 52dB at 1kHz, so from the math, my simulated sound system must generate 49dB of signal at 1kHz, giving a perfect value of 52dB. Now, what would happen if we moved the same simulator into the ASTi server room? Remember the background noise measured 55dB at 1kHz in the server room, so now, the resulting signal with the simulator sound system running, will equal 56dB, and increase of 4dB over the original case.
So we need to be aware of the importance of the background noise in what we will measure whenever we take a spectral plot, and be cognizant that any change in the background noise profile can, and will have a direct effect on the measured response of any sound system operating within that environment. However what may not be so obvious without some thought, is that the effect of the change in noise will not be linear for the various test cases on the simulator, since the average levels for these test cases vary by some considerable amounts.
To be sure we understand this, let's take another example test case – Steady Climb – in this condition the aircraft measured 65.7dB at 1kHz. Now reverting to our example, if I tuned the sound system to match the plot in the ASTi engineering area (SPL 49dB at 1kHz), then my sound system will be generating 65.6dB at 1kHz (note the background noise only contributes 0.1dB to the measured value). Now, after moving the simulator back to the server room (background noise 55dB at 1kHz), we would see the measured 1kHz band report 66dB, a change of just 0.3dB over the original installation.
So, this is a real problem, and not one that can be accommodated by simple overall level changes. To summarize, if we had moved the simulator from the ASTi engineering area to the server room, the quieter test case (Ready For Engine Start) would have been affected by 4dB, while the much louder overall Climb condition, would have only seen a change of 0.3dB. Such a change would necessitate re–tuning the simulator for the Ready For Engine Start test case, but would require no changes for the Climb condition.
So far this discussion has discussed what would happen if a simulator were moved from Site A to Site B, but the exact same set of problems occur if the background noise changes at the same location. This could happen for many reasons, but let's run through a few that are plausible: an adjacent simulator is running versus not running during original testing, another simulator is powered off for maintenance shutting down a noisy hydraulic pump, the building air–conditioning switches from heating to cooling, an external door is open to allow delivery of another simulator, etc. All of these situations have the potential to greatly change the background noise floor as measured inside the simulator cabin.
Is there a solution – what about sound insulation? We are faced with several problems when attempting to insulate any area acoustically, but there are two techniques; the first is to block the sound waves from entering the area in the first place (sound blocking), while the second is sound absorption where audio energy is reduced by conversion into heat energy. Blocking sound usually involves dense, thick materials that simply do not pass the sound energy. Fundamentally the materials used in a simulator are generally chosen for their relative lightweight, and structural characteristics, since the motion system intends to move the simulator around, at times at quite high accelerations. So we're probably not going to see a concrete simulator anytime soon.
This leaves us the option of sound absorption, and this can have some effect, even considering the requirement to keep the weight load to a minimum. There are a number of off–the–shelf materials that may be applied to large panel areas to reduce resonances, and sound transmission. Also sound absorbing materials may be packed into all open voids in the structure, and an attempt can be made to real all unnecessary openings in the simulator structure. But there will be a limit to how effective this can ultimately be, and it should be noted that it is very difficult to provide any effective absorption in the low frequency range (< 200Hz).
Download a discussion paper on the impact of background noise on sound systems. (.pdf)