Temporal averaging just means averaging a measurement over some period of time. Typically this is done at a single measurement point or microphone position, although moving-microphone measurements utilizing temporal averaging are sometimes used for specialized applications.
In acoustical measurements, a significant amount of noise from various sources gets mixed in with the Signal we are trying to measure. The noise components fluctuate randomly from each individual “frame” of incoming measurement data to the next. This tends to make the charts “jump around” a lot, appearing visually noisy and hard to read. This is where temporal averaging can be useful.
Averaging over time increases the signal-to-noise ratio of a measurement through a process known as regression to the mean. The noise in incoming data, being more random than the signal component, tend to cancel each other out when aggregated over time. Since the signal components, are either stationary features (in the case of steady state system measurements) or less random than the noise (when analyzing dynamic signals) tend to average out to themselves, becoming smoother and easier to see.
The trade-off for doing this is responsiveness. When analyzing the spectral content of dynamic signals, too much averaging can mask fluctuations that are part of the actual signal (and may be indicative of a problem you’re trying to find or solve). In system response measurements, excessive averaging makes the measurement slow to respond to changes in system settings such as equalization and delay adjustments. The trick is to find a balance and use “just enough” averaging for what you’re measuring.
For electronic measurements, you can typically get away with very little averaging. For acoustical measurements, however, averaging varies with background noise levels and user preference. When measuring in a noisy environment, it can be helpful to use longer averaging time and utilizing the refresh averager/flush data buffer command by pressing the [V] key after making a settings change. This flushes and restarts the averaging buffers so that you don’t have to wait for the oldest data to fall out of the measurement before you can begin to see the result of your changes.
Averaging is present in the real-time mode of Smaart Suite, RT, and LE, using the numbers 1 through 10 to indicate an ascending order of “less” to “more” averaging.
.png?sv=2026-02-06&spr=https&st=2026-11-18T17%3A24%3A04Z&se=2026-11-18T17%3A37%3A04Z&sr=b&sp=r&sig=E2V8h1snlQsFHkoOqPHnUB%2FpPiAnv3yaC8N0AEWupCI%3D)
Less averaging = more responsiveness, more vulnerable to noise, less stability.
More averaging = more stable, better noise rejection, less responsive.
Settings 1-3 are FIFO averages, with 4-10 being weighted variable averaging (like an S-curve where time constants are shorter in the HF than the LF). We’ve chosen to use this numerical coding and not literal labeling to help make the averaging settings feel more like turning a knob, from “less” to “more”. With transfer function measurement, the default setting of “6” represents a reasonable starting place that provides both stability and responsiveness. From there, you can choose to increase or decrease that value to whichever of those two trade-offs you’d prefer to take. With spectrum measurement the default setting in “None” - displaying the raw FFT values calculated with banding applied to allow you to see instantaneous results. Transfer Function measurements do not have the “None” option, as without averaging, we are unable to calculate coherence, and the measurements would just be far too erratic to be useful.