Smaart can average transfer function measurements in two places: in Real Time mode, using saved transfer function (TRF) files, and in Impulse mode, using impulse response (WAV) files. The two methods use different math and give different results, so an average of a set of TRF files will not necessarily match an average of the same measurements saved as WAV files - even if you are averaging data creating in Real Time mode in IR Mode (as of v9.7 L-IFFT data can be saved as .wav data and viewed in IR mode). This article covers what each method does, why the results differ, and how to choose one.
What’s Inside a TRF File
Each TRF file holds four sets of data: Polar and Complex magnitude averaged data, each for both MTW (Multi-Time Window) and fixed FFT resolution. You can switch between them in the Trace Info window using two settings:
Averaging Type | FFT Type |
|---|---|
Polar | MTW |
Complex | MTW |
Polar | FFT (32k by default) |
Complex | FFT (32k by default) |
The Average value displayed in Trace Info is the temporal averaging that was applied when the measurement was captured. It describes how the live data was averaged over time. It has nothing to do with averaging several saved files together and once a measurement is captured, that value is fixed.
Saving a TRF File as a WAV File
When you save a TRF file as a WAV file, Smaart converts the file’s FFT complex magnitude averaged data set into an impulse response (.wav) using only the complex single time record portion of the capture. This uses the same calculation method used for the L-IFFT display. The polar and complex magnitude averaged MTW date sets, and the polar magnitude averaged FFT data set are not used.
Comparing Real Time and Impulse Mode Views
If you compare a TRF file in Real Time mode with its saved IR in Impulse mode, the magnitude traces match only when the TRF’s Trace Info is set to Averaging Type: Complex, FFT Type: FFT. That is the data the IR was made from. With any other setting, such as the common Polar/MTW view, you’ll see differences between the two displays even though both came from the same measurement.
How Impulse Mode (WAV) Averaging Works
Averaging impulse responses is mathematically equivalent to a full complex average. Magnitude and phase are combined together, so where the inputs differ a lot in phase, they partially cancel and the average loses energy at those frequencies. This works like an acoustic sum: several sources arriving at one microphone with different phase add constructively at some frequencies and destructively at others.
This is why an averaged IR often shows less level than the individual measurements. The drop is expected and comes from phase differences between the inputs.
Before averaging, Smaart time-aligns the impulse responses. The averaged IR has its peak at the average time position of the input IRs. Without this alignment, differences in arrival time would cause heavy cancellation and the average would be unusable.
How Real Time Mode (TRF) Averaging Works
TRF averaging is more involved than IR averaging. There are two methods, dB and Power, and each handles the Polar and Complex data in the files differently:
dB Averaging | Power Averaging | |
|---|---|---|
Level adjustment before averaging | None | Inputs are level-adjusted to a common mean dB between 255 Hz and 8.8 kHz |
Polar data | dB averaged, with optional coherence weighting | Power averaged |
Complex data | Complex averaged, with optional coherence weighting; the averaged magnitude is then rescaled to the average magnitude of the input files | Complex averaged; the averaged magnitude is then rescaled to the average magnitude of the input files |
The rescaling step is the main difference from Impulse mode. A TRF average’s Complex data still reflects phase relationships in the averaged phase, but its magnitude is brought back to the average level of the inputs, so it does not show the energy loss that a pure complex average does.
Before averaging, Smaart time-aligns the transfer functions so their phase responses fall roughly within ±90° of each other. This puts the averaged phase roughly in between the phases of the inputs. As with IR averaging, the result would be meaningless without this step.
Why the Results Differ
TRF averaging and WAV averaging are different by design, and their results can’t be compared directly. To make them match, TRF averaging would have to be limited to a pure complex average with no rescaling, level matching, or Polar averaging. That would remove the features that make TRF averaging useful for its main purpose.
Choosing an Averaging Method
No single averaging method is correct absolutely for all applications. The right one depends on your goals, and any average is an approximation. Consider, ketchup and Italian marinara are objectively both tomato based sauces, but they have different applications. Two common use cases:
Understanding how multiple loudspeakers interact at one location. Either measure with all the loudspeakers on at once, or measure each one separately and combine them with a complex average in Impulse mode. The complex average reflects how the sources actually sum at that point, including cancellation from phase differences.
Getting a general picture of frequency response across a coverage area, for example to set system EQ. Use TRF averaging in Real Time mode, with either dB or Power, and look at the Polar view. Polar averaging ignores phase differences between positions, so the result shows the overall tonal trend instead of position-specific cancellations that EQ can’t fix.
Summary
TRF files contain four views (Polar or Complex, MTW or FFT), set in Trace Info, and those settings describe how the measurement was captured, not how files are averaged. Saving a TRF as a WAV uses only its FFT data, so a TRF and its IR match only when the TRF is viewed as Complex/FFT. Impulse mode averaging is a true complex average and can lose energy where phases differ. Real Time mode averaging offers dB and Power options with level handling that a pure complex average lacks. Choose Impulse mode when you need to know how sources combine at one point, and TRF Polar averaging when you need a general picture of response across an area.