Transfer Function Instructions
Hardware Setup
Transfer Function is a powerful two-channel acoustic measurement tool. To use it, you need an audio interface that supplies two discrete channels of audio.
The Basics
Transfer Function is an acoustic measurement tool that mathematically compares a reference signal, which can be pink noise or program material, with the same signal after it has gone through a signal chain and speakers. Since it is looking at the difference between the signals, factors such as room reverberation and noise are minimized.
To do this, you will route a mono line-level source signal to the right channel input, send the same signal to your speakers, and connect a microphone to the left input, as the test signal input. With this configuation, the app can subtract the source from the test signal, and show both the magnitude and phase of the transfer function, so that you can correct system EQ and other parameters for optimum system performance.
Since the source signal is always changing, the app needs to know very accurately the time delay between the source and test signals. Transfer Function includes a sample-accurate Impulse Response-based Delay Finder for this purpose.
The first step is to setup the hardware. As stated, you need a two-channel interface, such as iAudioInterface2, or another high-quality two channel iOS audio interface. You cannot use the built-in mic, since it is routed to both channels in mono.
For this example, we will assume that you are using music as your program source material, although you can also use pink noise. Send the music, in mono, to the speaker system that you wish to test, and also split the same signal out and route it to the line input on iAudioInterface2, or the right input on other hardware.
Set the gain range on your hardware to avoid clipping. Then, set the line input gain range to match by tapping the “Ref Source” butotn on the Setup page. That means using the low range for signals that are below -10 dBu, and the middle range for signals up to +15 dBu.
Connect your microphone, and if possible set it up on a mic stand so that its position will not change during the test. Set the mic input range to avoid clipping.
That’s it. You are ready to setup the software for the Transfer Function measurements.
Using the Stereo Line Input of iAudioInterface
If you are using one of the iAudioInterface models, you can use the line input in stereo mode, to route both the measurement signal and the reference signal to the transfer function.
To do this, tap the “Ref Source” button on the Setup page to bring up the Reference Source page, and then turn on “Use Stereo Line Input”. This will switch the hardware so that the line input will be feeding the transfer funciton. The left input will be the measurement signal, and the right input will be the reference signal in this mode.
Measuring Time Delay
Accurate time delay is the key to good Transfer Function results. Time delay offsets will be a funciton of distance to the speakers from the microphone, and also may be introduced by DSP processors in the signal chain. It is important to set the time delay before performing any Transfer Function measurements.
To set the time delay, tap the time delay value, which appears on the top right of the iPhone display, and the bottom right of the iPad display. In either case, the Time Delay setup screen appears:
First, select an FFT size that will accomodate the longest delay that you expect. The maximum delay time is shown and changes as you change the FFT size. Remember that delay is approximatley equal to a foot per ms, or a meter per 3 ms.
Next, select the number of averages you wish Delay Finder to take for a reading. For pink noise, a couple of averages are plenty, although you may want more for music, or if the noise level in the area is an issue.
Now tap the Find Delay button, and wait for the results to be computed. The result will be shown in the text field on the top of the screen.
If the result looks reasonable to you, tap Apply to save it and to put that delay time into effect.
If you know the actual delay time from another method, you can enter it directly in the field.
Delay Finder resolves delays to 1 sample at 48kHz, or about 21us.
The setup screen has these fields:
- Meas Source – Select the microphone and gain range for the measurment input.
- Ref Source – Select the reference signal source. In most cases this will be the right channel input, but you can also set Stereo Input mode from this screen.
- FFT Type
Select the size of the FFT to use. Larger FFTs resolve frequencies better, but take longer to update. Number of Hz = Sample Rate / FFT Size. - Display Type
Select Magnitude only, Phase only, or both. - Peak Tracking
Show the frequency of the highest point on the magnitude graph, either on the plot itself, or in large numbers. - FFT Windowing
Turn on to enable windowing. Windowing an FFT reduces edge effects, and results in a more accurate display. - Coherence Trace
Turn on to show the coherence trace in red on the graph. - Coherence Blanking % – Magnitude and Phase plot values that are below this level of coherence will not be shown.
- Audio Pass Through
Turn on to listen to the microphone input. - Difference Mode
Used to show the difference between the current plot(s) and a recalled plot. - Lock Graph Scale
When turned on, the screen will not respond to touch GUI commands.
Operating Transfer Function
Transfer Function is a more powerful measurment than an RTA, since it has the ability to analyze a system even in a reverberant space. However, it is more complex to set up and use correctly.
The most critical factors involved in using transfer Function are delay, averaging, and coherence. If these are setup correctly, you will likely get good results.
Other factors that are important are input levels, FFT settings, and graph smoothing. We will look at all of these, and explain how to use them to get the best results.
Delay
Delay is very important, since the algorithm is comparing two signals that are usually shifted in time. Use the Delay Finder, by tapping on the delay time number on the main screen, to set the delay. If the mic moves in relation to the speaker even small amount, it will effect the results. Initially, you will see the highest frequencies bending up or down.
See the section on Measuring Time Delay
for more information.
Averaging
Averaging, which is set be tapping the button in the lower right corner, is useful to get the plots to settle down, and show more stable results.
Transfer Function supports averaging up to 16 FFTs, or you can also select Fast or Slow, which are exponential decay averages, or Avg mode, which is a continuous running linear average of all FFTs. To reset Avg mode, just re-select averaging.
You will want more averaging with music than pink noise. With pink noise, 4-8 averages or Fast mode are reccommended, while 8-16 averages or Slow work better for music.
Coherence
Coherence is a measure of how much one set of FFTs is like another. Coherence is what you can use to determine if you have the other settings, such as delay and gain, set correctly. A coherence of 100% would be perfect, but if you can get an average above 70% that’s fine.
While you are getting setup, and learning how to use Transfer Function, it’s a good idea to leave the coherence trace turned on (Settings screen).
In fact, it is so important that we have implemented Coherence Masking, to essentially hide results that don’t meet a minimum coherence level. Try setting Coherence Masking on the Setup Screen to at least 50%, or just touch the right edge of the display, and slide the Masking marker up or down.
Time and Data Windows
Transfer function employs both time and data windows.
The time window is a side effect of the FFT length and sample rate. Since we run at 48kHz, the time window in effect is simply the FFT length / 48000. Ideally, the first reflections from the speaker would lie outside that time window, to remove the effects of the room on the measurement.
The data window is set on the Setup screen, as the FFT Window.
Input Levels
As in any other test measurement, if your input gain is not set correctly, you will not get usable results. We have two small level meters visible at all times, to help you see the input levels.
Try to get the level at the half to three-quarters area, and make sure you avoid clipping.
You can use the microphone gain settings, and the line level input gain setting (available on the Levels screen) to get solid input levels.
Smoothing
Smoothing can help you visualize overall eq or phase patterns, by removing some of the detail from the graphs.
On iPad, you can set the magnitude and phase smoothing independantly, and on iPhone, one setting controls both plots.
FFT Settings
There are two settings that control the FFT itself. FFT size, and FFT Windowing. Both are set from the Setup screen.
FFT size is a tradeoff between speed and frequency resolution. For most cases, we would suggest 8192 or 16384 points. Windowing helps remove artifacts and increases dynamic range, and we suggest using it in most cases. We have 3 different windowing algorithms, Hann is our personal favorite.
