Getting started with the Transfer Analyzer

This should give an introduction to the Transfer Analyzer and help to understand it's theoretical background and practical use with LAMA.

Theory

Transfer Analyzer

The Transfer Analyzer in LAMA works Source Independent, this means that no specified test signal is required, the transfer and coherence function are calculated with a reference signal and a measurement signal. While the kind of source signal will not affect the result being calculated, the source signal affects how fast and how good a measurement can be performed - this is obvious: imagine the references signal would be a sine-wave with 1 kHz. The only frequency on which the system can be determined would be 1kHz and nowhere else.

But don't worry: common music signals are broadband enough to make good measurements, when your program is speech, results will be good in the area of the human voice, which is usually the interesting area in that case.

However, with white or pink-noise it is possible to achieve results in the fastest way - so this might be the signal of choice when doing system-tuning.

Coherence

The 'quality' of the measurement is reflected in the Coherence Trace. Coherence ranges from 0 to 1, higher values reflecting a better measurement. LAMA automatically fades the areas with low coherence values so that they are not displayed - so you do not need to watch the coherence trace, but of course you can if you want to. When coherence is low in some ares, there might be the following reasons for that:

  • The reference lacks signal in a specific spectrum.
  • Noise in the measurement signal: every kind of signal in the measurement path that is uncorrelated to the reference signal is noise and degrading coherence (but is not included in the transfer calculation) In common language: every other sound source then the reference signal (like a loud audience or machines during setup) will lead to lower coherence, but it will not lead to wrong results.
  • Missing frequencies in the measurement signal: Especially in the low-mid area comb-filtering of the measurement signal can occur. As these frequencies are "missing", coherence will be low for these frequencies.
  • Wrong or bad time alignment: When the measurement signal arrives later then the reference signal, Time Alignment of the two signals must be done. LAMA has a Delay Finder integrated which does this job quite automatically for you. This will be discussed later.
  • Bad Signal/Noise ratio: this is the same as above, and it should be obvious why... no measurement can be done without signal.

Averaging Time

When coherence is low due to reasons of the current audio program, it might help to increase the Averaging Time of the transfer analyzer. This leads to a less responsive measurement but with better quality.

Hooking Up

Signal Flow for the Transfer Analyzer

Now it is time to connect your signals to your audio interface. For a typical live-sound setup the reference signal should be picked up before the main system-equalizer - otherwise the corrections being done with the equalizer will not be reflected in the measurement (because then the path Amp-Loudspeaker-Room-Microphone is measured and that path remains the same).

After having set up the system as described above, it is time to route corresponding signals into a new instance of the Transfer Analyzer:

Input Settings for the Transfer Analyzer

  • Create a Transfer Analyzer Instrument
  • Set the Source & Response input (Note that routing can be changed with the Patch Matrix)
  • When changing input routing, delay measurement will automatically be invoked to assist with proper time-alignment - see the Delay Finder page for details.

Usually, the transfer analyzer should be up and running now - check the coherence trace to see if you are getting proper results.

Multiple Traces

Of course you can add multiple measurements to the same window - either having more than one Measurement Point in your venue, also measuring your System EQ - what ever is preferred

Snapshots, Averages & the Accumulator

When doing system tuning, someone might measure multiple points in the room and find an average transfer response of those points.

  • Snapshot: a Snapshot of a transfer measurement can be taken in the Add menu on top of the Transfer Analyzer Instrument.
  • Average: After having taken some snapshots, it is possible to select two or more of them and select Average in the Add menu. A now trace will show up, which holds the average of those snapshots.
  • Accumulator: The Accumulator is a growing average buffer - someone can build averages of various measurements without creating a snapshot: Just select Push to Accu, and the value of the selected trace (Which is the one that has been pushed into it) will taken to display the average of all traces that have been pushed into the accu.

Reading the Results - and don't forget to hear!

Ok, this is a kind of personal experience: No measurement system should tell you what to do. The final decision is done by your ears, your feelings and your experience! But a measurement system can give you some great assistance - and that is what it should do. If someone is at the start of the learning curve with measurements, I would suggest to "work" with ears only, but to read the results too. Do some experimenting, and see how changes you perform on the system are reflected in the analyzer. It is not very difficult to learn what it is that LAMA tells you - but certainly it won't tell something like "cut 315Hz by 1.5 dB on the Main EQ and it will sound perfect".

One of the first thing someone will discover is that it is good to take care of drops in the Frequency Response - your ears simply don't get them as good as peaks (and you should not need a measurement system to discover this +10dB peak at 400Hz...)

Speaking of drops: bear in mind, that it is not necessarily a drop - it might also be some comb-filtering in the Measurement Mic Signal Path (especially in Low Mids) - but after having read this, you should know that it is possible to separate these by reading the Coherence trace.

Some things more...

There are some more parameters that can be adjusted in the Transfer Analyzer - more experienced users might want to know what they are about:

  • FFT Size: A bigger FFT-Size gives more granular results (especially in the Low Frequency Area) but makes them slower. When working with a Sample Rate of 44k1 or 48k, 32768 is a very good value to start with.
  • Window Type: Usually a "Hanning" window works best with music or speech - you might want to check out this article to lean more about the math behind it.