Impulse Response vs Transfer Function
Q: Wondering if I could get a hand with using Impulse Response to tune PA as opposed to using TF with Pink Noise.
I’m aware of the “Common way” blasting Pink noise, Finding the delay and then storing traces and averaging and making the adjustments. I’m just not familiar with using Impulse Responses. Is there an easy tutorial to follow or an article that I’ve missed that could help me understand it’s purpose? For instance using sine sweep etc….
A: In Smaart, IR mode is conceptually very similar to RT mode – in fact, you’ll notice it uses the same TF engine configuration (the TF engines you have configured in RT will follow you into IR mode). The acquisition is the same basic idea, just with an IFT at the end so we end up with time domain data rather than frequency domain data.
IR mode tends to use much larger FFTs (seconds vs milliseconds) and as a result, has much longer acquisition times. The main benefit is producing a measurement with much higher SNR than a RT measurement.
IRs are most useful if we want to investigate the acoustic properties of an environment (studying the decay of sound in the space). Doing this accurately requires us to expose 50 dB+ of the room’s reverberant decay, which is aided by longer FFTs.
Just as with RT mode transfer functions, IR mode can acquire a dual channel measurement using any broadband stimulus you like (pink noise, music, etc). It is not the case that pink noise must be used for RT measurement and sweeps must be used for IR. Smaart can acquire measurement data independent of test signal choice in both modes (they are both dual channel, and therefore signal-independent). It’s just that in IR mode, certain period-matched test signals offer some benefits.
Choosing period-matched pseudorandom noise allows us to drop the data window and further lower the noise floor of the measurement, and a period-matched log sweep (pink sweep) increases the dynamic range further mostly due to the signal’s low crest factor.
You can, of course, use these same stimuli in RT mode, however there is typically not much benefit from doing so – generating Magnitude, Phase and Live IR in realtime doesn’t require having such a high SNR. Likewise, if you acquire a large (high SNR) IR measurement and view the resulting magnitude response (Frequency plot) you will see that it gives you pretty much the same answer that the RT measurement gave you, it just took longer to there. For this reason, IR mode is not the best choice if your goal is to tune a PA system (IR mode also doesn’t produce a phase plot).
To tune a PA, a realtime measurement with Live IR, Magnitude, Phase and Coherence would be the preferred approach. For acoustic analysis, when long acquisition times are not a hinderance, IR measurements provide the mechanism by which we can penetrate deeper into the noise to better characterize the reverberant decay, and the high SNR data that results is important for accurate calculation of the Schroeder curve and the associated RT metrics.
Take a look at the user guide – there are two chapters dedicated to working with IR mode. You can also find some video content on our YouTube page if you feel like listening to me talk: https://youtu.be/brEgFZpsqCs
2 Replies to “Impulse Response vs Transfer Function”
I think the title of this article is kind of misleading. As you said the IR mode uses the same TF engines. IR mode measurement is TF measurement just like Real-time mode measurement is TF measurement. In Smaart you can’t do “one channel TF measurements” but in many other softwares it is possible. In that case the reference signal is not fed to the system via extra channel but is mathematically known in the computer hard drive as binary data (wave file). Actually you can compensate the reference delay and get the phase data in some softwares even if it’s “one channel”. All this kind of measurements are transfer function measurements. Dual-channel is problematic term because all TF measurements are not dual-channel measurement but the dual-channel term is often used to describe all TF measurement. “True one-channel measurement” is another story (SPL, RTA, Spectrogram) and has nothing to do with TF measurements. But once again the the term “one-channel” is problematic as I explained earlier (that’s why I called it “real one-channel” haha). Should there be some common standard to define measurement methods? Now it’s very confusing with all these different measurement softwares.
Well… they are the same data displayed different ways, but they are not the same thing. Transfer function = system response in the frequency domain. Impulse Response = system response in the time domain. They are related by T = 1/f and, in the general sense, can produced in the same fashion. In Smaart, they are performed in two distinct operational modes.
There is some amount of disagreement in the measurement community, but typically speaking the phrase “two port measurement” can describe both a realtime dual channel measurement such as the type performed by Smaart’s realtime and impulse response modes, and an “offline” response measurement as performed by some platforms, which are dependent on a specific stimulus. A “single channel” measurement is typically reserved for signal analysis measurements such as spectrum and SPL. However, the scope of this post is specifically the terminology used within the Smaart platform and not intended to be an exhaustive look at what every measurement platform is calling things. Same thing with “Control Group” vs “VCA” and some consoles labeling their polarity switches “phase.” It’s a lot to keep up with!