Author: Guido Helbling, Avosound - last updated June 20, 2017
Don't worry, I won't bore you with a treatise on acoustics. But multi-channel impulse responses can often create friction between the individual channels (e.g. over-emphasised frequencies and phase cancellation), so it is worthwhile to analyze one or two such situations
Let's recap: with a speaker we generate reflections of a sweep signal. Fortunately, this sweep signal is more than just a mere annoyance; a closer inspection of the signal reveals all sorts of information about the acoustics of a room - or its 'soul', if you will. :-)
On the page create impulse responses with sweeps I have demonstrated what a sweep is. Let us look at the spectrums and waveforms of recorded sweeps to get an idea:
A sweep signal is a signal that rises uniformly over the frequency range of 20 Hz - 20 kHz with the same audio level. If the signal is now played back and recorded in a room, you can see what the acoustics of the location have done to our beautifully clean sweep:
The perfectly shaped computer-sine-body has turned into a fully grown sine sweep, warts and all. Seriously, Sweep 1 is a very even and balanced sweep with beautiful reverberation, which you can clearly hear and see from the signal in the spectrum.
Good to see and hear are the resonances in the bass range, particularly the clanging at 180 Hz. As the dimensions of every room can become problematic for acoustics, we can safely say that our area produces a distinct resonance at 120 Hz. This does not mean anything negative for our impulse response - on the contrary, in fact. Our impulse response is a fingerprint of the room we want to copy (and in which these room modes are included). The rumbling frequencies are not very strongly developed, indicating that they belong to the natural peculiarity of this room. However, if they occur too strongly, a shifting of the microphones is recommend - or, alternatively, reducing the playback volume of the sweep or tweaking the sweep in post-production might also help.
The typical behaviour of room resonances is very peculiar. The increases that we see in the spectrum (clearly audible in the recording) are in proportion to each other: the first resonance occurs at approximately 45 Hz to 50 Hz, the second at 90 - 100 Hz, followed by 120 Hz, 180 Hz and 250 Hz.
90 Hz is a frequency doubling, i.e. an octave, of 45 Hz in the sub-region. A new, particularly strong resonance appears at 180 Hz, which is one octave again. Also, the resonances of 250 and 120 Hz are in proportion to each other and are one octave apart.
Sweep 2 is very quiet in the bass region, but sports clear resonances in the midrange. This sweep hints at a rather small space, whereas sweep 1 originates from a much larger space.
It is unfortunate that you can only do these analyses after the sweep has been recorded. More importantly, you can generally only do these analyses in the studio after the fact. And that is usually when problems occur and strange phenomena begin to happen. Then it's too late for adjustments and corrections, as they quickly become time-consuming and complicated, resulting in unusable sweeps. There goes all your hard work!
But if you take the right precautions, this can be avoided.
When impulse responses produce reverb with negative phase angles, it should be used with caution. Especially in audio post production and surround sound applications, this is undesirable since down-mix versions annihilate negative and positive areas, so sounds may disappear. This is an unpredictable phenomenon and must be prevented; hence negative phase angles should be avoided.
I have despaired over phase problems between the individual channels many times in post-production. Dozens of tracks ended up in the bin because they were simply not phase-stable enough.
Basically speaking, if phase problems occur between the individual sweep channels, you can be sure that the same problem will happen with the converted impulse responses. In the example of Sweep 3, you can clearly see how the phase meter moves into the negative range in the bass region. It is very important to prevent this during recording.
The causes of these problems had to be investigated and their origins found. Since the phase problems occured randomly and were frequency-dependent, a list of possible suspects was made:
Phase problems can occur through unfavourable microphony. There are microphone arrays that are generally more stable and less sensitive to phase problems than others. As the microphone capsules in the XY or MS stereo method are directly above each other, these systems produce their stereo image from the level differences. In contrast to AB microphones that are used in arrays for surround recordings, you can assume that stereo recordings with MS or XY produce no phase problems.
For Figure 18: Vertical arrangement of microphones results in additional phase equality, even if omnidirectional microphones are used. The two Neumann MS stereo microphones can be seen clearly above.
When the phase problems persisted, other possible culprits had to be excluded.
I never reverse channels! And anyway, MY cables are soldered properly! Or are they ...?
Nevertheless, it pays to be critical of oneself and to check one’s setup.
Since I had a large setup with ten audio channels for the Auro-3D® 11.1 impulse responses, I had sufficient channels to make comparisons. As front and rear each form individual stereo pairs in a surround setup, I was able to compare the behaviour of the phase relations between the microphone system. When the phase problems occurred independently of the microphone array, my long-held suspicion was confirmed: the phase problem was of 'natural' origin and had thereby effectively been recorded.
As I had a laptop with an audio interface for the playback of the sweeps with me, I decided to display the phase relations with a phase meter. This proved to be very helpful, since naturally occurring acoustic phenomena due to phase problems can be displayed easily. Naturally, using a correlation level meter is an additional step in the recording of impulse responses and it certainly does not make the workflow any faster. Ultimately, though, it was worth the effort to find out where problems occur and what not to do.
Here I have listed a few points that should help avoid certain problems. Simply because sweeps are out of this world! Sweeps are about as natural as the cosmetic surgery of certain rock stars. Sweeps have no resemblance to a violin or some other instrument. Because they cover the entire frequency spectrum, they stimulate all the resonances in a room due to their loud volume and bring them to life – with a roar! Thus, some points deserve special attention:
This was the first part of the tutorial on impulse responses. A second part is in the planning stage and will shed light on comparing impulse responses and how to make a final setup. Unfortunately, I have yet to find the time for it.
I hope this tutorial will help those who want to create impulse responses themselves. Those who would rather skip all the frustration and anger should look at my IR1 Impule Response Set. I also hope that the tutorial is not difficult to understand or too long-winded. And, naturally, I appreciate any feedback and comments from you.
Author: Guido Helbling, Avosound