Hello! I'm Ulla from Denmark.

This lesson is for week 5 of “Introduction To Music Production” at Coursera.org, and the subject is:

Create a recording that demonstrates comb filtering in an acoustic space.

Describe the recording situation and show proof of the comb filtering.

For this little experiment, I have used: a 15 m2 room with nice, hard surfaces, a singing bowl on a wool felt mat, a wooden mallet, a Zoom H2N portable recorder on a small tripod, a yoga mat, 3 woolen blankets, 3 dining room chairs and a handful of clothes pegs to hold the blankets in place.

As we know, comb filtering appears when a delayed version of a signal is added to the original signal.

It is easy to understand with a simple example such as two sine waves:

If the original signal and the delayed signal are perfectly in phase (i.e. their waveform peaks and troughs coincide exactly), they will add to each other and the resulting amplitude will be higher; if they are exactly opposite phases, i.e. the peaks of one of them coincides with the trough of the other one, they will more or less cancel out each other and the amplitude will be lower; if they have exactly the same amplitude, silence will be the result:

If the signal is more complex with many different frequencies, as our instructor demonstrated with white noise, a series of peaks and troughs will be created, extending up through the frequency spectrum and resulting in the characteristic “comb look”, something like this:

A delayed signal will be mixed with the original signal if a recording is made e.g. in a place where there are hard surfaces from which the sound waves will bounce back and thus reach the microphone a little later than the sound waves that traveled directly from the sound source to the microphone.

I you would like see a better diagram and to know about the math involved, look up comb filtering in wikipedia1.

I thought I might be able to demonstrate comb filtering if I could create such an environment, so from my small collection of musical instruments, I chose a singing bowl (because it gives a quite simple sound that goes on a long time), and I set up a nice, bouncy recording environment in a corner of my home office:

The room is about 15 m2, and there is linoleum on the floor. In the corner where I put the singing bowl and the recorder, there is a wall with a window and an electric radiator to the left, and to the right there are the built-in cupboards extending up to the ceiling. The singing bowl is a good half a meter away from the walls, and the recorder is a good half meter away from the singing bowl.

In order to show the comb filtering I needed to compare this with a recording of the same sound in a space without the feedback, and this is where the blankets etc. came in; I built this little hut out of dining room chairs etc.:

1 http://en.wikipedia.org/wiki/Comb_filter

The blanket on top can be folded down so that I have a closed room with only soft surfaces:

And here are the two recordings, one made inside the hut, one without the hut:

https://soundcloud.com/ulla-peterse/singingbowl-no-acoustic-feedback

https://soundcloud.com/ulla-peterse/singingbowl-acoustic-feedback

On the next page are two screen shots from “Sonic Visualiser”2; there is a screen shot for each recording, and the hut-recording (the one with no feedback) is a the top.

The top half of each screen shows a spectrogram depicting phases, where phase angles (a full cycle being 360 degrees) are shown with colors ranging from black over blue, green, yellow to red, and the lower half of each screen shows a spectrum for a particular moment of the recording.

For comparison I have tried to choose two similar moments from the recordings (it is difficult to strike with exactly the same force); you can see an overview of the recordings at the bottom of the screen shots.

The strongest frequency response from the singing bowl is a little below A at 440 Hz for both recordings; this is the pitch of the singing bowl.

The hut-recording shows only two really prominent frequencies, the 440 Hz A, and an octave and a fifth higher a little over D or just below E flat. There is a minimum of response around A of 880 Hz.

In the feedback recording, the minimum at around 880 Hz now has a prominent peak, and there are many more peaks now, giving the characteristic “comb filter” look to the spectrum.

These differences are also visible in the patterns in the corresponding frequency bands in the phase spectrograms.

In the hut-recording, there is less than one full cycle in the 880 Hz A area per 440 Hz A cycle, and about 3 cycles (in a funny repeating pattern of two “strong”, two “weak”) of the D frequency band per 440 Hz A cycle.

In the feedback-recording, there are two cycles in the 880 Hz A area per 440 Hz A cycle, and about 5 equally strong cycles in the D frequency band per 440 Hz A cycle.

2 Developed at Queen Mary University in London, see http://www.sonicvisualiser.org/

Well, that's all from me for this time. I had great fun preparing this presentation (when building that hut, I felt like a kid again!). I hope that you enjoyed reading about it, and I also hope that some of you know more about this subject than me and that you will take some time to comment on this little experiment. And anyway,

Thanks for reading! ☺