Article

Monday, January 23, 2012

Measurements Without Computers: Leave The PC At Home And Get Better Data

Combining the hearing process with the measurement process, allowing the benefits of each to be exploited

It is difficult to describe how something sounds.

The terms are vague and subjective, and often have different meanings to different people.

Sound is meant for listening, and the listening process is the most power diagnostic tool for the audio technician.

Measurement systems yield lots of data, but it can be difficult to interpret.

It makes perfect sense to combine the hearing process with the measurement process, allowing the benefits of each to be exploited.

The process of convolution allows measured data to be evaluated by listening. The method that I use requires the room impulse response (RIR) to be saved as a Wave file, the default file format for several measurement applications. The RIR WAV file can be listened to in a wave editor, or convolved with anechoic program material in a special program (read on!).

Of course, all of this requires a personal computer. This can be a hassle, as computers can be awkward and unwieldy when used to evaluate different listener locations. The setups can be time consuming, requiring very long cables, interface boxes, etc.

Ironically, the power of the personal computer can allow the RIR to be gathered without it. Here is one method to gather the system RIR without a PC, and how to post-process the data to evaluate it and listen to it.

The Signature
The RIR is without a doubt the single most important acoustical measurement.

It’s also one of the oldest, having been implemented by hand claps, balloon bursts, starter’s pistols and even yachting canon fire. Within it lies a wealth of information about the sound system and acoustic environment.

The principle is simple – whatever the room does to your hand clap, it will also do to your voice, or to a musical instrument played at that location in the space. The RIR is the “signature” that the room places on the sound.

Once the RIR is gathered at a listener position, it can be analyzed to reveal information about the direct sound, early and late reflected fields, and the diffuse field of the room.

If the sound system is used to excite the room, the RIR will also contain information about the loudspeaker’s response and resultant performance as it relates to speech intelligibility and music clarity.

In short, most system/room characteristics that can be heard at a listener position will be included in the RIR for that position.

The freeware GratisVolver (www.catt.se) can be used to convolve the RIR with anechoic music or speech.

“Convolution” means to take one file (time or frequency data) and encode it with the characteristics of another (time or frequency data).

If one file is dry program material, and the other is an RIR gathered in a room, the convolution will yield what the dry program material would sound like when encoded with the acoustic characteristics of the live space for that specific position.

Ideally, the RIR should be two channels gathered with a stereo microphone. This preserves many of the localization cues required by listeners to pinpoint a sound in three-dimensional space.

The benefits of convolution are obvious:
• Speech or music can be evaluated without making actual recordings of talkers or instruments.

• The RIR can be modified to simulate acoustical changes to the space. Convolution can then be used to evaluate the effects of the changes.

• There are signal-to-noise advantages over recordings. If the RIR is gathered in a noise-immune manner, then the convolutions will have a lower noise floor than actual recordings made in the space.

• The RIR is a complete documentation of the room/system response at one listener position. It can be reprocessed in the future as other processing algorithms become available.

Regardless of what measurement platform you use, a properly gathered RIR should have the following characteristics:

• When displayed on a dB vertical scale, the data should fill the screen from top left to bottom right.

• The full decay of the system must be measured, meaning that the decaying tail must not be truncated.

• For convolution purposes, good signal-to-noise ratio is required. I like to have at least 90 dB of dynamic range, and see the decaying energy fade into the noise floor at -80 to -90 dBFS. Less can work, but there may be some artifacts in the playback.

Applications & Processes
Deconvolution is the opposite of convolution.

A known stimulus is played into a room and the response is recorded.

The difference between the two files is the RIR, which represents what the room/system did to the sound as it passed from source to receiver.

One way that the RIR can be obtained is by a complex mathematical process called deconvolution. A number of software applications exist that can deconvolve two files.

GratisVolver can yield the RIR by convolving a sweep recorded in the room with an inverse version of the original sweep. This is deconvolution.

Using deconvolution, you can get the RIR without using an impulsive stimulus.

A slowly swept sine wave offers some significant advantages:

• Significantly more energy is fed to the room. This offers dramatic signal-to-noise benefits.

• It’s less likely that you will overload your recorder, since the level of a sine wave doesn’t fluctuate like a noise or impulsive stimulus.

• The simple nature of the waveform allows the recording to be done with most file formats (i.e., WAV, MP3). Complex waveforms can be significantly altered by the compression schemes used by some formats.

• It is relatively easy to set recording levels.

Beware. Since sine waves generate significant power (that’s why you get a better signal-to-noise ratio), be careful not to burn up the loudspeakers used for the test.

The Equipment
There are lots of possibilities here. My setup consists of things that I already had. GratisVolver includes the forward and inverse sweeps needed for the test. You will need to supply a few hardware items.

For my stimulus file, I created a WAV file that includes the 14-second forward sweep and a short speech track. The total duration is about two minutes.

I use a media player (i.e., iPod) to play the file into the system. As a separate unit, it eliminates the need to string long cables in the venue to excite the system. The file is loaded on to a stereo portable handheld recorder. My mic is a Crown SASS, ideal for this application since it emulates a human listener. All of this fits neatly in the SASS case.

The data is gathered by playing the sweep file into the system and simply recording it. The 14-second log sine sweep is recorded at each desired measurement position. The speech track is used to provide a reference for what the measurement position sounded like. It can later be used as a reference to compare the convolved files to. The whole process takes about two minutes per listener location.

The Possibilities
Now the fun begins! The RIR can be opened in any acoustics measurement package that recognizes the WAV file format, which is about all of them and processed to yield the various acoustic measures.

It can be convolved with dry program material in GratisVolver. The user is free to exploit the strengths of all of these platforms (plus their ear-brain system) to analyze the data. It also allows for easy comparison of the measurement platform algorithms such as reverb time and clarity scores.

You will learn which measures are more meaningful to the way that you think about sound, and that help you draw meaningful conclusions from the data. Such files can be exchanged with colleagues for consultation, and can serve to document the before - and - after the performance of a sound system renovation.

They are also excellent for archival purposes. How would you like to have an RIR of the Fogg Lecture Hall where Sabine derived his famous equation? We could return to that space virtually and listen to it, as well as many others.

RIR testing is nothing new. We just have a few new tools that help us collect and analyze. All serve to make us better measurers and better listeners, ultimately resulting in better sounding systems.

Pat & Brenda Brown lead SynAudCon, conducting audio seminars and workshops around the world in addition to providing web-based training at www.synaudcon.com.

{extended}