A good recording?

Every now and then, I get approached by students interested in phonetics, as well as by colleagues who are not phoneticians, about how to obtain data that would be useful for phoneticians. Here are some brief answers based on my personal experience.

What sampling rate do I choose?
The sampling rate of 44.1 is very good for phonetic studies that involve a wide range of phonetic creatures. I’d say that anything higher isn’t necessary. 44.1 is already fantastic and the higher the rate, the more space the files end up taking up on disks.

The machine gives me channel choices?
The channel choice depends on the purposes of the study and some practicalities. Extremely simply put, will you have one microphone (mono) or two (stereo)?
Practical tip number one: stereo always takes up more space because it records the “same” thing twice. If the recording is done with a single speaker, or if the quality of the data of anyone other than 1 of the participants isn’t crucial, I’d definitely pick mono and make sure the recorder is closer to the relevant speaker rather than e.g. the interviewer (purely acoustic thinking here, not socio thinking).
Stereo makes sense if we really want highly reliable and good quality phonetic data from two individuals not sitting/standing/doing other things right next to each other. Different recorders have different built-in mics – for some, stereo probably won’t even make sense if there’s just one little mic whose positioning wouldn’t enhance speech coming from two different directions. That’s what stereo is all about – making sure that signals coming from two different directions get captured, each captured better by one of the two mics.
To me personally stereo makes a LOT of sense if two external microphones are used, one directed at speaker A and the other one at speaker B, because even if the built-in mic has two channels rather than one, in my experience they tend to be so close to each other that in the end it doesn’t make that much of a difference (and thus unnecessarily takes up valuable space).

Think about getting an external microphone
But what matters a lot to me is whether one uses a built-in mic or an external mic. I prefer the latter because external mics (most, not all) can prevent some acoustic mishaps. For instance, built-in mics usually produce decent quality for vowel analyses and for things such as glottalling, but not for things involving noise, where acoustic noise includes aspects of fricatives and some voice quality profiles. This is because built-in mics have to deal with all sorts of echoing in the room (well, more so than the external head-mounted ones at least). It can make quite a lot of difference whether the recorder is placed on a wooden table or a piece of cloth in between the table and the recorder, of if the room has a lot of textiles (which absorb echoing). What gets affected most is noise-related info because then the echo of the (high-intensity) vowel overlaps with the consonant/voice quality-relevant noise. I myself am very particular though, but that’s because I focus on voice quality and noise-related consonantal details. I’m sure that phoneticians focusing on vowels and glottalling would give you way more relaxed suggestions.
If you use the built-in microphone, I’d place the recorder on a piece of cloth – or some textile surface, and ideally position the speaker in a more textile corner of the recording space. If you’d like, you can also record yourself in the environment in which you’ll be recording and ask a phonetician you know what they think of the quality for phonetic purposes.
If you happened to have an external mic such as the head-mounted AKG C520, this will produce AMAZING quality no matter where. But head-mounted mics get us to the issue of attention paid to speech.

I’m not a phonetician but am involved in collecting a lot of data which I’d love for phoneticians to be able to use. How do I know if the quality is ok?
That really depends on the phenomenon of interest, and also on the specific aspect of this phenomenon that the phonetician may be interested in. As a rule of the thumb, if you want the data to be useable also for analyses of voice quality and obstruent variation, you will need fairly good quality. If you’re genuinely serious about the usability of your future/current dataset by phoneticians, don’t be afraid to approach some of the phoneticians you know with a sample recording to see if the quality is suitable (and what for) and to get tips on how to improve the quality – we tend to be a friendly bunch!
%d bloggers like this: