DIGITAL AUDIO
Going digital gets rid of the noise inherent in using tape. However, if you
are going digital, there are several issues you should consider that will
impact audio quality: sampling frequency, resolution, compression.
If you want to capture a frequency accurately then you need to have a
sampling frequency that is at least double. The limit of human hearing is
around 20KHz hence the sampling rate on CD quality audio is 44.1kHz. (If you
want to know more about this do a Google search for "Nyquist frequency").
But, assuming you are not interviewing opera stars who will give impromptu
performances during the interview, you don't care about recording the entire
range of frequencies that the human ear can hear. Normal speech frequencies
are in the mid-range i.e 250 Hz to 8kHz. If you are just interested in
speech you probably don't want the low frequencies or high frequencies
outside this range. If you are recording speech for transcription the low or
high frequency stuff is just noise and it will impact the intelligibility of
the speech you are recording. In previous posts when people mentioned noise
reduction all they mean is that the microphone isn't sensitive to
frequencies at the upper and lower ends or the recorder doesn't encode those
frequencies. You can also get this kind of effect after the fact by using an
audio editing program like Audacity that can be used to filter out entire
frequency bands. Given the Nyquist issue, this means one probably should use
a sampling frequency of at least 16kHz for the best quality. I use 22.05kHz.
Now you can get away with less. Telephone systems generally limit
frequencies to the most critical frequencies for intelligibility, usually
400Hz and 3.4kHz. Similarly with a lot of digital voice recorders--you'll
find many have fairly low sampling rates and will only be good at recording
frequencies up to 3kHz or 5kHz. Personally, if you are going to listen to
this material over and over or transcribe it, I'd want more than the minimum
quality for intelligibility.
Practically all digital recorders use at least 16 bits to encode audio (i.e.
the resolution, the value that can be given to each sample). I would avoid 8
bit. The devices that plug into the iPod are 8 bit. It may get the job done
but don't expect wonders.
Compression is another issue. Uncompressed audio (PCM) takes up a lot of
space. So people use codecs like MP3, ATRAC (what Minidisc uses) etc. to
compress the audio. If you are a linguist you might desire to use
uncompressed audio but MP3, ATRAC, WMA, OGG and other popular codecs will
work very well at encoding speech even at fairly high compression ratios.
The sample rate and resolution are generally more important considerations.
MICROPHONES
You probably want an omnidirectional microphone as those are the easiest to
use for this type of work. Some mics are designed for recording speech, i.e.
have a frequency sensitivity geared to the mid-range. You'll also find
boundary or pressure zone microphones that are good for recording meetings
and other events where the mic can be positioned on a large flat surface.
These boost direct sound waves by 6db over reflected sound waves and thus
aid intelligibility in situations where there might be a lot of reflected
sound from walls and other surfaces. In some situations using multiple mics
and stereo recording might prove very helpful.
TECHNIQUE
The number one key to intelligibility regardless of what you are using is to
make sure that the speech you are recording is at least 30db louder than any
background sounds or noise (40-50db is better) while avoiding 'clipping'.
Microphones by themselves won't perform miracles in this regard. Lots of
microphones will work reasonably well. What's important is that you get the
microphone reasonably close to the speaker or speakers. Every time you
double the distance of the microphone from the source you drop the audio
relative to background noise by 6db. Move the microphone far enough away
(and it doesn't have to be very far) and the speech you are recording will
sink into the sludge of background noise. And you'd be surprised how much
noise there is in a 'quiet' room. A previous poster suggested lapel mics.
Good idea if they aren't perceived as intrusive and you don't run into a
clothes noise issue. The other strategy is to limit ambient noise as much as
possible.
RECORDERS
Voice Recorders ($200-$300). Some new models have much better sample rates
and record in formats that can be readily used by many types of software.
Check out the newest Olympus models (DM-10, DM-20, DS-2200). They have a
best quality mode that samples at 44.1kHz, has a frequency sensitivity of
300 to 8,000 Hz, and encode in WMA.
Minidisc ($200-$400). New Hi-MD models will record in ATRAC at various
compression ratios as well as PCM. You'll get 44.1 kHZ, 16 bit, stereo
quality audio. Very good quality audio. Downside is that Sony still hasn't
allowed rapid transfer to PC in a format that can be unlocked. Apparently
they will release some type of converter that will allow this in the fall.
As a previous poster pointed out, if you are going to a remote location MD
may be the way to go as you can take lots of discs (which are fairly cheap)
and don't have to bother immediately offloading your recordings as you would
with some other types of digital recorders.
Marantz PMD-670 ($600-$700). Uses CF cards. Sophisticated professional
recorder. Records in PCM, MP3 and MP2 at sample rates up to 48kHz. Supports
professional mics, selectable frequency band filtering, etc. Well designed,
easy to use. Very high quality audio.
|