Recording Quality and Editing
Recording quality is indicated in five categories, from poor to excellent, in the “grade of sonagram portion of cut” field. It is surprisingly rare under most circumstances to obtain truly excellent recordings of most species. Below average or even poor recordings can be useful, particularly for rare species, unusual vocalization types, or species which occur in difficult situations, such as by waterfalls or windy areas.
The type of recording equipment and the settings used are also pivotal to success in recording. Note that recording quality only pertains to the portions of original files presented on the AVoCet database, only to the target species, only to the best portion of the target species, and only to the recording after editing (if any). The “comments on grade” field typically spells out the reason why a recording isn’t listed in a higher category, for example, because of heavy insect noise.
Although the quality categories are somewhat subjective, the following serves as a guideline: “Excellent” recordings typically have minimal background or competing noise; they are very clear and recorded at an appropriate level so they stand out very well from background noise; they are sustained, e.g. usually including multiple iterations of the sound type; and they are made with good equipment and saved at high sample rate and bit rate, as wav files.
“Good” recordings typically lack one or two of the above attributes, and yet are fine for most purposes. Some “good” recordings could be further edited in other software to the point of excellence. Most recordings on the database are in this rather broad category.
The “Average” rating refers not to average quality among those presented in the database, but average among most recordists’ field recordings.
“Below average” refers to recordings with serious flaws such as great distance, other noise overlapping the sound, etc. These recordings still may be useful for some purposes.
“Poor” recordings are rarely presented in the database except in the case of a species or sound type that is rare or otherwise not present in the database. These recordings typically are too faint relative to background noise to be amplified to the usual levels, and/or are heavily overlain by other noise.
See “hints on recording” page for further information.
In many cases, a recording can be much more useful or can rise to a higher category with a limited degree of editing of background noise. This may involve replacement of segments of noisy background with a quieter segment with a similar profile (e.g., the same type of insect noise), while taking care not to affect target species vocalizations or their spacing.
If a target species vocalization is overlain with undesirable noise such that the noisy segment requires removal, this is clearly indicated by a blank segment. While care is taken to avoid affecting the target sound, should a scientific user require unedited sequences e.g. for analysis of insect sounds, the original unedited sounds can be requested from the recordists.
It should also be noted that many sounds are amplified for presentation on the website, and this is not generally indicated unless target and background sounds are differentially amplified, in which case this is mentioned in the “Addl. Info” field.
Another form of editing commonly done on AVoCet recordings involves filtering of non-target frequencies. Ideally just the lowest frequencies are filtered, but in a noisy recording it may be necessary to filter much more, particularly if the target sound is very high-pitched.
In Raven, which is a sound analysis program, not a typical sound editing program, the entire specified frequency band is filtered, rather than simply reduced. Hence, reduction of e.g. pervasive insect sound to dramatically improve a given recording could be done by end users in sound editing software.
More effective editing can be done on specialized sound editing programs, and this most commonly includes high- and low-pass filtering (high-pass filtering removes much of the noise below frequencies specified while low-pass that above specified frequencies), and less often fade in, fade out, and noise reduction. Programs such as Audacity (downloadable for free) easily allow for this type of editing, but one has to take care to avoid unintentionally altering sound quality.
For the most part, we have taken the position that it is better to present the recording as originally made and leave editing to the user, if required. For example, for purposes of playback loud background roar may need to be toned down in order to bring out the target vocalization.