Recording - July 2010 - 23

snare tracks, it was difficult to separate the two simply by level. When this happens, it can help to process the tracks with an eq or filter—usually the desired drum is brighter and more present, and the leakage is thinner and duller, and adjusting the tonal balance of the track (only during the detection phase) can further reduce the level of the unwanted hits. Drum replacement plug-ins often build this feature in—both apTrigga and Drumagog have filtering that can be applied to the track as a sidechain (without altering the tone of the audio) to aid in isolating the desired hits. In this case, however, the leakage was almost as bright in tone as the main drum, and so we turned to some of the fine-tuning adjustments to further help distinguish between them. Note detection There are a number of parameters that make adjustments to the note-detection algorithm(s). In the Logic Audio-to-Score parameter window (Figure 4), the main settings for this function, as far as drum tracks are concerned, are Granulation and Attack Range. The Granulation setting fine-tunes the algorithm’s ability to correctly identify the transient (impact) portion of each new note; the Attack Range setting further refines the determination of what should be considered an Attack Transient, to avoid false detection of unintended “notes”. All the drum replacement solutions have some degree of similar user adjustments. Drumagog offers Simple and Advanced detection modes, which trade off latency for more accurate detection, and it also includes a Resolution control, which prevents another note from being detected until a certain interval (milliseconds, usually) has passed since the last one. This prevents a flurry of false notes being seen because of a ragged transient (the Logic settings seem to address this as well). This last feature should be familiar to anyone who uses a MIDI drum kit like Roland’s V-Drums, where it is usually called Trigger Masking, and is designed to prevent false triggering of note clusters when the player’s stick bounces on the drum a little after a hit. As I began making adjustments to these settings, I discovered that no one set of adjustments for Velocity Threshold, Granulation, and Attack Range worked for an entire track, especially the Snare track (not surprisingly). Kick notes were detected fairly easily, once I got a handle on the right combination of settings, but in the Snare track, when the drummer switched from solid backbeats to fills or figures which included more subtle snare techniques like drags or ghost notes (rapid, quiet hits), different settings worked best for isolating the leakage while preserving all the nuances of performance.
RECORDING JULY 2010

Cut to pieces I began cutting the tracks up into separate Regions, based on the drummer’s playing techniques, and processing them separately, to use optimum settings for the different playing styles incorporated in the different sections of the various songs. Finding the right balance allowed the detection engine to (mostly) retain softer hits from the main drum, while (mostly) rejecting hits at similar levels from leakage drums or cymbals. Most likely, the difference in transient characteristics between the drum that was closer to the mic and those picked up from farther away helped this, as I found the right combination of settings for each section. Of course it never worked

100%—there were always some “good” hits missed and some “bad” hits detected. Since we needed to catch even the subtlest notes while rejecting all leakage, this meant that there was always some manual fine-tuning required. This was why I chose to use the older Audio-to-Score function over the newer dedicated Drum Replacement feature in Logic 9—the new feature allowed for manual fine-tuning after the audio track was converted to a MIDI track, but the Audio-to-Score allowed for it while I still had access to the display of the detected hits aligned against the audio wave (Figure 5), and I found this more flexible—easier to see/hear what was going on—before committing to the conversion.

23
http://www.neumannusa.com http://www.neumannusa.com
Recording - July 2010

Table of Contents for the Digital Edition of Recording - July 2010
