What the hell does Quantisation think it is?
Quantisation: the sequencer's revenge on the trained musician. Vic Lennard looks at this much-used but little-understood feature and finds that it has feelings, too.
Almost all of us use quantisation, but few of us get the most from it. This article takes you through the finer points of one of sequencing's most powerful features.
All sequencers are pretty much the same: they "record" the notes you play in from a keyboard and get it to play them back. Of course, we know that it isn't sounds being recorded, only the digital information created by pressing a key, moving a pitchbend wheel or stepping on a sustain pedal. But by recording in this way, we can keep our options open - we can change the sounds when the sequencer plays back, for instance, by using patch change messages. Sometimes notes are played in from a keyboard for other reasons than their actual pitch - most multitimbral synths dedicate a MIDI channel to the percussion sounds and then map them out along the keyboard. So Bottom C may be a bass drum, the D a tone above may be the snare drum and so on. The key to all this flexibility is our ability to edit the MIDI information in one way or another.
One of the most musically powerful forms of editing is quantisation. Let's take an example: you want to record 16th-note hi-hats, so you repeatedly press the key assigned to the hi-hat in time to the metronome. On playback, the chances are that the notes aren't quite where you intended them to be. You go to the quantise function on the sequencer and voila - it sounds even worse than it did before! But what does quantising actually do?
Quantising is what happens when a sequencer moves notes to the nearest division of a bar; you decide what the value of that division is. If you quantise to 8ths, all notes will be moved to the nearest quaver, or eighth note. It's like having a grid where all events have to lie on the vertical lines. For instance, Figure 1 shows a C major chord played on the downbeat of the bar (or at least that was the intention). Each crotchet, or quarter note, is divided into four so the vertical lines represent 16ths.
All the various types of quantising rely on this "movement" of events - if you choose the wrong quantise value, the result is not going to be what you wanted. No matter how intelligent the algorithms within a computer program, no computer (to date) can read your thoughts - so it doesn't know where you intended to play the notes. Figure 2 shows what happens if you select 16th-note quantise for the C major chord. Because the selected quantise value is wrong, one note has been moved to the desired position while the other two have been moved to the next 16th. However, quantising to 8ths gives the correct result (see Figure 3).
Quantising to the correct value may still leave some notes sounding as though they were played in the wrong place. This is because they have been played too far away from the desired point in the bar and, consequently, have been moved to another quantise position. Unless you can individually edit these notes - as you can on a "micro-edit" screen on a hardware sequencer, or via a note display page on a computer sequencer - you'll probably need to play them in again.
The secret is to figure out what is the smallest note value you intend to play. To do this, check the note value that your sequencer's metronome is ticking along to and listen to it. It's likely to be crotchets (quarter notes), for which it will sound four to each bar. Quite often, the metronome on the downbeat of the bar will be either louder than the rest, or a note of a different pitch. Tap the notes you have played against the metronome and decide how many times more frequently you're tapping; if it's twice, then you're using eighth notes; if it's four times, then you're using 16ths (semiquavers).
A helpful trick is to slow down the tempo when playing in fast passages such as certain drum parts. It's like trying to jump into a moving car - you have to judge the position of the open door, which is easier to do if the car is moving slowly. This is an effective practice, but it tends to remove the more subtle emotive qualities from the music. Many musicians would argue that playing should never be quantised, but more about this later.
The most common type of quantising is one where the Note On is moved to the nearest quantise position. However, there are three possible effects on the Note Offs:
The Note Off doesn't move: This method of quantisation changes the note length. While there may be circumstances where you may want to keep the Note Off in its played position, these are rare. This is shown in Figures 2 and 3.
The Note Off moves with the Note On: This keeps the note length the same. In the playing of a drum part, note lengths are likely to be irrelevant, as most drum sounds work on a "one-shot" principle, where the sound plays for its full length irrespective of when the Note Off is received. See Figure 4 for the C major chord example.
The Note Off is quantised as well: This not only changes the note length but as each Note Off will be moved differently, it does so in an inconsistent way (Figure 5). For chords, this is rarely used, but for a solo it might be useful. If the intention was to play the solo in a legato manner (where each Note Off is butted up to the next Note On), then this type of quantisation will produce the required result. Unfortunately, the same quantise value is usually used for Note On and Off so unless the part has been played in accurately, the Note Off quantise is unlikely to produce correct legato - some Notes Off will be quantised to the wrong position.
Most of the more expensive computer sequencers offer different types of quantisation. These do more than just move Notes On and Off, and so are flexible and often more useful.
The Over Quantise in Steinberg's Cubase uses the quantise value you select but takes into account whether you play in front of, or behind the beat. It then attempts to retain the "feel" of your music and will even line up the Notes On of chords. The Musical Quantise II feature found on C-Lab's Creator/Notator is a similar function, and uses the timing of notes around the quantise positions to decide how best to move note events. Notes played behind the beat will be pulled forward and vice versa - this produces the possibility of actually changing the playing style of someone who wants to sound "lazy". Consequently, intelligent quantisation functions often have problems handling your style if your playing is too erratic.
Another quantisation option is to leave notes which are close to the quantise position unchanged, and to move notes which are further away towards the quantise position. Cubase's Iterative Quantise works in this way: you set a number of ticks (192 ticks = a quarter note) distance from the quantise position within which notes are not moved, and then set a percentage movement towards the quantise position for the remaining notes. A setting of 50% will move a note halfway between its current position and the quantise position. Creator/Notator's Capture Quantise works the other way round: if a note is within the capture range it is moved, otherwise it's left untouched. This function uses percentage movement in the same way as Iterative Quantise.
Sometimes, you know what feel you want but can't play it in. Both Creator/Notator and Cubase offer a Groove Quantise option in which a preset "template" is used. This template is a pattern where the quantise positions are in odd and exotic places. The success of the result still depends largely on how accurately you have played in the original part, however, because various types of tuplet (triplets, 5ths, 7ths) feel may have to be dealt with, and misinterpretation of the original musical intention is all too easy.
Now let's take a situation where a pianist has played in a part which has excellent feel and where you want to take that feel and impose it on a string part playing alongside. Creator/Notator's User-defined Groove is intended specifically for this. It analyses the timing of the piano part, uses it as a template and attempts to quantise the strings using the same parameters as for Capture Quantise (see above). It also has the ability to change the velocities of the notes in a similar fashion. Cubase's Match Quantise is a simpler function which is intended primarily for use with drum parts. The notes in the reference track need to be pretty regular (like a hi-hat for instance), and these "pulses" are then used as the quantise positions. The quantise value you set then determines which notes of another part are moved to the reference positions.
Creator/Notator allows you to take this idea further with its Adaptive Groove function. This uses seven parameters to seriously reshape your playing, although the muso fraternity would certainly argue that the time taken to get to grips with a function like this would be better spent brushing up your playing technique.
Deciding whether or not to hit the quantise button depends largely on whether your sequencer also gives you the option to "undo" quantising. Most computer sequencers do whereas most hardware sequencers don't. If you can't undo an operation which can have drastic effects on your sequence, you're well advised to make a copy of the part you're about to alter before proceeding. If you can select and change quantise values while your sequencer is playing back, this will give you an immediate idea of whether or not you're doing the right thing. If not, then it's play, stop, quantise, play, stop, change, play, stop...
Another approach to quantising is to use an auto-quantise function. These are most useful when recording drum parts in "loop in overdub" mode, because there are few things worse than trying to build up a drum part when each part is out of sync with every other. Each time the part loops round, what you hear has already been quantised. You can usually change the quantise value from one cycle to the next, but make certain that this too can be undone, otherwise the wrong quantise value will ruin your "take". You won't get a second chance.
Finally, remember that the timing inaccuracies caused by the queueing up of MIDI data (due to MIDI's serial nature) and the reaction time of your multitimbral synths is likely to adversely affect the nuances which can be achieved with some of the complex quantisation methods. Sometimes the simplest solutions are the best; Note On quantise with no change of note length will do the job in most cases. But then, technology is there to be used...
Feature by Vic Lennard
Previous article in this issue:
Next article in this issue:
mu:zines is the result of thousands of hours of effort, and will require many thousands more going forward to reach our goals of getting all this content online.
If you value this resource, you can support this project - it really helps!