Introduction

One of the features of Csound which I find very useful in my own work is the tied notes capability in Csound score. By using the tied notes feature, the user can add expressive possibilities to their music, such as legato phrasing and continous pitch glissandi. By designing instruments that can work with both tied and non-tied notes, we will be creating instruments that can interpret musical ideas in more than simply one way.

Setting up your scores and instruments does require understanding how to write scores with tied notes as well as designing instruments to work with these notes. This article will cover how to write tied-note scores, how to design instruments to work with both tied and non-tied notes, and provides examples for the reader to study and experiment with.  The user is encouraged to take the ideas from this article and expand upon them for their own personal musical needs.

I. Tied Notes in Csound Score 

Usage

To use the tied notes feature of Csound is fairly simple: all that one needs to do is make the p3 field of a note(i-statement) negative. What this will do is tie a note to the next note following it that has the the same instrument number.  You can tie a group of notes together by making all of the p3 fields negative except the last note.

For example, the passage here consists of four notes that are not tied:

i1 0 1 8.00 80
i1 + . 8.02 80
i1 + . 8.04 80
i1 + . 8.04 80

To tie the first three notes together, simply change the above score to the following:

i1 0 -1 8.00 80
i1 + . 8.02 80
i1 + 1 8.04 80
i1 + 1 8.04 80


Technical Details

In Csound, everytime a note is processed, a chain of data structs are put together for each opcode in the instrument. These structs hold information that the opcodes uses when processing. For example, for the oscil opcode, the struct will hold what ftable to use, what frequency the opcode runs through the table, what amplitude to generate at, as well the phase (where in the table it currently is in its processing).

Normally for every note, when Csound sees the note, it checks to see what instrument it is for, find that instrument's definition from the orchestra, and then allocates a fresh chain of structs for each of the opcodes in that instrument.

When two notes are tied together, the second note does not get a fresh chain of structs. Rather, it is given the previous note's chain of structs. By having the previous note's information, the new note can continue to process and generate sound where the previous note left off, or reference the information from the previous note to help determine what to do in the current note. 

Also, a key point, a flag is set in the note's instance that marks it as being tied or not. We'll be using Csound opcodes in the next section to check that flag to determine if a note has been tied or not.  

II. Some Essential Information

From the previous section we find that tying notes together in the score is not so difficult to do, and it allows for notes of an instrument to inherit a previous note's data. With that information we'll be able to add some expressive possibilities to our instrument designs, but first we'll need to cover a few essential concepts before we can move on to learning how to design instruments to take advantage of Csound's tied-notes feature.

Tied Notes and Duration

One thing to be very careful about when working with tied notes is that your instrument code does not assume p3 is positive. If your instrument has references to p3 with that assumption, when you go to build tied-note instruments, you may find that your instruments don't produce sound correctly because they assume the note's duration is a positive value.  For instrument code that depends on the duration of the note, It is recommended to always first assign p3 to an i-time variable and to use the absolute value of p3 to guarantee that the instrument will read the duration correctly. For example, if your instrument contained code such as this:

kenv    linseg    0.0, .05, 1, p3 - .1, 1, .05, 0.0
Your instrument may not work as envisioned. Instead, use code like this: 
idur    = abs(p3)
kenv    linseg    0.0, .05, 1, idur - .1, 1, .05, 0.0

This code is guaranteed to work regardless if the note is tied or not.


Csound Tie Opcodes

For our instruments to take advantage of the tied notes feature of Csound, they'll first have to know if their note is tied or not. To do this, we use the tival opcode:


ir tival

which returns either a 1 or 0 if the note has been tied or not, repectively.  Csound also has one other opcode that works with tied notes, tigoto:


tigoto label

This opcode is a part of the goto family of opcodes and will go to the label given if the note has been tied.  This is generally used to skip a block of initialization code which may follow the tigoto line and the label that tigoto points to.  This might not make sense at the moment but in context of instruments designed for tied-note usage that we'll explore later in this article, knowing how and why to use this opcode will hopefully be made clear.


Fractional Instrument Numbers

To help with using the tied notes feature in Csound, it is possible to use fractional instrument numbers when writing your i-statements in your SCO so that if you would like two have multiple lines of tied notes using the same instrument but would like to have them separated, you can add a fraction to the instrument number. For example:

i1.1 	0 	-.25 	8.00 80
i1.1 	+ 	. 	8.02 80
i1.1 	+ 	. 	8.04 80
i1.1 	+ 	.25 	8.00 80
i1.2	0	1	7.00 80

In the above score fragment, there are two musical ideas, one consisting of four tied notes of duration .25 each, and one note of duration 1.  The notes for instrument 1.1 are all tied to each other, and since the second musical idea of the note of duration1 is labeled inststrument 1.2, the notes with 1.1 will not tie to it.

This feature is used prominently in the glissandi instrument example.

Understanding Instrument Initialization Time

To understand how to take advantage of the tied notes information, I'd like to discuss a little bit about how Csound initializes instrument instances when it reads in notes, as I think it's important to clarify at this point how this works in case there is any questions the reader may have.  

When Csound goes to to perform a note, the instrument for that note is looked up and the instrument definition is found.  From the instrument definition, a chain of data structs--one data struct for each opcode--is allocated for the instrument and then attached to the note instance.  For notes which are untied, these structs usually come in uninitialized with any preset values.  After the data structs are allocated, an initialization time pass is done and all opcodes with i-time functions will run.  Afterwards, the note is then added to the set of active notes, and when the engine goes to perform, the note will only be running at control (k-) and audio (a-) rates until the note is finished and deallocated.  (This is unless a reinit is called an another i-time pass is explicitly invoked.  This is beyond the scope of this article and if interested the reader is encouraged to look up the reinit opcode in the manual to further explore this idea.)

For opcodes in Csound, they are able to operate at either initialization time (i-rate), or performance time (k- and a-rate), or any combination of those three rates.  Most every opcode in Csound has an init-time phase, even those which generally run at k- or a-rate.  Those opcodes which are generally used for performance-time operations--such as oscilators and envelope generators--generally use their initialization time phase to do basic setup of their data struct which later gets used when they are running at performance-time. This may include setting what frequency to use, resetting the current phase to 0, allocating memory to use for delayed samples (such as used in a filter), etc.

For tied notes, two key aspects initialization time play a part in how to design instruments to take advantage of the tied note feature in Csound. When notes are tied together, the chain of data structs from the previous note is reused, but all initialization that an opcode normally goes through will occur as usual, unless the opcode supports skipping initialization and is set to do so. This means that for our instruments, if we are using opcodes that support a skip intialization flag (one can look at the user's manual to see if the opcodes being used in your instrument support this), we will most likely be utilizing that flag to skip initialization.  If we don't do so, then opcodes that run through their normal initialization won't pick up where they were in the previous note, so we may get phase discontinuities in oscilators or cleared out delay lines, both of which may produce audible pops as well as simply unintended sonic results.

The second aspect of initialization time to keep in mind for tied notes is that since initialization time runs for every opcode unless told otherwise, and performance time runs as it normally would if the note was tied or not, then it is at initialization time when we need to do all our work for the tied note.  An example of things we'll be doing at initialization time are evaluating what type of note is being performed and conditionally using different amplitude envelopes as well as setting up portomento pitch glides (as is commonly used in monosynth mode in MIDI synthesizers).

Now that we've had a brief overview of Csound's instrument initialization time, let's see what we can do with the opcodes mentioned above and the information we've covered to create some instruments that can take advantage of Csound's tied-notes feature.

III. Instrument Design

To take advantage of Csound's tied notes, our instruments will be using the tival and tigoto opcodes to detect what type of note we have as well as to conditionally intialize what our note will do, depending on the context of the note.  By doing this, we'll be able to do things such as emulate MIDI monosynth mode as well as to create continous glissandi sound masses.  

Design Goal

For this article, we'll be making two different types of instruments, described below.

Type 1 - Monosynth Instruments

This type of instrument will emulate MIDI synthesizers which support monosynth mode (which in themselves are a mode to emulate older analog synths which only supported one voice at a time). For those unfamiliar with monosynths and MIDI's monosynth mode, these are instruments which generally:

While this is a simplified explanation, I think it is enough to get an idea of what we'll be aiming for in designing our tied note instrument.

Type 2 - Glissandi Instruments

This type of instrument will be based on the type 1 instrument from above, but will differ mainly in its treatment of pitch.  Rather than a fixed time for shifting of pitch when a new note is played, a duration dependent glissandi time will be implemented.  This will allow us to express continous pitch glissandi and create sound masses common in avant-garde musical techniques (for example, those found in music by Iannis Xenakis or Krzysztof Penderecki). 

General Instrument Design

For tied note instruments, we'll follow a general template of design.  The sections of code will be:

There are three instruments in the example CSD file.  The first of which is a non-tied-note instrument which we'll be using as the basis for the other two instruments.  This instrument is designed pretty much in a way that is commonly found in many Csound examples.  The audio generation code is not too remarkable, a VCO2 running through a moogladder filter and also a feedbacked delay for a bit of body and color. It does not follow the plan above as it is not a tied note instrument.

The second instrument will take the design from the first instrument and modify it to handle tied notes and do so in a way to behave like traditional monosynths.  The third instrument also builds on the first instrument but is modified to behave in a way to allow continous glissandi.

Determining Type of Note

Besides knowing if a note has been tied to by the previous note or not, I've found it useful to know the "type" of notes and instrument is getting.  The type of note is something I have classified in the following table:

Note Type Type Code Explanation
Normal Note -1 This note is a free-standing note and is neither tied-from or tied-to any other notes.
Initial Note  0 This note is the first note in a group of tied notes. This note is tied to the next note but is not tied from any other note.
Middle Note 1 This note is contained within a group of tied notes and is neither the initial or ending note of the group. This note is both tied to the next note and tied from previous note.
End Note 2 This note is the final note in a group of tied notes.  This not is tied from a previous note but is not tied to any other notes.

To determine the type of note a note is, you will both have to get it's tie status by using tival as well as looking at the p3 value of the note. tival tells us if the note is currently being tied from a previous note, while looking to see if p3 is negative will tell us if the note is tied to the next note for the instrument.

How to use this information is summarized in the following table:

Note Type How to Detect Note Type
Normal Note tival is 0 and p3 > 0
Initial Note  tival is 0 and p3 < 0
Middle Note tival is 1 and p3 < 0
End Note tival is 1 and p3 > 0

I have written a small User-Defined Opcode to aid in developing tied-note instruments.  It uses the above scheme to determine the type of note currently being processed and returns the type code as defined in the first table.  The code is listed below:

    opcode tieStatus,i,0

itie tival

if (itie == 0 && p3 < 0) ithen
    ; this is an initial note within a group of tied notes
    itiestatus = 0
       
elseif (p3 < 0 && itie == 1) ithen
    ; this is a middle note within a group of tied notes 
    itiestatus = 1

elseif (p3 > 0 && itie == 1) ithen
    ; this is an end note out of a group of tied notes
    itiestatus = 2

elseif (p3 > 0 && itie == 0) ithen
    ; this note is a standalone note
    itiestatus = -1

endif  

    xout    itiestatus

    endop

We will be using the note's type information to design our instrument to work with scores that contain notes both tied and untied.

Implementing the Monosynth Instrument

For the monosynth instrument, we'll want it to perform like a normal instrument when there is no tied notes. When there is a series of tied notes, we have two main concerns for the instrument, one of which is our treatement of pitch, the other other of amplitude.  

For pitch, we will use a simple scheme of using the linseg opcode to move linearly from the previous notes value to the new notes value (I used a .05 second time for transitioning to the new pitch value), then to hold on the new pitch value for the duration of the note.  The way this instrument is coded, the pitch line will always use the same scheme regardless of note type.

For amplitude, our concerns are a little tricker.  Depending on the note type, we will generate amplitude envelopes differently:

Amplitude Envelope Properties
Note Type Type Code Handling of Amplitude
Normal Note -1 Use ADSR
Initial Note  0 Use same attack and decay time as ADSR, holding sustain value until end of note
Middle Note 1 Linearly move from old amplitude value to new amplitude value in same amount of time as portamento time
End Note 2 Like middle, move to new amplitude in same amount of time as portamento time, but also do a release to zero in the same amount of time as ADSR's release time

In the instrument, the tieStatus User-Defined Opcode is used get the type of note. From there, the type of note is used in conditional code (a chain of if-then's) to determine what manner of amplitude envelope to generate.

After this, the instrument pretty much operates like the non-tied-note version, except that opcodes which support skipping initialization do so when in a series of tied notes.


Implementing the Glissandi Instrument

For the glissandi instrument, our amplitude envelope scheme is similar to the one used in the monosynth example, though we've also made linear changes for the amplitude value given in, such that the overall amplitude of the note is determined as a product of both the ampltiude envelope (which is a value between 0 to 1) and the line from the old amplitude value to the new amplitude value (which is a value in decibels).  This give a smooth amplitude change over time.

For pitch, the implementation of this instrument for tied notes is to glide from the old pitch to the new pitch in the duration of the note.  Now, with this example, the instrument has an issue in that the first note of a tied-note group will stay on its initial note and there is no method to have a glissandi on the first note.  The design can be modified to take in to use two pitch fields, only using the second field if the first of a group of tied notes.  This is done already in the author's personal instruments, but for this article it is left as-is for the sake of simplicity.


IV. Examples

Download the examples here:

Notes on the Examples

The examples include three instruments: a normal non-tied-note instrument, a monosynth instrument, and a glissandi instrument.  The score sections are as follows:

  1. Melodic Line for Instrument 1 using regular notes
    A standard score as would be used in normal Csound usage

  2. Melodic Line for Instrument 1 using tied notes
    A tied note score but using a normal non-tied-note instrument.  Notice there are click in the notes, demonstrating that simply tying the notes in the score is not enough and may cause inadvertent problems. The melodic line is given a fracitional instrument number so as note to interfere with the held chords.

  3. Melodic Line for Instrument 2 (monosynth) using regular notes
    A standard score used with our Type 1 instrument. The audio result is the same as score example 1, demonstrating that this instrument is capable of working with non-tied notes.

  4. Melodic Line for Instrument 2 (monosynth) using tied notes
    A tied note score used with our Type 1 monosynth instrument.  The melodic line is phrased together using groups of tied notes, and the audible results resemble those of monosynths, using portomento between note pitches.

  5. Sound Mass for Instrument 3 (glissandi) using tied notes - This example uses the Type 2 glissandi instrument to create sound masses of microtonally inflected glissandi clouds.  There are many lines and different fractional instrument numbers are used to allow for distinct lines to be created without the lines getting tied to each other inadvertently. The score was generated using the author's Orchestral Composition library, written in Python and included with the author's blue music composition environment.

Some other notes:

Acknowledgements

I would like to thank Richard Dobson for his chapter in The Csound Book, "Designing Legato Instruments in Csound", as it helped a great deal in understanding Csound's tied-notes functionality.