Experimental "Staff" Notation



 

    This graphical display is a result of trying to compress a lot of information into a small space, with the goal of showing tuning information from the separate voices of a small ensemble all on one screen. I don't have any idea how readable other people will find it. Currently, nothing more than two channels is operating, awaiting your feedback.

    First, setting all the parameters is more tedious than it can be made, but I don't have any incentive to develop a final form until I get user feedback. So be patient. Once the settings are made for any session, of course, they are usually not changed - your main interaction with the program is observing the tuning feedback. I would anticipate an Ensemble menu list that you could create, a multiple-channel analog to the Instrument menu.

    Select View
| Staff Notation, which will reveal all the menu items for this mode, at the moment all under View. The range and pitch etc.that are already selected will become those for Channel 1; use the familiar dialog to change that when you need to. Now you can select the Clef for Channel 1. The Key Signature (with either major or minor mode as appropriate) operates for all channels; however there is no way to extend note spelling beyond twelve preset names according to the key. (That is to say, if the music modulates to far distant keys, the note names may show as their enharmonic equivalents). If you need to transpose an entire piece, set its key under Transpose (right now, this operates separately from the Transpose information under the Pitch & Range dialog). This might be a common need for an a capella vocal ensemble.

    Size of Staff allows you to control the vertical magnification of the graph. The program will fit however many systems it can onto the screen, given the size you select. Make it large enough to be comfortably readable.

    So now, what exactly is being displayed on the graph? Essentially two separate images are being superimposed for each channel. First, the staff lines are read like any normal music notation; and the notes being heard are drawn out over time as light gray bars. Think perhaps of the information encoded on a player-piano roll, which also moves linearly in time. As long as a note lasts, a bar is extended on the representative place on the staff. Since time is spatially represented, there is no place to put accidentals in front of a note - they are put above at the beginning of a note-bar, just as editorial accidentals are added in printed music. (One experiment, I have added a small color-coded beginning to the note, depending on its being sharp, flat, or natural).

    Here is the beginning of the c minor Sarabande for solo cello by J.S. Bach, in this spatial "piano-roll" notation,

notebars1

notebars2

    which you can compare with the score for these four measures.

sarabande score

[We can note one major discrepancy in the last beat of bar 3, where the lower G seems to sound much too long while the upper g begins very late. This happens in the recordings I have used because the lower G string continues to vibrate after the bowing has stopped, so even though the player actually began the g on time the continuing sound of the lower note influences the tuning algorithm.]


    But we want the tuning information as well! Superimposed over this chart of the notes being played, we reproduce the fine-tuning information that is found in the lower graph in the form you are already familiar with. As before, the same twelve colors help in distinguishing the different semitones. But now you need to read the five lines of the staff as the center five lines of that lower graph - that is, in bass clef, the line for d becomes the zero, the center of the pitch of each note, the f line represents 10 cents sharp, a line represents 20 cents sharp, and so on. [The scaling is labeled on the right-hand edge of the page. And remember, the scaling can be magnified under View | Lower Graph Scaling, where you can select 2, 5, or 10 cents per graph division. In the case of 2 or 5, however, notes far out of tune will be off-scale and will be drawn at the edge of the space allotted to the staff.] Admittedly, this is not quite as easy to read, as the off-center lines are not dashed, but shouldn't be any more difficult than distinguishing a b' from a g'.
    Late addition : the off-center lines can be set to a gray, to contrast with a black center line. This somewhat reduces the illusion of a normal musical staff, but the improvement to the fine-tuning trace is most helpful.  If you wish to experiment with varying degrees of contrast,
play around with setting a new palette [Parameters | Palette Colors] for yourself, where you can change the staff gray lines field. This contrast is not illustrated in these images.

    Here is the tuning trace for the same cello piece, with the note-bars removed for clarity. [Sorry, somehow I changed the reference pitch a little since the screen shot of this on the first page, so the notes all appear sharper than in that image.] Compare the first line with the image on the 
introductory page that you have already seen.
   

trace1
trace2

    And combining the two:
   
combined1
combined2

   
    [The trace becomes very wide at the end of some notes, when the string has stopped being bowed but the tuner is picking up the room echo in the recording.]



    With a little practise, does this become readable? Part of the success in viewing this, I find, comes with adjusting the gray value of the note bar light enough not to be too prominent in the background of the colored trace, yet dark enough to be read easily. After all, ultimately the colored traces of the pitch information is what you will react to.  Play around with this by setting a new palette [Parameters | Palette Colors] for yourself. [the bars are too dark in these final images, I've decided.] Can your brain switch between Color mode, reading the tuning information, and Gray mode, reading the bars as a "score"?


    Some of the difficulties I am struggling with:
    The analysis often has points where it is unable to reach a decision. These "dropouts" are not too noticeable in the overall trace graph - the eye easily connects the lines, and little blips (where the analysis has made a mistake) are easily ignored. But in this "staff" notation, dropouts cause the note bars to appear discontinuous, which I find disturbing [they have been edited out of the images above]; naturally, I'm working to try to improve the pitch-tracking. Even worse, sometimes these discontinuities come from thinking the note is somewhere else, and very short note bars get drawn at different octaves than the correct one. The graph would be much more readable if most of these dropouts can be eliminated. That is one of the main thrusts of my efforts.

    Large vibratos (vocal) go beyond the range of a semitone, making this notation (and the lower graph notation itself) useless for them.

    Not a struggle, but just time-consuming work to be done, includes improving the accidental images for larger sizes (now, Windows does the magnification, and as it does it badly, I will have to create images for each size). (also for clefs, but the ugliness of the magnified ones is only cosmetic, not interfering with the readability of the graph).
  



    After experimenting reading this notation alone, if you wish to try a duet you can select
Options | DualChannel Input before you begin Tuning. Set the Range and Clef separately for the instrument on Channel 2. But you must also be aware - most, probably all, sound cards have only a monaural microphone input. To obtain a true stereo input, 2 separate channels, you will need two microphones connected to a pre-amplifier, whose output then goes to the soundcard's Line-In connector. (Some sound devices, particularly on laptops, have only one connection functioning for both microphone and line-in; you switch between them through the recording controls. Consult your computer's manufacturer for details).


    Two approaches for ensemble tuning occur to me - currently both channels are shown relative to an absolute pitch, but showing the upper voice (in the future, voices) in relationship to the bass line should have at least as much utility. In that approach, if the bass were sharp but the treble equally sharp, the treble trace would indicate exactly in tune. Your thoughts?

    A major caveat - this idea of ensemble tuning will work only if the input to the separate microphones is as isolated on each instrument or singer as feasible. Suggestions include using contact microphones for instruments, and a microphone on a headset for singers. If the signal from the desired performer does not overwhelm the signal from the rest of the ensemble, the tuning will not be successful. Remember, the algorithm works only for a single note at a time; so each channel m




Back to the Index