Teletype dev help - Scale quantizer [SOLVED]

Monome
#1

I wrote a couple scale quantizer OPs for Teletype based on @discohead’s N.S lookup table. These are my first OPs and my C is rusty so I’m sure I’m missing something, but I can only get one of my two OPs (QT.S) to work correctly. nb: these OPs are just working with semitones; I might convert to volts later, I dunno.

QT.S simply checks the input semitones against all semitone values on the specified row of table_n_s until it finds the highest value less than the input. I feel like my code’s a bit messy (I wanted to wrap all the input values in useful ways) but it’s working great.

QT.SC is similar but the allowed values are selected by giving scale (0-8), degree (0-6) and chord type / voices (1-7). 1 voice gives a single note, 2 is a third, 3 is a triad, 4 is a seventh, etc. My problem is that my wrapping math is not working when a given degree + chord exceeds the length of the scale. This line: dix = (2 * i + degree) % 7; is supposed to pick the table index and wrap gracefully while checking the input semitones against the table_n_s value but when I test on TT it’s not selecting any values which occur after the wrapping starts. I could use a fresh set of eyes!

QT.S input root scale

static void op_QT_S_get(const void *NOTUSED(data), scene_state_t *NOTUSED(ss),
                     exec_state_t *NOTUSED(es), command_state_t *cs) {
	
	int16_t note_in = cs_pop(cs);
    int16_t root = cs_pop(cs);
    int16_t scale = cs_pop(cs) % 9;
    if (scale < 0) scale = 9 + scale;
	
	note_in = note_in - root;	
	int16_t octave = (note_in >= 0) ? (note_in / 12) : ((note_in + 1) / 12 - 1);

	if (note_in >= 0) {
		note_in = note_in % 12;
	}
	else {
		if (note_in % 12 == 0) {
			note_in = 0;
		}
		else {
			note_in = 12 + (note_in % 12);
		}
	}

	int16_t note_out;
	
	for (uint8_t i = 0; i < 7; i++) {
		if (note_in >= table_n_s[scale][i]) note_out = table_n_s[scale][i];
	}
	
	cs_push(cs, 12 * octave + note_out + root);
}

QT.SC input root scale degree voices

static void op_QT_SC_get(const void *NOTUSED(data), scene_state_t *NOTUSED(ss),
                     exec_state_t *NOTUSED(es), command_state_t *cs) {
	
	int16_t note_in = cs_pop(cs);
    int16_t root = cs_pop(cs);
    int16_t scale = cs_pop(cs) % 9;
    if (scale < 0) scale = 9 + scale;
	
    int16_t degree = cs_pop(cs);
	int16_t degree_octave = (degree >= 0) ? (degree / 7) : ((degree + 1) / 7 - 1);
	if (degree >= 0) {
		degree = degree % 7;
	}
	else {
		if (degree % 7 == 0) {
			degree = 0;
		}
		else {
			degree = 7 + (degree % 7);
		}
	}
	
	int16_t voices = cs_pop(cs);
    if (voices < 1) voices = 1;
	if (voices > 7) voices = 7;
	
	note_in = note_in - root;	
	int16_t octave = (note_in >= 0) ? (note_in / 12) : ((note_in + 1) / 12 - 1);

	if (note_in >= 0) {
		note_in = note_in % 12;
	}
	else {
		if (note_in % 12 == 0) {
			note_in = 0;
		}
		else {
			note_in = 12 + (note_in % 12);
		}
	}

	bool quant = false;
	int16_t max_n_s_val = 0;
	int16_t n_s_val;
	int16_t note_out;
	int16_t dix;
	
	for (int16_t i = 0; i < voices; i++) {
		dix = (2 * i + degree) % 7;
		n_s_val = table_n_s[scale][dix];
		if (n_s_val > max_n_s_val) max_n_s_val = n_s_val;
		if (note_in >= n_s_val) {
			note_out = n_s_val;
			quant = true;
		}
	}
	
	if (!quant) note_out = max_n_s_val - 12;
	
	cs_push(cs, 12 * (octave + degree_octave) + note_out + root);
}
1 Like
#2

I think I mostly get the idea and nothing jumps out to me, can you give some input / output examples of what you expect vs what you’re getting?

1 Like
#3

Example: QT.SC 9 0 0 5 3 should return 9 but returns 4. Allowed values are 9, 0, 4.

QT.SC 11 0 0 4 4 returns 5 but it should return 11. Input 12 returns 11. Allowed values 7, 11, 2, 5.

#4

So for QT.SC do you also want the highest note less than or equal to the input? I apologize that my understanding of music theory is all but nonexistent so I don’t entirely follow the calculation here but it seems to me that this condition:

		if (note_in >= n_s_val) {
			note_out = n_s_val;
			quant = true;
		}

will select whichever value less than or equal to note_in gets hit last in the iteration, not necessarily the largest one. The final value of max_n_s_val is equal to the desired result in both the examples you gave but maybe this is not the correct answer in all cases? I also don’t quite understand this adjustment

if (!quant) note_out = max_n_s_val - 12;

since it seems like this would always give a negative value. What’s the desired behavior when note_in % 12 is 0? Find the least note in the chord?

1 Like
#5

Ohhh you definitely hit the nail on the head. It’s supposed to pick the highest note among the eligible notes, not the last eligible note.

If note_in isn’t bigger than any of the eligible notes in this octave, (!quant) is true and it picks the highest note from the lower octave. So that will always be a negative number of semitones when it needs to dip below the current octave.

1 Like
#6

I used QT.BX for the first time today. Am I right the docs have x and i in the wrong order:

Doc says:
QT.BX x i
quantize 1V/OCT signal x to scale
defined by N.BX in scale index i

But it should be
QT.BX i x

#7

yeah the docs are backwards. good catch.

#8

QT.B and QT.BX are broken/bonkers with negative input voltages

n.b 10 1189 // scale = Bb, C, Eb, F, Ab or N=10,12,15,17,20
vn n.b 1 // return 10 OK
vn n.b -1 // return 5 OK
vn qt.bx 0 n 0 // return 0 OK
vn qt.b n 1 // return 0 OK
vn qt.b n 2 // return 3 OK
vn qt.b n 4 // return 3 OK
vn qt.b n 5 // return 5 OK
vn qt.b n 6 // return 5 OK
vn qt.b n 7 // return 8 OK
vn qt.b n 8 //return 8 OK
vn qt.b n 9 //return 10 OK
vn qt.bx 0 n -1 // return 0 OK
vn qt.bx 0 n -2 // return -3 should be -2
vn qt.b n -2 // return -3 should be -2
vn qt.b n -5 // return -5 should be -4

I hope whoever wrote this will look into it, not sure if it’s @desolationjones or someone else…

1 Like
#9

There is a bug with negative voltages because I dumbly mirrored the voltage around 0 :upside_down_face: Fix incoming.

#10

Cool, looking forward to this! Thanks!