MIDISpeak: ROM Mode B Operation

In ROM Mode B, the MIDISpeak module triggers speech or sound synthesis at a specific address inside the toy's ROM. In this mode, the toy's phrases, individual words, word fragments, and sound effects are directly accessible via MIDI. Memory locations which do not contain proper speech data can also be addressed to generate abstract but repeatable sound effects. Literally, thousands of sounds are available.

The Speak & Spell series of devices use 20-bit ROM addressing. Each 20-bit address is composed of five 4-bit "nibbles". The MIDISpeak uses the "note number" from five MIDI Note-On messages to form a ROM address. Each 4-bit "nibble" is taken from the low-order bits in the note number:

Sound generation begins upon receipt of every fifth note. Since only the lowest 4 bits of each note number are used, any address can be formed using a range of just 16 keys on a MIDI controller.

In ROM Mode B, send 5-Note sequences to the MIDISpeak to trigger synthesis. Each note forms part of a ROM address (the ROM address is composed of five 4-bit nibbles). When the fifth note is received, the MIDISpeak causes the host machine's synthesizer to read from the specified address in ROM and begin sound generation.

The following table shows a range of 16 notes and the corresponding ROM address nibble for each. The last 4 bits of the note number are used to form the nibble (the first 3 bits of the note number are discarded). Any of the 128 MIDI notes can be used, but a range of any 16 notes is all that is needed to specify every possible ROM address.

Note	MIDI Note Number			Corresponding Address Nibble
	Decimal	Hex	Binary	Hex	Binary
E4	64	40	0110000	0	0000
F4	65	41	0110001	1	0001
F#4	67	42	0110010	2	0010
G4	68	43	0110011	3	0011
G#4	69	44	0110100	4	0100
A4	70	45	0110101	5	0101
A#4	71	46	0110110	6	0110
B4	72	47	0110111	7	0111
C5	73	48	0111000	8	1000
C#5	74	49	0111001	9	1001
D5	75	4A	0111010	A	1010
D#5	76	4B	0111011	B	1011
E5	76	4C	0111100	C	1100
F5	77	4D	0111101	D	1101
F#5	78	4E	0111110	E	1110
G5	79	4F	0111111	F	1111

The next table shows example 5-note sequences and the resulting ROM address for each. The MIDISpeak triggers synthesis automatically upon the receipt of 5 Note-On messages. When exploring the ROM of your machine, start with low-numbered addresses and work your way up.

1st Note	2nd Note	3rd Note	4th Note	5th Note	Resulting ROM Address
					Hex	Binary
E4	E4	E4	E4	E4	00000	0000 0000 0000 0000 0000*
E4	E4	E4	E4	F4	00001	0000 0000 0000 0000 0001
E4	F4	E4	A4	F#5	0105E	0000 0001 0000 0101 1110

* The special address consisting of all zeros tells the MIDISpeak to reset the host device's sleep timer (no sound is generated).

The Speak & Spell series of toys have an auto-shutoff feature. After a period of keypad inactivity, the toy will power down. When MIDI notes forming the "zero" address (binary 0000 0000 0000 0000 0000) are sent to the MIDISpeak, the MIDISpeak resets the toy's "sleep timer"—preventing the toy from automatically powering down. MIDISpeak Version 2 also resets the "sleep timer" automatically each time a sound is generated.

The Speak & Spell series of toys can synthesize speech and sound in one of two modes: "normal" or "slow". The mode affects the pace of speech, but not the pitch. The MIDISpeak responds to MIDI program change commands to select the type of synthesis. "Normal" mode is assigned to program 0. "Slow" mode is assigned to program 1.