Sound in Basic I - Programmable Sound Generator PSG

At the time of writing this post, it is planned for Commander X16 has three main ways to make sound:

Programmable Sounds Generator (PSG) in VERA
Pulse-Code Modulation (PCM) playback system in VERA
Yamaha YM2151 sound chip

PCM playback is not very useful for BASIC programmers but both PSG and YM2151 can be used effectively in BASIC programs. This post is about PSG.

! Important Update !

Before we go too deep into this post it is important to emphasize that ROM version R42 introduced dedicated BASIC commands to deal with sound. Specifically for using PSG sound system we have following "new" commands:

PSGCHORD - Starts or stops simultaneous notes on VERA PSG

PSGFREQ - Plays a frequency in Hz on VERA PSG

PSGINIT - Stops sound and reinitializes VERA PSG

PSGNOTE - Plays a musical note on VERA PSG

PSGPAN - Sets stereo panning on VERA PSG

PSGPLAY - Plays a series of notes on VERA PSG

PSGVOL - Sets voice volume on VERA PSG

PSGWAV - Sets waveform on VERA PSG

The document below was written before those commands existed. It might still be useful to read it for better understanding on how PSG works but it is probably easier to use new commands instead of VPOKE-ing directly into VERA PSG Registers. Of course teh below examples still work just fine.

PSG Capabilities

Let’s summarize capabilities of Programmable Sound Generator

16 independent channels or voices
Each voice can have any of four available waveforms
Each voice can be played in stereo with on/off control over left and right side
Each voice has independent volume control from 0 - silent to 63 - Max volume
16 bit frequency value

We can also list some of the features that PSG can’t do, at least not directly in hardware.

It does not support ADSR envelopes
It doesn’t have timers to turn sound off, that has to be done programmatically

Registers in VRAM

VERA is responsible for PSG playback so we have to send data to VRAM where VERA can read it from. We already know that command to write to VRAM directly is VPOKE so where do we write?

Space reserved for PSG is from $1F9C0 - $1F9FF so total of 64 bytes. With 16 voices that means each voice has four bytes or registers to configure it. Or in other words following are memory locations per voice:

$1F9C0 - $1F9C3 Voice 0
$1F9C4 - $1F9C7 Voice 1
$1F9C8 - $1F9CB Voice 2
...

$1F9FC - $1F9FF Voice 15

Now let’s look at what those four registers look like:

Register	Bit 7	Bit 6	Bit 5
0	Frequency Bits 0-7
1	Frequency Bits 8-15
2	Right	Left	Volume
3	Waveform		Pulse width

I think it is becoming quite clear that programming sound using PSG is not very difficult. Let’s describe each of the registers in more detail. I think it makes most sense to go from back to front.

Register 3 - Waveform

We have two bits (Bit 6 and 7) to choose between four available Waveforms. The values are:

00 - Pulse
01 - Sawtooth
10 - Triangle
11 - Noise

Or to represent them graphically:

There are plenty of resources online that describe the characteristics of these waveforms and that is far beyond the scope of this post. However it is important to highlight a few things.

PSG does not generate Sine waves, however Triangle is very similar and produces very warm sound and for most cases can be used instead of sine wave.
Noise is not completely random and also changes with frequency so we can have different pitch noise which is very handy.
Square wave is the sound we most commonly associate with the 8bit home computer era, however Pulse is not the same as square and we will discuss differences in the next paragraph.

Six bits of this register (Bits 0 to 5) are used to determine the width of the Pulse wave. Again let’s use graphics to illustrate how it works.

As we see all these three examples have the same frequency and therefore the same pitch but the width of hills and valleys is not which causes it to sound differently.

With 6 bits values can be between 0 and 63 and value 0 has a very narrow hill and therefore sounds very “thin”. Maximum valley of 63 will cause that hills and valleys are of the same width and the wave of that shape is what we call “square wave” and is much fuller and stronger.

Register 2 - Volume

Very simple and straightforward register. Bit 6 and 7 determine which side of stereo output the particular voice will sound. As expected 0 means off and 1 means on but left and right are switched if we assume that bit 7 is leftmost:

00 - no sound
10 - sound on the right side only
01 - sound on the left side only
11 - sound on both sides

The remaining 6 bits are used to set volume. 6 bits have range 0-63 where 0 means sound does not generated and 63 is the loudest. Note that we can’t define volume per side so both left and right side are played at the same volume or not at all.

Registers 0 and 1 - Frequency

Two bytes are available to determine the frequency and there is a formula that can be used to calculate the values in order to hit the exact frequency. Here is copy from the official manual

sample_rate = 25MHz / 512 = 48828.125 Hz
output_frequency = sample_rate / (2^17) * frequency_word

To make your life a bit easier I calculated most of the music notes in hearable range:

Note ID	Note	Frequency	Hi	Lo	Note ID	Note	Frequency	Hi	Lo	Note ID	Note	Frequency	Hi	Lo
1	C0	16.35	0	44	37	C3	130.81	1	95	73	C6	1046.50	10	249
2	C#0/Db0	17.32	0	46	38	C#3/Db3	138.59	1	116	74	C#6/Db6	1108.73	11	160
3	D0	18.35	0	49	39	D3	146.83	1	138	75	D6	1174.66	12	81
4	D#0/Eb0	19.45	0	52	40	D#3/Eb3	155.56	1	162	76	D#6/Eb6	1244.51	13	13
5	E0	20.60	0	55	41	E3	164.81	1	186	77	E6	1318.51	13	211
6	F0	21.83	0	59	42	F3	174.61	1	213	78	F6	1396.91	14	166
7	F#0/Gb0	23.12	0	62	43	F#3/Gb3	185.00	1	241	79	F#6/Gb6	1479.98	15	133
8	G0	24.50	0	66	44	G3	196.00	2	14	80	G6	1567.98	16	113
9	G#0/Ab0	25.96	0	70	45	G#3/Ab3	207.65	2	45	81	G#6/Ab6	1661.22	17	107
10	A0	27.50	0	74	46	A3	220.00	2	79	82	A6	1760.00	18	116
11	A#0/Bb0	29.14	0	78	47	A#3/Bb3	233.08	2	114	83	A#6/Bb6	1864.66	19	141
12	B0	30.87	0	83	48	B3	246.94	2	151	84	B6	1975.53	20	183
13	C1	32.70	0	88	49	C4	261.63	2	190	85	C7	2093.00	21	242
14	C#1/Db1	34.65	0	93	50	C#4/Db4	277.18	2	232	86	C#7/Db7	2217.46	23	64
15	D1	36.71	0	99	51	D4	293.66	3	20	87	D7	2349.32	24	162
16	D#1/Eb1	38.89	0	104	52	D#4/Eb4	311.13	3	67	88	D#7/Eb7	2489.02	26	25
17	E1	41.20	0	111	53	E4	329.63	3	117	89	E7	2637.02	27	167
18	F1	43.65	0	117	54	F4	349.23	3	169	90	F7	2793.83	29	76
19	F#1/Gb1	46.25	0	124	55	F#4/Gb4	369.99	3	225	91	F#7/Gb7	2959.96	31	10
20	G1	49.00	0	132	56	G4	392.00	4	28	92	G7	3135.96	32	226
21	G#1/Ab1	51.91	0	139	57	G#4/Ab4	415.30	4	91	93	G#7/Ab7	3322.44	34	215
22	A1	55.00	0	148	58	A4	440.00	4	157	94	A7	3520.00	36	233
23	A#1/Bb1	58.27	0	156	59	A#4/Bb4	466.16	4	227	95	A#7/Bb7	3729.31	39	27
24	B1	61.74	0	166	60	B4	493.88	5	46	96	B7	3951.07	41	110
25	C2	65.41	0	176	61	C5	523.25	5	125	97	C8	4186.01	43	229
26	C#2/Db2	69.30	0	186	62	C#5/Db5	554.37	5	208	98	C#8/Db8	4434.92	46	129
27	D2	73.42	0	197	63	D5	587.33	6	41	99	D8	4698.63	49	69
28	D#2/Eb2	77.78	0	209	64	D#5/Eb5	622.25	6	134	100	D#8/Eb8	4978.03	52	51
29	E2	82.41	0	221	65	E5	659.25	6	234	101	E8	5274.04	55	77
30	F2	87.31	0	234	66	F5	698.46	7	83	102	F8	5587.65	58	151
31	F#2/Gb2	92.50	0	248	67	F#5/Gb5	739.99	7	194	103	F#8/Gb8	5919.91	62	19
32	G2	98.00	1	7	68	G5	783.99	8	57	104	G8	6271.93	65	196
33	G#2/Ab2	103.83	1	23	69	G#5/Ab5	830.61	8	182	105	G#8/Ab8	6644.88	69	173
34	A2	110.00	1	39	70	A5	880.00	9	58	106	A8	7040.00	73	210
35	A#2/Bb2	116.54	1	57	71	A#5/Bb5	932.33	9	199	107	A#8/Bb8	7458.62	78	54
36	B2	123.47	1	75	72	B5	987.77	10	92	108	B8	7902.13	82	220

Armed with those values we can simply VPOKE them into registers and get the correct tone.

Example code

Ok enough theory let’s make some noise. Below are a few examples from simplest to more complicated.

Beep

Single beep in square wave by setting four registers of Voice 0, waiting in empty FOR loop and then setting volume to 0 to turn it off:

Falling

In the following demo we will be changing the frequency or pitch as we play the sound and make it sound like something is falling. We are using triangle waveform:

I used very common code to split value to high (Most significant) and low (Least significant) bytes:

MSB = INT(VALUE/256)
LSB = VALUE AND 255

Engine throttling

This demo will be a little more interactive and we will use Sawtooth waveform. The rest is very similar to previous one except that we are reading Joystick (or Keyboard mapped to Joystick keys) to increase the frequency and with that simulating higher RPMs and when released the engine RPM is slowly dropping.

Explode

The last demo will use the final waveform, the noise. To achieve a good explosion sound this time we will not be changing frequency but instead the volume to sound explosion slowly fading.

What next

One of the problems of programming sound in BASIC is that in BASIC we don’t have very precise control over timing so music is not easy to program or at least to make it sound great is not easy.

For that reason I wrote couple of libraries in Assembly to make including sound effects and background music to BASIC programs easier, links below:

Search This Blog

8bit Coding