Sound Controller

15.1 Overview

The Zeal Video Board includes a powerful sound subsystem with five voices:

• 4 standard voices that generate basic waveforms: square, triangle, sawtooth, and noise.
• 1 sample-based voice capable of playing arbitrary wave data from a 256-byte FIFO buffer.

The 4 standard voices share a single 16-bit Linear Feedback Shift Register (LFSR) for noise generation. The sample-based voice is independent but mixed together with the other voices before output.

Mixing is done by averaging the four standard voices and then adding the current sample from the sample table voice. This approach keeps the audio output balanced but requires care when playing both types of voices simultaneously to avoid saturation or clipping.

Like other controllers, the DMA Controller can be mapped into the device bank, starting at address 0xA0 on the I/O port. Its device index is 3.

15.2 Voice Control and Synchronization

• The hold register allows muting voices by freezing their output at the current sample value, effectively silencing them without stopping waveform generation.
• This hold feature enables atomic start/stop of multiple voices simultaneously to avoid timing glitches or artifacts. The register MASTER_HOLD can be set to choose the voices to hold and unhold.
• The sound select register determines which voices are affected by writes to frequency and waveform configuration registers, allowing flexible per-voice setup. The voice to select/enable can be choosen in the MASTER_ENABLE register.

For example, if voices 0 and 1 must be set to frequency 0x1000, the following pseudo code can be used:

// Enable both voices 0 and 1 (1 bit per voice)
MASTER_ENABLE  = (1 << 1) | (1 << 0);

// All the selected voices will listen to the the following registers changes.
FREQUENCY_LOW  = 0x00;
FREQUENCY_HIGH = 0x10;

// Set voice 0 to SQUARE wave, 50% duty cycle
MASTER_ENABLE  = (1 << 0);
WAVEFORM = 0x80;

// Set voice 1 to SAWTOOTH wave
MASTER_ENABLE  = (1 << 1);
WAVEFORM = 0x02;

15.3 Voices and Waveforms

Each of the 4 standard voices can produce one of the following waveforms, selectable per voice:

Setting the waveform of one or multiple voices can be done by setting their WAVEFORM. The voices to configure must first be selected in the MASTER_ENABLE register.

The duty cycle of the square wave is controlled by a 3-bit field, allowing for these settings:

Setting the duty cycle, only applicable when setting a voice to square wave, can be done by setting the Duty cycle fields in the WAVEFORM register.

⚠️ Changes to frequency, waveform, or duty cycle are latched and only take effect at the end of the current waveform period, preventing clicks or pops.

15.4 Frequency and Pitch

The frequency of each voice is set by a 16-bit register. The resulting frequency in Hz is calculated as:

Where the fixed sample rate is 44,091Hz for the standard voices.

The hardware does not limit frequency values, so very high or low values may produce inaudible sounds.

The frequency for the enabled voice(s) can be written by setting the FREQUENCY_LOW and FREQUENCY_HIGH for the LSB and MSB respectively.

15.5 Volume and Mixing

• Each voice has a disable bit and a 2-bit linear volume control allowing five levels: 0%, 25%, 50%, 75%, and 100%. Enabled voice(s) volume can be set via the VOLUME register.
• The master volume controls output volume per channel (left and right) separately, also with 2-bit linear scaling. The left/right channels volumes can be set via the register MASTER_VOLUME
• Voices can be assigned individually to the left and/or right audio channels, including the sample table voice, to do so, the registers MASTER_LEFT_CHANNEL and MASTER_RIGHT_CHANNEL are provided.
• When mixing, the outputs of the four standard voices are averaged before adding the sample table voice's current sample.

This scheme keeps the output balanced and avoids simple sum clipping, but be aware that playing the sample voice alongside all four standard voices can cause saturation since its sample is added after the averaging.

15.6 Sample Table Voice

The fifth voice uses a 256-byte FIFO buffer to play arbitrary digital samples. It can be selected by setting SAMPLE_TAB bit in MASTER_ENABLE register. This special does not use the same WAVEFORM, FREQUENCY and VOLUME registers as the other voices. The registers offsets are repurposed as the ST_FIFO, ST_SRATE_DIV and ST_STATUS registers.

• The following sample formats are supported, which can be set in register ST_SRATE_DIV:
  • 8-bit unsigned
  • 16-bit signed or unsigned (little-endian)
• The Sample playback rate is programmable via a divider, set in ST_SRATE_DIV, and is interpreted as:

• The FIFO has full and empty status flags:
  • Playback halts if the FIFO is empty.
  • Writing stops when FIFO is full.

The current FIFO status is readable from the ST_STATUS register.

This voice allows playback of complex sounds or sampled audio alongside the standard synthesized voices.

15.7 Programming the Sound Controller

The SDK provides a straightforward API to control voices and samples:

• Initialize and reset the sound subsystem.
• Configure voices' frequency, waveform, volume, and channel assignment.
• Manage the sample table voice: configure sample format, sample rate divider, and write sample data to the FIFO.
• Use the hold register to mute or start voices atomically.

For detailed examples and complete API usage, check out the SDK sound example.

15.8 Registers Summary

15.9 Registers