466 TT6581 :: Quicker, easier and cheaper to make your own chip!

How it works

Inspired by the legendary MOS6581 Sound Interface Device (SID) chip used in retro computers such as the Commodore 64, the Tiny Tapeout 6581 (TT6581) is a original digital interpretation supporting nearly the entire original MOS6581 feature set, implemented in 2x2 tiles for Tiny Tapeout.

All configuration is done via a 4-wire SPI interface, and the audio output is a 1-bit PDM signal at 10 MHz.

Features

Full control through a Serial-Peripheral Interface (SPI).
Three independently synthesized voices.
Four supported waveform types (triangle, sawtooth, square and noise).
Attack, decay, sustain, release (ADSR) envelope shaping.
Chamberlin State-Variable Filter (SVF) for low-pass, high-pass, band-pass and band-reject.
Second-order Delta-Sigma DAC.

Architecture

TT6581 Architecture

The diagram above shows the datapath in the TT6581. A tick generator triggers the generation of a single audio sample at 50 kHz.

Voice Generation: One 10-bit voice at a time is generated. Internal phase registers keep track of each voice's state while inactive. The frequency and waveform type are set by the programmed values in the register file. Supported waveforms are triangle, sawtooth, pulse or noise.
Envelope: An ADSR envelope generator produces an 8-bit amplitude value per voice. The envelope is applied to each voice by multiplication.
Wave Accumulation: The three voices are accumulated (mixed) by addition. Depending on the filter enable bit of each voice, they are accumulated in one of two registers: one that will be passed through the SVF, or one that will bypass it.
Filter: A Chamberlin State-Variable Filter (SVF) processes the filter accumulator. It supports low-pass, high-pass, band-pass and band-reject modes with tuneable frequency cutoff and resonance (Q).
Global Volume: The SVF output is summed with the bypass accumulator. A global 8-bit volume is applied by multiplication resulting in the final mix.
Delta-Sigma PDM: An error-feedback Delta-Sigma modulator converts the final mix to 1-bit PDM output at 10 MHz (OSR = 200).

To fit the strict 2x2 tiles area requirement, the entire synthesis is time-multiplexed meaning most modules are finite state machines. A single 24x16 multiplier is shared for all modules. The 50 kHz sample tick wakes up a master controller (FSM) that then initiates the synthesis of a single sample.

Pin Mapping

The TT6581 uses the bidirectional IO pins for SPI and a single dedicated output for the PDM audio signal. All dedicated inputs are unused.

Pin	Direction	Function
`uio[0]`	Input	SPI Chip Select (active low)
`uio[1]`	Input	SPI MOSI
`uio[2]`	Output	SPI MISO
`uio[3]`	Input	SPI SCLK
`uio[4:7]`	-	Unused
`uo[0]`	Output	PDM audio output
`uo[1:7]`	-	Unused
`ui[0:7]`	-	Unused

The PDM output should be passed through a 4th order Bessel filter for the best reconstruction of the analog waveform.

Programming

The TT6581 is programmed in much the same way as the original MOS6581. The register layout mirrors the original SID, three voice channels followed by filter and volume registers and the same ADSR, waveform selection and filter concepts apply. The main differences are:

Registers are accessed through an SPI interface.
The filter coefficients are pre-calculated and written directly as fixed-point values, rather than the raw 11-bit FC value used by the MOS6581.

Keeping the changes in mind, the original MOS6581 datasheet is a good reference for how to use the TT6581.

SPI Protocol

The SPI interface uses CPOL=0, CPHA=0 (data sampled on the rising edge of SCLK). Each transaction is a 16-bit frame while CS is held low:

Bit	15	14:8	7:0
Field	R/W	Address [6:0]	Data [7:0]

Bit 15 = 1 for write, 0 for read.
Bits 14:8 = 7-bit register address.
Bits 7:0 = write data.

Data is transmitted MSB first.

Playing a tone

Set volume: Write 0xFF to register VOLUME for max volume.
Set frequency: Compute the 16-bit frequency control word and write it to FREQ_LO / FREQ_HI.
Set ADSR: Write attack/decay to AD and sustain/release to SR.
Select waveform and gate on: Write the CONTROL register with the desired waveform bit and GATE=1.

For example, to play a 440 Hz sawtooth on Voice 0 with instant attack and full sustain:

SPI Write: addr=0x1A, data=0xFF    # Volume = max
SPI Write: addr=0x00, data=0x05    # FREQ_LO = 0x05  (FCW for 440 Hz = 0x1205)
SPI Write: addr=0x01, data=0x12    # FREQ_HI = 0x12
SPI Write: addr=0x05, data=0x00    # AD = 0x00 (attack=0, decay=0)
SPI Write: addr=0x06, data=0xF0    # SR = 0xF0 (sustain=15, release=0)
SPI Write: addr=0x04, data=0x21    # CONTROL = sawtooth + gate on

To release the note, write CONTROL again with GATE=0:

SPI Write: addr=0x04, data=0x20    # CONTROL = sawtooth + gate off

Formulas

Frequency Control Word (FCW):

$ FCW = \frac{f_{\text{desired}} \times 2^{19}}{F_s} $

where $F_s = 50$ kHz (sample rate). The 16-bit FCW is split across FREQ_LO (bits 7:0) and FREQ_HI (bits 15:8).

The maximum representable voice frequency is limited by the 16-bit FCW:

$ f_{\max} = \frac{65535 \times F_s}{2^{19}} \approx 6250 \text{ Hz} $

Filter Cutoff Coefficient (Q1.15 signed):

$ \text{FCC} = \left[ 2 \cdot \sin \left( \frac{\pi \cdot f_c}{F_s} \right) \cdot 32768 \right] $

where $f_c$ is the desired cutoff frequency in Hz. The 16-bit result is split across F_LO and F_HI.

Filter Damping Coefficient (Q4.12 signed):

$ \text{FDC} = \left[ \frac{1}{Q} \cdot 4096 \right] $

where $Q$ is the desired resonance. The 16-bit result is split across Q_LO and Q_HI.

Register map summary

The TT6581 has a 7-bit address space and 26 8-bit registers. Three identical voice register groups are followed by a filter/volume group.

The full register map is described in regs.yaml.

Voice Registers

Each voice occupies 7 consecutive 8-bit registers. Voice 0 starts at 0x00, Voice 1 at 0x07, and Voice 2 at 0x0E.

Offset	Name	Bits	Description
0x00	FREQ_LO	7:0	Frequency control word - low byte
0x01	FREQ_HI	7:0	Frequency control word - high byte
0x02	PW_LO	7:0	Pulse width - low byte
0x03	PW_HI	3:0	Pulse width - high byte
0x04	CONTROL	7:0	Waveform select and voice control
0x05	AD	7:0	Attack (7:4) / Decay (3:0)
0x06	SR	7:0	Sustain (7:4) / Release (3:0)

CONTROL register bit fields:

Bit	Name	Description
7	NOISE	Select noise waveform
6	PULSE	Select pulse (square) waveform
5	SAW	Select sawtooth waveform
4	TRI	Select triangle waveform
3	-	Reserved
2	RING_MOD	Enable ring modulation
1	SYNC	Enable oscillator sync
0	GATE	Gate (1 = attack, 0 = release)

Filter and Volume Registers

Address	Name	Bits	Description
0x15	F_LO	7:0	Filter cutoff coefficient - low byte
0x16	F_HI	7:0	Filter cutoff coefficient - high byte
0x17	Q_LO	7:0	Filter damping coefficient - low byte
0x18	Q_HI	7:0	Filter damping coefficient - high byte
0x19	EN_MODE	5:0	Filter enable and mode select
0x1A	VOLUME	7:0	Global volume (0x00–0xFF)

EN_MODE bit fields:

Bit	Name	Description
5	FILT_V2	Route Voice 2 through filter
4	FILT_V1	Route Voice 1 through filter
3	FILT_V0	Route Voice 0 through filter
2:0	MODE	Filter mode: `001`=LP, `010`=BP, `100`=HP, `101`=BR

How to test

Connect an SPI master to the bidirectional IO pins:
- uio[0] = CS (active low)
- uio[1] = MOSI
- uio[2] = MISO
- uio[3] = SCLK
Connect uo[0] (PDM output) through a low-pass reconstruction filter (e.g. 4th-order Bessel, cutoff ≈ 20 kHz) to an amplifier or speaker.
Program a voice. Minimal example for a 440 Hz sawtooth:
- Write 0xFF to 0x1A (volume = max).
- Write 0x05 to 0x00 and 0x12 to 0x01 (frequency control word for 440 Hz).
- Write 0x00 to 0x05 (attack=0, decay=0) and 0xF0 to 0x06 (sustain=15, release=0).
- Write 0x21 to 0x04 (sawtooth waveform + gate on).
A 440 Hz sawtooth tone should be playing. Write 0x20 to 0x04 to release the note.

The project also includes a CocoTB test suite that runs automatically via GitHub Actions on both RTL and the synthesized gate-level netlist.

External hardware

SPI master.
Low-pass 4th order reconstruction filter for the PDM output (Ideally Bessel).
Audio amplifier and speaker/headphones.

#	Output	Bidirectional
0	pdm	cs_n
1		mosi
2		miso
3		sclk
4
5
6
7

466 TT6581

466 : TT6581