Vs 1103
Vs 1103
VS1103b
mic
audio
VS1103 Mono Stereo Stereo Ear−
audio
L
MIC AMP MUX R
ADC DAC phone Driver
line
audio output
GPIO 4
GPIO
X ROM
DREQ
SO
SI Serial X RAM
Data/
SCLK Control 4
XCS
Interface VSDSP
XDCS Y ROM
RX
TX UART
Y RAM
Clock
multiplier
Instruction Instruction
RAM ROM
Contents
1 Disclaimer 8
2 Definitions 8
4.1 Packages . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.1.1 LQFP-48 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
4.1.2 BGA-49 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
5 SPI Buses 15
5.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
5.4.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
5.5.1 General . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
6 Functional Description 23
7 Operation 37
7.1 Clocking . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
7.6 Play/Decode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
8 VS1103b Registers 45
8.10.1 Registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
8.11.1 Registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
8.12.1 Registers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
8.12.6 Interrupts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
10 Contact Information 61
List of Figures
1 Disclaimer
2 Definitions
B Byte, 8 bits.
b Bit.
W Word. In VS DSP, instruction words are 32-bit and data words are 16-bit wide.
1 Must be connected together as close the device as possible for latch-up immunity.
2 The maximum sample rate that can be played with correct speed is XTALI/256.
Thus, XTALI must be at least 12.288 MHz to be able to play 48 kHz at correct speed.
For other implications rising from not using a 12.288 MHz clock, see Chapter 7.4.5.
3 Reset value is 1.0×. Recommended SC MULT=4.0×.
1 3.0 volts can be achieved with +-to-+ wiring for mono difference sound.
2 AOLR may be much lower, but below Typical distortion performance may be compromised.
3 Harmonic Distortion increases above typical amplitude.
TBD
Both LPQFP-48 and BGA-49 are lead (Pb) free and also RoHS compliant packages. RoHS is a short
name of Directive 2002/95/EC on the restriction of the use of certain hazardous substances in electrical
and electronic equipment.
4.1.1 LQFP-48
48
1
4.1.2 BGA-49
C
4.80
7.00
0.80 TYP
1.10 REF
0.80 TYP
1.10 REF
4.80
7.00
TOP VIEW
Pin types:
Type Description
Type Description
AO Analog output
DI Digital input, CMOS Input Pad
AIO Analog input/output
DO Digital output, CMOS Input Pad
APWR Analog power supply pin
DIO Digital input/output
DGND Core or I/O ground pin
DO3 Digital output, CMOS Tri-stated Output Pad
CPWR Core power supply pin
AI Analog input
IOPWR I/O power supply pin
In BGA-49, no-connect balls are A7, B7, D4, E5, F1, G1.
In LQFP-48, no-connect pins are 11, 12, 25, 35, 36.
5 SPI Buses
5.1 General
The SPI Bus - that was originally used in some Motorola devices - has been used for both VS1103b’s
Serial Data Interface SDI (Chapters 5.4 and 6.4) and Serial Control Interface SCI (Chapters 5.5 and 6.5).
The DREQ pin/signal is used to signal if VS1103b’s SDI FIFO is capable of receiving data. If DREQ is
high, VS1103b can take at least 32 bytes of SDI data or one SCI command.
Because of a 32-byte safety area, the sender may send upto 32 bytes of SDI data at a time without
checking the status of DREQ, making controlling VS1103b easier for low-speed microcontrollers. If
SARC DREQ512 is set, the safety area is 512 bytes (see Chapter 6.6.13).
Note: DREQ may turn low or high at any time, even during a byte transmission. Thus, DREQ should
only be used to decide whether to send more bytes. It should not abort a transmission that has already
started.
Note: In VS10XX products upto VS1002, DREQ was only used for SDI. In VS1103b DREQ is also
used to tell the status of SCI.
5.4.1 General
The serial data interface operates in slave mode so the DCLK signal must be generated by an external
circuit.
Data (SDATA signal) can be clocked in at either the rising or falling edge of DCLK (Chapter 6.6).
VS1103b assumes its data input to be byte-sychronized. SDI bytes may be transmitted either MSb or
LSb first, depending of SCI MODE (Chapter 6.6.1).
In VS1002 native modes (SM NEWMODE is 1), byte synchronization is achieved by XDCS. The state of
XDCS may not change while a data byte transfer is in progress. To always maintain data synchronization
even if there may be glitches in the boards using VS1103b, it is recommended to turn XDCS every now
and then, for instance once after every flash data block or a few kilobytes, just to keep sure the host and
VS1103b are in sync.
If SM SDISHARE is 1, the XDCS signal is internally generated by inverting the XCS input.
BSYNC
SDATA D7 D6 D5 D4 D3 D2 D1 D0
DCLK
When VS1103b is running in VS1001 compatibility mode, a BSYNC signal must be generated to ensure
correct bit-alignment of the input bitstream. The first DCLK sampling edge (rising or falling, depending
on selected polarity), during which the BSYNC is high, marks the first bit of a byte (LSB, if LSB-first
order is used, MSB, if MSB-first order is used). If BSYNC is ’1’ when the last bit is received, the receiver
stays active and next 8 bits are also received.
BSYNC
SDATA D7 D6 D5 D4 D3 D2 D1 D0 D7 D6 D5 D4 D3 D2 D1 D0
DCLK
If SM NEWMODE is 0 and SM SDISHARE is 1, the operation is otherwise like the VS1001 compat-
ibility mode, but bits are only received while the BSYNC signal is ’1’. Rising edge of BSYNC is still
used for synchronization.
5.5.1 General
The serial bus protocol for the Serial Command Interface SCI (Chapter 6.5) consists of an instruction
byte, address byte and one 16-bit data word. Each read or write operation can read or write a single
register. Data bits are read at the rising edge, so the user should update data at the falling edge. Bytes are
always sent MSb first. XCS should be low for the full duration of the operation, but you can have pauses
between bits if needed.
The operation is specified by an 8-bit instruction opcode. The supported instructions are read and write.
See table below.
Instruction
Name Opcode Operation
READ 0b0000 0011 Read data
WRITE 0b0000 0010 Write data
Note: VS1103b sets DREQ low after each SCI operation. The duration depends on the operation. It is
not allowed to finish a new SCI/SDI operation before DREQ is high again.
XCS
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 30 31
SCK
3 2 1 0
SI 0 0 0 0 0 0 1 1 0 0 0 0 don’t care don’t care
15 14 1 0
SO 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 X
execution
DREQ
VS1103b registers are read from using the following sequence, as shown in Figure 5. First, XCS line is
pulled low to select the device. Then the READ opcode (0x3) is transmitted via the SI line followed by
an 8-bit word address. After the address has been read in, any further data on SI is ignored by the chip.
The 16-bit data corresponding to the received address will be shifted out onto the SO line.
XCS should be driven high after data has been shifted out.
DREQ is driven low for a short while when in a read operation by the chip. This is a very short time and
doesn’t require special user attention.
XCS
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 30 31
SCK
3 2 1 0 15 14 1 0
SI 0 0 0 0 0 0 1 0 0 0 0 0 X
execution
DREQ
VS1103b registers are written from using the following sequence, as shown in Figure 6. First, XCS line
is pulled low to select the device. Then the WRITE opcode (0x2) is transmitted via the SI line followed
by an 8-bit word address.
After the word has been shifted in and the last clock has been sent, XCS should be pulled high to end the
WRITE sequence.
After the last bit has been sent, DREQ is driven low for the duration of the register update, marked
“execution” in the figure. The time varies depending on the register and its contents (see table in Chap-
ter 6.6 for details). If the maximum time is longer than what it takes from the microcontroller to feed
the next SCI command or SDI byte, status of DREQ must be checked before finishing the next SCI/SDI
operation.
XCS
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 29 30 31 32 33 m−2m−1
SCK
3 2 1 0 15 14 1 0 15 14 1 0
SI 0 0 0 0 0 0 1 0 0 0 0 0 X X
execution execution
DREQ
VS1103b allows for the user to send multiple words to the same SCI register, which allows fast SCI
uploads, shown in Figure 7. The main difference with a single write is that instead of bringing XCS up
after sending the last bit of a data word, the next data word is sent immediately. After the last data word,
XCS is driven high as with a single word write.
After the last bit of a word has been sent, DREQ is driven low for the duration of the register update,
marked “execution” in the figure. The time varies depending on the register and its contents (see table
in Chapter 6.6 for details). If the maximum time is longer than what it takes from the microcontroller
to feed the next SCI command or SDI byte, status of DREQ must be checked before finishing the next
SCI/SDI operation.
XCS
tXCS
0 1 14 15 16 30 31
SCK
SI
tH
tSU
SO
tZ
tV tDIS
1 25ns is when pin loaded with 100pF capacitance. The time is shorter with lower capacitance.
Note: As tWL and tWH, as well as tH require at least 2 clock cycles, the maximum speed for the SPI
bus that can easily be used is 1/6 of VS1103b’s internal clock speed CLKI. Slightly higher speed can be
achieved with very careful timing tuning. For details, see Application Notes for VS10XX.
Note: Although the timing is derived from the internal clock CLKI, the system always starts up in 1.0×
mode, thus CLKI=XTALI.
Note: Negative numbers mean that the signal can change in different order from what is shown in the
diagram.
XCS
0 1 2 3 30 31 32 33 61 62 63
SCK
1 0 2 1 0
SI 0 0 0 0 X 0 0 X
DREQ
Figure 9 shows two consecutive SCI operations. Note that xCS must be raised to inactive state between
the writes. Also DREQ must be respected as shown in the figure.
SDI Byte 1
SDI Byte 2
XCS
0 1 2 3 6 7 8 9 13 14 15
SCK
7 6 5 4 3 1 0 7 6 5 2 1 0
SI X
DREQ
SDI data is synchronized with a raising edge of xCS as shown in Figure 10. However, every byte doesn’t
need separate synchronization.
XCS
0 1 7 8 9 39 40 41 46 47
SCK
7 6 5 1 0 7 6 5 1 0
SI 0 0 X
DREQ
Figure 11 shows how an SCI operation is embedded in between SDI operations. xCS edges are used to
synchronize both SDI and SCI. Remember to respect DREQ as shown in the figure.
6 Functional Description
VS1103b is based on a proprietary digital signal processor, VS DSP. It contains all the code and data
memory needed for WAV PCM + ADPCM audio decoding, MIDI synthesizer, together with serial in-
terfaces, a multirate stereo audio DAC and analog output amplifiers and filters. Also ADPCM audio
encoding is supported using a microphone amplifier and A/D converter. A UART is provided for debug-
ging purposes.
General MIDI and SP-MIDI format 0 files are played. Format 1 and 2 files must be converted to format
0 by the user. The maximum simultaneous polyphony is 40 (peak polyphony 64). Actual polyphony
depends on the internal clock rate (which is user-selectable), the instruments used, and the possible
postprocessing effects enabled, such as bass and treble enhancers. The polyphony restriction algorithm
makes use of the SP-MIDI MIP table, if present. MIDI implements click-avoiding smooth note removal.
When run in Real Time RT-Midi mode without other signal paths, 36.86 MHz (3.0× input clock)
achieves 16-26 simultaneous sustained notes. The instantaneous amount of notes can be larger. 36 MHz
is a fair compromise between power consumption and quality, but higher clocks can be used to increase
polyphony. They are also needed if multiple signal paths are used.
Reverb effect can be controlled by the user. In addition to reverb automatic and reverb off modes, 14
different decay times can be selected. These roughly correspond to different room sizes. Also, each
midi song decides how much effect each instrument gets. Because the reverb effect uses about 4 MHz of
processing power the automatic control enables reverb only when the internal clock is at least 3.0×.
VS1103b supports unique instruments in the whole GM1 instrument set and one bank of GM2 percus-
sions.
Gain 1
UART MIDI stream
Buffer 1
SCI
Gain 2
SDI ADPCM stream Audio stream
Buffer 2 Mixer
SM_ICONF
AGC/Gain 4 Gain 3
A/D stream
A/D GAIN3 != 0
44.1 kHz
SM_RECORD_PATH=1
Generation of the Audio stream and recording A/D stream is presented in Figure 12.
Stream 1, which is the MIDI stream, may be fed either through SDI, SCI or UART. If it is fed through
UART, real-time MIDI, or RT-MIDI is assumed. The buffer size is 1024 bytes.
Stream 2, which is the ADPCM stream, may be fed either through SDI or SCI. The buffer size is 1024
bytes.
Stream 3, which is the A/D stream, running always at 8 kHz, is active if register SM ADPCM is set.
The outputs of the three streams are forwarded to the Mixer, which resamples all data to 44.1 kHz, and
forwards the data.
Either one of the A/D stream and the output of the Mixer can be fed to ADPCM encoding. If the data
is read from the A/D stream, it will be encoded as 8 kHz mono and if it is read from the Mixer, it will
encode as 44.1 kHz mono. The ADPCM compressed data may be read from SCI registers SCI IN0 and
SCI IN1. The buffer size is 1024 bytes.
Audio
stream User Bass Treble
application enhancer enhancer Earspeaker
L
Volume Audio [Link]
control FIFO and DAC R
Figure 13 presents the data flow of the Audio stream generated in Figure 12.
If SCI AIADDR is non-zero, application code is executed from the address pointed to by that register.
For more details, see Application Notes for VS10XX.
Then data may be sent to the Bass and Treble Enhancer depending on the SCI BASS register, followed
by Earspeaker Spatial Processing, depending on ST EARSPEAKER.
After that the signal is fed to the volume control unit, which also copies the data to the Audio FIFO.
The Audio FIFO holds the data, which is read by the Audio interrupt (Chapter 8.13.1) and fed to the
sample rate converter and DACs. The size of the audio FIFO is 1024 stereo (2×16-bit) samples, or
4 KiB.
The sample rate converter converts all different sample rates to XTALI/2, or 128 times the highest us-
able sample rate. This removes the need for complex PLL-based clocking schemes and allows almost
unlimited sample rate accuracy with one fixed input clock frequency. With a 12.288 MHz clock, the DA
converter operates at 128 × 48 kHz, i.e. 6.144 MHz, and creates a stereo in-phase analog signal. The
oversampled output is low-pass filtered by an on-chip analog filter. This signal is then forwarded to the
earphone amplifier.
If GPIO1 is 1 and GPIO0 is 0 at startup, RT-Midi Mode is activated. In this mode RT-Midi data is read
through the UART at the default MIDI speed 31250 bit/s. The generated audio is sent to the audio path
as shown in Figure 13.
When RT-MIDI mode is activated, GPIO2 and GPIO3 are read and their contents are copied to register
bits SCIMB EARSPEAKER0 and SCIMB EARSPEAKER1, respectively. This way it is possible to
activate EarSpeaker in this mode without writing to any SCI registers. Also, if SCI CLOCKF has not
been set to a non-zero value, the clock multiplier is automatically set to 3.5X.
The serial data interface is meant for transferring ADPCM or MIDI data.
If the input of the decoder is invalid or it is not received fast enough, analog outputs are automatically
muted.
Also several different tests may be activated through SDI as described in Chapter 7.
The serial control interface is compatible with the SPI bus specification. Data transfers are always 16
bits. VS1103b is controlled by writing and reading the registers of the interface.
1 This is the worst-case time that DREQ stays low after writing to / reading from this register. The user
may choose to skip the DREQ check for those register writes that take less than 100 clock cycles to
execute.
2 In addition, the cycles spent in the user application routine must be counted.
3Firmware changes the value of this register immediately after reset to 0x38, and in less than 100 ms to
0x30.
4 When mode register write specifies a software reset the worst-case time is 16600 XTALI cycles.
5Writing to this register may force internal clock to run at 1.0 × XTALI for a while. Thus it is not a
good idea to send SCI or SDI bits while this register update is in progress.
Note: it is not allowed to do an SCI operation while DREQ is low. If this is done, however, DREQ stays
low even after the SCI operation has been processed.
SCI MODE is used to control the operation of VS1103b and defaults to 0x0800 (SM SDINEW set).
When SM DIFF is set, the player inverts the left channel output. For a stereo input this creates virtual
surround, and for a mono input this creates a differential left/right signal.
If SM RECORD PATH is set, ADPCM recording is performed from the A/D stream at 8 kHz, otherwise
the Mixer output is recorded at 44.1 kHz. This bit is only valid if SM ADPCM is set.
To stop decoding a MIDI file set SM OUTOFMIDI, and send data until SM OUTOFMIDI has cleared. If
SM OUTOFMIDI is set while MIDI decoding has not been going on, the register bit will not be cleared
before the few first words of the next MIDI file (or zeros) have been sent to the decoder.
Bit SM PDOWN sets VS1103b into software powerdown mode where the only operational software
part is the control bus handler. Note: software powerdown is not nearly as power efficient as hardware
powerdown activated with the XRESET pin.
If SM TESTS is set, SDI tests are allowed. For more details on SDI tests, look at Chapter 7.8.
SM ICONF specifies the configuration of the data input streams. The following table shows its bits.
SM ICONF S1 Port Stream1 S2 Port Stream2
0 SDI MIDI SCI ADPCM
1 SCI MIDI SDI ADPCM
2 UART/SDI RT-MIDI/RT-SDI SCI ADPCM
3 UART RT-MIDI SDI ADPCM
When SM ICONF is set to 2, Real Time MIDI messages can be sent either through the UART or SDI.
If sent through UART, the standard MIDI protocol and date speed (31250 bit/s) is used. If send through
SDI, the protocol is otherwise the same, but every byte must either be preceded or followed by a zero
byte (but only one of these two alternative zero byte orders may be used at a time). So, a message that
would be sent as 0x92 0x37 0x73 through normal MIDI, would become 0x92 0x00 0x37 0x00 0x73 0x00
if sent through SDI.
NOTE! If you change SM ICONF, a software reset is performed as if you had also set SM RESET!
SM DACT defines the active edge of data clock for SDI. When ’0’, data is read at the rising edge, when
’1’, data is read at the falling edge.
When SM SDIORD is clear, bytes on SDI are sent as a default MSb first. By setting SM SDIORD, the
user may reverse the bit order for SDI, i.e. bit 0 is received first and bit 7 last. Bytes are, however, still
sent in the default order. This register bit has no effect on the SCI bus.
Setting SM SDISHARE makes SCI and SDI share the same chip select, as explained in Chapter 5.2, if
also SM SDINEW is set.
Setting SM SDINEW will activate VS1002 native serial modes as described in Chapters 5.2.1 and 5.4.2.
Note, that this bit is set as a default when VS1103b is started up.
Bits in SM EARSPEAKER control EarSpeaker spatial processing. They are used as follows:
SM EARSPEAKER Setting
0 Off
1 Minimal
2 Normal
3 Extreme
EarSpeaker uses approximately 6 MIPS at 44.1 kHz sample rate.
SM LINE IN is used to select the input for ADPCM recording. If ’0’, microphone input pins MICP and
MICN are used; if ’1’, LINEIN is used.
When SM ADPCM is turned on, ADPCM encoding is activated (see Image 12 at Page 26).
SCI STATUS contains information on the current status of VS1103b and lets the user shutdown the chip
without audio glitches.
SS VER is 0 for VS1001, 1 for VS1011, 2 for VS1002, 3 for VS1003, 4 for VS1053, 5 for VS1033, and
7 for VS1103.
SS APDOWN2 controls analog driver powerdown. Normally this bit is controlled by the system firmware.
However, if the user wants to powerdown VS1103b with a minimum power-off transient, turn this bit to
1, then wait for at least a few milliseconds before activating reset.
SS APDOWN1 controls internal analog powerdown. This bit is meant to be used by the system firmware
only.
SS AVOL is the analog volume control: 0 = -0 dB, 1 = -6 dB, 3 = -12 dB. This register is meant to be
used automatically by the system firmware only.
The Bass Enhancer VSBE is a powerful bass boosting DSP algorithm, which tries to take the most out
VSBE is activated when SB AMPLITUDE is non-zero. SB AMPLITUDE should be set to the user’s
preferences, and SB FREQLIMIT to roughly 1.5 times the lowest frequency the user’s audio system can
reproduce. For example setting SCI BASS to 0x00a6 will give 15 dB enhancement below 60 Hz.
Note: Because VSBE tries to avoid clipping, it gives the best bass boost with dynamical music material,
or when the playback volume is not set to maximum. It also does not create bass from nothing: the
source material must have some bass to begin with.
Treble Control VSTC is activated when ST AMPLITUDE is non-zero. For example setting SCI BASS
to 0x7a00 will give 10.5 dB treble enhancement above 10 kHz.
Bass Enhancer uses about 3.0 MIPS and Treble Control 1.2 MIPS at 44100 Hz sample rate. Both can be
on simultaneously.
SC MULT activates the built-in clock multiplier. This will multiply XTALI to create a higher CLKI.
The values are as follows:
SC FREQ is used to tell if the input clock XTALI is running at something else than 12.288 MHz. XTALI
is set in 4 kHz steps. The formula for calculating the correct value for this register is XT ALI−8000000
4000
(XTALI is in Hz).
Note: As opposed to some other VS10XX chips, a software reset must be performed after SCI CLOCKF
has been set. It is recommended that SCI CLOCKF is set only after each hardware reset / startup.
XT ALI
Note: Because maximum sample rate is 256 , all sample rates are not available if XTALI < 12.288
MHz.
Example: If SCI CLOCKF is 0xC3E8, SC MULT = 6 and SC FREQ = 0x3E8 = 1000. This means that
XTALI = 1000 × 4000 + 8000000 = 12 MHz. The clock multiplier is set to 4.0×XTALI = 48 MHz.
When decoding correct MIDI data, current decoded time is shown in this register in full seconds.
The user may change the value of this register. In that case the new value should be written twice.
SCI DECODE TIME is reset at every software reset and also when MIDI decoding starts or ends.
The current sample rate and number of channels can be found in bits 15:1 and 0 of SCI AUDATA,
respectively. Bits 15:1 contain the sample rate divided by two, and bit 0 is 0 for mono data and 1 for
stereo. Writing to SCI AUDATA will change the sample rate directly (not recommended for VS1103b!).
As VS1103b always runs in stereo mode at 44100 Hz, contents of this register is always 0xAC45 (44101).
SCI WRAM is used to upload application programs and data to instruction and data RAMs. The start
address must be initialized by writing to SCI WRAMADDR prior to the first write/read of SCI WRAM.
As 16 bits of data can be transferred with one SCI WRAM write/read, and the instruction word is 32 bits
long, two consecutive writes/reads are needed for each instruction word. The byte order is big-endian
(i.e. most significant byte first). After each full-word write/read, the internal pointer is autoincremented.
SCI WRAMADDR is used to set the program address for following SCI WRAM writes/reads. Address
offset of 0 is used for X, 0x4000 for Y, and 0x8000 for instruction memory. Peripheral registers can also
be accessed.
Only user areas in X, Y, and instruction memory are listed above. Other areas can be accessed, but should
not be written to unless otherwise specified.
SCI IN0 and SCI IN1 are used for offering SCI stream data and for reading encoded ADPCM Stream 4
data.
Example: If reading SCI IN1 returns 0x0312, then 0x03×8 words = 24 words = 48 bytes can be read
from SCI IN0 and 0x12×8 words = 144 words = 288 bytes can be written to SCI IN0.
SCI AIADDR indicates the start address of the application code written earlier with SCI WRAMADDR
and SCI WRAM registers. If no application code is used, this register should not be initialized, or it
should be initialized to zero. For more details, see Application Notes for VS10XX.
SCI VOL is a volume control for the player hardware. For each channel, a value in the range of 0..254
may be defined to set its attenuation from the maximum volume level (in 0.5 dB steps). The left channel
value is then multiplied by 256 and the values are added. Thus, maximum volume is 0 and total silence
is 0xFEFE.
Example: for a volume of -2.0 dB for the left channel and -3.5 dB for the right channel: (4*256) + 7
= 0x407. Note, that at startup volume is set to full volume. Resetting the software does not reset the
volume setting.
Note: Setting SCI VOL to 0xFFFF will activate analog powerdown mode.
Gain values are defined in 1 dB steps so that 25 corresponds to 0 dB (signal is passed on as is) and 31 is
+6 dB (signal is doubled). 0 means the channel is disabled.
If SMV ACTIVE is 0, then Gain 1 is set to 25 (0 dB), and both Gain 2 and Gain 3 are set to 0 (mute).
See Figure 12 on page 26 for more details on where gains are applied.
Note: The polarity of the gains are opposite to registr SCI VOL: higher means a higher gain, not higher
attenuation.
SARC DREQ512 affects how the DREQ pin works. If not set, when DREQ is active there is at least 32
bytes space to write to. If set, DREQ is set only when there is at least 512 bytes of free space in the SDI
input buffer.
SARC OUTOFADPCM does same to ADPCM playback as SCI MODE register bit SM OUTOFMIDI
does to MIDI playback. Thus, if you want to stop decoding an ADPCM file, set SARC OUTOFADPCM,
and send data until SARC OUTOFADPCM is cleared.
If SARC MANUALGAIN is set to 1, SARC GAIN4 sets Gain 4. Otherwise SARC GAIN4 sets the
maximum gain allowed for the automatic gain control. The value is set at 1 dB steps and value 25 means
0 dB gain (signal is passed on without change). 31 is equal to to +6 dB gain, etc. 0 disables the signal
path completely.
SCI AICTRL[x] registers ( x=[0 .. 3] ) can be used to access the user’s application program.
Note: VS1103b reservs AICTRL0 as SCI MIXERVOL and AICTRL1 as SCI ADPCMRECCTL. They
can, however, also be used for user applications if the applications don’t conflict with the originally
intended register contents.
7 Operation
7.1 Clocking
VS1103b operates on a single, nominally 12.288 MHz fundamental frequency master clock. This clock
can be generated by external circuitry (connected to pin XTALI) or by the internal clock crystal interface
(pins XTALI and XTALO).
When the XRESET -signal is driven low, VS1103b is reset and all the control registers and internal
states are set to the initial values. XRESET-signal is asynchronous to any external clock. The reset mode
doubles as a full-powerdown mode, where both digital and analog parts of VS1103b are in minimum
power consumption stage, and where clocks are stopped. Also XTALO is grounded.
After a hardware reset (or at power-up) DREQ will stay down for at least 16600 clock cycles, which
means an approximate 1.35 ms delay if VS1103b is run at 12.288 MHz. After this the user should set
SCI CLOCKF, perform a software reset, and then set other basic software registers as e.g. SCI MODE,
SCI BASS, and SCI VOL before starting decoding. See section 6.6 for details.
The internal clock can be multiplied with a PLL. Supported multipliers through the SCI CLOCKF regis-
ter are 1.0 × . . . 4.5× the input clock. Reset value for Internal Clock Multiplier is 1.0×. If typical values
are wanted, the Internal Clock Multiplier needs to be set to 4.0× after reset. Wait until DREQ rises, then
write a proper value to SCI CLOCKF, followed by a software reset. See section 6.6.4 for details.
In some cases the decoder software has to be reset. This is done by activating bit 2 in SCI MODE register
(Chapter 6.6.1). Then wait for at least 2 µs, then look at DREQ. DREQ will stay down for at least 16600
clock cycles, which means an approximate 1.35 ms delay if VS1103b is run at 12.288 MHz. After DREQ
is up, you may continue playback as usual.
If GPIO0 is set to 1, Spi Boot is performed (Chapter 7.5). If GPIO0 is set to 0 and GPIO1 to 1, RT-MIDI
Mode is activated (Chapter 6.3.2).
As opposed to some earlier VS10XX products, VS1103b has been designed so that using software resets
during normal operation shouldn’t be necessary.
This chapter explains how to create RIFF/WAV file with IMA ADPCM format. This is a widely sup-
ported ADPCM format and many PC audio playback programs can play it.
IMA ADPCM recording gives roughly a compression ratio of 4:1 compared to linear, 16-bit audio. This
makes it possible to record approx. 8 kHz audio at approx. 32.44 kbit/s or 44.1 kHz audio at 178.85
kbit/s.
IMA ADPCM recording mode is activated by setting bit SM ADPCM in SCI MODE. Before activating
ADPCM recording, user must see to it that SCI ADPCMRECCTL has been properly set.
After IMA ADPCM recording has been activated, results can be read through registers SCI IN0 and
SCI IN1.
The IMA ADPCM sample buffer size is 512 16-bit words, or 1 KiB. If the data is not read fast enough,
the buffer overflows and returns to empty state.
Each IMA ADPCM block consists of 128 words, i.e. 256 bytes (or 505 mono audio samples). If you
wish to interrupt reading data and possibly continue later, please stop at a 128-word boundary. This way
whole blocks are skipped and the encoded stream stays valid.
Note: if SCI IN1[15:8] ≥ 60 (i.e. there are more than 60 × 8 = 480 words waiting), wait for the buffer
to overflow and clear before reading samples to avoid buffer aliasing.
To make your IMA ADPCM file a RIFF / WAV file, you have to add a header before the actual data.
Note that 2- and 4-byte values are little-endian (lowest byte first) in this format:
S = n × 505. D = n × 256
If you know beforehand how much you are going to record, you may fill in the complete header before
any actual data. However, if you don’t know how much you are going to record, you have to fill in the
header size datas F , S and D after finishing recording.
The 128 words (256 bytes) of an ADPCM block are read from SCI IN0 and written into file as follows.
The high 8 bits of SCI IN0 should be written as the first byte to a file, then the low 8 bits. Note that this
is contrary to the default operation of some 16-bit microcontrollers, and you may have to take extra care
to do this right.
A way to see if you have written the file in the right way is to check bytes 2 and 3 (the first byte counts as
byte 0) of each 256-byte block. Byte 2 should always be less than 90, and byte 3 should always be zero.
In order to play back your IMA ADPCM recordings, you have to have a file with a header as described
in Chapter 7.4.3. If this is the case, all you need to do is to provide the ADPCM file to VS1103 as an
ADPCM stream.
VS10xx chips that support IMA ADPCM playback are capable of playing back ADPCM files with
any sample rate. However, some other programs may expect IMA ADPCM files to have some exact
sample rates, like 8000 or 11025 Hz. Also, some programs or systems do not support sample rates below
8000 Hz.
If SM RECORD PATH is set, recording is performed from the mixer output at exactly 44.1 kHz from
XT ALI
the mixer output. If SM RECORD PATH is not set, recording is performed at 8kHz × 12.288M Hz from
the microphone or line input. From the formula it can be seen that the nominal 8 kHz sample rate can
only be obtained if XTALI = 12.288 MHz. Example: If you have a 12 MHz clock, you will get a sample
rate of 7812.5 Hz, which should be recorded to the file.
The following code initializes IMA ADPCM encoding on VS1103 and shows how to read the data.
/* Synchronize */
do {
n = 8 * ((ReadMp3SpiReg(SCI_IN1) >> 8) & 0xFF);
Yield(1); /* Give control to other processes for 1 ms */
} while (n >= 480); /* whole buffer size = 512 words */
/* Record loop */
while (recording_on) {
while (idx < 512) {
do {
n = 8 * ((ReadMp3SpiReg(SCI_IN1) >> 8) & 0xFF);
Yield(1); /* Give control to other processes for 1 ms */
} while (n < 16); /* Only load data if >= 16 words available */
while (n--) {
w = ReadMp3SpiReg(SCI_IN0);
db[idx++] = w>>8;
db[idx++] = w&0xFF;
}
}
idx = 0;
write_block(datasector++, db); /* Write one disk block */
adpcmBlocks+=2; /* Disk block contains 2 adpcm blocks */
}
if (adpcmBlocks >= 0) {
/* The previous algorithm will always write an unfinished ADPCM block.
It doesn’t matter as we consistently only tell of the data until
the last completely written block. */
dataSizeD = adpcmBlocks*256;
chunkSizeF = dataSizeD+52:
numOfSamplesS = adpcmBlocks*505;
... /* Fix WAV header information */
}
}
If GPIO0 is set with a pull-up resistor to 1 at boot time, VS1103b tries to boot from external SPI memory.
The memory has to be an SPI Bus Serial EEPROM with 16-bit addresses (i.e. at least 1 KiB). The serial
speed used by VS1103b is 245 kHz with the nominal 12.288 MHz clock. The first three bytes in the
memory have to be 0x50, 0x26, 0x48. The exact record format is explained in the Application Notes for
VS10XX.
7.6 Play/Decode
This is the normal operation mode of VS1103b. MIDI and ADPCM are decoded, mixed and converted
to analog domain by the internal DAC.
When there is no input for decoding, VS1103b goes into idle mode (lower power consumption than
during decoding) and actively monitors the serial data input for valid data.
All different formats can be played back-to-back without software resets in-between. Send at least 4
zeros after each stream.
VS1103b can be used as a PCM decoder by sending to it a WAV file header. If the length sent in the WAV
file is 0xFFFFFFFF, VS1103b will stay in PCM mode for a long time (or until SARC OUTOFADPCM
has been set). 8-bit linear and 16-bit linear audio is supported in mono or stereo.
There are several test modes in VS1103b, which allow the user to perform memory tests, SCI bus tests,
and several different sine wave tests.
All tests are started in a similar way: VS1103b is hardware reset, SM TESTS is set, and then a test
command is sent to the SDI bus. Each test is started by sending a 4-byte special command sequence,
followed by 4 zeros. The sequences are described below.
Sine test is initialized with the 8-byte sequence 0x53 0xEF 0x6E n 0 0 0 0, where n defines the sine test
to use. n is defined as follows:
n bits
Name Bits Description
F s Idx 7:5 Sample rate index
S 4:0 Sine skip speed
F s Idx Fs
0 44100 Hz
1 48000 Hz
2 32000 Hz
3 22050 Hz
4 24000 Hz
5 16000 Hz
6 11025 Hz
7 12000 Hz
S
The frequency of the sine to be output can now be calculated from F = F s × 128 .
Example: Sine test is activated with value 126, which is 0b01111110. Breaking n to its components,
F s Idx = 0b011 = 3 and thus F s = 22050Hz. S = 0b11110 = 30, and thus the final sine frequency
30
F = 22050Hz × 128 ≈ 5168Hz.
To exit the sine test, send the sequence 0x45 0x78 0x69 0x74 0 0 0 0.
Note: Sine test signals go through the digital volume control, so it is possible to test channels separately.
Pin test is activated with the 8-byte sequence 0x50 0xED 0x6E 0x54 0 0 0 0. This test is meant for chip
production testing only.
Memory test mode is initialized with the 8-byte sequence 0x4D 0xEA 0x6D 0x54 0 0 0 0. After this
sequence, wait for 500000 clock cycles. The result can be read from the SCI register SCI IN0, and ’one’
bits are interpreted as follows:
Sci test is initialized with the 8-byte sequence 0x53 0x70 0xEE n 0 0 0 0, where n − 48 is the register
number to test. The content of the given register is read and copied to SCI IN0. If the register to be
tested is SCI IN0, the result is copied to SCI IN1.
Example: if n is 48, contents of SCI register 0 (SCI MODE) is copied to SCI IN0.
8 VS1103b Registers
User software is required when a user wishes to add some own functionality like DSP effects to VS1103b.
However, most users of VS1103b don’t need to worry about writing their own code, or about this chapter,
including those who only download software plugins from VLSI Solution’s Web site.
VS DSP is a 16/32-bit DSP processor core that also had extensive all-purpose processor features. VLSI
Solution’s free VSKIT Software Package contains all the tools and documentation needed to write, sim-
ulate and debug Assembly Language or Extended ANSI C programs for the VS DSP processor core.
VLSI Solution also offers a full Integrated Development Environment VSIDE for full debug capabilities.
SCI registers described in Chapter 6.6 can be found here between 0xC000..0xC00F. In addition to these
registers, there is one in address 0xC010, called SCI CHANGE.
0000 0000
0030 System Vectors 0030
User
Instruction
0500 RAM 0500
X DATA Y DATA
RAM RAM
1800 1800
User User
1880 Space Space 1880
Stack Stack
1940 1940
1C00 1C00
1E00 1E00
4000 4000
8000 8000
C000 C000
Hardware
Register
C100 Space C100
Every fourth clock cycle, an internal 26-bit counter is added to by (DAC FCTLH & 15) × 65536 +
DAC FCTLL. Whenever this counter overflows, values from DAC LEFT and DAC RIGHT are read and
a DAC interrupt is generated.
GPIO DIR is used to set the direction of the GPIO pins. 1 means output. GPIO ODATA remembers its
values even if a GPIO DIR bit is set to input.
Note that in VS1103b the VSDSP registers can be read and written through the SCI WRAMADDR and
SCI WRAM registers. You can thus use the GPIO pins quite conveniently.
INT ENABLE controls the interrupts. The control bits are as follows:
Note: It may take upto 6 clock cycles before changing INT ENABLE has any effect.
Writing any value to INT GLOB DIS adds one to the interrupt counter INT COUNTER and effectively
disables all interrupts. It may take upto 6 clock cycles before writing to this register has any effect.
Writing any value to INT GLOB ENA subtracts one from the interrupt counter (unless INT COUNTER
already was 0). If the interrupt counter becomes zero, interrupts selected with INT ENABLE are re-
stored. An interrupt routine should always write to this register as the last thing it does, because in-
terrupts automatically add one to the interrupt counter, but subtracting it back to its initial value is the
responsibility of the user. It may take upto 6 clock cycles before writing this register has any effect.
By reading INT COUNTER the user may check if the interrupt counter is correct or not. If the register
is not 0, interrupts are disabled.
AD DIV bits
Name Bits Description
ADM POWERDOWN 15 1 in powerdown.
ADM DIVIDER 14:0 Divider.
ADM DIVIDER controls the AD converter’s sampling frequency. To gather one sample, 128 × n clock
cycles are used (n is value of AD DIV). The lowest usable value is 4, which gives a 48 kHz sample rate
when CLKI is 24.576 MHz. When ADM POWERDOWN is 1, the A/D converter is turned off.
The watchdog consist of a watchdog counter and some logic. After reset, the watchdog is inactive.
The counter reload value can be set by writing to WDOG CONFIG. The watchdog is activated by writ-
ing 0x4ea9 to register WDOG RESET. Every time this is done, the watchdog counter is reset. Every
65536’th clock cycle the counter is decremented by one. If the counter underflows, it will activate vs-
dsp’s internal reset sequence.
Thus, after the first 0x4ea9 write to WDOG RESET, subsequent writes to the same register with the
same value must be made no less than every 65536×WDOG CONFIG clock cycles.
Once started, the watchdog cannot be turned off. Also, a write to WDOG CONFIG doesn’t change the
counter reload value.
After watchdog has been activated, any read/write operation from/to WDOG CONFIG or WDOG DUMMY
will invalidate the next write operation to WDOG RESET. This will prevent runaway loops from re-
setting the counter, even if they do happen to write the correct number. Writing a wrong value to
WDOG RESET will also invalidate the next write to WDOG RESET.
8.10.1 Registers
Start Stop
bit D0 D1 D2 D3 D4 D5 D6 D7 bit
When the line is idling, it stays in logic high state. When a byte is transmitted, the transmission begins
with a start bit (logic zero) and continues with data bits (LSB first) and ends up with a stop bit (logic
high). 10 bits are sent for each 8-bit byte frame.
8.11.1 Registers
A read from the status register returns the transmitter and receiver states.
UART ST RXORUN is set if a received byte overwrites unread data when it is transferred from the
receiver shift register to the data register, otherwise it is cleared.
UART ST TXFULL is set if a write to the data register is not allowed (data register full).
A read from UARTx DATA returns the received byte in bits 7:0, bits 15:8 are returned as ’0’. If there is
no more data to be read, the receiver data register full indicator will be cleared.
A receive interrupt will be generated when a byte is moved from the receiver shift register to the receiver
data register.
A write to UARTx DATA sets a byte for transmission. The data is taken from bits 7:0, other bits in the
written value are ignored. If the transmitter is idle, the byte is immediately moved to the transmitter shift
register, a transmit interrupt request is generated, and transmission is started. If the transmitter is busy,
the UART ST TXFULL will be set and the byte remains in the transmitter data register until the previous
byte has been sent and transmission can proceed.
The same as UARTx DATA, except that bits 15:8 are used.
The divider is set to 0x0000 in reset. The ROM boot code must initialize it correctly depending on the
master clock frequency to get the correct bit speed. The second divider (D2 ) must be from 6 to 255.
fm
The communication speed f = (D1 +1)×(D2 ) , where fm is the master clock frequency, and f is the
TX/RX speed in bps.
Transmitter operates as follows: After an 8-bit word is written to the transmit data register it will be
transmitted instantly if the transmitter is not busy transmitting the previous byte. When the transmission
begins a TX INTR interrupt will be sent. Status bit [1] informs the transmitter data register empty (or
full state) and bit [0] informs the transmitter (shift register) empty state. A new word must not be written
to transmitter data register if it is not empty (bit [1] = ’0’). The transmitter data register will be empty
as soon as it is shifted to transmitter and the transmission is begun. It is safe to write a new word to
transmitter data register every time a transmit interrupt is generated.
Receiver operates as follows: It samples the RX signal line and if it detects a high to low transition, a
start bit is found. After this it samples each 8 bit at the middle of the bit time (using a constant timer),
and fills the receiver (shift register) LSB first. Finally if a stop bit (logic high) is detected the data in
the receiver is moved to the reveive data register and the RX INTR interrupt is sent and a status bit[2]
(receive data register full) is set, and status bit[2] old state is copied to bit[3] (receive data overrun). After
that the receiver returns to idle state to wait for a new start bit. Status bit[2] is zeroed when the receiver
data register is read.
RS232 communication speed is set using two clock dividers. The base clock is the processor master
clock. Bits 15-8 in these registers are for first divider and bits 7-0 for second divider. RX sample
frequency is the clock frequency that is input for the second divider.
There are two 32-bit timers that can be initialized and enabled independently of each other. If enabled,
a timer initializes to its start value, written by a processor, and starts decrementing every clock cycle.
When the value goes past zero, an interrupt is sent, and the timer initializes to the value in its start value
register, and continues downcounting. A timer stays in that loop as long as it is enabled.
A timer has a 32-bit timer register for down counting and a 32-bit TIMER1 LH register for holding the
timer start value written by the processor. Timers have also a 2-bit TIMER ENA register. Each timer is
enabled (1) or disabled (0) by a corresponding bit of the enable register.
8.12.1 Registers
TIMER CF CLKDIV is the master clock divider for all timer clocks. The generated internal clock
fm
frequency fi = c+1 , where fm is the master clock frequency and c is TIMER CF CLKDIV. Example:
With a 12 MHz master clock, TIMER CF DIV=3 divides the master clock by 4, and the output/sampling
clock would thus be fi = 12M Hz
3+1 = 3M Hz.
The 32-bit start value TIMER Tx[L/H] sets the initial counter value when the timer is reset. The timer
fi
interrupt frequency ft = c+1 where fi is the master clock obtained with the clock divider (see Chap-
ter 8.12.2 and c is TIMER Tx[L/H].
Example: With a 12 MHz master clock and with TIMER CF CLKDIV=3, the master clock fi = 3M Hz.
If TIMER TH=0, TIMER TL=99, then the timer interrupt frequency ft = 3M Hz
99+1 = 30kHz.
TIMER TxCNT[L/H] contains the current counter values. By reading this register pair, the user may get
knowledge of how long it will take before the next timer interrupt. Also, by writing to this register, a
one-shot different length timer interrupt delay may be realized.
8.12.6 Interrupts
Each timer has its own interrupt, which is asserted when the timer counter underflows.
The System Vector Tags are tags that may be replaced by the user to take control over several decoder
functions.
jmpi DAC_INT_ADDRESS,(i6)+1
The user may, at will, replace the first instruction with a jmpi command to gain control over the audio
interrupt.
jmpi SCI_INT_ADDRESS,(i6)+1
The user may, at will, replace the instruction with a jmpi command to gain control over the SCI interrupt.
jmpi SDI_INT_ADDRESS,(i6)+1
The user may, at will, replace the instruction with a jmpi command to gain control over the SDI interrupt.
jmpi MODU_INT_ADDRESS,(i6)+1
The user may, at will, replace the instruction with a jmpi command to gain control over the AD Modu-
lator interrupt.
jmpi EMPTY_INT_ADDRESS,(i6)+1
The user may, at will, replace the instruction with a jmpi command to gain control over the UART TX
interrupt.
jmpi RX_INT_ADDRESS,(i6)+1
The user may, at will, replace the first instruction with a jmpi command to gain control over the UART
RX interrupt.
jmpi EMPTY_INT_ADDRESS,(i6)+1
The user may, at will, replace the first instruction with a jmpi command to gain control over the Timer
0 interrupt.
jmpi EMPTY_INT_ADDRESS,(i6)+1
The user may, at will, replace the first instruction with a jmpi command to gain control over the Timer
1 interrupt.
jr
nop
If the user wants to take control away from the standard decoder, the first instruction should be replaced
with an appropriate j command to user’s own code.
The system usually activates the user program in less than 1 ms. After this, the user should steal interrupt
vectors from the system, and insert user programs.
The System Vector Functions are pointers to some functions that the user may call to help implementing
his own applications.
VS DSP C prototype:
void WriteIRam(register i0 u int16 *addr, register a1 u int16 msW, register a0 u int16 lsW);
VS DSP C prototype:
This is the preferred way to read from the User Instruction RAM.
VS DSP C prototype:
u int16 DataBytes(void);
If the user has taken over the normal operation of the system by switching the pointer in UserCodec
to point to his own code, he may read data from the Data Interface through this and the following two
functions.
This function returns the number of data bytes that can be read.
VS DSP C prototype:
u int16 GetDataByte(void);
Reads and returns one data byte from the Data Interface. This function will wait until there is enough
data in the input buffer.
VS DSP C prototype:
Read n data byte pairs and copy them in big-endian format (first byte to MSBs) to d. This function will
wait until there is enough data in the input buffer.
VS DSP C prototype:
void Reboot(void);
Causes a software reboot, i.e. jump to the standard firmware without reinitializing the IRAM vectors.
This is NOT the same as the software reset function, which causes complete initialization.
• Added Buffer 1 and 2 sizes. Now both are 1 KiB (Chapter 6.3).
• Size of ADPCM encode Buffer 3 increased to 1 KiB (Chapter 6.3.1).
• Writing to SM ICONF also causes a software reset (Chapter 6.6.1).
• SCI MODE’s SS VER is now documented as 7 for VS1103 (Chapter 6.6.2).
• New register bit SARC DREQ512 (Chapter 6.6.13).
• Minor modifications.
• First release.
10 Contact Information
VLSI Solution Oy
Entrance G, 2nd floor
Hermiankatu 8
FIN-33720 Tampere
FINLAND
Fax: +358-3-3140-8288
Phone: +358-3-3140-8200
Email: sales@[Link]
URL: [Link]