Alpus is a library of VHDL components, promoting good design practices for FPGA designs. Alpus is:
- Easy to use
- High quality for professional use
- Technology independent, but optimizable for most vendor architectures
The components are mostly implemented as single files, including a package for easy instantiation. Often there are function implementations as well for flexible use.
Reset synchronizer. Arst and locked signal inputs are first synchronised to slow_clk, guaranteeing a minimum reset length for all clocks. This common reset is then synchronized separately to multiple reset outputs, one for each clock.
rstsync: alpus_resetsync generic map (
NUM_CLOCKS => 1
) port map (
slow_clk => clk, -- connect to slowest clock
arst => rst_ah, -- optional
locked => locked, -- optional
clk(0) => clk_i,
rst(0) => rst_i );
Unified interface for arhcitecure specific instantiation for a simple PLL. The architecture can be chosen from following values:
- "SIMULA": simple simulation model
- "DUMMYX": dummy clock connection with no pll functionality
- "X7MMCM": 600 - 1600MHz (Xilinx 7 series MMCM, integer dividers only supported)
- "X7PLLE": 800 - 1600MHz (Xilinx 7 series PLL)
- "GWRPLL": 400 - 1200MHz (Gowin rPLL, limited outputs available)
- "ALTPLL": 600 - 1600MHz (Altera PLL)
Supports four output clocks, frequencies selectable by input divider, vco multiplier, and output dividers. Supports phase shifing for outputs.
pll: alpus_pll generic map (
ARCH => alpus_pll_arch_synth_or_sim(PLL_ARCH, "SIMULA"), -- selects automatically "SIMULA" in simulator
IN_FREQ_MHZ => 12.0,
IN_DIV => 1,
IN_MUL => 100,
OUT0_DIV => 12,
OUT1_DIV => 24
) port map (
in_clk => clk,
in_rst => rst_ah,
out_clk0 => clk0,
out_clk1 => clk1,
out_locked => locked );
Blinks a led. Every design starts with blinking a led to see that configuration succeeds and clocks and reset are working.
blink: alpus_led_blinker generic map (
PERIOD_LEN => 24
) port map (
clk => clk_i,
rst => rst_i,
led => led );
For filling a chip with dummy logic for evaluating clock frequency in a full chip, power consuption etc.
fill: alpus_filler generic map (
ADDER_LEN => 16,
ADDER_NUM => 1000
) port map (
clk => clk_i,
rst => rst_i,
o => o );
Template for starting a new project.
Easy to use, high-performance and flexible VHDL implementation of Pipelined Wishbone B4 bus interconnect. Intended to provide similar easyness and functionality to generation based tools, but without generating.
Record types alpus_wb32_tos_t and alpus_wb32_tom_t are named according to signal direction
to master (tom) or to slave (tos). This makes signal naming simple and consistent.
Consists of components for connecting multiple bus masters to one bus, multiple bus slaves to one bus and pipeline bridges and adapters:
_____________ ________ ____________
[master0]<->| | |pipeline| | |<->[slave0]
|master_select| <-> |_bridge | <-> |slave_select|<->[slave1]
[master1]<->|_____________| |________| |____________|<->[std_slave_adapter]<->[slave2]
More details in its own repo.
Instantiation of an open-source RISC-V soft CPU, a tightly coupled memory and wishbone peripheral bus interface. Currently supports SERV and VexRiscCPU instantiation. An example design with a minimal C++ environment is included in alpus_riscv_cpu repo.
VHDL SPI slave providing memory mapped access to a 32-bit pipelined Wishbone bus. Use with alpus_wb Wishbone interconnect framework to connect any wishbone ip to an external CPU using a SPI bus. Features include:
- Easy to use synchronous design - component package included
- Provides access to Pipelined Wishbone B4 bus
- SPI burst access support with >10MHz spi bus clock
- All four SPI modes supported
- 32-bit data bus with byte select
- 24-bit address, auto-incrementing within bursts
- Address high bits (31:24) can be preset when instantiating
- Includes SPI master model for simulation in VHDL testbench
- Includes C code for register access from MCU
use work.alpus_wb32_pkg.all;
use work.alpus_spi_slave_phy_pkg.all;
spim: alpus_wb32_spi_slave port map (
clk => clk,
clk_sampling => clk_sampling,
rst => rst,
-- SPI slave
ncs => ncs,
sclk => sclk,
mosi => mosi,
miso => miso,
-- Wishbone master
wb_tos => spim_tos,
wb_tom => spim_tom );
More documentation in alpus_spi_slave repo.
A Simple Wishbone to asynchronous SRAM memory chip controller. Supports 32/16/8 bit memory bus by splitting the 32-bit wishbone transfer in multiple phases as needed. Supported memory configurations:
1/2/4 x 8bit; 2/4 x 16bit; 1 x 32bit (single rank)
- Serialized transfers for narrow chip buses
- Configurable timing (one clk resolution), min 2 clk per transfer
- Full SRAM bus utilization for pipelined transfers
- IOB Registered outputs
ramif: alpus_asram_controller generic map(
SRAM_RD_CLKS => 3,
SRAM_RDEND_CLKS => 1,
SRAM_WR_CLKS => 2,
SRAM_WREND_CLKS => 1,
SRAM_AWID => 19,
SRAM_DWID => 32,
SRAM_CEWID => 1,
SRAM_WRWID => 1,
SRAM_BEWID => 1
) port map (
clk => clk_i,
rst => rst_i,
sram_a => sram_a,
sram_d => sram_d,
sram_nce => sram_nce,
sram_noe => sram_noe,
sram_nwr => sram_nwr,
sram_nbe => sram_nbe,
wb_tos => wb_asram_tos,
wb_tom => wb_asram_tom );
Access timing is configured by SRAM_*_CLKS generics:
______________
addr XXX______________XXX
_________
data(wr) --<_________>-----XXX
__ ____
nce/nwr/nbe \___1*____/ 2* \XX
1* SRAM_WR_CLKS: Active write (must be > T_WC_SRAM)
2* SRAM_WREND_CLKS: Rising edge for nwe, bus turnaround
______________
addr XXX______________XXX
____
data(rd) ---------<____>-----
__ ____
nce/noe/nbe \___1*____/ 2* \XX
1* SRAM_RD_CLKS: Read setup (must be > T_CO + T_AA_SRAM + T_IDELAY)
2* SRAM_RDEND_CLKS: Bus turnaround, skipped during read bursts
Quadrant-flipping sin(x)/cos(x) lookup, using a rom lookup of one quadrant of pre-calculated values. Instantiate as function calls embedded in your pipeline or as a component.
Optional linear interpolation stage gives a better SNR/SFDR at the expense of two additional multipliers. Current approximation works for 4 to 6 additional phase bits.
dut: alpus_sin_lookup generic map (
PHASE_WID => 12,
D_WID => 16,
) port map (
clk => clk,
phase => phase_acc,
sin => sin_val_out,
cos => cos_val_out );
Includes Simulation testbench calculating SNR. Performance examples for some parameter combinations:
Lookup only:
8 10 11 12 14 16 phase bits
7 40.5 dB
8 42.4 48.8 49.6 dB
10 42.9 54.2 58.4 60.8 dB
12 55.0 60.9 66.2 72.8 dB
14 55.0 67.1 78.4 84.9 dB
16 67.1 79.1 90.4 dB
Lookup with 4 bit interpolation (2pi=6.0):
8 10 12 14 16 20 phase bits total
10 48.9 56.3 58.5
12 60.4 68.3 70.9
14 72.6 80.5 84.5
16 72.8 84.4 94.7
Lookup with 8 bit interpolation (2pi=25/8):
16 18 20 22 phase bits total
12 70.0 70.3
14 80.0 82.4 82.6
16 83.2 91.7 94.3 94.6
18 95.2 103.9 106.4
Sin only 12bit phase x 10bit data: 60.8 dB
Artix7: 16 LUT, 22 FF, 1 MEM18K, >300MHz
Cyclone10LP: 24 LUT, 14 FF, 1 MEM9K, >250MHz
CertusNX: 32 LUT, 12 FF, 1 MEM18K, >200MHz
Gowin GW2A: 25 LUT, 12 FF, 1 MEM9K, >250MHz
Efinity T13: 32 LUT, 32 FF, 4 RAM5K, >200MHz
Sin only 9bit phase x 8bit data: 46.3 dB
Artix7: 26 LUT, 25 FF, 0 MEM18K, >500MHz
Cyclone10LP: 80 LUT, 25 FF, 0 MEM9K, >250MHz
CertusNX: 79 LUT, 24 FF, 0 MEM18K, >200MHz
Sin/cos interpolated with 12+4 phase bits, 16 data bits: 84.4 dB
Artix7 -1: 67 LUT, 93 FF, 1 MEM18K, 2 DSP, >280MHz (REG:0111110)
Cyclone10GX -6: 100 LUT, 142 FF, 1 MEM20K, 1 DSP, >300MHz (REG:1101100)
Cyclone10LP -7: 174 LUT, 94 FF, 5 MEM9K, 2 DSP, >200MHz (REG:0111110)
LFD2NX-17 -8hp: 187 LUT, 142 FF, 2 EBR, 2 DSP, >180MHz (REG:1101100)
GW5A -C0: 167 LUT, 115 FF, 2 BSRAM, 2 DSP, >130MHz (REG:1111101)
Efinity T13 -4: 150 LUT, 189 FF, 8 RAM5K, 2 DSP, >250MHz (REG:1111100)
Efinity Tz50 -3: 139 LUT, 134 FF, 4 RAM10K, 2 DSP, >350MHz (REG:1101100)