IP Core Modules for FPGA use
Intona offers sophisticated modules for audio digital signal processing and multimedia data transmission purposes. The IP cores have been successfully deployed in real devices using Xilinx Spartan, Artix and Kintex FPGA. Some modules where also deployed on Lattice ECP5 using the Yosys/Nextpnr open source toolchain. All modules are written in Verilog.
Asynchronous Sample Rate Converter
Ultra-low Latency, Small Footprint
- 32 audio channels per stride
- 24 bit fixed point
- 30 to 230 kHz input sample rate
- 30 to 230 kHz output sample rate
- < -135 dB THD+N typical
- qualifies as replacement for CS8422
- industry-wide lowest latency (0.3 ms easily possible)
- smallest footprint: 32 channels in XC6SLX4, 96 in SLX9, 224 in SLX16
- occupied Spartan 6 slices: 430 for 32 channels, 510 for 64 channels
- User Guide: PDF HTML
Lattice Ladder IIR Filters
Lowest Distortion, Analog Feeling
- 32, 40, 48 or 64 bit floating point
- eight channels à 20 filters (32 on 7-series FPGAs) or
- sixteen channels à 10 filters (16 on 7-series FPGAs)
- chainable
- filter types (excerpt): parametric eq, shelving eq, low/high/all-pass filters
- less recursion than direct form (although more complex)
- separate coefficients for Q and Fc
- precise feedback quantization: excellent bit utilization even at low signal frequencies and high sample rates
- very low distortion and analogue sound
- four modules (= 640 filters) fit easily in XC3S500E
- simple bus interface for coefficient loading
- coefficient calculation LUT based or using separate MCU (C code provided)
- highest stability even in modulated systems like VCF
- lowest quantization noise industrial-wide
True-RMS Compressor/Limiter
plus zero-delay Peak Limiter
- 32, 40, 48 or 64 bit floating point
- parameters: threshold (dBFS), attack (ms), release (ms), knee (dB), crest (dB)
- sophisticated peak limiter with guaranteed voltage protection
- sixteen channels with a latency of only one sample at 96 kHz
- simple bus interface for coefficient loading
- coefficient calculation using separate MCU (C code provided)
- ideal supplement to the Lattice Ladder Core
I2S/TDM Transceiver
Very Small Footprint
- save 90% of flip-flops compared to conventional serial/parallel conversion
Ethernet Transceiver MAC
Low Latency Media Transport
- connects UDP packets directly to FPGA resources
- deterministic latency, no external buffers needed
- Linux kernel driver available
Gigabit Ethernet Transceiver MAC
Small Footprint
- PTP hardware time stamps
- AVB-Ready
- Linux kernel driver available
Gigabit AVB Endpoint
Audio Video Bridging
- up to 420 audio channels
- clock recovery or low-latency ASRC
- requires MCU (Microblaze, ARM, Risc-V) for house keeping
- Linux-based control software
Precision lin/log functions
log10(x) and 10x
- 32 bit floating point
- ultra small footprint
- 40 million operations per second at 120 MHz on Spartan
- 80 million operations per second at 240 MHz on 7-series
- standard deviation less than 0.01dB
AES67 audio streaming endpoint
Project name: AVALOT
- receive streams with any number of channels and pick 1..4 for play out
- optional transmission of a stream
- clock recovery using internal PLL
- optional upsampling to 96k
- JSON-based API over UDP and serial UART
- includes a compact RISC-V soft MCU
- reference implementation for ECP5 using Yosys/Nextpnr
.
.