UK microcontroller startup finds alternative ways to make chip variants
UK microcontroller startup finds alternative ways to make chip variants
By Chris Edwards, EE Times UK
February 11, 2002 (4:10 a.m. EST)
URL: http://www.eetimes.com/story/OEG20020207S0015
UK startup Cyan Technology aims to use flash memory and packaging to let the company build variants and application-specific standard products based on its 16bit microcontroller using a comparatively small number of silicon implementations. The company disclosed the design of its first 16bit microcontroller yesterday (6 February), having secured the funding needed to commercialise the product and build variants. Cyan has raised £2m in a second funding round. The round completes at the end of March, potentially raising another £1m. The company raised £1.5m in its first-stage funding round. The new money will be used to help develop variants of the first chip and provide the tools and development boards seen as necessary to break into a highly competitive market. David Griffiths, CEO of Cyan, said: "Spinning a new product at the silicon level is expensive. But we can do a lot with packaging options." Chief t echnology officer Dr Chris Davies said: "We recognised the product is not just defined by the chip or package. It's the software subroutines that we can store in the flash [memory]. "We have found ways to extend the microcontroller, defining new products through flash software." The eCog1 is a 16bit microcontroller aimed at communications-intensive designs. The company has gone for a design to act as a cheaper and less power hungry alternative to 32bit architectures. The company has raised £1m from local venture capital firm The Cambridge Gateway Fund, which provided half of Cyan's initial funding and was the fund's first investment. Cambridge Gateway has provided money to a number of Cambridge technology companies, including ActiveRF, Advanced Rendering Technologies, Antenova and Toshiba medical-imaging spinout Teraview. The other £1m came from Japanese VC Nippon Investment Fund, which has also invested £2m in Cambridge Gateway. Alan Barrell, head of Cambridg e Gateway said: "We liked the fabless business model. It is a complex product technically but a simple business model. The second round was not too difficult a decision." Although the 16bit market is small in comparison to the 8 and 32bit sectors, the company reckons it has identified a need for a new 16bit microcontroller architecture by focusing on power consumption and protocol-handling speed. William Giovino, executive vice-president of marketing, said: "16bit has always been the orphan child [of the microcontroller market]. It is generally as easy to build a 32bit processor as a 16bit. But a lot of the external bit twiddling needed for communications is 16bit. If I throw a '32bitter' in there, I am probably wasting power." David Griffiths, CEO of Cyan, said: "There is a gap that we see in the market. There is no easy way to add connectivity to devices at sensible price points today." Cache saves flash The ECog1 is based on a Harvard architecture a nd uses a number of techniques to cut power consumption and improve the micro's ability to handle communications. The company has used TSMC's 0.35um flash process to put 64Kbyte of flash memory on-chip. At the peak frequency of 25MHz, the processor consumes 30mW. Unusually for a 16bit microcontroller, the eCog1 has a 1Kbyte code cache. "We put the cache on because flash is power hungry and slow when you access it," said Griffiths. With the cache turned on, typical power consumption falls to a little over 20mW at 25MHz. The device has 16 clock domains to let designers turn off sections of the chip separately to cut power. A number of the peripherals are designed to offload the processor core from dealing with the large number of interrupts typical of serial interfaces. Peripherals, such as a two-channel serial port, contain additional hardware, programmed through registers, to perform much of the bit and byte level operations that are normally left to software running on the pro cessor core. "We have tried to build the timing and protocol support into hardware. For example, the dual UART does collision detect," said Davies. Similarly, the external memory interface can be configured to support a variety of memory types, including SDRAM. The processor architecture was originally developed by Cambridge Consultants for its own internal use and then spun out through Cyan. It uses a 16bit datapath with 16bit registers in a Harvard architecture that supports up to 16Mword of linear program memory through byte-wide address-register extensions and 64Kword of data. A memory management unit (MMU) gives the eCog1 virtual-memory support and the ability to map data held in code memory into the data space. This allows lookup tables to be put into on-chip flash memory and then used by instructions by making the page available as data. The MMU allows a variety of mappings but the company has been careful not to make too many obscure modes publicly available to avoid potenti al support problems caused by the use of self-modifying code techniques. The register map and instruction set has similarities with the Motorola 68HC05 and Zilog Z80 architectures, albeit with 16bit-wide base registers. It has two 16bit arithmetic registers that can be combined to form a 32bit accumulator. For addressing, the processor has two index registers, with one that can be extended for 24bit linear program addresses, plus a 24bit program counter. Memory models The company inherited a toolchain from Cambridge Consultants, which includes a C compiler. The compiler supports large or small memory models, similar to x86 compilers, to let it generate code for the extended program address modes. With the small model, it can save space by using only 16bit addresses. A scratchpad RAM area can be accessed using one-word instructions. For debug, the company has used an embedded in-circuit emulation core with a parallel bus interface based on a technique devised and patented by Cambridge C onsultants. It supports non-intrusive reads and writes to the registers, as well as access to memory locations. "We avoided JTAG because we have heard a lot of people complain about JTAG emulation. But there is a school of thought within the company that says JTAG is fine if you do it right. We may do JTAG on the next chip," said Griffiths. Although it is aimed primarily at communications control, the company has put a number of analogue functions onto the chip, including a 12bit A/D converter, a power sensor and a temperature sensor based on a bandgap reference. Davies said: "We tried to build on the analogue support products that usually surround the microcontroller on a board. If it saves space, we thought we might as well pop it onboard."
Related Semiconductor IP
- AES GCM IP Core
- High Speed Ethernet Quad 10G to 100G PCS
- High Speed Ethernet Gen-2 Quad 100G PCS IP
- High Speed Ethernet 4/2/1-Lane 100G PCS
- High Speed Ethernet 2/4/8-Lane 200G/400G PCS
Related News
- DRAM Alternative Spawns U.K. Startup
- UK startup unveils LTE transceiver IC
- UK Judge Finds Conversant Wireless' European Standard-Essential Patents Infringed by Huawei and ZTE
- UK startup is raising funds for Open Power processor
Latest News
- HPC customer engages Sondrel for high end chip design
- PCI-SIG’s Al Yanes on PCIe 7.0, HPC, and the Future of Interconnects
- Ubitium Debuts First Universal RISC-V Processor to Enable AI at No Additional Cost, as It Raises $3.7M
- Cadence Unveils Arm-Based System Chiplet
- Frontgrade Gaisler Unveils GR716B, a New Standard in Space-Grade Microcontrollers