Process leaps drive power and memory in micros
Manufacturers of 8bit microcontrollers are beginning to take advantage of process technologies until recently reserved for more advanced products to lower power consumption and boost memory capacity, with Embedded World in Nürnberg this week providing the focus for their efforts.
Silicon Laboratories has chosen the 0.18µm process for the first 8bit microcontrollers that can be powered directly from a single AA battery while STMicroelectronics has leapt ahead to the 0.13µm generation to push on-chip flash memory capacity to 256KB.
Gary Franzosa, marketing manager for interface products and microcontrollers at Silicon Labs, said the company decided that off-chip power converters were too inefficient to drive a growing body of consumer electronics products that run off single AAA or AA cells. He pointed to wireless mice as typical targets where off-chip DC/DC converters are unable to reduce power consumption to levels designers are happy with. “We went through and did the analysis on five or six external DC/DC converters. If they are able to achieve a low-power sleep mode, then active mode current is very poor,” he said.
So the company tuned the transistors on the 8051-compatible microcontroller to run at a slightly higher voltage than a single AA cell and built a boost DC/DC converter onto the die. To allow use with two cells connected in series, the microcontroller also has a low-dropout regulator that steps the voltage down to the required voltage of the core transistors. “The sweet spot for the threshold voltage of the 0.18µm transistors for a microcontroller that runs at 20MHz or so happened to be at that point,” said Franzosa.
Because the company moved to a comparatively advanced process, the designers had to deal with higher leakage current. “Most low-power devices are done in 0.25µm or 0.35µm because they have very low leakage,” said Franzosa.
To take power consumption down to 50nA while the microcontroller is sleeping, there is a separate power manager able to work with an input voltage from 0.9V up to 3V that removes power from all the other non-essential circuitry. The contents of registers are shifted to shadow registers implemented using low-leakage transistors to allow the microcontroller to maintain its state.
To avoid the need to add an off-chip DC/DC converter for external circuitry, the on-chip power circuitry can provide 65mW of power, which should be enough to power LEDs or simple liquid-crystal displays (LCDs), Franzosa claimed.
The use of a 0.18µm process allows for larger on-chip memory than other microcontrollers in the low-power market, such as Texas Instruments’ current generation of MSP430 devices. The Silicon Labs C8051F900-series micros can hold up to 64KB of flash memory the base device with 32KB of flash memory costs just under $2 in 10,000-off quantities.
ST’s decision to move to a 0.13µm process will allow the company to provide up to 256KB of on-chip memory on members of its STM8 series of microcontrollers. The move to 0.13µm was helped by the availability of an electrically erasable non-volatile memory (EEPROM) option. Alexander Czajor, marketing manager for STs automotive body-electronics microcontroller group, said most competitors use standard flash memory to hold application data but that demanded emulation techniques, which proved troublesome for users.
ST has decided to make the STM8 family compatible with the 32bit STM32, which is based on ARM’s Cortex-M3, in terms of peripherals the same strategy taken by Freescale Semiconductor with its ColdFire 32bit and 68HC08 8bit microcontrollers.
Instead of encouraging customers to simply move to 32bit to get bigger memory arrays, Czajor said the company saw a need for an 8bit line as well, primarily in the Chinese market. Customers there are cost-sensitive enough to take the difference in core size between 8bit and 32bit processing into account. However, realising that paging is difficult to use, ST designed the STM8 so that 8bit registers could be combined to generate the 24bit addresses needed to address up to 16MB.
Czajor said the initial parts are aimed at the automotive market but added that the STM8 will be a broadly used family, with industrial-market parts following on. “Each application area will have its own sub-portfolio,” he noted.
In the 32bit space, NXP Semiconductors is aiming at even larger memory arrays, with plans to move below 0.18µm and 0.14µm processes as flash memory becomes available on the 90nm processes. The shrink will make the bigger flash sizes economic. “The biggest trend we are seeing is customers going straight in with large code sizes and using 512KB flash parts in volume. That was non-intuitive to us: we thought the smallest memory would have the highest volume. The 32bit space seems to be encouraging larger code sizes,” said Geoff Lees, vice president and general manager of the microcontroller product line at NXP. “We are looking at going to the most aggressive node we have for flash and we are evaluating 90nm. We have test chips [on 90nm] but the flash has different characteristics to existing technologies. The flash can be larger but the module has significantly worse power-down characteristics. However, it would, theoretically, make 2MB or 4MB flash options affordable.
“Cost reduction through shrinks will allow us to provide 256KB or 512KB at last year’s 64KB prices,” Lees added.
Lees said a lot can be done architecturally to make 90nm more attractive to microcontroller users in terms of power consumption. “Going to 90nm, we go in the wrong direction for the microcontroller market. But we can work together with ARM to improve on that with more advanced power management,” said Lees.
Image: Silicon Labs has earmarked wireless mice as potential targets for its 0.9V microcontrollers