EET-i Home Page

Embedded Systems news

8-, 16-bit MCUs are the stars at Embedded Systems Conference

The new family came from Hitachi America Ltd. A giant in the seething Japanese microcontroller market, Hitachi is less well known in the homeland of Intel and Motorola. But the company has a dedicated--if small--U.S. following for its H8 family of 8-bit and 16-bit controllers.

The parts have several advantages. While the most obvious are Hitachi's modern processes, other benefits are modern, symmetric architectures with decent amounts of accessible SRAM. In the view of some software engineers, those characteristics make the new devices better suited than older architectures to C-level programming.

While the register-level architecture has been quite modern, the H8 family has still bee n very much a conventionally microcoded execution architecture. That was fine for moderate-speed operation; but as customers began to push the limits of H8 performance, Hitachi felt the need for change. It would have been too much to ask most customers to jump from the H8/300s to the incompatible, SH 32-bit processors, so the company opted to apply what it had learned in the RISC business to a new H8 execution unit.

The result was the 16-bit H8S series. The H8S is upward-compatible with the H8/300s at the object code level, but it executes a key subset of the H8 instructions in a single cycle. That allows a 20-MHz H8S chip to benchmark fully five times faster than the high-end H8/300H.

The new architecture is also far more power-efficient, according to the company. Hitachi claims that in an apples-to-apples comparison, the H8S chips will consume about one-fifth the power of the H8/300H. Helping reduce the power are not only new process stuff but also significant reorganization of the bus and clock loads o n the chips. Many of the peripheral modules in the H8 family have been redesigned as static, event-driven circuits, further reducing clock power.

The first member of the series will be the H8S/2653. Hitachi claims 10-Mips operation for the device, which includes 64 kbytes of ROM or EPROM and 4 kbytes of SRAM. A 16-bit multiply-accumulator with a 42-bit accumulator width gives the chip a significant boost in signal-processing operations.

Otherwise, the 2653 has the rich set of peripherals usually found on 16-bit MCUs, including both a 10-bit A/D and an 8-bit D/A converter; direct-memory-access (DMA), bus, timer and pulse controllers; and a serial interface. The part will operate with as little as 2.2 V. Since Hitachi doesn't intend to sample it until the first half, pricing is not yet available.

Extending the limits

There is also significant activity among older 16-bit families. Both IBM Corp. and Advanced Micro Devices Inc. are pushing out the limits on the older Intel 16-bit architec tures: the 80196 and 80186.

Advanced Micro Devices has added low-voltage parts to its line of am186EM and am188EM embedded-system processors. The two new devices operate on 3.3-V power supplies. AMD claims the benefits of that extend beyond reduced power consumption.

"There are a lot of users, like the disk-drive guys, who want 3.3 V for temperature concerns. The low-voltage parts run cooler because they consume less power,'' said Mark Mills, division marketing manger at AMD's Embedded Processor Division.

The parts run at 20 and 25 MHz. The line has a bus design that lets it work with DRAMs that run at 70 or 110 ns, so designers can use comparatively low-cost memory chips.

The low-voltage parts augment a line of parts that run software written for early members of the 80186 family. The AMD line includes DMA and PSRAM controllers, interrupt controllers, timers and other circuitry, in addition to the 186 CPU core, helping designers shrink system size. But the additional circuitry means that some of the pins normally associated with the discrete parts have been added to the AMD chips, increasing the pin count to 100 leads.

Of those pins, 32 are programmable I/O lines. In addition, an asynchronous serial port permits full-duplex 7- or 8-bit data transfers. A synchronous serial interface handles bidirectional half-duplex data transfers to and from ASICs.

The parts provide up to 3.3 VAX Mips. In lots of 10,000, pricing for both parts begins at $8.86.

"Users can get the performance of a 25-MHz 386SX without spending much money," Mills said. "Using a 386SX would push the cost up to $12 to $15."

AMD also added extended-temperature-range models, running at-40ýC to +85ýC. Pricing begins at $9.67 in lots of 10,000.

IBM, meanwhile, is soldiering away at both the 80186 and the 80196 families. The company announced its own low-power version of the AMD 186EM, using its superior process skills to pretty much blow the doors off of AMD's part. The MC18621 operates at 3.3 V internally, with either 3.3-V or 5-V I/O . But it does that at 40 MHz, yielding what IBM claims to be by far the highest 186 performance available on the market.

The speed will cost, though. The part is scheduled for late 1995 sampling and mid-1996 production, at $12 each for a 100-pin quad-flat-pack version and $12.50 each for a 100-pin SQFP, both in quantities of 10,000.

On the 196 front, the company unveiled the MC19622. For slightly less money--$11 each or $11.50 each, depending on your choice of package, in 10,000-piece quantities--you get the same, 40-MHz clock as in the 186, but with the more recent 196 architecture.

The 19622 is sampling now and is scheduled for production in the fourth quarter.

Another company strong in the Japanese MCU market but not heard from much in U.S. controller circles is Toshiba Corp. Toshiba markets both 8-bit and 16-bit MCUs here, and both families announced new members at the conference.

In the 16-bit world, Toshiba showed two new TLCS-900L family members. The 900 family in general is tuned for relativ ely large codes and C programming. It offers large numbers of register banks for Sparc-style context switching, lots of addressing modes for efficient compiled code, and a 32-bit stack pointer.

Low power

The 900L series, in particular, brings those features to a low-power environment. Operating from 2.7 to 5.5 V, the devices consume under 20 mA at 5 V and only 7 mA at 3 V.

The new chips are the TMP93PS40F and CS40F. Both have 64 kbytes of ROM--masked for the CS and one-time programmable for the PS--and 2 kbytes of SRAM.

The CS is in full production and is priced in the $8 ballpark in quantity. The PS is scheduled for production next month and costs something over $17 each in quantity. Packaging is a 100-pin QFP.

On the 8-bit front, Toshiba's TLCS-870 family is pretty much an 8-bit image of its 16-bit relative. The smaller architecture also includes lots of register sets, a 16-bit stack pointer and a rich CISC instruction set. The two new members are both low-voltage, 28-pin SOP chip s with 256 bytes of SRAM and operating voltages ranging from 2.7 V to 5.5 V.

The TMP87C408M comes with 4 kbytes of masked ROM for $2.95 each in 10,000-pieces volumes, with a $3,000 front-end charge. The TMP87P808M gives you 8 kbytes of one-time-programmable EPROM for $5.90 each in thousands. Samples are available now, with production slated for October.

Free Subscription to EE Times First Name Last Name Company Name Title Email address   Click here for your Free Subscription to EETimes Europe