Engine-control units, airbag-deployment systems, antilock brakes, and
audio systems can all contain flash memory, but you won't hear much about these
systems. Car makers still regard the use of flash as a company secret. The
advantages are that great.
Fine tuning engine control. Delco Electronics Corp., Kokomo, IN, uses flash memory in its engine-control unit (ECU). An ECU electronically controls the ratio of air and fuel during combustion. It optimizes fuel economy and performance while also reducing fuel emissions.
"Our first ECU with flash was a mainstream General Motors application first introduced in 1993. Flash is or will be in all our current and future ECU products for GM," says Dick Harrison, a technical fellow at Delco.
Delco engineers previously used a UV-erasable ROM for the engine-control unit. It held the software and calibrations for the engine-control computers, but was only updatable in special remanufacturing shops. Technicians had to replace a repluggable assembly in the ECU--they could not erase or reprogram the memory at the dealership. Flash proved a cost-effective approach because it eliminated access covers so that dealers wouldn't have to open up the engine controller to upgrade the memory contents--they could do it electronically via a cable.
Why update the memory? One reason: GM engineers might determine they can solve a driveability complaint or problem by releasing new engine calibrations. With flash, all technicians have to do is plug a connector in a vehicle's test port. With the proper equipment, they can electronically reprogram the flash without touching the engine-control module.
Another reason Delco moved to flash, says Senior Development Engineer Mark Lowden, was that it kept them abreast of where memory suppliers told them the technology was going. "We're always needing higher densities and higher performance," he says, "and most people told us they were discontinuing that work with UV EPROMs and were going to focus on flash." Delco's choice: Intel's 2-Mbit Boot Block flash chips.
Technical challenges involved in redesigning the ECU were numerous. "They went way beyond adding the memory to the engine-control unit itself all the way to our data-information systems at Delco and GM," says Harrison. This involved a complete upgrade of all dealership equipment and the creation of new information systems so the dealers could get the latest memory releases.
Before, if dealers wanted an upgrade, they had to order a new part. Now they can solve complaints right away without having to tell the customer to be patient and come back in a couple of days.
Customer complaints. Typically, customers have driveability complaints. They might have a rough start or shifting problem, or may find acceleration sluggish in a particular region of the U.S. With the flash ECU, the dealer goes into an electronic service system and looks at all the calibrations available for any particular vehicle to solve the problem.
In addition, flash gives Delco manufacturing economies. Before flash, for a particular product, the company had 20 to 40 different variations of the memory.
"There is virtually no end to the amount of memory we will use in our products," says Harrison. "Our goals are to make the engine run as smoothly as possible, to improve emissions, to improve fuel economy, and to introduce other functions we hadn't been able to take on because of lack of processor speed. As memory increases, there's more demand for speed. As microprocessor speed increases there's more demand for memory--they work hand in hand."
Cost is bottom line. "There's not anything magical about flash relative to EPROM or EEPROM except it is going to be a lower-cost solution to do the same type of memory alterations," claims Mike Stewart, automotive DSP marketing manager at Texas Instruments. The reason? Flash takes less silicon area to store a given amount of code or data.
Texas Instruments is sampling what it claims is the industry's first fixed-point digital signal processor (DSP) with on-chip flash memory. The new 16-bit DSP offers flexible reprogramming to help OEMs offer new or user-specific features. Designated the TMS320F206, the device integrates 32k words of flash memory. Another new device, the TMS320F207 DSP, has added peripherals.
Both U.S. and foreign car companies are working with flash DSPs, ex-plains Stewart.
Why have the flash on the DSP instead of on a separate memory chip? "You have to look at the system solution," explains Stewart. "If your system cost is less by integrating it on chip, then you will do it. If not, then you won't."
System cost includes the price of both components and board space, which has a price per area--the less board space the cheaper the overall cost. The interface between the flash and the DSP tends to cause EMI problems, so integrating them on the same chip reduces electrical noise. Other integration benefits: You're dealing with only one manufacturer and one part rather than two or three.
Stewart notes that flash memory is finding its way into car systems that have different features. For example: audio systems. One radio might have a CD player; another might have a tape player. The controls would be slightly different, but you could use the same device--such as a flash DSP--and just change the software in the flash memory during production to customize each audio system.
DSPs are making their way into power-train control, airbag control systems, and electric-motor use in internal-combustion vehicles. For example, a DSP could control motors for power steering, AC compressors (instead of using power from the engine), and antilock brakes (instead of using hydraulic systems).
Another key application: "smart" airbag deployment. Right now, airbags are one-size fits all. You don't want that airbag going off if you've got a rear-facing child seat or a small child. A flash DSP with sensors could detect who occupied a car seat and not deploy an airbag or deploy one in stages.
DSPs are typically more expensive than microcontrollers, but they have more capability. If you're adding features to airbag deployment, you'd need a higher-performance microcontroller or a DSP. The best choice depends on the overall system and how it's implemented.
"I'd like to say that DSPs are the world's best solution for everything," says TI's Stewart. "But that's not true. The engineer will use the lowest cost solution, period."
Flash microcontrollers. Motorola is one of several companies adding flash memory to its microcontrollers, including versions of the 16-bit 68HC16 family. Sampling this quarter are flash versions of the 16-bit 68HC12 microcontroller family. One of these, the B32, has 32 kbytes of flash and 768 bytes of byte-erasable EEPROM. These chips are code-compatible with the 8-bit 68HC11 architecture. Also available, and designed originally for BMW, is the 32-bit 68F333--Motorola's first microcontroller with integrated flash.
Integrating flash onto a microcontroller reduces the number of external components, which reduces board space and complexity and increases system reliability, say Motorola officials. Integrated flash also improves memory access times by up to 50%, improving overall system performance.
"The 68HC16 family with integrated flash has been very successful in the Japanese auto market," says Bob Pinteric, automotive product manager in modular operations for Motorola. One customer uses an 68HC16 derivative with 100 kbytes of flash for a power-train application. The flash contains all the instructions for the CPU to control the different engine functions.
The customer likes the flash for several reasons, including field-reprogrammability. If the car maker had to upgrade the code in existing vehicles, it could do that very simply with flash. Also, it can use the flash to customize an individual module by doing a software change on the production line. One set of hardware could be used for several different applications by simply changing the code in the flash, which cuts down on inventory and manufacturing costs.
The next big application area for flash microcontrollers? "Antilock brakes," says Pinteric, who adds that Motorola has several customers evaluating the technology in Japan. "For these systems, size and space are a big constraint," he says, "so integrating everything on-to a single chip is a big benefit."
One such benefit for ABS is that the system algorithm might need fine tuning de-pending on the vehicle options--tire size, engine type, and power-train configuration. "If you have flash," says Automotive Market Development Manager Rudam Bettelheim, "it's very easy to customize the final code or algorithm for each application using just one hardware module. You don't have to stock 20 different ROM versions."
Body control also lends itself to flash memory. Not only do cars have many equipment options, but the legislation varies for each country. For example, some countries let you have the dim beam and the bright beam on at the same time, others don't. You could use the same body-control module and use flash to program it for the options on the car and the country it's going into.
'Hot ROMs.' Many car makers don't like to broadcast what processor or flash chip it uses to stop people from trying to tamper with the code, says Bettelheim. Current power-train systems tend to use EPROM, sometimes in a socket. Bettelheim says there's a memory aftermarket developed by people who have modified calibration data to get more performance out of a car--at the cost of wearing out the engine sooner and having higher emissions.
Somebody might put in a "hot ROM," damage the engine, and then make a claim under an extended warrantee program after plugging in the original memory part, Bettelheim adds. There's no way to prove what's been done.
Using these hot ROMs is illegal in many states. So manufacturers are under pressure to prevent code tinkering. If the public doesn't know which processor or memory a car uses, that makes it harder to go in and tinker with the programs.
Motorola says integrated flash--as well as secrecy--helps because a hot-ROM dealer would have to replace the entire CPU or reprogram the entire flash. A decidedly nontrivial task.
What is flash? Flash memory is a type of EEPROM, or electrically erasable programmable read-only memory. It's nonvolatile, unlike DRAM and SRAM (dynamic and static random-access memory), requiring no power to retain data. It's also electrically reprogrammable in the field--unlike EPROM, which has to be programmed off-line and soldered on a pc board.
Also unlike EEPROM, flash stores and erases data in blocks--EEPROM can change data bit by bit. Different types of flash memory have different types of blocking structures, block sizes, or block combinations.
ROM, another alternative to flash, is cheaper on a cost-per-bit basis. But it takes 12 to 16 weeks to manufacture such a device. And, if you have a bug in your code, you're forced to trash all the ROM devices you have and spend another 16 weeks making new devices. Flash memory would let you make those changes on the production line. The code is not hard-wired into the device, it's programmed in.
Flash brings flexible reprogramming, so products can be updated easily with new features, standards, or user information at any time--during development, production, and even after the product is in the customer's hands.
BOOT BLOCK INTEGRATES FLASH TYPES
T he Intel flash-memory chips Delco uses in its engine-control units feature the company's Boot Block architecture. Essentially, the chip combines three types of flash memory into one device, optimizing the chip for storing embedded code.
The main block--in 96- or 128-kbyte sections--functions as an EPROM replacement for application code that requires occasional updates. The midsize 16-kbyte block serves as a ROM for storing boot code. This block is protected against accidental erasure to store a small amount of execution code required to boot the system.
The third block type emulates byte-alterable EEPROM, which many designers use to store operational parameters that change or age, historical information, constants, or user data. Intel's part has two 8-kbyte parameter blocks. When one block is full, stored bytes are written to an unused block and the old block is erased to be used later. This provides EEPROM-like byte-alterability at the system level without the added chip count or cost.
Texas Instruments: http://www.ti.com/