Percent
`01
`Application
`
`Electronic
`Trans-
`mission
`
`ABS
`
`Traction
`Control
`
`Multi-
`Cellular Alraags Active
`Suspension
`plexing Phones
`
`l ; Delphi V study
`Fig.
`reveals future trends in
`automotive electronics
`features for North-American-
`made passenger cars.
`
`In addition. many sophisticated
`"adaptive" control systems such as
`those found in anti-lock brakes and
`
`injection. will require reliable,
`fuel
`real—time diagnostic monitoring for
`functional verification.
`Likewise.
`
`more sophisticated passenger com-
`partment electronics will evolve and
`become more common as comfort
`
`and convenience options become
`more affordable (see Fig. l].
`Multiplexing. or
`in—vehicle net-
`works. will be required to intercon-
`nect many of these electronic subsys-
`tems as automakers try to reduce
`costs, improve reliability and serv-
`iceability, and increase system effi-
`ciency.
`It is not clear yet which sys-
`tems will be interconnected. What is
`clear is that the semiconductor in-
`
`dustry must provide the capability to
`assist in the implementation with
`more cost-effective transistors for
`
`components.
`These transistors will yield higher-
`performance. more highly integrated
`controllers and processors for auto-
`motive applications. denser
`and
`more unique nonvolatile memories.
`and specialized communication de-
`vices for in—vehic1e networks.
`
`Power-train applications lead
`technological demands
`
`The most demanding application
`
`
`
`Cam
`
`se
`for
`been (.
`time) 1
`phlstli
`
`the er
`need
`
`
`Flash Memory
`
`Today we see advanced electronics
`used in engine management. elec-
`tronic transmission and anti-skid
`
`braking systems. shock damping.
`four—wheel steering. climate and
`cruise control. infrared door lock-
`
`ing. air bags. and service diagnostic
`computers.
`The 1990's will bring power-
`trains with continuously variable
`transmissions.
`traction control.
`
`and ultimately fully active suspen-
`sion and sophisticated information
`systems such as navigation. colli-
`sion avoidance. and “head—up" dis-
`plays.
`These applications all utilize
`various memory technologies.
`Static RAM handles the need for a
`
`data manipulation “scratchpad."
`The density requirement is often
`small enough to be integrated onto
`the CPU. Most of the external
`
`memory. on the other hand, pro-
`vides nonvolatile code storage. The
`
`growth in code density in the auto-
`motive world drives large. olf—chip
`memory needs. Also. increasing
`controller complexity translates to
`larger CPU die size and even more
`memory usage: functionality off-
`sets integration galned from ad-
`vances
`in semiconductor pho-
`tolithography. Thus. the nonvola-
`tile memory remains off—chip. The
`designer. in turn. can choose the
`code memory type which best suits
`the application needs.
`Next—generation control mod-
`ules will not only need field—repro-
`grammable memories. but very
`dense memories as well. New auto-
`
`motive memory density targets are
`64 kbyte and 128 kbyte [one—mega—
`bit). While 128 kbyte EPROMS have
`been shipped for automotive appli-
`cations for some time. they are not
`electrically erasable.
`EEPROMS,
`on the other hand. are electrically
`erasable and have just recently
`
`reached the 32 kbyte density in
`automotive versions. and as such.
`
`fall short of emerging needs.
`Flash is the newest nonvolatile
`
`Intel's
`memory technology.
`ETOXTM [EPROM tunnel oxide]
`flash memory technology is based
`on the company‘s EPROM process
`and is comprised of a single—tran-
`sistor cell. As such. it provides high
`device density. reliabilty. manufac-
`turability, plus electrical erasure.
`In
`contrast.
`conventional
`
`EEPROMS are typically two to three
`times larger than EPROMS or flash
`because their memory cells are
`made up of two or more transistors.
`The result is that flash adds the
`
`critical in—system electrical altera-
`bility of EEPROM with the storage
`capacity and reliability of EPROMS.
`Today. where EPROMS provide
`flexibility in the factory and allow
`manufacturers
`to differentiate
`
`engine control options
`
`among
`
`Page 3
`
`Automotive Engineering. August 1989
`
`
`
`Volu
`
`
`
`

`
`cam pick up
`91 34 B5 B6 87 88 89 90
`
`
`
`
`
`
`
`
`82
`B3
`92
`93
`94
`95
`96
`97
`98
`
`
`7.2=
` 720°
`
`100 team ' tooth
`Missing tooth means TDC #1
`
`nic subsya.
`
`' t0 Ted“
`
`A.
`
`,
`
`H
`ev-H-
`Tl."
`.F
`In:
`.'.I1-in
`
`.'.'iL-u-Ml
`um
`III!
`
`._II
`
`
`ield higher-
`I integrate - 1
`rs for auto ';
`
`
`enser an
`memories .,3
`lication dei
`rks.
`
`devices escalates. An order of magni-
`tude's performance increase can be
`expected in the 1990s versus today's
`production engine controllers. Sub-
`one-micron geometries will produce
`these new higher—performance de-
`vices at costs comparable to today's.
`
`
`
`5 lead
`
`applicaflo
`
`technology has
`fr semiconductor
`‘a em {and will likely remain for some
`:1»
`u e} powertrain control. As more so-
`lo sticated “adaptive” control algo-
`a thms are developed which optimize
`‘- e engine combustion process, the
`.u.- for higher performance silicon
`
`
`
`
`
` .: duct line offerings, flash mem— maintenance service. Emission
`of the assembly line. Flash—based
`systems can immediately use this
`lgegsgzcw‘
`.
`. offers flexibility beyond the as» measurements might be taken.
`eeds
`‘
`.
`_ mbly line. In fact. flash memory
`then custom—tailored code revi— same capability for field service.
`nonéolatfle
`a viable alternative for many of
`sions could be developed by a host Rapid factory reprogramming also
`Intersz
`WI automotive EPROMS in use
`computer.
`enhances testing and quality.
`mel oxide’
`ay.
`Code modification is a
`This host would have the same
`Future implementations of flash
`gy is based
`V’
`erful and virtually essential ca-
`capabilities as standard EPROM will further exploit the technology’s
`)M pmcess_
`billty for post-sales service. The
`programming equipment. Simply
`capabilities. Simple circuits can
`;ingle_U_an_
`I
`ber of parameters requiring
`plugging this machine into the handle the need for special repro-
`Ovides high
`dlvidual byte manipulation in microcomputer permits flash re— gramming voltages, all fromafixed
`I manufac_
`H time, while the vehicle is in
`programming. Removing the pro— programming power supply. The
`$1 erasure.
`.» a ration.
`is often quite small.
`gramming voltage source gives
`local microprocessor, rather than
`wenuonar
`_- cordingly,
`byte-alterable
`absolute insurance against code
`an external EPROM programmer,
`two to three
`L PROMs have been employed in
`disruption. All system modifica-
`can administer the flash repro-
`MS or flash
`is: cases where that functional— tion is done ina controlled, factory— gramming. This opens two ave-
`7 cells are
`T
`is not really needed. Flash al— authorized environment.
`nues: 1) the ability to update code
`Iansistors.
`- more cost—effective cmploy—
`Present manufacturing tech— from a serial communication link,
`h adds the -
`nt of EEPROM, concentrating it niques lay the foundation for flash.
`and 2) the ability for the vehicle to
`_iCa1 altera_ _
`ere it is truly needed.
`Compact, surface—mount packag— routinely update itself. This sec-
`the Storage
`!_ in addition,
`today's electronic
`ing is accelerating a trend toward
`ond area is becoming increasingly
`)fEPR0Ms'
`gdules are quite capable of repro— progra-mming EPROMS “on-
`important as environmental regu-
`MS provide
`M” ming themselves, given that
`board." On—board programming
`latory agencies will likely require
`I and allow‘
`:~ ey incorporate flash memory.
`allows full module assembly with
`alterable code to accommodate the
`
`ifferentiate.
`‘gu first automotive microcompu-
`noncoded memories streamlining performance changes that occur
`ns amon ~
`_' _rs to employ flash memory will
`factory output since coding can be within an engine as it wears.
`g
`l=‘ ely receive new code during done "just in time," i.e., at the end
`
`
`
`_
`
`
`
`
`
`
`August 1989
`alum: 97. Number 8
`
`
`
`
`
`Fig. 2 — Camshaft signal pickup to
`the ECU is shown over 720“ of
`crankshaft rotation together with
`spark and injection output signals
`
`based oncam-inputsignal.
`
`'_
`' and 's_
`iystem cm‘
`which sys a
`3d. What :1
`
`
`
`cam
`
`n
`
`apability
`_l U Ll: U U l...............
`I
` 1d‘-lctor 111-
`
`
`
`,
`;ation wi
`sistors fo
`
`
`
`Injection I Ignition
`
`Interrupt Processor and
`Set - up Fall Time
`
`

`
`2
`
`3
`
`4
`
`5
`
`Each edge Increments a
`counter used for an
`index Into RAM Array
`
`3rd Cam edge creates
`values for
`Cyllnderl4
`+!OFFSET I=
`
`Fig. 3 - Automatic Input
`and output engine control
`signals use OFFSET and
`DURATION registers to allow
`maximum pulse-placement
`flexibility and reduce
`processor overhead.
`
`30
`
`Input/Output critical to
`system performance
`
`Increases in computational speed
`must be coupled with an equal im(cid:173)
`provement in the microcontroller's
`input/output performance. Many
`microcontrollers used in powertrain
`control applications today have imple(cid:173)
`mented a method to detect when ris(cid:173)
`ing or falling edges occur on an input
`line. This is typically referred to as
`winput capture" and records the "real
`time" (in processor units) when an
`edge is detected {see Fig. 2).
`From this information the automo(cid:173)
`tive engineer programming the micro(cid:173)
`controller can determine engine speed
`or position in degrees. Usually, reso(cid:173)
`lution errors occur in the process due
`to the microcontroller's inability to
`detect an edge accurately.
`wlnput capture" determines the
`point in time when a crankshaft sen(cid:173)
`sor edge is detected. Based on this in(cid:173)
`formation, an output signal is gener(cid:173)
`ated to drive the fuel injectors or spark
`plug coil(s) . Figure 2 shows how an
`output signal is generated from a cam
`sensor signal. Outputs are typically
`based on one timer I counter and are
`not based on both the crankshaft po(cid:173)
`sition and the microcontroller's inter(cid:173)
`nal "real time," causing the edge place(cid:173)
`ment resolution errors.
`New microcontrollers with automo(cid:173)
`tive application-specific 1/0 will be
`available shortly to alleviate these
`problems. With the input signals
`
`'
`
`programmed toincrementanintemal
`counter automatically, these control(cid:173)
`lers will be able to keep track of the
`current tooth (engine position in
`degrees (one tooth= 7.2 degrees). The
`same input signal will be captured in
`order to translate engine position (in
`increments of 7.2 degrees) informa(cid:173)
`tion into its processor's wreal time"
`terms. This will allow easy processor
`access and calculations for output
`signals. Output signals may be pro(cid:173)
`grammed to be "automatic." Loca(cid:173)
`tions in RAM will be used to contain
`both OFFSET and DURATION values
`(from the engine position in wreal(cid:173)
`time" units). This will allow maxi(cid:173)
`mum pulse-placement flexibility with
`little or no processor overhead.
`When the OFFSET value is equal to
`zero, the rising-edge placement needs
`to occur within one microsecond. If
`every rising edge of the cam signal is
`"time stamped" into a RAM location
`and indexed via the edge number for
`easy access. an array of 100 numbers
`can be generated (one for each cam
`rising edge). The array pointer will be
`reset when the wmissing tooth" is
`found .
`This table of"real-time" values will
`be useful for generating output sig(cid:173)
`nals. OFFSET and DURATION values
`will b e added to these numbers and
`placed in their respective control reg(cid:173)
`isters. Figure 3 illustrates such a
`task.
`Decentralized control still
`important
`
`Even with improvements in proc(cid:173)
`essing performance, decentralization
`of powertrain electronics seems to be
`a trend due to cost and size con(cid:173)
`straints, environmental factors. as
`well as diagnosability and servicea(cid:173)
`bility. These distributed systems may
`perform transmission control. dis(cid:173)
`tributorless
`ignition control.
`fuel
`control, throttle control, and various
`other auxiliary engine control func(cid:173)
`tions (Figure 4).
`Distributed systems will require
`more highly integrated single-chip
`microcontrollers to meet perform(cid:173)
`ance, space, and cost requirements.
`Many new electronic systems will
`have increased subsystem capability
`while mechanical parts are reduced.
`Distributors are a good example of
`
`Automotive Engineering. August 1989
`
`

`
`
`this. With further decreases in the
`
`
`
`L:-'UJt'D3(I1"I'-303-—hCI'1O
`
`cost of single chip microcontrollers for
`a given level of performance, new dis-
`tributed systems like distributorless
`ignition may become commonplace.
`
`Application example:
`distributorless ignition
`
`affordable
`
`Let’s take a look at how a single-
`chip, 8-bit microcontroller can be
`used in a distributorless ignition sys-
`tem.
`
`_
`
`All distributorless ignition systems
`employ some type of toothed or slotted
`wheel attached to a crankshaft or
`camshaft. These wheels are used to
`l ‘calculate engine speed and to deter-
`mine engine position. The teeth are
`spaced at regular intervals or ar-
`ranged in patterns. The mechanical
`to electrical conversion is made by a
`Hall—effect sensor with additional sig-
`' nal conditioning to generate a square
`wave output.
`The microcontroller
`measures the square wave period.
`and in most applications, counts the
`number of pulses.
`One way of measuring the square
`wave period is with a 16-bit hardware
`By using a
`timer in capture mode.
`timer in capture mode to measure
`engine speed,
`the microcontrol1er’s
`- dedicated high—speed input] output
`channels are free for other tasks.
`in
`
`.
`
`the negative
`this configuration. at
`transition of the square wave pulse,
`the current value of the 16-bit timer is
`
`captured into a dedicated register.
`There is no software interrupt associ-
`ated with this timer value if it is exe-
`cuted in hardware.
`Some microcontrollers such as In-
`
`flexible high-
`tel’s 80C5lGB, offer
`speed l/O. These I/O come in the
`form of two five-channel program-
`mable counter arrays [PCA].
`PCA
`channels may be used in capture
`mode. like the timer. for measuring
`the pulse period. One of their primary
`advantages is that they can be trig-
`gered on either the negative or positive
`edge.
`It is possible to perform relatively
`simple table "1ook—ups" and interpola-
`tion [16 X 16 data points]. Total access
`time is approximately 1.3 millisec-
`onds [at 12 MHZ], permitting one read
`per revolution of the crankshaft. The
`majority of this time is spent perform-
`ing division operations to convert
`
`Volume 97. Number 8
`
`Analog Inputs
`
`Engine Sensors
`-coolant temp.
`-MAP
`-battery volt.
`-barometric
`pressure
`-air mass
`sensor
`
`Crankshaft
`Sensor toothed
`wheel with
`
`magnetic sensor
`
`CONTROL
`UNIT
`
`B7C51GB
`
`II.
`
`FUEL
`CONTROL
`UNIT
`
`8TC51GB
`
`high-speed
`pulse signals
`
`Ignition Coils
`(one per
`
`cylinder)
`
`high-speed
`pulse
`feedback
`
`OUTPUTS
`
`cylinder) high-speed
`
`Injection
`Driver
`Modules
`(one per
`
`pulse
`signals
`
`THFt01'|'LE
`CONTFIOL
`
`throttle bypass
`(idle control)
`
`throttle position
`
`Digital
`Signal
`(Ftich or Lean)
`
`; ABS
`5 CONTROL
`5 UNIT
`-_
`
`VEHICLE
`
`Exhaust Gas
`Recirculaion
`
`Exhaust Gas Sensor
`
`(Catalytic Converter)
`
`Fig. 4 — "typical engine
`control system shows the
`distributed nature of engine
`control function.
`
`degrees to microseconds: table look—
`up and interpolation itself requires
`only 150 microseconds.
`Once the ignition point and dwell
`time have been calculated, pulses
`must be sent to each pair of ignition
`coils or to individual coils. Ten PCA
`
`channels on the 80C51GB support up
`to eight cylinders with individual
`coils. The channels are divided into
`
`two sets of five, with each set associ-
`ated with its own hardware timer.
`
`Each channel has a compare / capture
`register and an output pin.
`Fuel injection is another good ap-
`plication example where distributed
`processing can make sense. Injection
`control is similar to ignition control
`because the outputs are pulses and
`the duration of the outputs are calcu-
`lated from a look-up table. As in