|
by Tom Cantrell
Start ę Little
Network ę Mini-Message ę LO
BAT ę Caravan ę Sources
and PDF
LITTLE NETWORK
Actually, LIN stands for Local Interconnect
Network, not little network, but you get the idea. Itęs an econobox
Class A network for all those proliferating lights, switches, and
motors at a low cost that CAN canęt match.
LIN is the proposal of a largely European
consortium of automakers (Audi, BMW, Volvo, and VW) with some U.S.
participation from recently married DaimlerChrysler and U.S. automotive-electronics
kingpin Motorola. The specs and first generation tools were originated
by the final consortium member, Volcano Communications Technologies
AB.
LIN incorporates many design features
and eschews others, in a manner unique to the task at hand. Itęs a
simple UART-based (ISO 9141 NRZ, in auto spec terms) scheme that sacrifices
speed and fancy desktop network wannabe pretensions in favor of simplicity
and low cost.
Electrically, LIN is a single-wire (up
to 40 m), wired-AND arrangement (like I2C), where each
node is connected by a pull-up resistor to the bus and any node can
transmit by pulling the line low (see Figure 2).
 |
Figure 2ęThe basic wired-AND
configuration is similar to I2C, but LIN is even
simpler. (enlarge) |
The LIN bus runs off the car battery
(12 V), thus, eliminating many electrical and wiring hassles right
up front. The downside is that the bus needs to handle power surges
up to 40 V, polarity reversal, and such, but that dirty work is handled
by off- and on-chip transceivers.
The single-wire configuration and large
voltage swing dictate a limited slew rate (typically 2 V/ęs) to minimize
EMI concerns. That, in turn, limits the maximum data rate to 20 kbps,
which is just as well. Itęs fast enough to handle lights-and-switches
duty, but unlike J1850/CAN, can easily be bit-banged by any 8-bit
micro.
Itęs more than the speed limit that differentiates
LIN. After all, itęs roughly similar to current J1850 nets. Rather
than raw bandwidth, itęs the way the LIN bandwidth is used that sets
it apart.
The major distinction is that LIN is
a polled, rather than a contention, network. In LIN, slaves speak
only when spoken to. Yes, LIN does require a master node to run things
(i.e., poll), which represents a potential single point of failure.
However, thatęs also true of many other electrical components in modern
cars, and silicon is more reliable than most. If thatęs not good enough,
a combination of explicit design redundancy and limp-home backup can
be applied.
In recompense, the master-slave approach
has some practical benefits. The master controls all communication.
There is no arbitration or automatic retry. The message duration is
either bound by the LIN specification or under the softwareęs control.
Put it all together and itęs possible to achieve accurate timing and
guarantee the worst-case service interval for each node.
In fact, what timing uncertainty LIN
does have is largely a by-product of the data rate being allowed to
drift quite a bit (ę15%). Indeed, all nodes re-synchronize their timing
to the master on every message. The benefit is that, unlike the typical
fixed data rate setup, the nodes donęt need expensive, fragile crystals,
but can instead use simple (albeit sloppy) RC clocks.
PREVIOUS
NEXT
Circuit Cellar
provides up-to-date information for engineers. Visit www.circuitcellar.com
for more information and additional articles.
For subscription information, call (860) 875-2199, subscribe@circuitcellar.com
or subscribe
online. ęCircuit Cellar, the Magazine for Computer Applications.
Posted with permission.
|