Skip to content

Latest commit

 

History

History
150 lines (136 loc) · 8.15 KB

README.md

File metadata and controls

150 lines (136 loc) · 8.15 KB
Raw

OlcbRefresh

WARNING This code is in development. TivaCAN appears to work, but not AT90CAN (it transmits, but does not receive). Others not tested.

This is a refresh of the Arduino base libs, ie OlcbStarLibraries.

It is meant to simplify and extend the Arduino code.

Changes:

  1. Added support for multiple processors: Arduino, Teensy, Tiva.
    • Each set of files specific to a CAN-processor is kept in its own directory.
    • The processor is automatically selected in the processor.h file.
  2. A sorted Index[] is used to speed eventID processing, using a hash of the eventid.
  3. Simplified the definition of CDI/xml for the node by matching a struct{} to the xml structure, see the example below.

e.g.: This CDI/xml, which self-describes the node to the system:

    <cdi>
        ...
        <group replication='8'>
        <name>Channels</name>
            <eventid><name>event0</name></eventid>
            <eventid><name>event1</name></eventid>
        </group>
        ...
    </cdi>

parallels this program structure:

    typedef struct {
        ...
        struct {
            EventID event0;
            EventID event1;
        } channels[8];
        ...
    } MemStruct;

Memory Models:

The code can be written to be maximize space or speed, and the user can choose from three memory models:

  1. Small: All operations are from EEPROM
    Minimizes RAM size, but is slower. If many eventIDs are needed, it may be the only choice.
  2. Medium: only eventIDs are copied to RAM into eventids[]
    Quite good compromize between size and speed.
  3. Large: The whole of EEPROM is mirrored to RAM as mem[]
    Faster but larger. This is best on larger processors.

In Flash:

eventidOffset[] stores the offset of each eventID into the EEPROM or RAM struct{}.
It is initialized at compile-time.

In RAM:

eventidIndex[] contains a hash of each eventID, and an associated sequential index into eventidOffset[].
It is sorted on the hash values.

In the medium model only, eventids[] contains a copy of the node's eventIDs, and is indexed by eventidIndex[].

This restated as a diagram:

  • In all models:
        eventidIndex[]--(index)-->eventidOffset[]--(offset)-->mem[] or EEPROM[]
        eventidIndex[]--(index)-->Events[].flags
  • In the Medium model only, eventIndex also indexes the eventIDs array:
        eventidIndex[]--(index)-->eventids[]

More about OpenLCB/LCC - what is it?

OpenLCB/LCC is a set of heirarchical protocols to let nodes talk to each other.
For more information on OpenLCB, see: OpenLCB.org
For protocol descriptions, see: Adopted LCC Documents

These protocols consist of:

  • Systems/Application Messaging
    These are the 'workhorse' messages on which most applications will be built, they are useful for sytems-messaging as well, and for building further systems-protocols.
    • PCE - Produver/Consumer Event Messages
      • These are unaddressed EventID messages.
      • They implement Producer/Consumer Events (64-bit)
      • EventIDs are globally unique 64-bit numbers.
      • These are one-to-many messages.
    • Datagram Messages
      • These are addressed messages containing up to 70-bytes of data in a single message.
      • These are one-to-one messages.
    • Stream Messages
      • These are addressed messages carrying unlimited data in multiple messages.
      • These are one-to-one messages.
  • Systems/Housekeeping
    These are the 'behind-the-scenes' protocol that enables and ensures the system's integrity.
    • Link - establishes and maintains the node's link to the network
      • Announces state of Node
      • Announcement of Intialization Complete
      • Announcement of Consumed- and Produced-eventIDs
      • NodeID reporting on request.
      • EventID reporting on request.
      • On the CAN-implementation, this maintains alias assignment and maintenance;
    • SNIP - Simple Node Information Protocol
      • Provides a brief description of the node for UI Tools to use.
    • PIP - Protocol Identification Protocol
      • Defines which protocols the node uses, and it is reported as a bit-map.
    • CDI - Configuration Description Information
      • Reports the node's CDI/xml on request.
    • Memory Configuration
      • Reading and writing to the node's memory spaces, including Configuration, RAM and EEPROM spaces.
  • Additional Protocols
    These protocols extend the base-system.
    • Teaching -- teaching an eventID from one node to one or more others.
    • Traction Control -- train control.
  • Additional Utility-Libraries
    These libraries implement useful functionality.
    • BG - Blue/Green -- node health indicators and system buttons.
    • ButtonLed -- implements controlling a button and LED from a single processor pin.

How the Above Translates to the Codebase

The 'codebase' is a set of libraries and functions that implement the basic protocols of OpenLCB/LCC.
Each protocol has corresponding code, usually in the form of a class, and implemented as a pair of *.h and *.cpp files.
The codebase tries to hide some of the complexity in #include files.

However, each protocol needs to have:

  • initialization, and
  • processing

For example there are some selected lines of code from the OlcbBasicNode example used for initialization:

  NodeID nodeid(5,1,1,1,3,255);    // This node's default ID; must be valid 
  const char SNII_const_data[] PROGMEM = "\001OpenLCB\000DPHOlcbBasicNode\0001.0\000" OlcbCommonVersion ; 
  uint8_t protocolIdentValue[6] = {0xD7,0x58,0x00,0,0,0};
  ButtonLed* buttons[] = { &pA,&pA,&pB,&pB,&pC,&pC,&pD,&pD };
  PCE pce(&nodal, &txBuffer, pceCallback, restore, &link);
  BG bg(&pce, buttons, patterns, NUM_EVENT, &blue, &gold, &txBuffer);

Most of the processing is hidden as functions in the #include files.

How Does the Application Interact with the Codebase?

The programmer of the Application must:

  • Decide what and how the new Application works, ie how eventids and other node variables are used to build the Application.
  • Choose the NodeID - this must be from a range controlled by the manufacturer - ie you.
  • Write the CDI/xml describing the node and its node-variables, including its eventIDs.
  • Write a MemStruct{} that matches the xml description.
  • Choose the Memory Model, one of: SMALL. MEDIUM, or LARGE. A good first choice is MEDIUM.
  • Write code to flag each eventID as a Producer, Consumer, or Both.
  • Write pceCallback(), which processes received eventIDs, ie eventIDs to be consumed, and causing whatever action is required, eg a LED being lit or extinguished.
  • Write produceFromInputs() which scans the node's inputs and, if appropriate, flags an evenItD to be sent.
  • Write userConfigWrite() which is called whenever a UI Tool writes to the node's memory. This code can then compare the memory address range of the change to the node's variables, and take whatever action is appropriate, e.g. update a servo position.
  • Write additional support and glue code for the Application.

Example Applications

The provided examples will give some ideas of how to accomplish sample projects. They can form the basis of, or be adapted to, a new Application, or just used for inspiration.

  • OlcbBasicNode
    Implements a simple node which exercises most of the protocols. It has two inputs and two outputs. Each input has two Producer-eventIDs and each output has two Consumer-eventIDs, so 8 eventIDs in total. This Application makes use of the ButtonLed library to control two buttons and two LEDs. In addition, it implements the BG (Blue-Gold) protocol to allow the teaching of eventIDs between this node and others.
  • OlcbServoPCA8695
    Implements driving a number of servos from a PCA8695 PWM chip. It shows how to write a different pceCallback(). It also uses userConfigWrite() to allow real-time updating of a servo positions from a UI Tool, such as JMRI or Model Railroad System.