data.doc

一个C style Assembler的source code

                      The 8051 Data Collection Unit

(0) Introduction
   This reference details the implementation and operation of the data
collection software included in this package, and its required hardware
configuration.

(1) Microprocessor-Based Data Collection
   The original purpose of this software was to carry out a set of fluid flow
tests on fluid meters in parellel.  Two of these units could be controlled from
a PC over a RS-485 serial link.  Each unit could test up to 6 meters at a time.

   Each meter has inside it a magnet which rotated in response to the fluid
flowing through it.  At a given flow rate, this magnet will rotate a constant
amount for each unit of fluid volume passing through the meter.  This constant
will be approximately, but not exactly, the same for different flow rates and is
an inherent characteristics of the meter's construction.

   The magnet internal to the meter is divided into a fixed number of sectors
of alternating magnetic field.  As the magnet rotates, it can be used to
generate an square electrical waveform which is ideally suited for measurement
by a microprocessor.
   The microprocessor-based data collection unit will mark off a set number
of square waves (corresponding to one full rotation of the magnet), measure
its width, and collect the information for successive measurements.  It will
use this information to determine the number of complete rotations of the
meter's magnet and to estimate the number of fractional rotations, thus
obtaining a highly accurate measurement of the meter's behavior.
   For future reference, the term NUTATION is also used to mean one full
rotation of the meter's internal magnet.

   The system is set up in such a way that each unit will receive a signal at
the start of a test and a signal at the end of the test.  The function of these
signals is to delimit a test and mark off a fixed volume of fluid.

   The remaining part of this reference will refer to this specific
configuration, but it should be noted that this software is completely generic.
It can be used to measure and perform statistical processing on any set of up
to 6 square wave inputs using a 7th falling edge input to mark the start and end
of a test.

   (a) Hardware Interface
   Data collection is done using an Intel 8051FA microprocessor running
dedicated stand-alone software.  Each microprocessor is connected to 6 pulse
inputs, to a sensor signal input, and to a RS-485 serial communications
connection.  The pulse inputs enable the microprocessor to perform measurements
on each of the 6 sources it is monitoring, as explained above.  The sensor
input is responsible for signalling to microprocessor when to start and stop
the measurement and counting.  Finally, the communications link enables the
micrprocessor to communicate with the master PC so that its data can be
transferred to the central controller.
   The inputs and outputs are set up as listed below:

         Function                          Location
      Pulse input 0                 T2EX: Timer 2 External
      Pulse input 1                 CEX0: Counter External 0
      Pulse input 2                 CEX1: Counter External 1
      Pulse input 3                 CEX2: Counter External 2
      Pulse input 4                 CEX3: Counter External 3
      Pulse input 5                 CEX4: Counter External 4
      Sensor input                  INT0: External Interrupt 0
      Serial Communications         RxD: Receive input
                                    TxD: Transmit output
                                    T0: RS-485 transmit enable
 
   (b) Internal Software Architecture
   This is a multi-tasking software architecture.  Several independent tasks
will be running simultaneously on the one 8051FA chip.  They communicate with
each other and with the outside world using the 8051 multitasking library.
Because of the large number of processes and the limit on stack space, several
compromises were made when integrating this library with the rest of the data
collection software.

   The microprocessor is responsible for monitoring 6 counters, and performing
statistical operations for each counter concurrently as it is collecting the
data.  A processing loop is set up for each counter and is synchronized by the
counter input.

   A process is set up to monitor a test, and is synchronized to the sensor
input.  When the first signal is receives, this process will Spawn the 6
counter processes (and also 2 timer processes).  When it receives the second
signal, it will disable the counter inputs (and clocks) and thus stop the
processes it spawned.

   Finally, the microprocessor will monitor the serial communications link,
waiting for commands from the master PC.  These commands serve three purposes:
(a) the enable/disable the master test process, (b) the signal the
microprocessor to transfer its data over the serial communications link, and
(c) for diagnostic purposes to test the microprocessor's status and proper
functioning.

   (c) Data Collection Method
   The data collection method used measures the times for full meter nutations
during a test between the two sensor signals

   Suppose these events take place, as illustrated in the time-line below:

 *---------------------------------------------------------------------------*
 | N = Meter Nutation                                                        |
 | S = Sensor Signal                                                         |
 |                                                                           |
 |                  S                                         S              |
 |        N   N   N     N      N      N      N      N     N     N    N       |
 |                  *-----------------------------------------*              |
 |        |   |   | | F |  w0  |  w1  |  w2  |  w3  |  w4 | L | |    |       |
 |                  *-----------------------------------------*              |
 |                  ^   ^                                 ^   ^              |
 |                  |   |   <--- N full nutations --->    |   |              |
 |                  S   B                                 C   T              |
 |                                                                           |
 |                  Events: S = First Sensor Time,                           |
 |                          B = First Meter Nutation Time,                   |
 |                          C = Last Meter Nutation Time,                    |
 |                          T = Last Sensor Time                             |
 |                                                                           |
 |   First Partial Pulse Width: F = B - S                                    |
 |   Last Partial Pulse Width:  L = T - C                                    |
 |   Average Pulse Width:      <w> = (C - B)/N (= (C - B)/5 in this example) |
 |                                                                           |
 |   Estimated First Partial: F/<w>                                          |
 |   Estimated Last Partial:  L/<w>                                          |
 |   Estimated Total Nutations:      N + F/<w> + L/<w>                       |
 |     which actually comes out to:  (T - S)/<w> =  N (T - S) / (C - B)      |
 *---------------------------------------------------------------------------*

Then the tester will only be actively counting BETWEEN events S and T.  The
data collection method will calculate the nutation widths w0, w1, w2, w3, w4
(thus counting 5 full nutations in this example); and F and L -- the first and
last partial nutations.  It will then ESTIMATE what fraction of full nutations
F and L represent by dividing F and L respectively by the average value of the
w's.

   To ensure that a steady flow condition is valid, we also measure the VARIANCE
of the pulse widths, which is defined here as:

          Variance = Standard-Deviation of w's / Average of w's  x 100%

Typical values in the actual test system this software was used for could vary
between 2% and 3% for high flow rates, down to 0.5% for low flow rates.  A high
variance (say, above 5%) invalidates the extrapolation method we use, which is
premised on a constant flow rate.  High variances can happen for two reasons:

      (1) Fluid flow is not in a steady state at the test boundaries.
          This can occur if the underlying system is still in a transient state
          as a test starts.
      (2) The unit undergoing test itself is not operating reliably.

   (d) Counter Processing
   To be able to deal with 6 counters with one microprocessor requires some
special measures be taken.  On average, for our actual test conditions (20
nutations per second received from each of the 6 inputs) the 8051FA will spend
30% of its time during the high flow test doing actual statistical calcultions.
However, the demand on the processor can vary between 5% and 200% depending on
when the 6 inputs delimit full nutations relative to one another.  The maximum
load will occur when the majority of the 6 inputs have full nutations close to
one another in time.

   What we do is the following.  When an input has signalled one full nutation,
the time of this event is saved in a protected location.  A completely
independent process, called the Scheduler, started at the beginning of the test,
will periodically remove one AND ONLY ONE of these saved items, passing it on
to the corresponding statistical processing routine.  This simple, but
effective, strategy will keep the processor active at about a constant 30% rate
(for the test conditions indicated above).

   (e) Communications
   The details of serial commuications will be described in more detail in
the section dealing with the MicroNet.  The underlying protocol used for this
system is: 9 data bits, 1 stop bit, no parity, and 9600 baud (9600,N,9,1).
The 9th data bit is used for networking purposes to distinguish commands from
data.  The centrally located PC will always send commands with the 9th bit high,
and the microprocesors will always send data, with the 9th bit low.  This
protocol will help protect the network from inadvertent access by two or more
processors at once.

   (f) Command Reference
   As a slave unit to the central PC, each microprocessor will respond to a
set of commands.  These commands are described in brief below and will be
detailed in the section on the MicroNet.

   STATUS: A diagnostic command that queries the current state of the data
           collection unit.  It will respond to the PC that it is in one of
           the following modes:

         ACTIVE: In active command mode, ready to start a new test.
         WAITING: In test mode, waiting on the first sensor.
         TESTING: In test mode, collecting data, waiting on the second sensor.

   TEST:  A command telling the microprocessor to pass into the waiting mode.
          It is ignored unless the microprocessor is active.
   ABORT: A command that forces the processor back into the active mode and
          cancels the current test.