CHANGELOG.txt

============================================================
Version numbers and features of caSCADA 
(in reverse chronological order)

Wish-list for future versions:
--> General:
-----> modbus_open_close function should be capable
-----> Fix the "PID scan rate" display so that it reads correctly when scans are skipped in pid.c or looptune.c
--> looptune:
-----> Model processes with noise and no load
-----> Model processes with specific lag orders (e.g. 1-order, 2-order) and deadtimes
-----> Multiple lag times for sim_process() 
--------> Recursive function of the integral?
-----> Control quality metrics:
--------> Percent-minutes total error accumulation
--------> Maximum valve velocity
--------> Valve reversals per minute
--------> Porpoising detection
--------> Overshoot percentage
--> pid:
-----> Series algorithm option
-----> Ratio control strategy (type=2)
-----> Cascade control strategy (type=3)

May 10, 2019 (version 3.7)
Changed license to CC0 (Public Domain).

March 21, 2018 (version 3.6)
* Added an "if" conditional to the pid.c
and looptune.c source code that allows the 
PID algorithm to be skipped every "n" scans.  
This is useless for real control applications, 
but very useful as an educational tool to
show the effects of slow PID scan rates on
control quality.  Operator interface and
trend graph scan time is unaffected by 
this conditional.

March 15, 2018 (version 3.5)
* Changed "algorithm" to "equation" where it 
represented Ideal versus Parallel, because 
the term "algorithm" is more accurately 
representative of position vs. velocity, which
"pid" currently only implements the position
algorithm.
* Edited instructions for choosing simulation
valve type to be clearer on the fact that the 
valve action (ATC vs ATO) is random, and must 
be empirically determined upon start-up of the 
simulation.
* Instructions for modifying the trend size in 
"pid" now come after the trend, so that if it 
is too large the instructions don't get blown
off the screen!
* Maximum integral action is now 400.0 repeats
per minute rather than 99.
* A text banner now appears in the upper border
of the trend screen showing whether this is a
SIMULATION or a SINGLE-LOOP control application.

December 29, 2017 (version 3.4)
* Fixed the bug in "looptune" causing a floating-
point exception error every few startups when 
selecting the "Surprise me!" option for simulated
process type.  The load interval parameter's range
was 0 to 5 when it should have been 1 to 5.  A 
value of 0 created a divide-by-zero error when run. 
* Modified the "splash" screen function for the
pid controller application, so that the initial
network and PV limit settings will be on their 
own page regardless of the height of the trend
graph.

November 16, 2017 (version 3.3)
* Eliminated unnecessary #include statement in pid.c.

July 28, 2017 (version 3.2)
* Consolidated all Modbus open/close functions
within its own source file (modbus_open_close.c)
and made this a dependency for the pid, poll,
and modprobe applications.
* Edited a typo in README.txt as well as added
another function programming example using
a static local variable.

July 27, 2017 (version 3.1)
* Added a useful Modbus/TCP testing utility of
my own called "modprobe".  This consists of a
stand-alone C-language source file that opens
a Modbus/TCP connection and then queries a 
Modbus slave device according to register
numbers the user enters in a repeating loop.
If you are setting up a Modbus/TCP device and
need to experiment with it by "probing" various
registers (i.e. reading those registers' 
contents), this utility is very useful.

June 1, 2017 (version 3.0)
* Cleaned up some trailing whitespace problems
with the ncurses display, where the controller
tuning parameters are shown.  Rather than pad
each printw() function text with blank spaces
at the end, I simply added a nested for() 
routine that "paints" a solid backdrop of color
with proper dimensions for that area of the
display.
* Fixed a floating-point exception error that
resulted in occasional crashes upon start-up.
If the lag time got randomly set to zero, it
would cause a divide-by-zero error!
* Add split-ranged output sequencing, making use
* Altered the LOAD trend recorder pen to be a
letter "w" rather than a letter "l".  This stands
for "wild" and will make more sense when I add
ratio control capability and that pen represents
the wild variable.

May 24, 2017 (version 2.9)
* Added feedforward control capability to the
controller.c file and also added code to pid.c
to read AIN1 as the feedforward load variable.
Feedforward gain and bias parameters have been
added to the ncurses display (FFG and FFB), 
and output tracking code in controller.c has
been updated to achieve bumpless transfer 
switching from manual mode to automatic mode
despite any changes which may have occurred
with feedforward action while in manual mode.
The ncurses-based display shows the load 
variable as a percentage and also as a trend
pen (l) whenever FFG is set to a practical
value.  When running as the "pid" process,
this load display is the AIN1 variable from
the feedforward transmitter.  When running as
the "looptune" process, this load display is
the randomly-walking load variable created by
the sim_process() function.

May 1, 2017 (version 2.8)
* Added the nanosleep() function to pid.c
so that the PID controller doesn't use up 
100% of the CPU's processing power when it 
runs.  Based on the success of using
nanosleep() to slow down the "looptune"
application, I decided to use it here as well.
Without nanosleep() delay, the PID controller
typically scans at several hundred times per
second on a desktop CPU and about 60 times 
per second on a Raspberry Pi model B+ (!!). 

April 11, 2017 (version 2.7)
* Added the following functionality to "looptune":
*** Created a new general-purpose randomvalue() function
returning a random floating-point value between lower- 
and upper-range values passed to the function as 
arguments.  This will be useful for randomizing all 
manner of process parameters upon start-up!
*** High, Low, High-High, and Low-Low alarm values are
*** Control valve gain is now randomized at start-up.
*** Control valve stiction ("stick" and "slip") is now 
randomized at start-up.
*** Process lag time is now randomized at start-up.
*** Process noise is now randomized at start-up.
*** Load amplitude is now randomized at start-up.
*** More comments added in controller.h regarding good
values to use for TRENDWIDTH and TRENDHEIGHT.
*** Graphical trend now begins with PV, SP, and OUT
all having a history of 50%.
*** Added a dead time function for the simulated
process, with randomized start-up value.


March 29, 2017 (version 2.6)
* Added HH, H, L, and LL indicators to the right of
the trend graph showing the hi-hi, hi, lo, and lo-lo
alarm points, respectively.
* Eliminated initializing values within pid.h which
was causing compilation errors whenever multiple 
source files included this header file.  Instead,
all initializations happen inside the set_defaults()
function.  
* Eliminated declaration of the modbus_t pointer 
"Device" from the pid.h header file, instead 
locating it at the beginning of pid.c.
* Removed PID controller code from pid.c and put
it in its own separate file (controller.c), leaving
pid.c to do the setup and initialization necessary
for real process control.  This let me write a new
program called "looptune" (looptune.c) which does
process simulation for PID tuning, using the exact
same code module (controller.c) as pid.
* Modified the project Makefile to support compilation
of "looptune" as its own program, as well as better
organize dependencies for all the other programs.
* Created two additional functions (sim_valve and
sim_process) used to simulate a physical process for
the new "looptune" program.
* Simulated loop execution time is set using the
nanosleep() function, defaulting at 0.2 seconds.
This helps conserve processing power by causing
the program to briefly halt rather than needlessly
consuming CPU cycles with delay loops.  Running on 
a base-model Raspberry Pi, the looptune process 
only demands 1.7% of processor capacity and 0.3% 
of system memory!
* Set the default PID controller gain to 0.5 instead
of 1.0 as it was before, since a gain of 1 may be
too much for some processes!
* Changed the minimum bound for scans_per_second
from 0.1 to 5.0 which is much more realistic.  A
value as small as 0.1 was causing the sim_process() 
function's numerical integration to begin much
too aggressively!  Since the "looptune" program
sleeps every 0.2 seconds anyway, a starting value
of 5.0 scans per second is perfect.


October 27, 2016 (version 2.5)
Added sawtooth step function to the formerly
"random-walk" function for cascada_poll_simulate.


October 19, 2016 (version 2.4)
Fixed error whereby the "simulate" executable was 
relying on libmodbus when it really didn't need to.
A compilation error was being generated in version
2.3 when running "make simulate", because t_modbus
declared in cascada.h header file was not a defined 
variable type.  The fix required updates to both
cascada_poll_simulate code as well as cascada_poll
and the cascada.h header file and the Makefile.
This was basically a bug that was untested in 
version 2.3.  My bad!


April 24, 2016 (version 2.3)
Re-wrote Modbus read/write functions for floating-point
variables to be generic rather than specific to the 
LabJack model T7 DAQ.  Now you just specify the absolute
Modbus address and byte-swapping order, and the read/write 
function does all the relative address arithmetic and
swaps bytes as ordered.  Also, the PID executable now only
reads one analog input (rather than all 14 inputs on the
LabJack T7), speeding up loop execution by almost one 
whole order of magnitude (600 scans/sec versus 63 
scans/sec on one machine).


March 11, 2016 (Version 2.2)
Modularized the PID functionality to support multiple
loops, each one called by passing an integer number
to it (e.g. pid(0); or pid(1);).  Also eliminated time
delay in network error recovery, so it returns to normal 
sooner.  Added GNU General Public License text and
notices to the code.  Added integral deadband, so that
we can turn off integral action whenever the PV gets to
within some specified margin of SP.  Modified derivative
calculation, so that dPV/dt is calculated over a timespan
of just a few scans (rather than over several trend ticks)
for a more prompt determination of rate-of-change.
Added comments explaining the Modbus register ranges read
by each "modbus_read_" function.


February 5, 2016 (Version 2.1)
Added basic PID control functionality on analog input
AIN0 and analog output DAC0 on the LabJack model T7
DAQ.  The new executable (pid0) is a stand-alone
process containing all the code for communication with
the LabJack as well as the PID algorithm and operator
interface (based on ncurses).  The trend display's
width and height are both set by a #define term 
(default = 78 wide to accomodate an 80-column display).
Also switched on the built-in network error recovery
of libmodbus, so that it handles events such as a 
temporarily unplugged Ethernet cable.


December 8, 2015 (Version 2.0)
Added analog output capability for the LabJack model T7
DAQ, which has two 12-bit 0-5 VDC analog outputs.  The
two analog outputs are driven by two floating-point
variables in the caSCADA memory: aout[0] and aout[1]
whose raw values are to be interpreted as volts DC.


November 17, 2015 (Version 1.1)
The random walk function in cascada_poll_simulate.c now
has easily-configurable step span and hi/lo bound limit
variables.


November 8, 2015 (Version 1.0)
The two data output files, data.txt and data.html, now 
contain status values.