Transcript Practicalities of Digital Control

Practicalities of Digital Control
A Survey
The Overall System
•
•
•
•
The individual controllers
The interconnection
The central computer or computers
The transducers and actuators
The individual controllers
• PLC
• General-purpose controllers
• Purpose-built
More on the P L C
• Probably still ‘ladder logic’ -- good for onoff control but cumbersome for analogue -but ...
• Can incorporate analogue I/O
• Can often include p.i.d. controller blocks
and routines in high-level languages
• Most modern PLC types can be interfaced
to a SCADA system
Traditional example -- Mitsubishi
F1/F2
•
•
•
•
•
Basically ladder logic
Can do analogue quantities, but awkward
Good at on -- off control
An analogue ladder example follows
More modern networkable PLCs will be
described in a later lecture
General-purpose Controllers
• Most common type is PID but other
strategies possible
• Parameters can be changed/downloaded
by/from SCADA system
• Self-tuning types becoming more popular
but care is still needed in their use
• Less good at on-off than the PLC
• Better at ‘analogue’ control
DSP Possibility
• Can be based on DSP Chips
• Very fast micros optimised for multiplyand-add ... the sums of digital control
• Some can operate in floating-point
• Serial interface to I/O
How fast shall we sample ?
• Shannon/Nyquist Theorem -- we must
sample at least twice the highest frequency
present if we are not to lose information
• Actually we need to sample faster than this
to avoid aliasing and because of noise
• 10 - 20 x highest frequency of interest is
usual
Why ?
• Less than 10 x means it is difficult to
produce an effective anti-aliasing filter
• More than 20 x leads to a double penalty ...
• We have to do the sums faster ...
• ... and more accurately if they are to work !
• But what is this aliasing thing ?
Suppose we sample this signal
every 14 s
x
10
x
x
5
0
x
x
-5
-10
x
x
0
20
40
60
80
100
“I’ve got aliasing, doctor.”
• We have ‘found’ a sine wave of much lower
frequency than the actual one.
• The system may be able to respond to the
lower frequency one ...
• ... even if the original was too fast for it to
respond to.
“I’ll give you a prescription for
...”
• Some effective screening (the highfrequency signal is likely to be the mains or
Radio 1/2/3/4/5/etc)
• An analogue low-pass filter on the inputs
The anti-aliasing filter
• Has to be analogue (it would itself be at the
risk of aliasing if it were digital !)
• It must not appreciably affect signals within
the normal frequency range of the controller
but it must effectively remove everything
above half the sampling frequency.
• The faster we sample, the easier it is to
remove the aliasing signals.
A Sampling-rate Example
• Controller 0.1s + 1 +2/s
• We have already digitised with a sampling
interval of 0.05 s
• We will see what happens with 0.25 s ...
• ... and 0.01 s.
• Using the simple substitution.
We obtain with Ts = 0.25 s ...
• 1.9 - 1.8z-1 + 0.4z-2
• ----------------------•
1 - z-1
• This one causes very serious
degradation of performance -- if
not actual instability
What is happening ?
• The problem is that by sampling we are
producing a Transport lag
• We remember from Analogue Control that a
Transport Lag is a pure time delay ..
• ... and that it reduces system stability by
increasing the phase lag in the loop.
• We introduce one by sampling ...
What is happening -- Continued
• ... because an event happening during a
sampling interval is only detected at the
next sampling instant.
• So the delay in detecting it can be anything
between zero and a full sampling interval ..
• .. so it is Ts/2 on average.
• This is the effective extra transport lag
introduced by sampling.
...So let us use Ts = 0.02 s ...
• 11.02 - 21z-1 + 10z-2
• --------------------------•
1 - z-1
• If we do not do the sums very
accurately, we entirely lose the
integral term !
Ts = 0.02 s .. A Consequence
• We will lose the integral term entirely if we
use 8-bit arithmetic.
• I will do the sum ...
• ... in which we are only allowed integer
numbers between 0 and 255 (or probably
between -128 and +127 in practice)
Interconnection
•
•
•
•
Multi-line bus (VME etc)
Parallel or serial
Two-wire (FIELDBUS etc)
Systems often combine hardware and
software
Analogue Interfaces
• Voltage ranges (often 0 -> 10 V or -10 -> 10
V)
• Current loop (usually 4-20 mA)
• So e.g. for 8-bit, 4 mA converts to 0 and 20
mA converts to 25510
• What would the values be for 12 and 16 bits
?
Arithmetic
• Now normally floating-point within the
controller
• Fixed-point arithmetic is still used in some
low-cost (often mass-produced) equipment
• It saves hardware cost but incurs extra
development time
• Input and output are still fixed-point
Precision
• 8-bit I/O restricts us to 0-255 decimal
• 12-bit often used in ‘good’ systems
Supervisory Control -- SCADA
• Central computer (or network) connected to
local controllers, PLCs and data loggers
• Data recording as well as control -- often
now with an economic process optimisation
overtone
• Central control of parameters and setpoints
but the local controllers and PLCs do the
actual controlling
SCADA Continued
• Often able to do statistical analysis on the
data collected
• Especially ‘trending’ to see if quantities are
changing when they should be constant (or
vice versa)
SCADA Continued
• Upmarket PCs often used now instead of
minis/mainframes/workstations
• Examples follow ....
First -- just a PLC !
•
•
•
•
Canal-lock control panel
Controlling two sets of gates and ..
... two sets of paddles.
Needs to detect gate position and water
level on each side (done via pressure)
• Hydraulics to operate gates and paddles
Again not SCADA -- Disk Head
Drive
• Linear motor plus drive electronics
• Must be fast, so DSP chip used
• Position feedback from format track pattern
on disk
A Glassworks
• Central Computer -- high-spec PC
(duplicated)
• “Hot End” -- GP Controller for zone
temperatures and feed + PLC for batching
• “Cold End” -- PLC (mostly on-off)
• Transducers -- mostly of the “on-off” type
apart from temperature
Transducers
• Analogue then A - D ...
• or...
• ... direct to digital
Example -- Position or Angle
• We can use a potentiometer or LVDT
• to give a voltage dependent on the position
or angle to be measured
• then digitise it
Position or Angle -- Continued
• Or we can use a Gray-coded disc or strip to
give a digital reading directly.
Precision
• The control is only as accurate as our
measurement of the quantity being
controlled
• Our transducer must be accurate enough to
fulfil the specification
Timing
• Interrupts
• Real-Time Clock
• Watchdog Timer
Real-Time
•
•
•
•
Operating System or Language ?
Hierarchy of interrupts
Solves the sampling-interval problem
May need an arrangement for immediate
action in the event of problems during an
interval
• Local or central ?