"Andy Hall" wrote in message
...
On Sun, 8 Feb 2004 19:32:48 -0000, "Set Square"
wrote:

In an earlier contribution to this discussion,
Andy Hall wrote:

On Sun, 8 Feb 2004 15:22:09 -0000, "Set Square"
wrote:

Thankyou for explaining cycles per hour too. Presumably with the
default 10 minute cycle (6 per hour) once it has done its thing
within the current cycle, it won't turn the boiler on again until
the start of the next 10 minute cycle even if the house has cooled
rapidly in the meantime because (say) some fool has opened all the
doors and windows?

That would be true unless the temperature has fallen so rapidly that
you reach the edge of the proportional band. At that point, the
controller will go to 100% on. Obviously this is another factor that
you could trade in deciding on the bandwidth setting. If the sudden
loss of heat situation happens a lot, then setting the bandwidth
narrower might be in order.

Ah, now I understand the proportional band width a bit better - I had
previously assumed that it had something to do with the gain - but it is
actually the temperature band within which the controller exercises what
might be called "time division proportionality".

Proportional control has no time division. That would be PI control, which
the CM67 has. Most commercial controllers now tend to have PID control.

Full explanation.


PROPORTIONAL

Proportional control is a corrective action that is proportional to the
error. Remember that error is defined as the difference between the desired
value (setpoint) and the actual measured value (process variable).
Proportional control is referred to as either "gain" or "proportional band".
Proportional band is the percent of change in the controlled variable that
causes the controller output to move over 100% of its range. In other words,
a proportional band of 100% means that if the measured variable moves over
100% of its range, the controller output signal will move over 100% of its
range; or, if the measured variable moves 1% of its full range, then the
control output moves 1%. In general, as the proportional band percentage
gets smaller, the controller becomes more sensitive. It is possible to make
it so sensitive that a very small change in the process variable measurement
will result in very large controller output. For those who like to work in
gain instead of proportional band, the relationship between these terms is
simple: gain is the reciprocal of proportional band. If the proportional
band is 100%, then gain is 1/100%, or 1. If proportional band is 50%, then
gain is 1/50%, or 2.

A proportional-only controller cannot control most processes well because it
never reaches the desired setpoint. This offset, defined as the difference
between the desired setpoint and the actual process value that the
controller can achieve for the process, is due to the mathematical nature of
the proportional control mode. Because there is only one controller signal
output value associated with the particular process variable value the
controller is reading, the controller sticks to its rigid mathematical
association between input and output values. It cannot recognize that there
is a problem of not achieving the desired setpoint value with
proportional-only control. Increasing the controller gain will reduce, but
not eliminate, the offset, but too much gain will result in unstable control
that will oscillate endlessly above and below the desired setpoint.

INTEGRAL
Integral settings are defined in repeats per minute or minutes per repeat. A
setting of 10 repeats per minute will instruct the controller to repeat the
action 10 times each minute. In general, the more often the repeats, the
faster the offset will be eliminated. Integral control is a control mode
that acts only as long as an offset exists.


There are three basic modes of control:

1. Proportional - operates on off-set and never reaches setpoint
2. Integral - time.
4. Derivative - anticipation, measurting rate of change.

PID controllers, which incorporate all three, are generally better.
Certainly more flexible. In short, the PI of the CM67: proportional gets it
near the setpoint, but never reaches setpoint because it operates on
"offset". A proportional controler can sit there all day and not move and
the setpoint will not be reached. They need a prod. The Integral mode of
control operates on time and pulses the offset (the differnce bewteen the
setting the proportional mode has decided to stay on an the setpoint) up to
setpoint, giving high accuracy.

Be careful not to have the proportional band too tight as "hunting" will
occur, known is short cycling to you.

In a CM67, the control aspect that times the delayed start is separet from
the PI temperature control.

The time proportions of the CM67, are a crude way of keeping an on-off
boiler within the proportional band.

The CM67 can control an electric actuator. A better system would be a heat
bank, heated by a basic on-off control boiler (these are cheaper and have
simple more reliable basic controls. The CH flow and return pipes would
have a modulated 3-way valve, moved up and down by and electric actuator.
This can stop at any point in the travel. This way only PI control would be
used and guarantee much more accuracy of setpoint temperature control. As
many heat bank and thermal stores are recommended to be on 24/7, the boiler
can be controlled only by the heat bank and the CM67 moves the modulated
valve and switches on the pump when it calculates the start time.

Bandwidth is a bad word to use but is the usual one. Normally it's
used in connection with radio and network engineering to mean
something completely different.

It is not "bandwidth". It is "Proportional Band - width". The "width" is
the range of the band, e.g., 2C. That is the setpoint and 1C either side.

The way you've described it for
this application is exactly right, though.

Outside that band it is
either fully on or fully off (if I'm right).

Yes. Also within the band, don't forget that there is the minimum
on-time which has an influence.


I presume that there isn't any
user control over the gain applied to either the proportional or integral
feedback?

Not directly. There will be several factors, including the house
behaviour which will affect behaviour.


[I must admit that, having only worked in the past on analog
controllers for mechanical systems, I'm not too clear on how the
principles
apply to something which has only an on/off output].


If you want to think of it electronically, in the proportional range
it is roughly equivalent to a switch applying charge or discharging a
capacitor and being switched each way to set a fixed voltage. If you
were to take the voltage on the capacitor and filter it to remove the
short term variation, you will end up with more or less DC voltage
which increases and decreases according to required heat.

This is really all that is happening. The house has a much longer
time constant than the cycle rate of the controller and so there is a
smoothing effect.

The proportional switching is only there because most boilers don't
have an analogue input to control the power level.

This is a very common method of industrial temperature control where
the process and the heater have a long time constant.

snip crap about snakes, which should all be banned