World Patents Pending ©AirSense Technology Ltd. 1999
ISSUE 2.4
Page 17
T E C H N I C A L
.
M A N U A L
Stratos is a relatively scaled detector designed around a modern microprocessor of the
same family as that used in Personal Computers. The detector chamber is based on the
light scattering principle with a high power semiconductor laser used as the light
source. The detector operates on the forward scattering detection principle and is
responsive to a wide range of particle sizes. Both dust discrimination and filtration
techniques are employed, which makes the system virtually immune to problems
caused by dust pollution. The problems encountered in its design, their solutions and
the general method of implementation, are listed below.
Histograms and Learning
Readings from the detector chamber(s) are obtained by the micro-processor at the
average rate of once per second. A continually updated histogram is generated from
the detector(s) output, in which the histogram classes contain the percentage of
readings taken between different signal levels. The classes cover all possible readings
that may be obtained, and it shows the distribution of signal levels. From the histogram
the Standard Deviation and Mean of the distribution are continually calculated. A
decaying factor is used in the build up of the histogram. The effect of the decaying factor
is to limit the past time period or historical period that is represented by the histogram,
it also sets the time which the histogram requires to become truly representative of the
distribution. The method for doing this is an AirSense Technology Limited patented
process called ‘
ClassiFire
’. If the ClassiFire factor is set so that the a fully maintained
histogram always represents a history of the last ten hours, then it will take ten hours
running from the initial set up to build up a representative histogram. Similarly, a factor
used to make the histogram represent a ten minute history will require a ten minute
period from the initial set up to become fully representative. This is called the Learning
Time. Since the ideal should be a seven to eight hour history, this is a problem for initial
installation. The Learning Time is also a problem for a change in use of the protected
area such as a day-to-night or a night-to-day change. Both of these problems are
overcome by maintaining
two
histograms; one with a seven hour learning time (the
standard histogram) and the other with a fifteen minute learning time (the fast
histogram). On initial power up the alarm is disabled for fifteen minutes while the fast
histogram becomes representative of the signal distribution during this period. The
histogram so obtained is then used as the standard histogram which then has the decay
factor set to cover a multi-hour historical period. Since the historical period is initially
only fifteen minutes, the spread of the distribution will be smaller than if the historical
Stratos-HSSD
®
Description
