Product Overview

FlexStat (General) 10 Application Guide, Rev S
INPUTS
1: SPACE TEMP
2: DISCH AIR TEMP
3: MIXED AIR TEMP
4: OUTSIDE AIR TEMP
5: SPACE HUMIDITY
7: REMOTE SPACE TEMP
8: AI_08
SPACE TEMP SENSOR
VALUE: 74.0° F
CAL. OFFSET: 0.0
OUT-OF-SERVICE: FALSE
FILTER WEIGHT:
6
Filter Weight (Input “Smoothing”)
Introduction
Filter Weight smooths (dampens) input value readings, reducing sensi-
tivity to signal noise and temperature spikes. Smaller lter weight val-
ues (more sensitive) cause the sensor readings to react more quickly to
sudden changes, and larger values (more stable) cause sensor readings
to react more slowly to sudden changes or noise. (See Response Time
Examples on page 12.)
Filter Weight values can be changed from the defaults in the FlexStat
menu (in rmware version R2.0.0.21 or later) or in BACstage or TotalC-
ontrol.
Areas subject to sudden swings of temperature from factors such as
door openings and closings may benet from adjusting the lter weight.
Temporarily reducing lter weight may also be helpful for some diag-
nostic or commissioning purposes (but the defaults should be restored
for normal operation). Although knowing “what’s under the hood” is
not necessary for adjusting lter weight, this section gives a brief de-
scription of the sophisticated processing involved, followed by addition-
al details and examples.
After the analog sensor signal is converted to a digital value, the value is
processed through two types of ltering, and the resulting value is ap-
propriately scaled. Additional details are:
1. After the Analog-to-Digital Converter (ADC) converts the voltage
to a digital value, the rst stage of ltering is a ve-sample median
lter. (See Median Filter on page 10.)
2. The second stage of ltering is a rst-order Innite Impulse Re-
sponse (IIR) lter, using a factor that is calculated based on the lter
weight. (See Innite Impulse Response (IIR) Filter on page 11.)
3. The nal displayed temperature value is calculated using the voltage
scalar, multiplier, and oset. (See Final Scaling on page 12.)
NOTE: These details are specic to the FlexStat, but the general lter
weight principles apply to all digital controllers from KMC
Controls.
Median Filter
The rst stage of ltering is a ve-sample median lter, which is de-
signed to discard fast, false, extreme values caused by glitches, such as
transient spikes, digital noise, and quantization outliers.
The rmware reads a 16-bit A/D sample every 250 milliseconds. The
device remembers the last ve Analog-to-Digital Converter samples,
sorts them lowest to highest, and picks the sample in the middle. This
is the median value. When the next sample is taken, the oldest sample
of the previous ve samples is removed from the buer, and the new
sample is added. The sort is done again, and the median sample is used.
See Median Filter Process on page 11.