Ametek Western Research IPS-4, User Manual

Page: 21 / 238

Overview | 1-3
STEP 4: Filter output of Step 3.
A median filter is applied to the current analyte value reading
1
V
incorpo-
rating the previous analyte value reading
0
V
.
The size of the median filter
is set in parameter MEDSZ. The median filter is follow by an IIR filter
(Equation 1.3) weighted by a non-linear function defined by ALPHA and
BETA parameters. The weight, wt, is calculated using the following equa-
tion, where
avg
V
is the average of the last AVERAGE readings (Equation
1.4). The values for MEDSZ, ALPHA, BETA, and AVERAGE dictate the re-
sponse lag due to filtering. There is a trade-off between response time and
signal noise. Decreasing the amount of filtering (e.g. decreasing MEDSZ)
will speed up the response time of the analyzer, but will also increase the
noise on the analyte reading. Typically, the filtering is adjusted at the fac-
tory to provide a T
90
response of less than 30 seconds (from the time the
sample reaches the Measuring Cell
(s)
). T
90
is defined as the time required
to see 90% of a step change in the analyte value.
( )
wt V V V V
IIR
• − + =
0 1 0
Equation 1.3
( ) ( )
0
1
V V
abs BETA
avg e ALPHA wt
−
• −
• − =
Equation 1.4
STEP 5: Apply linearization.
For some applications, linearization is then applied to the resulting fil-
tered value. If linearization is turned on, the value is checked against
linearization breakpoints. The response between any two consecutive
breakpoints is assumed linear. The slope and intercept are calculated for
each segment and used to calculate the linearized value.
STEP 6: Convert to desired units of measure.
STEP 7: Apply calibration span and offset corrections.
Finally, calibration span and offset corrections are applied to the resulting
filtered (and potentially linearized) value,
v
(see Equation 1.5).
FFSET
O S v V
soc
+ • =
PAN
Equation 1.5