beautypg.com

Summed averaging, Continuous averaging – Teledyne LeCroy WaveRunner XI SERIES Operator’s Manual User Manual

Page 130

background image

W

AVE

R

UNNER

X

I

S

ERIES

130

WRXi-OM-E Rev C

S

UMMED

A

VERAGING

Summed Averaging is the repeated ad

on, with

weight, of successive source waveform records. If a

stable trigger is available, the resulting a

e ha

se component lower than that of a single-shot

record. Whenever the maximum number of swee

he averaging process stops.

An even larger number of records can

umu

hanging the number in the dialog. However, the

other parameters must be left unchang

ne

ion will be started. You can pause the

averaging by changing the trigger mod

NO

The instrument resumes averaging when

you change the trigger mode back to N

M/AUT

You can reset the accumulated averag

push

CLEAR SWEEPS button or by changing an acquisition

parameter such as input gain, offset, co

ng, tri

ondition, timebase, or bandwidth limit. The number of

current averaged waveforms of the fun

, or it

, is shown in the acquisition status dialog. When summed

averaging is performed, the display is u

ted at

uced rate to increase the averaging speed (points and

events per second).

C

ONTINUOUS

A

VERAGING

Continuous Averaging, the default setti

he r

ted addition, with unequal weight, of successive source

waveforms. It is particularly useful for r

cing no

gnals that drift very slowly in time or amplitude. The

most recently acquired waveform has m

eig

all the previously acquired ones: the continuous average

is dominated by the statistical fluctuatio

e

cently acquired waveform. The weight of ‘old’ waveforms

in the continuous average gradually ten

zero

ng an exponential rule) at a rate that decreases as the

weight increases.

The formula for continuous averaging is

new average = (new data + weight *

rag

This is also the formula used to compute summe

ing. But by setting a

sweeps value, you establish a fixed

weight that is assigned to the old avera

e th

sweeps is reached.

For example, for a sweeps (weight) val

:

1

st

sweep (no old average yet): new a

e =

old average)/(0 + 1) = new data only

2

nd

sweep: new average = (new data +

+ 1) = 1/2 new data +1/2 old average

3

rd

sweep: new average = (new data +

+ 1) = 1/3 new data + 2/3 old average

4

th

sweep: new average = (new data +

av

ata + 3/4 old average

5

th

sweep: new average = (new data +

old av

+ 1) = 1/5 new data + 4/5 old average

6

th

sweep: new average = (new data +

ld av

1) = 1/5 new data + 4/5 old average

7

th

sweep: new average = (new data +

old av

) = 1/5 new data + 4/5 old average

In this way, for sweeps > 4 the importa

the

egins to decrease exponentially.

diti

equal

verag

s a random noi

ps is reached, t

be acc

lated simply by c

ed or a

e from

w averaging calculat
RM/AUTO to STOP.

OR

O.

e by

ing the

upli

ction

gger c

s zoom

pda

a red

ng, is t

epea

edu

ise on si

ore w

ns of th

ht than
most re

ds to

(followi

old ave

e)/(weight + 1)

d averag

e number of

ge onc

ue of

4

verag

(new data +0 *

1*old average)/(1

2 * old average)/(2

3 * old

erage)/(3 + 1) = 1/4 new d

4 *

erage)/(4

4 * o

erage)/(4 +

4 *

erage)/(4 + 1

nce of

old average b

Note: The number of sweeps used to compute the average will be displayed in the bottom line of the trace
descriptor label: