4 Signal Conditioning and Data Acquisition

Recall from Section 1 of the lecture notes that a generalized

measurement system can be divided into:
sensing element, signal
modification system
indicator or recorder

However, many, if not most (but not all) modern engineering

measement systems can be made using sensors which have
electronic outputs. In this case, the parts of an electrical signal
measrement system are sometimes described in the following way,
using an alternate terminology

We discuss the different types of sensors/transducers later on.

In this section, which pertains to data acquisition, we discuss both
i) Signal conditioning, and
ii) Data acquisition

60

4.1 Signal Conditioning

The object of signal conditioning is to
improve, change or analyze the sensor output before being acquired in a way that
it improves the quality of the data

Emphasis is often placed on the removal of

noise

from a signal.

Some of the most common functions of a signal conditioner are:

1)

amplification

2)

attenuation

3)

filtering (low-pass, high-pass,

band-pass, band-stop)

4)

Computation : differentiation,
integration, comparison, etc.

5)

signal conversion

We discuss these in differing amounts of detail in the subsequent

subsections.

61

4.1.1 Amplification
The most common signal processing operation is, arguably,
amplification of

weak signals

Special care is required in the processing of such low-level (e.g. V)

signals, which are typical of strain gauges, thermocouples and other
transducers.
Furthermore, amplification of signals can also be used to maximize
the resolution of data acquisition systems.
We have seen that the performance of an amplifier (with respect to
its capability to amplify) is quantified by the gain, G:
G=Vo/Vi
GdB =20logG=20log(Vo/Vi)
However, it is important to note that, when used with time-varying
signals, amplifiers do not have the same gain for input signals of
different frequencies.

62

A perfect amplifier will have a

In reality, an amplifier will have

frequency response.
constant

intermediate range of frequencies, being

gain over an
atttenuated

at low and

high frequencies.
Some terminology:
bandwidth
3dB

fc1 , fc2 are known as:

cut off or corner
frequencies

fc1

Why -3dB?

fc2
-3dB=20log(Gc/Gmax) --> Gc=0.7099Gmax ~ 1/sqrt2*Gmax

Most modern instrumentation amplifiers have constant gain at low

frequencies such that

fc1=0

(Note: f=0 corresponds to a

bandwidth =fc2-fc1

constant

(i.e.

Also, for most op-amp-based amplifiers:

(i.e. larger gains imply

smaller bandwidth

63

DC

) signal).

Gx*fc=constant

).

An amplifier with a narrow bandwidth will change the shape of a

time-varying input signal called

frequency distortion

Frequency distortion of a square wave due to high-frequency attenuation (from Wheeler and Ganji)

The phase angle of the output signal from an amplifier can also be
non-zero, i.e.
if

Vi = Vmaxi sin(t)

phase angle

then Vo = GVmaxi sin(t + )

Typical phase response:
log (freq)

Note the (amplitude) frequency response + the phase response of

a system are called the

bode diagram

64

The performance of amplifiers is also quantified by the

common-mode rejection ratio (CMRR)
CM RR 20 log10

Gdif f
Gcm

Gdif f = gain when different voltages are applied to the 2 terminals

Gcm= gain when the same voltage (w.r.t. ground) is applied to
the 2 terminals
For an ideal amplifier, Gcm=

. Large CMRR is

desirable .

4.1.2 Attenuation
If a signal is larger than the input range of the next component (e.g.
the

), then it must be attenuated.

data acquisition

A voltage divider can be used, e.g.

Vout = Vin

R2
R1 + R2

but this may be susceptible to

loading

can be solved by use of high-input impedance

amplifiers.

65

4.1.3 Filtering
As you may already know, and as we will discuss in more detail,
later on in this course, any time-varying signal (voltage, in most
cases) can be decomposed into a sum of (possibly an infinite number
of) sine/cosine waves of different frequencies and amplitudes.
Consideration of a time-varying signal as a sum of trigonometric
functions of different frequencies leads to the notion of the
frequency domainand Fourier analysis.

y(t)=Asin(2*pi*f*t)

y(t)=Asin(2*pi*f1*t)+Bsin(2*pi*f2*t)

A>B, f1<f2

66

Filtering is used to eliminate undesirable elements from a signal,

particularly

noise

. Generally, this is best done in the frequency

domain.
As we shall see later on in the course, filtering is also required
during data acquisition, when the signal is being sampled at
discrete times (as in the case in digital systems). This can prevent a
serious problem called aliasing, which can result in entirely
erroneous/misleading data being collected.
Filtering acts by modifying (generally

) certain

components of a time-varying signal that are of (a) certain

frequencies. There are 4 types of filters:

low pass

high pass

band pass
band stop

Categories of electrical filters: (a) lowpass; (b) highpass; (c) bandpass; (d) bandstop (from Wheeler and
Ganji)

67

fc is called the
amplitude drops by

or

frequency, where the output

.

Low-pass filters

High-pass filters

Band-pass filters

Band-stop filters

(Notch filters

).

68

There are 4 main classes of filters:

Butterworth: Pro:
For a low-pass Butterworth filter:
G=
For a high-pass Butterworth filter:
G=
(will discuss this in more detail at the end of the course).

Gain of lowpass Butterworth filters as a function of order and frequency (from Wheeler and Ganji)

n is the

. For n = 1 (first order filter), when the

frquency doubles, G decreases by a factor of

corresponds to a

, which

dB attenuation.

In general 6n dB/octave roll-off. (48 dB/octave @ 8th order).

(octave= a doubling of the frequency).
69

(a decadeis an increase of the frequency by a factor of

).

Con:
Chebyshev filters:
Pro:
Con:

Gain of lowpass Chebyshev filters as a function of order and frequency (from Wheeler and Ganji)

Higher-order Chebyshev filters are better than Butterworth filters

for use as a notch filter.
Elliptic filters:
Pro:
Con:

70

Bessel filters:
Pro:

Con:

Comparison of Butterworth and Bessel phase-angle

Gain of lowpass Bessel filters as a function of order

variation with frequency

and frequency
(from Wheeler and Ganji)

Engineers generally specify, rather than design, a filter (EEs being

possible the most likely exceptions), which can be commercially
purchased. However, their basic, underlying theory must be
understood.

71

DC Source/Calibrators, Tunable Electronic Filters,

Wideband Power Amplifiers
Precision Phasemeters, Distortion Analyzers
Function Generators, RC Oscillators

Quality in Test and Measurement Since 1949

Model 3384
0.1Hz to 200kHz
Four Channel 8-Pole Filter
Four Independent Filter Channels

Two Independent Band-Pass Channels

Attenuation: 48dB/Octave
Functions: High-Pass, Low-Pass, Band-Pass
Response: Butterworth and Bessel
Input: Differential and Single-Ended
Input Gain: 0dB to 50dB in 10dB steps
Output Gain: 0dB to 20dB in 0.1dB steps

DESCRIPTION
The Krohn-Hite Model 3384 Four Channel Filter provides ease of
operation and the reliability Krohn-Hite has been known for since
1949; along with a price that will fit any budget.
The 3384 four independent channels provide a tunable frequency
range from 0.1Hz to 200kHz; and with the 002 option, the range is
extended to 0.005Hz. The frequency response characteristic is
selectable to either maximally flat (Butterworth) for clean filtering
in the frequency domain, or linear phase (Bessel) for complex
filtering.

Low-Pass Amplitude Response

Each channel of the 3384, also offers selectable functions of

low-pass and high-pass 8-pole filtering with selectable input gains
to 50dB in 10dB steps, and output gains to 20dB selectable in
0.1dB steps. The 3384 will accept input signals to 10V peak at
0dB gain and has selectable ac or dc coupling. Memory is
available for storing a set-up of the instrument which can be
recalled with a simple command.

BAND-PASS/BAND/REJECT OPERATION
With the Model 3384 the user can simply connect two channels in
series to achieve band-pass operation or connect the two in
summed parallel for band-reject operation. A total of two
band-pass/band-reject channels are available in the 3384.

APPLICATIONS
Applications of the Model 3384 are anti-aliasing, ultra-sound
measurements, random noise testing, sound recording,

High-Pass Amplitude Response

Krohn-Hite Corporation, 15 Jonathan Drive, Unit 4, Brockton, MA 02301

Tel: 508-580-1660; Fax: 508-583-8989; www.krohn-hite.com; email: sales@krohn-hite.com

71.1

Page 1 of 2
3384Data.vp

0.1Hz to 200kHz Four Channel Filter

suppressing interference in audio communications and related

fields of medical, geological, geophysical, oceanographic,
military and many more.

Model 3384

Output:

SPECIFICATIONS
Specifications apply at 25C, 5C.

Post-Filter Gain: 0dB to 20dB in 0.1dB steps, 0.2dB.

FUNCTIONS
Low-pass filter, high-pass filter. Provides two channel of
band-pass or band-reject via external connections.

Maximum Current: 80mA peak.

FILTER CHARACTERISTICS
Type: 8-pole, selectable Butterworth and Bessel.

Maximum Voltage (open circuit): 10V peak.

Impedance: 50 ohms.
DC Offset: Adjustable to zero volts.
GENERAL

Attenuation Slope: 48dB/octave.

Crosstalk Between Channels ( input source !50 ohms):

80dB for fsig !200kHz, 70dB for fsig >200kHz.

Tunable Frequency Range fc: 0.1Hz to 200kHz; (option

002, 0.005Hz).

Memory: 9 stored set-ups.

Frequency Resolution: 0.001Hz, 0.1Hz to 0.999Hz; 3

Digits, 1Hz to 200kHz, (option 002, 0.001Hz from 0.005Hz
to 0.1Hz).

Self-Test Diagnostics: MPU checks unit upon power-up.

Display indicates failure mode.
Displays: 7 segment, green, LED; 0.3" high.

Cutoff Frequency Accuracy: 3%.

Operating Temperature: 0C to 50C.

Relative Gain at fc: 3dB, Butterworth; 12.6dB, Bessel.

Isolation to Chassis: 200Vdc.

High-Pass Bandwidth (0dB Gain): >2MHz.

Input/Output Connectors: BNC.

Stopband Attenuation: >80dB.

Power Requirements: 90-132/180-264 volts ac,

50Hz-400Hz, 10 watts (3361), 15 watts (3362), 30 watts
(3364).

Wideband Noise (2MHz bandwidth detector): 0dB gain,

<400Vrms. Max. gain, <25Vrms RTI.
Harmonic Distortion (1V input, 0dB gain): 60dB (0.1%)
to 10kHz; 50dB (0.3%) to 100kHz.

Dimensions and Weights: 3.5 (9cm) high, 14" (36cm)

wide, 12.5 (32.13cm) deep; 12 lbs (5.4kg) net, 14 lbs
(6.3kg) shipping.

DC Stability: Typically 1mV/C.

Accessories: 3-terminal line cord; operating manual.

Input: Differential or single-ended.

OPTIONS
002: extends low end cutoff to 0.005Hz.

Pre-Filter Gain: 0dB, 10dB, 20dB, 30dB, 40dB, 50dB,

0.2dB.
Impedance: 1 megohm in parallel with 25pf.

BK-330: Line/battery operation.

Maximum Input: 10V peak at 0dB gain, reduced in

proportion to gain setting.

Rack Mount Kit: Part No. RK-314, permits installation of

the Model 3384 into a standard 19" rack spacing.

CMRR: >60dB to 10kHz; >50dB to 100kHz.

Extended 1 Year Warranty: Part No. EW3384.

Coupling: ac (0.16Hz) or dc.

OPTIONAL ACCESSORIES
CAB-025: Cable, BNC, 3ft, Low Noise

Sensitivity: 3mV peak with 70dB total gain for 10V

peak output.
Maximum DC Component: 100V in ac coupled
mode.

Specifications subject to change without notice.

Krohn-Hite Corporation, 15 Jonathan Drive, Unit 4, Brockton, MA 02301

Tel: 508-580-1660; Fax: 508-583-8989; www.krohn-hite.com; email: sales@krohn-hite.com

71.2

Page 2 of 2
3384Data.vp

[from manual, not included in the specification sheet]

Model 3380 Series

Section 2 - Operation

Figure 2.3 Phase Response

2.6.3 Group De lay
Group delay1 , shown in Figure 2.4, on opposite page, is defined as the derivative of radian phase with respect to radian
frequency, which is the slope of the phase curve. A flat group delay is considered a linear phase response which corresponds to a constant slope of the phase curve. With linear phase response, the distortion of complex data signals will be
minimized because their various frequency components, due to constant time delay, will not shift relative phase.
In numeric terms, the zero frequency phase slope is 293.7/Hz for Butterworth and 351.9/Hz for Bessel, when normalized for a cutoff frequency of 1Hz. This will be 2 times greater in /Hz for a cutoff of 1 radian/sec or 1845/Hz and
2211/Hz respectively. Dividing by 360 converts /Hz to radians/radians-per-sec yields a group delay time of 5.13s for
Butterworth and 6.14s for Bessel.

[1] IEEE Standard Dictionary of Electrical and Electronic Terms, Institute of Electrical and Electronic Engineers,
IEEE-STD 100-1977, Second Edition, 1977, page 296. , shown in Figur , shown in Figur

2-7

71.3

Butterworh filters using op-amps

As compared with passive filters (constructed using
and

), active filters, are most constructed using

op-amps, due to their ease of construction and better performance.

The simplest example: (a modified

Lowpass Butterworth filter using op-amp: (a) op-amp circuit; (b) frequency response (from Wheeler and
Ganji)

We will show later on that

 
G
1
 =p
 G 
1 + (2f CR2)2

 
f
& = tan1
fc

where G is the low-frequency gain (-R2/R1) (i.e. the gain

0Hz=DC).
Note G(fc)=0.707G

fc =

72

1
2CR2

For large f :

 
G
 =
 G 

a doubling of the frequency reduces the gain by a factor of

.
High-pass filter (1st-order Butterworth)

First-order Butterworth highpass filter using an op-amp: (a) op-amp circuit; (b) frequency response (from
Wheeler and Ganji)

Band-pass filter (1st-order Butterworth)

Bandpass filter using op-amp: (a) op-amp circuit; (b) frequency response (from Wheeler and Ganji)

In general, higher-order filters are required/prefered, and, obviously,

involve more complex circuitry.
73

4.1.4 Differentiation, Integration, Comparison

Op-amp circuits for (a) integration and (b) differentiation (from Wheeler and Ganji)

Op-amp comparator: (a) circuit; (b) output voltage (from Wheeler and Ganji)

See discussion at the end of Section 3.

4.1.5 Signal Conversion
Certain devices output a given type of signal (e.g. current) that
needs to be converted to another type of signal (e.g. voltage) to be
recorded. Various types of circuits can be designed to this effect.

74

4.2 Noise & Signal-to-Noise Ratio

Even a constant (DC) signal can be shown to fluctuate in time (if
measured with enough precision). Such small fluctuations in a
signal are generally characterized as noise.
Noise is generally

and can never be precisely known, nor

entirely eliminated.
The amount of noise in an electrical signal is only truly relevant
when compared to the

Hence, we define the

Types of noise:
1) Thermal or Johnson noise:

75

or

).

2) Shot noise:

Other types of noise also exist, but will not be discussed here.
Some other factors to be aware of when measuring electric signals
to be (signal-conditioned and) acquired:
1) Thermoelectric voltages:

2) Ground loops/Grounding currents:

3) Electromagnetic interferences:

4) Crosstalk:

76

4.3 Data Acquisition

Given a sensor that produces an electric signal, one either wants to
display or record the signal, for subsequent analysis (possibly after
having been conditioned, to optimize the signals quality of
usefulness).
4.3.1 Indicating & Recording Devices
If the signal (conditioned or not) is a constant, analogue voltage, it
can be read with a simple digital voltmeter (DVM).
These can vary greatly in their accuracy & precision, e.g.

(from www.sciencebuddies.org)

(from www.agilent.com)

- The DVM is generally one function of a digital multimeter (DMM)

- DVMs have very high input impedances to minimize
- DMMs can also measure other types of input signals (e.g.
,

, etc.)
77

Oscilloscopes:

(from www.directindustry.com)

(from www.tek.com)

The primary function of an oscilloscope is to

(although they also have
many other functionalities).
Some useful (secondary) functions of an oscilloscope include:

78

Strip-Chart Recorders:

(from Wheeler and Ganj)

(from www.woodleyequipment.com)

- Used to plot V(t) by moving a pen in a rectilinear motion as a

strip of paper moves past the pen at a right angle to its motion.
- Somewhat replaced by computerizes data acquisition systems.

79

4.3.2 Computerized Data Acquision (DAQ) Systems

(from Wheeler and Ganj)

(from www.ni.com)

Like other aspects of technology, the advent of increasingly small

and increasingly powerful computers has drastically affected and
improved the task of data acquisition.

80

As you know, within computers, numbers are represent in a binary

(base 2) fashion, instead of using a decimal representation (base
10), as is typically the case by people. The use of base 2 is also
relevant to computerized DAQ, and thus bears a quick review:
Decimal

Binary

# of bits

1
10
10
16
10001
10101
Components of a DAQ system
1) Multiplexers
2) Analog-to-Digital Converters (A/D converters)
3) Digital-to-Analog Converters (D/A converters)
4) Simulation Sample-and-Hold Subsystems
A brief overview of these components follows, but will principally
focus on A/D converters.
81

Multiplexers
are required when the outputs of multiple sensors are to be
recorded, given that most computers perform instructions
sequentially.

Four-channel analog multiplexer (channel 1 connected) (from Wheeler and Ganji)

A multiplexer selects one channel at a time and feeds the result to a

sample-and-hold circuit which:
i)

ii)

thus maintaining the voltage constant over a time that is long

enough for the conversion to take place.
82

Analog-to-Digital Converters
Although there are several A/D conversion methods, the A/D
converter produces a

value that

the voltage stored in the sample-and-hold circuit.

The discretized digital value is represented by a binary number.
Using a binary number with more bits will result in the digital value
.
An N-bit A/D converter will then discretize its input (analog)
voltage into one of 2N possible values.

(from Tavoularis, 2005)

See W&G p.86-90 for a discussion of different types of A/D

converters.

83

In general, any type of A/D converter can be classified according to

3 primary characterized:
1)

2)

3)

Note A/D converters discretize in

Errors in A/D conversion

Linearity, zero and sensitivity (see section 2 of notes)
Quantization error (i.e. resolution error)
- determined from the A/D converter range and # of bits
= 0.5(Vrangemax Vrangemin )/2N
- e.g. for a 5V , 16-bit AD converter, the (ideal)
resolution =
84

In reality, however, the actual resolution of an A/D converter will

be

due to

Aperture time

Thermal drift, etc.

Digital-to-Analog Converters
When acquiring data (with a computer) A/D conversion is most
common. However, when a computer is used to

a system

(measurement or otherwise), the computer output may need to be

converted to an (analog) voltage.

(from Wheeler and Ganji)

The output electrical power of a D/A converter is generally used to

control an amplifier, which will in turn operate a device.
85

Simultaneous Sample-and-Hold Subsystems

When more than one channel needs to be recorded simultaneously,
such simultaneous sample-and-hold subsystems need to be
employed.
(However, if the input signals change

relative to the time

it takes to complete all the readings sequentially, such a system

may not be required, especially since A/D converters with
sampling rates are readily available).

(from Wheeler and Ganji)

Lastly, varous configurations of DAQ systems exist, ones with

internal devices (e.g. cards in PCI slots), external devices
(connected via USB, fireqire, ethernet, wireless, etc.), etc.

86

NI 622x Specifications
Specifications listed below are typical at 25 C unless otherwise noted. Refer to the M Series User Manual for more information
about NI 622x devices.

Analog Input
Input impedance

Number of channels

Device on

NI 6220/6221.............................. 8 differential or
16 single ended

AI+ to AI GND ......................>10 G in parallel

with 100 pF

NI 6224/6229.............................. 16 differential or
32 single ended

AI to AI GND.......................>10 G in parallel
with 100 pF

NI 6225....................................... 40 differential or
80 single ended

Device off
AI+ to AI GND ......................820

ADC resolution ............................... 16 bits

AI to AI GND.......................820

DNL ................................................ No missing codes

guaranteed

Input bias current.............................100 pA

INL.................................................. Refer to the AI Absolute

Accuracy Table

Crosstalk (at 100 kHz)

Adjacent channels .......................75 dB
Non-adjacent channels ................90 dB1

Sampling rate
Maximum ................................... 250 kS/s single channel,
250 kS/s multi-channel
(aggregate)

Small signal bandwidth (3 dB)......700 kHz

Minimum .................................... No minimum

Scan list memory .............................4,095 entries

Timing accuracy ......................... 50 ppm of sample rate

Timing resolution ....................... 50 ns

Input FIFO size................................4,095 samples

Data transfers
PCI/PXI devices..........................DMA (scatter-gather),
interrupts,
programmed I/O

Input coupling ................................. DC

Input range ...................................... 10 V, 5 V,
1 V, 0.2 V

USB devices................................USB Signal Stream,

programmed I/O

Maximum working voltage for analog inputs

(signal + common mode) ................ 11 V of AI GND
CMRR (DC to 60 Hz) ..................... 92 dB

For USB-6225 devices, channel AI <0..15> crosstalk to channel AI <64..79> is 71 dB; applies to channels with 64-channel
separation, for example, AI (x) and AI (x + 64).

86.1

Typical Performance Graphs

Overvoltage protection (AI <0..79>, AI SENSE, AI SENSE 2)

Device on ....................................25 V for up to
two AI pins

Settling Error Versus Time for Different Source Impedances

Device off ...................................15 V for up to

two AI pins

Error (ppm of Step Size)

10 k

Input current during

overvoltage condition ......................20 mA max/AI pin

Settling Time for Multichannel Measurements

Accuracy, full scale step, all ranges
90 ppm of step (6 LSB) ..........4 s convert interval
30 ppm of step (2 LSB) ..........5 s convert interval

2 k

100

1 k
100

10

15 ppm of step (1 LSB) ..........7 s convert interval

10 k

5 k

1k

10
Time (s)

100

Normalized Signal Amplitude (dB)

AI <0..79> Small Signal Bandwidth

2
0
2
4
6
8
10
12
14
1k

10 k

100 k
1000 k
Frequency (Hz)

10000 k

AI <0..79> CMRR
120

0.2 V Range

CMRR (dB)

100

1 V Range

80
60
40

10 V Range
5 V Range

20
0
10

NI 622x Specifications

86.2

100

1k
10 k
Frequency (Hz)

100 k

ni.com

Analog Output

Settling time, full scale step

15 ppm (1 LSB)...............................6 s

Number of channels
NI 6220/6224.............................. 0

Slew rate ..........................................15 V/s

NI 6221/6225.............................. 2

Glitch energy

NI 6229....................................... 4

Magnitude ...................................100 mV
Duration ......................................2.6 s

DAC resolution ............................... 16 bits

DNL ................................................ 1 LSB

Calibration (AI and AO)

Monotonicity................................... 16 bit guaranteed

Recommended warm-up time .........15 minutes

Maximum update rate

Calibration interval..........................1 year

1 channel..................................... 833 kS/s

2 channels ................................... 740 kS/s per channel
3 channels ................................... 666 kS/s per channel
4 channels ................................... 625 kS/s per channel
Timing accuracy ............................. 50 ppm of sample rate
Timing resolution............................ 50 ns
Output range ................................... 10 V
Output coupling .............................. DC
Output impedance ........................... 0.2
Output current drive........................ 5 mA
Overdrive protection ....................... 25 V
Overdrive current ............................ 10 mA
Power-on state................................. 20 mV1
Power-off glitch .............................. 400 mV for 200 ms
Output FIFO size ............................ 8,191 samples shared
among channels used
Data transfers
PCI/PXI devices ......................... DMA (scatter-gather),
interrupts,
programmed I/O
USB devices ............................... USB Signal Stream,
programmed I/O
AO waveform modes:
Non-periodic waveform
Periodic waveform regeneration mode from onboard FIFO
Periodic waveform regeneration from host buffer including
dynamic update

For all USB-6221/6229 Screw Terminal devices, when powered on, the analog output signal is not defined until after USB
configuration is complete.

National Instruments Corporation

86.3

NI 622x Specifications

NI 622x Specifications

AI Absolute Accuracy Table

Nominal Range

Reference
Tempco

Residual
Offset
Error
(ppm of
Range)

Offset
Tempco
(ppm of
Range/C)

INL Error
(ppm of
Range)

Random
Noise,
(Vrms)

Absolute
Accuracy
at Full
Scale1 (V)

Sensitivity2
(V)

Positive
Full Scale

Negative
Full Scale

Residual
Gain Error
(ppm of
Reading)

10

10

75

25

20

57

76

244

3,100

97.6

85

25

20

60

76

122

1,620

48.8

95

25

25

79

76

30

360

12.0

0.2

0.2

135

25

80

175

76

13

112

5.2

Gain
Tempco
(ppm/C)

AbsoluteAccuracy = Reading (GainError) + Range (OffsetError) + NoiseUncertainty

GainError = ResidualAIGainError + GainTempco (TempChangeFromLastInternalCal) + ReferenceTempco (TempChangeFromLastExternalCal)
OffsetError = ResidualAIOffsetError + OffsetTempco (TempChangeFromLastInternalCal) + INL_Error
3
NoiseUncertainty = RandomNoise
----------------------------------------For a coverage factor of 3 and averaging 100 points.
100
4

Absolute accuracy at full scale on the analog input channels is determined using the following assumptions:
TempChangeFromLastExternalCal = 10 C
TempChangeFromLastInternalCal = 1 C
number_of_readings = 100
CoverageFactor = 3
For example, on the 10 V range, the absolute accuracy at full scale is as follows:
GainError = 75 ppm + 25 ppm 1 + 5 ppm 10
GainError = 150 ppm
OffsetError = 20 ppm + 57 ppm 1 + 76 ppm
OffsetError = 153 ppm
244 V 3
NoiseUncertainty = -------------------------100

NoiseUncertainty = 73 V

AbsoluteAccuracy = 10 V (GainError) + 10 V (OffsetError) + NoiseUncertainty AbsoluteAccuracy = 3,100 V

2

Sensitivity is the smallest voltage change that can be detected. It is a function of noise.

Accuracies listed are valid for up to one year from the device external calibration.

ni.com

86.4

 National Instruments Corporation

AO Absolute Accuracy Table

Nominal Range
Positive
Full Scale

Negative
Full Scale

Residual Gain
Error (ppm of
Reading)

10

10

90

Gain Tempco
(ppm/C)
10

Reference
Tempco

Residual
Offset Error
(ppm of
Range)

Offset Tempco
(ppm of
Range/C)

INL Error
(ppm of
Range)

Absolute
Accuracy at
Full Scale1
(V)

40

128

3,230

Absolute Accuracy at full scale numbers is valid immediately following internal calibration and assumes the device is operating within 10 C of the last external calibration.
Accuracies listed are valid for up to one year from the device external calibration.
AbsoluteAccuracy = OutputValue (GainError) + Range (OffsetError)
GainError = ResidualGainError + GainTempco (TempChangeFromLastInternalCal) + ReferenceTempco (TempChangeFromLastExternalCal)
OffsetError = ResidualOffsetError + AOOffsetTempco (TempChangeFromLastInternalCal) + INL_Error

5
NI 622x Specifications

86.5