Capgo datalogger, data logger, datalogging and data logging.




Understanding data acquisition specifications

This page is intended to help you understand the technical specifications of measuring equipment. Frequently, specification fail to tell the whole truth and are sometime (deliberately??) quite misleading. First, it is important that you understand your requirements, and then how they match equipment specifications. If a particular specification that is important to you is not adequately covered for a piece of equipment, it is most likely the equipment is not suitable. Most reputable manufactures provide very comprehensive specifications for measuring equipment.

A summary of the important specifications is presented in typical order of importance in the following table:

Specification Poor OK Good
Accuracy at 25°C
Accuracy over 0°C to 50°C
Resolution - voltage
50 µV
5 µV
1 µV
Resolution - bits
80 db
95 db
110 db
50/60 Hz Rejection
10 db
40 db
50 db
Common Mode Range
± 1V
± 10V
5 Hz
100 Hz
50 kHz
Settling Time (to 0.01%)
100 ms
5 ms
10 µS
Temperature Range
15 °C to 30 °C
0°C to 50 °C
-45°C to 70°C
Input Bias Current
1 µA
10 nA
100 pA
Input Impedance
100 kOhms
10 MOhms
100 MOhms
Channel Cross Talk
60 db
80 db
110 db
RF Susceptibility
50 V/m
500 V/m
2 kV/m
Input Noise
10 µVpk-pk
3 µVpk-pk
1 µVpk-pk

Old hands at measurement will dispute aspects of the summary, saying it is far too simplistic to generalize in such a way. While this is a valid point, the summary is a starting point to the following discussion.


Accuracy can mean different things to different people. Accuracy can also be specified in many different ways. There are many sources of error that combine to reduce total system accuracy. Most commonly, accuracy is specified as a percentage of full scale at 20°C or 25°C. In practice there are five primary error sources:

    • Scale calibration error
    • Nonlinearity errors
    • Zero or Offset error
    • Temperature caused drifts
    • Aging related calibration drifts

Some manufactures like to group these into a single system accuracy figure. This is ok if the assumptions are valid. For example 0.1% accuracy at 25°C says nothing about the equipment's performance at 15°C, but if the equipment is to be operated in an air conditioned room, then the simple specification may be ok. A more useful specification would be 0.13% of full scale over 10°C to 40°C temperature range.


Firstly, resolution has almost no relation to accuracy. Resolution is a measure of a system's ability to resolve two different but very similar inputs levels. The figure may be expressed in units of the parameter being measured (e.g. °C), in units of the measuring device's input at one or more gains (e.g. µV), as a percentage of full scale or in terms of the number of bits in the measuring device's analog to digital converter binary output word. The latter has become the most common and is a good all round indicator and is simple to determine.

Different systems have a different relationship between each type of resolution specification depending on sensor sensitivity, noise levels, amplifier gain, ADC span and word size and sample duration. The following table relates the various resolution figure for several particular full scale conditions and a commonly used sensor:

ADC Bits % Full Scale 5V Full Scale 0.5 V Full Scale 50 mV Full Scale Type T Thermocouple
(50 mV Scale)
6.3 %
310 mV
31 mV
3 mV
80 °C
1.6 %
78 mV
8 mV
1 mV
20.0 °C
0.4 %
20 mV
2 mV
200 µV
5.0 °C
0.10 %
5 mV
490 µV
49 µV
1.3 °C
0.02 %
1 mV
122 µV
12 µV
0.3 °C
0.012 %
610 µV
61 µV
6 µV
0.16 °C
0.006 %
310 µV
31 µV
3 µV
0.08 °C
0.003 %
150 µV
15 µV
2 µV
0.039 °C
0.0015 %
76 µV
8 µV
1 µV
0.020 °C
0.0004 %
19 µV
2 µV
0.2 µV
0.005 °C
0.00019 %
10 µV
1.0 µV
0.10 µV
0.0024 °C
0.00010 %
5 µV
0.5 µV
0.05 µV
0.0012 °C
0.000024 %
1.2 µV
0.12 µV
0.012 µV
0.0003 °C
0.000006 %
0.3 µV
0.03 µV
0.003 µV
0.0001 °C

The table ignores the contribution of noise in limiting the resolution. The effect of noise on resolution varies greatly, but generally only has an impact on high gain or high resolution (>18 bit) systems. On of the most troublesome sources of equipment noise is that known as transistor flicker noise generated in the input stage of an amplifier. This will cause two sequential reading on a steady input to be different, thus creating an uncertainly in measurement. If it is allowable to average readings over a period of time (a form of filtering) this noise can be reduced, but such averaging is not necessary possible due to the speed requirements of the system. See the page on noise and the section on input noise specifications below.

Another important question you need to ask is does the equipment have input full scale input ranges that match the sensor's likely output range? Measuring a thermocouple (typically 10 mV to 50 mV full scale output) on a 1 volt range is not likely to yield satisfactory results unless the equipment has very high resolution and a stable input offset.

Common Mode Rejection Ratio (CMRR)

The Common Mode Rejection Ratio is a misunderstood specification. It only applies to differential inputs and is good measure of the quality of a measuring systems input electronics. Ideally the figure should be very high 120 db or better.

The simplest definition of CMRR is:

CMRR = 20 log(differential gain / common mode gain)

where the gains are normally voltage gains. An example will illustrate the importance of the specification.

Assume we are measuring the temperature of a number of lithium cells in a battery. A Type T thermocouple is spot welded to the metal anode of each cell. Consider just two adjacent cells - the thermocouples will have about a 3V common mode voltage difference. What temperature error will this induce for various CMRR's?

Input Voltage Error
(VCM = 3V)
Temperature Reading
60 db
3000 µV
77 °C
70 db
949 µV
24 °C
80 db
300 µV
7.7 °C
90 db
95 µV
2.4 °C
100 db
30 µV
0.8 °C
110 db
9.5 µV
0.2 °C
120 db
3.0 µV
0.08 °C
130 db
0.9 µV
0.02 °C
140 db
0.3 µV
0.01 °C

Clearly, for an error of less than say 0.5°C, a CMRR of 105 db is required - and this is for just a 3V common mode voltage. If the battery's full 12V were allowed for, then the CMRR would need to be closer to 120 db.

Now it is possible to compensate for a less than perfect CMRR at the expense of other resources. If a zero reading is taken at the same common mode voltage as the measurement of interest and subtracted from this reading, then the impact of CMRR can removed. Thus a low CMRR (say 70 db) can be tolerated if there is capacity available for extra zero reading channels.

A complication to the CMRR specification is the frequency response of the rejection. Typically the CMRR falls away rapidly as the common mode voltage's frequency increases. The fall away is usually at a rate of 20 db per decade and often begins in the low 5 to 100 Hz range. The CMRR at 60 Hz (line frequency) is the most useful figure.

Hum rejection (50/60 Hz rejection)

One of the most common types of noise pickup is inductively or capacitively coupled from mains or line operated machinery and power line associated with them. The line frequency is 60 Hz in North America and 50 Hz in most other parts of the world. The induced noise therefore has a fundamental at 50 or 60 Hz but may also be rich in harmonics.

inset picture

The induced noise may present as a common mode signal - in which case the frequency characteristics of the common mode rejection are important. In addition, the noise may be superimposed on the sensor signal (or in series) in which case it needs to be filtered out of the signal path and the Series Mode Rejection figure become important.

Effective series mode rejection is only possible in relative slow systems where the effective sampling period is at lease one 50 or 60 Hz cycle period. Using an integrating ADCs such as multi-slope units or voltage controlled oscillators and sampling over an integral number of 50 or 60 Hz cycles, excellent rejection is possible. The technique will easily yield 40 db and with careful design 60-70 db is possible. It is common to use a 100 ms (or multiple of) period as it effectively deals with both 50 and 60 Hz frequencies by sampling over 5 or 6 cycles respectively.

For special applications it is possible to filter out a 50 or 60 Hz noise by injecting a signal with equal magnitude and opposite phase. This approach has been applied to electrocardiographs and is successful when the noise is clean and relatively constant, however it is not used in more general purpose instruments.

Input common mode range

Input common mode range is one of those things that is neglected until you find you don't have enough of it! More the better.

There are three aspects to the Common Mode Range specification. The first is related to the full operational range on an input channel, beyond which measurement errors are incurred due to limiting effects. The second aspect relates to the the input range over which a channel may range without impacting other channels than may still be within the full operational range. The Third common mode range refers to the safe range that the inputs may be exposed, beyond which failure of components may occur. An application example will help:

Consider a UPS (Uninterruptible Power Supply) battery. For large installations 200 cells may be placed in series. Assume each cell has a terminal voltage of 2V and that one end of the battery bank is grounded. Therefore the last cell in the bank will be 398V above ground. To measure the voltage of this cell, the measuring equipment will need to tolerated almost 400V on the measuring leads while measuring just 2V. In other words, this 400V is common to both measuring leads and is referred to as the common mode voltage. (More precisely, the common mode voltage is the average to the voltage on the two leads i.e. (398+400)/2 = 399V).

Single channel systems (e.g. a multimeter) can often have a very high common mode voltage range as the entire equipment can be operated at the common mode voltage due to its electrical isolation. With multichannel systems, the common mode voltages must be presented to the equipment simultaneously and therefore must be electrically isolated within the equipment. This isolation is provided by the input selector switch or multiplexer. Mechanical relays are typically used for common mode ranges above 30V and semiconductor multiplexers for voltages below 30V.

Speed and settling time

Your speed requirements need to be understood. Speed can refer to a number of related but different specifications. The rate at which an input is sampled, processed and then output is probably the most useful specification It is commonly refereed to as throughput. A slow system may have a throughput of less then 5 readings per second and a fast one 500,000 reading per second. Special purpose systems may achieve 2,000,000,000 samples per second (2 GHz).

Acquisition speed refers to the time a system takes to acquire a reading or the actual duration of time that the signal is sampled to gain a reading. A slow system may take 200 ms or longer while a fast system using a sample and hold circuit may take less than a nanosecond. Again special purpose systems may be faster.

Sometimes a system may utilize a high speed front end and ADC but the throughput can be relatively low. This may because the raw input data is processed (filtered) and reduced to compensate for errors in software.

While it is always possible to slow a fast system, it is generally not possible to speed up a slow system. This may lead one to think it is safer to go for speed, however there is often sacrifice in other specifications as the speed is increased.

Temperature range

Most measuring devices will operate correctly only over a limited temperature range. This limitation is due to fundamental properties of the components from which they are constructed. For example:

Component Temperature range
Alkaline batteries
-5°C to 50°C
LCD displays
-5°C to 50°C
-45° to 80°C
-15 to 70°C

In addition to function limits associated with temperature, variations in temperature typically cause measuring devices to change their reading. Most commonly the input zero will change, however scale calibration and linearity can also change. The magnitude of this effect is normally express in part per million per degree Celsius (ppm/°C).

Input bias current and input impedance

All amplifiers require an input bias current. This is a small current that must be able to flow into or out off an amplifier's input for it to function properly. This current is supplied by either the sensor or input bias resistors. Sometimes the input bias current can be so small that it can be ignored, however in lower cost systems provision must generally be provided.

Input bias current and input impedance are related but different specifications. Input impedance is a measure of how much the input current of an amplifier changes when the input voltage is changed. Using Ohms law the input impedance is calculated Z = V/I. Frequently the input bias current will change very little with a change of input voltage, resulting in a high input impedance.

Channel cross-talk

A signal on one channel of a multichannels measuring device can interfere with another channel. This is especially true for high frequency inputs, but it is also possible with high level DC signals where current leakage can occur. Cross talk is usually specified in decibels:

CrossTalk = 20 log (Vin / Vint) in db

where: Vin is the magnitude of the unwanted signal input equivalent, and Vint is the interfering channel signal magnitude. The impact is similar to common mode rejection.

RF susceptibility

Sensor leads can act as radio antenna, and unless the measuring device has an inbuilt filter to reject radio frequency energy, errors will be introduced. Some solid state amplifiers can be particularly sensitive to radio frequency energy over a wide 1 MHz to 1 GHz band. This is because of parasitic diodes (a side effect of the manufacturing process) rectifying the signal to form a DC offset.

Input noise

When low level signal are to be measured, it is important that a measuring devices input amplifier be of a low noise type. The difference between a low noise and a noisy amplifier can be a factor of 20. While noise can be suppressed by filtering, the cost is slower response to obtain the same signal to noise ratio.