Differential Readings (App Note)

Differential vs. Single-Ended

In this discussion, a voltage is the difference in electric potential between 2 points.  For a single-ended voltage reading, 1 point is an analog input terminal, while the other point is the common ground (GND).

For a differential voltage reading, the 2 points are 2 analog input terminals as shown in Figure 2.

Differential vs. Bipolar

Note that differential is not the same as bipolar, and they do not necessarily have anything to do with each other (but sometimes do).  Bipolar refers to a voltage that can be positive or negative, compared to unipolar which refers to a voltage which is positive only.  We use the term bipolar or true bipolar to describe a point that can be greater than or less than ground.  We use the term pseudo-bipolar to describe a voltage where the positive point can be greater than or less than the negative point (thus the difference is positive or negative), but neither point can be less than ground.

Differential vs. Isolated

Differential does not define anything about isolation. For example, assume you have a 90V battery pack built out of 10x 9-volt batteries in series. You might be tempted to use differential inputs to measure the voltage of each cell, but this will not work if your inputs are not isolated. You have to define ground somewhere and it will have large common-mode voltages when compared to LabJack GND. Conversely with an isolated input, you can define ground anywhere and therefore measure the difference between very high voltages as long as the difference between the voltages is still within the LabJack measurement range.

Bipolar vs. Pseudobipolar

Bipolar refers to a signal that can be plus and minus versus ground.  Pseudobipolar refers to a differential signal where the voltage difference (positive - negative) can be plus or minus, but neither the positive lead or negative lead can be minus versus ground.

Take a differential voltage of -2.2 volts.  For pseudobipolar inputs, a valid way to get this voltage is if the positive lead is at 0.2V versus ground and the negative lead is at 2.4V versus ground.  For bipolar inputs, that same scenario is valid, and it is also valid if the positive lead is at -2.4V and the negative lead is at -0.2.

Why use differential?

Reasons #1 & #2 are key reasons for differential measurements.  Reason #3 is good in theory for certain situations, but most of the time single-ended measurement performs as well as differential.

1. The signal is differential and the negative cannot be connected to GND:  For example, consider a DAQ monitoring a typical Wheatstone bridge circuit that is excited by 4V/GND from the DAQ and is outputting a 2 mV signal, which means that signal+ is about 2.001V and signal- is about 1.999V.  You cannot connect signal- to GND, because that would short out 1 leg of the bridge, so you must connect signal+ and signal- both to analog inputs and do 2 single-ended measurements or 1 differential measurement.  If the bridge was excited by a floating source (not referred to the DAQ device), you can define the common ground wherever you want so you could connect signal- to GND and just do a single-ended measurement of signal+.

2. Measuring a small difference between 2 large voltages:  For example, consider a DAQ with a simple 1% accuracy spec.  This DAQ is monitoring a typical Wheatstone bridge circuit that is excited by 4V/GND from the DAQ and is outputting a 2 mV signal, which means that signal+ is about 2.001V and signal- is about 1.999V.  You could take single-ended readings of signal+ and signal-, and subtract them in software to find the difference, but the 1% error of each single-ended measurement is about 20 mV which is very large compared to the 2 mV difference you are trying to measure.  A direct differential measurement of the 2 mV difference with the same 1% error has just 0.02 mV of error.

3. Rejection of common-mode noise:  Say you have signal+ and signal- coming from a floating AA battery through a long 2-wire cable, and you expect the long cable to pick up a lot of AC noise where the induced noise is the same on both wires.  Since the battery is floating, you could connect the wires to 2 differential analog inputs on the DAQ device, and then add a high-value resistor from the negative analog input to GND to provide a path for bias currents.  Since the noise at any point in time is the same on both wires, it will get subtracted out by a differential measurement.  Alternatively you could connect signal+ to an analog input and signal- to GND, for a single-ended measurement.  In theory, the wire connected to ground can't have noise because it is connected to ground, so you just have noise on the positive wire which all shows up in your measurement, but in reality it does not usually work exactly that way and you don't see 100% of the noise.  Rejection of common-mode noise is the main reason someone uses a differential connection when it is not required, but few systems see improvement due to this theoretical advantage, or more often than not the added complications lead to more problems than improvements.

Differential inputs must have a reference

The most common mistake when using differential inputs is connecting 2 signals that have no reference to the difference amplifier ground.

Consider an obviously floating voltage source such as a thermocouple or AA battery.  If you simply connect the positive and negative leads to 2 analog inputs on a U6/T7, or to IN+ and IN- on an LJTick-InAmp, there is no ground path for the bias currents that must flow in/out of the inputs.  The voltage source will try to properly hold the voltage difference between the leads, but the voltage of each lead compared to ground will likely be near one of the power rails and the common-mode voltage will not be valid.  A common solution is a resistor from the negative terminal to ground, which can be quite large if desired.  A typical resistor used with the U6/T7 would be 100k (e.g. CF14JT100K).

Another example is a bridge circuit excited by an external supply which is isolated from the U6 or LJTick-InAmp.  In this case the negative from the supply should be connected to GND (a series resistor can be considered if you don't want a direct connection between the supply ground and GND). Figure 2 a shows a broad overview of how a bridge circuit should be connected for differential measurements.

The common-mode voltage must be in range

Another common mistake is connecting voltages that are referenced to ground, but where the voltages compared to ground are not in the valid range.

For example, the LJTick-InAmp uses a pair of AD623 instrumentation amplifiers from Analog Devices with power rails at VS (~5 volts) and GND (0 volts).  Figures 22 and 23 of the AD623 datasheet show the common-mode range.  Note that the maximum under any condition is about 3.5 volts and the minimum is about -0.3 volts.  Signals with a common-mode voltage outside -0.3 to +3.5 volts will definitely not work, and for signals inside that range we recommend looking at the LJTIA signal range tables or online calculator from  Appendix A of the LJTick-InAmp Datasheet.

Say you have a 12 volt battery system where the battery negative is connected to LabJack/LJTIA GND.  You want to measure the current the battery is providing to some load, so you put a high-side shunt between the positive battery terminal and the load.  The shunt is providing a 100 mV signal, so the voltage compared to ground on each side of the shunt is 12.0 and 11.9 volts, and thus the common-mode voltage is 11.95 volts.  This is definitely too high for the LJTIA.  However, if a low-side shunt is used instead between the negative battery terminal and the load, the common-mode voltage is only 0.05 volts and the LJTIA is fine.

For signal ranges:

• U6: see Section 2.6.5 of the U6 User's Guide

• T4 or T7: see Appendix A-3-2-3 of the T-Series Datasheet.

Why don't I worry about ground when I measure voltages with a simple DMM?

So why can you just take the 2 leads from a simple battery-powered DMM and measure the voltage across a battery or thermocouple, regardless of what grounds might or might not be connected?  Because the DMM is isolated and is actually taking a single-ended reading.  The black lead is ground for the DMM, but since it is isolated that ground has no meaning to the battery or thermocouple, and wherever the black lead connects is defined as ground for the DMM.

How about a fancier DMM with 2 channels, and 2 pairs of red/black leads, powered by AC mains?  First the input channels are isolated (optically or galvanically) from AC mains, so there is no common ground there, and the input channels are also isolated from each other, so the black leads are ground for each channel but not the same.  Each black lead defines ground for each channel.

Measuring the voltage of each power supply with multiple supplies in series

This is covered by the above information, but is a common enough application to warrant its own section here.

Say you have a system-under-test (SUT) consisting of 10x floating 9-volt sources (batteries, isolated power supplies) all connected in series.  You define your common reference between the T7 and SUT with a connection from the low point of the SUT (negative of 1st battery) to T7-GND.  Thus the voltage of the SUT is 0 to 90 volts compared to T7 ground.  If you then connect AIN0/1 across the +/- terminals of the 10th supply, to do a differential measurement of its voltage, those connections would be about 90V/81V versus GND which violates the common-mode limits of the LabJack analog inputs (you want to be more like ±10V per signal range links above).

In an attempt to reduce the extreme common mode voltage, you could instead define your common reference between the T7 and SUT with a connection from the mid point of the SUT (between 3rd and 4th sources) to T7-GND, but you still have ±45 volts which is too much. 

Some Solutions:

1. Individual channel-to-channel isolation.  If the analog inputs have channel-to-channel isolation, you can connect the pair to 90V/81V and effectively have very little common-mode.  We don't have any devices with this feature at this time (written July 2019 ... watch for something within a year!), so you would have to use some sort of isolation amp on each.  One possibility is a 5B module from Dataforth.   When you have individual channel-to-channel isolation, you can think of it like using a bunch of DMMs where each DMM is connected across 1 of the supplies.

2. Use 1 LabJack for each differential measurement and keep the LabJacks isolated from each other (easy to do over Ethernet or WiFi).

3. Divide each signal down (e.g. LJTick-Dividers) before measuring, do single-ended measurements, and calculate the differentials in software.  This can have the problem #2 under "Why use differential" above, but by doing your own calibration on each signal you can improve things greatly.

4. Build an amp that handles high-common mode.  TI and others make nice simple chips that are good at this without requiring isolation.

Testing differential inputs

The T4 does not support differential measurements.

We always recommend testing with all external connections removed from the LabJack, and using our Kipling software (for T-series devices) or the Test panel in LJControlPanel (for UD devices).

To test differential inputs, securely clamp a wire between the differential AIN pair and VS or GND.  For example:

AIN0=VS,      AIN1=GND:    5 volts (2.4V on the U3*)
AIN0=VS,      AIN1=VS:       0 volts
AIN0=GND,  AIN1=VS:      -5 volts​​​​​​​ (-2.4V on the U3*)
AIN0=GND,  AIN1=GND:    0 volts

This simple test also works for other analog input ranges, the 5 V or -5 V readings will just hit the output limit of the range. This will not cause damage to the inputs.  For example, if your device analog input is configured to use a range of ±0.1 V then 5V and -5V will produce readings of 0.1 V and -0.1V, respectively.

As described earlier in this app note, isolated inputs are not quite the same as differential inputs. However, if your device has isolated inputs you can perform the tests above between AIN+ and AIN- terminals.

^{* Only the flexible IO on the U3 are differential and they have a 0-2.4V input range. The high voltage AIN0-3 on the U3-HV are inherently single-ended. See the U3 Datasheet for further information.}

Common amplifier types

Operational Amplifier (op-amp):  Single-ended input and output.

Instrumentation Amplifier (in-amp):  Differential input and single-ended output.

Difference Amplifier (diff-amp):  Differential input and output.