A floating source with no bias path drives a differential input into erratic or full-scale readings. Match the coupling, the source impedance, and the isolation to one published row, then copy the resistor and the common-mode limit from the device manual. Compiled from public sources by an independent publication (about this site). This site is not affiliated with National Instruments (NI) or Emerson.
Match the bias path to the coupling, the source impedance, and whether the source is isolated. National Instruments’ field-wiring note, updated 15 August 2024, says a floating source on a differential input needs a DC path for amplifier bias current. A missing path produces erratic or full-scale readings. For a low-impedance DC-coupled source that note uses one resistor from the negative input to measurement ground, and two equal resistors when the input is AC-coupled or the source impedance is high. LabJack’s differential note treats channel isolation as its own property, so each lead versus ground still has to sit inside that device’s common-mode range. Acromag’s 1 March 2017 article and Andy aka’s 23 January 2018 answer tie the cable shield at different ends. Copy the resistor and the common-mode limit from this input’s manual. Your ohms cell stays blank until the device manual states a value.
| Field | What you copy | Your job |
|---|---|---|
| Coupling | DC-coupled or AC-coupled | |
| Source impedance | Ohms, or the low/high class in NI’s transducer table | |
| Earth at the source | Floating, or tied to building ground | |
| Channel isolation | The manual’s words for this input | |
| Bias resistor | Ohms, count, and which input, from that manual | |
| Common-mode range | The limit, and the reference node the manual uses | |
| V+ and V− | Volts of each lead versus that same node | |
| Shield end | Which end is tied, and to what |
Your ohms cell stays blank until the device manual states a value.
Which bias path matches this source?
NI’s note, updated 15 August 2024, defines a floating source as one whose voltage is not referred to earth or building ground. Its examples are batteries, battery-powered sources, thermocouples, transformers, isolation amplifiers, and instruments that float their output. On a differential input, bias current from the amplifier can walk that source outside the input range. The note’s bias resistors are the DC path from the amplifier inputs to measurement-system ground.
The count depends on coupling and source impedance, in that note:
| Condition on the NI note | Resistor path that note describes | NI’s stated band for that path | Your count | Your ohms |
|---|---|---|---|---|
| DC-coupled, low source impedance | One resistor, from the negative input to measurement-system ground. The note names thermocouples as this class. | Typically 10 kΩ to 100 kΩ for low-impedance sources such as thermocouples and signal-conditioning outputs | ||
| DC-coupled, source impedance relatively high | Two equal resistors, positive input and negative input, each to measurement-system ground, so the impedances to ground match | The figure’s band is greater than 10 kΩ and under 100 kΩ | ||
| AC-coupled | Two bias resistors, and the figure sets R1 = R2 | Greater than 10 kΩ and under 100 kΩ |
A missing path, the note warns, produces erratic readings or saturation at positive full scale or negative full scale. In NRSE mode the same note puts the bias resistor or resistors between AI SENSE and AI GND.
NI’s transducer table is the low/high split used in the rows above. It is a class label. It does not fill the ohms cell.
| Transducer, as NI names it | Impedance characteristic in that table |
|---|---|
| Thermocouples | Low (under 20 ohm) |
| Thermistors | High (over 1 kohm) |
| Resistance temperature detector | Low (under 1 kohm) |
| Solid-state pressure transducer | High (over 1 kohm) |
| Strain gauges | Low (under 1 kohm) |
| Glass pH electrode | Very high (1 Gohm) |
| Potentiometer, linear displacement | High (500 ohm to 100 kohm) |
The same note says high-impedance, low-level sensor outputs should be processed by signal conditioning near the sensor. In one setup in that note the source impedance was 1 kΩ in both arrangements, and the balanced arrangement used 100 kΩ on each input. That pair loads the source with the sum of the two resistors. The note says a source impedance of 0 Ω makes the single-resistor arrangement balanced as well. Those ohms belong to that setup.
LabJack’s differential app note, which shows no publication date in the 11 October 2026 archive, uses the same idea for a thermocouple or an AA battery on a U6, a T7, or an LJTick-InAmp: with only the two signal leads connected, bias current has no path to ground, each lead can sit near a power rail, and the common-mode voltage is then invalid. The note’s usual fix is a resistor from the negative terminal to ground. It says the resistor can be quite large. A typical resistor it names for the U6 and T7 is 100 kΩ, example part CF14JT100K. That part is their example for those devices.
Andy aka’s answer on the Stack Exchange question, posted 23 January 2018, is for several floating sources at different distances. He keeps the resistors on each IN−, individually, to the DAQ earth point. He sets 1 MΩ as the highest value he would use. A high value that still holds the common-mode voltage can, in that answer, be a single resistor. A lower value is for the case with a lot of EMI, so the common-mode extremes stay smaller. A value low enough to attenuate the signal as a divider is the one he tells the reader to avoid. His 1 MΩ ceiling is that answer’s ceiling.
Common mistake: Copying 10 kΩ, 100 kΩ, or 1 MΩ into the ohms cell from a row whose device and premise are someone else’s.
Takeaway: Pick the row whose coupling and impedance match the source, then leave the ohms cell blank until this input’s manual states a value.
Where do the signal pair, the shield, and the bias return land?
The signal pair, the shield, and the bias return are three conductors. The pair carries the voltage being measured. The bias return is the resistor path in the section above. The shield is the noise conductor, and the pages opened here tie it at different ends.
The question that prompted this sheet was asked on 18 January 2018. The asker had DC-coupled floating sources, mostly DC and in some cases up to about 10 Hz, about 10 m from the board and about 10 m from each other, and four diagrams that disagreed. On 20 January 2018 a comment said case 3 had no DC path for bias current into the two amplifier pins. The question had labeled case 3 as the NI bias-resistor figure. NI’s note says those resistors are the DC path.
| Conductor | NI, updated 15 August 2024 | Acromag, Process Online, 1 March 2017 | Andy aka, 23 January 2018 | Your tie |
|---|---|---|---|---|
| Signal pair | Differential positive and negative inputs | Signal high, signal low, and a return as the third connection | The differential input pair. His resistor line names IN− | |
| Bias return | Resistors to measurement-system ground. NRSE: between AI SENSE and AI GND | For a source that is not already grounded, their instruction connects IN− to return and connects input return to earth. They say that tie makes the differential input single-ended, and that applying it to every channel stops the channels from measuring offset from return | Each IN− through its own resistor to DAQ earth | |
| Shield | For capacitive shielding, ground the shield at the source end only. Ground at both ends carries shield current. Their example is 1 V across a 0.5 Ω shield, which is 2 A | For an earth-grounded single-ended source, tie the shield to the negative lead and to earth at the source, and to the return and to earth at the input. They say that double earth connection reduced noise. Figure 3b in that article is an isolated source with no direct connection to earth | Bond the shields at a proper DAQ earth, which they distinguish from signal ground unless the manual says signal ground may be used for EMC. Leave the sources off the shield |
NI also says conductive metal near the signal path needs an electrical connection, because metal left floating can raise capacitively coupled noise. Twisted pair, or shielded twisted pair, is NI’s balanced cable. Coaxial cable is unbalanced on that page because the two conductors have different capacitance to ground. Acromag’s article says shielded twisted pair is the wiring that couples noise onto both leads so the amplifier can treat it as common mode. The cable type does not choose the shield end. The shield row above does.
If a source is specifically required to be grounded, Andy aka’s answer uses a capacitor of a few nanofarads to a few tens of nanofarads at the source for an EMC ground. If the source must be galvanically grounded, that answer moves the capacitor to the shield at the DAQ, in the same range. He also says to use one ground unless the consequences of a second ground are already known, and that common-mode chokes may be required at the DAQ. Leave any capacitor value blank until the manual for this source states one.
His separate worked example is an SMPS that puts 50 Hz on both output wires. He gives source-end capacitors of 10 nF to 10 µF, because many supplies have about 1 nF from the DC output back to the rectified bus. He states an RMS current of about 100 µA, and he calculates 50 µA through 10 kΩ as 0.5 V RMS of common mode. Capacitors at the DAQ end, in the same answer, cut bandwidth: 100 Ω with 10 µF at 160 Hz, and 100 Ω with 1 µF at 1600 Hz. Those figures stay in his example. The ohms cell is still the number in this input’s manual.
Takeaway: Write the shield end that matches the row’s premise. A shield tied at the source, at the DAQ earth, or at both ends is three different published premises.
What common-mode range does the manual allow?
Copy the limit, and the reference node that limit uses, from the device manual. NI defines common-mode voltage as Vcm = (V+ + V−) / 2, with V+ and V− each measured against measurement-system ground. Acromag defines the common-mode range as the maximum allowable swing of each input with respect to the signal return. A lead outside that window is an out-of-range input on both pages.
NI says violating the range produces measurement error and can damage components on the device. The same note says CMRR in decibels is 20 log (differential gain / common-mode gain), that CMRR falls as frequency rises, and that most data-acquisition devices specify CMRR up to 60 Hz. Acromag says to look for 100 dB or better, equates 100 dB with 100,000:1, and says common-mode rejection usually starts to roll off above 100 kHz. Those are their stated targets. The CMRR cell for this input stays blank until the manual states it.
Acromag also says an earth connection on the differential return keeps high-impedance inputs from floating in strong EMI, and gives the input a low-impedance path for transient energy. The same article says a second earth on that return can form a ground loop. Which of those two sentences applies is the shield-and-return row you already matched.
LabJack’s examples are for the devices that note names:
| LabJack example in that note | What the note states | Your device, if it is this one |
|---|---|---|
| LJTick-InAmp, AD623, datasheet figures 22 and 23 | Maximum under any condition about 3.5 V, minimum about −0.3 V. Leads outside −0.3 V to +3.5 V will not work on that module | |
| 12 V battery, negative on LabJack/LJTIA ground, high-side shunt of 100 mV | Leads at 12.0 V and 11.9 V, common mode 11.95 V, which the note says is too high for the LJTIA. A low-side shunt in the same example has a common mode of 0.05 V, which the note says is fine for that module | |
| Ten floating 9 V sources in series, low end on T7 ground, pair across the tenth source | About 90 V and 81 V versus T7 ground. The note says that violates the common-mode limits of the LabJack analog inputs, and that the reader wants to be more like ±10 V per the signal-range links in that note | |
| U6 and T4/T7 range links | U6 readers are pointed to section 2.6.5 of the U6 user’s guide. T4 and T7 readers are pointed to appendix A-3-2-3 of the T-series datasheet. Those guides were not the text used here |
The ±10 V phrase stays with that T7 series-string example. It does not fill the common-mode cell for a different device.
A battery-powered DMM can read a floating battery or thermocouple, on that LabJack note, because the meter is isolated and the black lead defines ground for the meter. A DAQ channel that shares the instrument ground still uses the bias path and the common-mode window above. The same note says the T4 does not support differential measurements.
A bridge excited from the DAQ is outside the floating-source rows. LabJack’s example is a Wheatstone bridge on 4 V and GND from the DAQ, with a 2 mV output, so the leads sit at about 2.001 V and about 1.999 V. The note says connecting the negative lead to GND would short one leg of that bridge. If the excitation itself is floating, the same note says the common ground can be defined at the DAQ and the positive lead read single-ended.
| Record | How you fill it | Your entry |
|---|---|---|
| Reference node the manual uses | Measurement-system ground, signal return, or the name printed in the manual | |
| Common-mode limit | Manual, with unit and the node above | |
| V+ versus that node | ||
| V− versus that node | ||
| Vcm = (V+ + V−) / 2 | Only after both voltage cells have numbers | |
| CMRR | Manual, with the frequency it states |
Takeaway: Fill V+ and V− against the manual’s node, then compute Vcm. A blank limit means the comparison waits.
Where does this sheet stop?
This sheet is the bias-return and common-mode record for a floating voltage source on a differential input. Chassis protective earth, a single resistor value for every instrument, a 4–20 mA loop supply, and a ghosting test sit outside it.
A negative reading is a polarity question on a different page. DATAQ’s post of 12 September 2011 says a single-ended input reproduces the negative half of a sine wave when the generator polarity matches the amplifier. Swapping that polarity on a single-ended input ties the generator high to ground, and the reading sits at zero. The differential connection in that post inverts the wave. Use that distinction when the only change is which lead is high. The resistor rows above stay on the floating-source conditions.
If the source is already referred to building ground, NI’s grounded-source section applies. The same note says the ground difference between two instruments on one building power system is typically on the order of 10 mV to 200 mV, and that it can be higher when the distribution circuits are wrongly connected. Differential or NRSE is the configuration that note uses so that difference appears as common mode. NI says differential connections are the ones to use when any of these fail: signals greater than 1 V, cabling that is short and either quiet or shielded, and a common reference shared at the source. The January 2018 question’s 10 m runs fail the short-cable line, which is why that asker was in the differential section.
A 4–20 mA loop’s wire count and shunt checks are on the current-input sheet. After this manual’s common-mode range is filled and the leads sit inside it, channel count and storage sit on the sample-rate worksheet. Other instrument notes are on the blog.
| Record to keep with the channel | Your entry |
|---|---|
| Coupling, DC or AC | |
| Source impedance, ohms or NI class | |
| Earth at the source | |
| Bias resistors: count, which input, ohms | |
| Shield end and the node it ties to | |
| Common-mode limit, V+, V−, Vcm | |
| Manual title and section |
Method and sources
Compiled on 14 October 2026 from the five public pages listed at the end of this article. The text used here is the 11 October 2026 archive of those pages. NI’s field-wiring note shows an update date of 15 August 2024 and is used for the floating-source definition, the bias-resistor conditions, the 10 kΩ to 100 kΩ band, the transducer impedance table, the 1 kΩ and 100 kΩ setup, the shield-at-the-source rule and the 1 V / 0.5 Ω example, the common-mode definition, the CMRR definition, the 60 Hz specification remark, the 10 mV to 200 mV ground difference, and the three single-ended criteria. LabJack’s differential app note shows no publication date in that archive and is used for the bias-path failure mode, the 100 kΩ U6/T7 example and part CF14JT100K, the LJTick-InAmp window, the 12 V shunt example, the series-battery example, the DMM comparison, the T4 statement, and the bridge example. Process Online’s article, by Bruce Cyburt of Acromag and dated 1 March 2017, is used for the third return connection, the IN− instruction, the shield connections in that article, the 100 dB remark, and the common-mode window against signal return. DATAQ’s post by Roger Lockhart, dated 12 September 2011, is used only for the polarity comparison. The Stack Exchange question, asked 18 January 2018, and Andy aka’s answer, posted 23 January 2018, are used for the conflicting-diagram case, the distances and the 10 Hz bound the asker stated, the shield-at-DAQ-earth instruction, the 1 MΩ ceiling, and the capacitor and SMPS figures in that answer. Example numbers stay in the example cells. This site did not wire a channel and did not measure a device. Product names appear only where those public pages name them. Nothing here is an NI or Emerson document, and nothing here is investment advice.
