The 4-20 mA loop: reading the signal itself

What does it mean when my 4-20mA transmitter is reading 3.8 mA?

3.8 mA is a live, valid signal at the bottom of the range, -1.25% of span. Under NAMUR NE 43 it is the lowest current a compliant transmitter will output before it declares a failure.

NAMUR NE 43, Standardisation of the Signal Level for the Failure Information of Digital Transmitters (current revision 2021), divides the loop into zones. 3.8 mA to 20.5 mA carries measurement. 3.6 mA or below is a low fault. 21.0 mA or above is a high fault. The gaps, 3.6 to 3.8 mA and 20.5 to 21.0 mA, are transition zones that a compliant transmitter will not sit in. A transmitter holding 3.8 mA instead of 4 mA is reporting a process at or below the bottom of its range. It has not failed.

The usual causes, in the order worth checking:

• The process is genuinely below the calibrated zero. A tank drained below the lower tap pins a level transmitter at 3.8 mA and holds there.
• The lower range value is set above the actual process minimum. A transmitter ranged 0 to 150 psi on a line sitting at atmospheric reads 3.8 mA all day and is working correctly.
• Zero has drifted low. On a DP transmitter with a zero-based range, opening the equalizer valve should give exactly 4.00 mA. Elevated or suppressed ranges are different, as in a wet-leg level application: 0 DP is not 4 mA there. Get the range values before you judge.
• The device is configured for NAMUR saturation limits. On a Rosemount 3051 ordered with no alarm option code, low saturation is 3.90 mA; ordered with the NAMUR option it is 3.80 mA. An exact 3.8 mA floor points at a NAMUR-configured device, not a drifting one.

Confirm the number at the source. Break the loop and read milliamps in series, or measure across a 250 Ω resistor, where 1.000 V is 4 mA and 5.000 V is 20 mA. A transmitter putting out 4.00 mA while the DCS displays 3.8 mA is a wiring drop, an input card, or scaling.

What does it mean when a 4-20mA signal reads 21.5mA?

21.5 mA is not a measurement. The transmitter has found an internal failure and driven the loop above the NAMUR NE 43 high alarm threshold of 21.0 mA to say so.

A NAMUR-compliant transmitter saturates at 20.5 mA for an over-range process and stays there. It will not output between 20.5 and 21.0 mA. At or above 21.0 mA, the device has declared a failure. Usual causes: an open or shorted sensor, lost communication between the sensor module and the electronics, an ADC or memory fault on the transmitter board, or on a temperature transmitter an open thermocouple or broken RTD lead with upscale burnout configured.

Pull the diagnostic message with a HART communicator or the asset management system before touching the wiring. The transmitter usually names the failure. A local display showing a normal process value while the loop reads 21.5 mA puts the fault downstream: a second power source backfeeding the loop, or moisture across a terminal block giving supply voltage a leakage path into the signal leg.

Check the configured alarm levels before you condemn the device. The exact fault current names the device and its option code. The Rosemount 3051 configuration data sheet gives four sets:

• No option code, standard high. High alarm 21.75 mA, high saturation 20.80 mA.
• Option CT, standard low. Low alarm 3.725 mA, low saturation 3.90 mA.
• Option C4, NAMUR high. High alarm 22.50 mA, high saturation 20.50 mA.
• Option CN, NAMUR low. Low alarm 3.575 mA, low saturation 3.80 mA.

Emerson's own constraint on custom values is that the high alarm must be at least 0.1 mA above high saturation, and low saturation at least 0.1 mA above low alarm.

On a 3051, an exact 21.5 mA is not a factory alarm current. It sits above the 21.0 mA NE 43 fault threshold and below both the 21.75 mA standard and 22.50 mA NAMUR alarm values. Three ways to land there: a custom high alarm setting, a different manufacturer, or something other than the transmitter driving the loop. And a device configured for downscale burnout reports that same internal fault as a low current instead. A dead-looking loop is not always a dead loop.

Why is my analog input reading 32767 (or -32768) in the PLC instead of a real value?

32767 and -32768 are the maximum and minimum of a signed 16-bit integer. The tag is telling you the channel has railed or is holding a fault value, not that the process is at some real number.

A signed 16-bit word runs -32768 to +32767. Over range, under range, open circuit: many modules drive the raw word to one of those limits. The count values are vendor-specific and configuration-specific. Read them off the module manual before you interpret them.

• Siemens S7, SIMATIC analog value representation. A channel configured for 4 to 20 mA gives 0 counts at 4 mA and 27648 at 20 mA. Over range is 27649 to 32511, the overflow band starts at 32512, and overflow is reported as 32767. Under range is -1 to -4864, where -4864 corresponds to about 1.185 mA, and anything below that is underflow, reported as -32768. A broken wire on a 4-20 mA channel rails to -32768, not 32767. The 5530 to 27648 figure people quote is where 4-20 mA lands inside a channel configured for 0 to 20 mA. Different configuration, and a common source of scaling errors.
• Allen-Bradley 1769-IF4, Raw/Proportional, 4-20 mA range. From the valid input data table in the Compact I/O analog modules manual (1769-UM002): 4.0 mA is 6241 counts and 20.0 mA is 31206 counts. The module clamps rather than wrapping: at and above 21.0 mA the data word pins at 32767, and at and below 3.2 mA it pins at 4993. A dead loop reads 4993, not 0. The module also sets general status, over-range flag and under-range flag bits per channel. Those are faster to read than the data word.

Work it in this order:

• Read the actual current at the field terminals. 0.0 mA means the loop is dead: check the 24 V supply, the fuse, and continuity.
• Check the channel fault and status bits. They call out over range, under range, and open wire directly and save you a trip to the field.
• Confirm the channel is configured for current, not voltage, and that the wiring is on the correct terminal pair. A 4-20 mA signal into a channel configured for ±10 V produces nothing meaningful.
• Confirm the channel is enabled. An unconfigured or disabled channel often reports the rail value, and on the 1769 modules the status bits are only valid when the channel is enabled.

One common cause is not in the field at all. Moving a scaled REAL into an INT, or into a 16-bit word without limiting, saturates or wraps at 32767 while the loop runs perfectly. Check the data types between the raw word and the faceplate before you pull a transmitter.

What formula do I use to scale a 4-20mA input to engineering units like 0-100 PSI in the PLC?

For a 0-100 PSI range, PSI = (mA - 4) x 6.25; if the PLC hands you raw counts instead of milliamps, PSI = (raw - raw_at_4mA) x 100 / (raw_at_20mA - raw_at_4mA).

It is a straight line: EU = (raw - raw_low) x (EU_high - EU_low) / (raw_high - raw_low) + EU_low. The rest is filling in the four endpoints your hardware actually uses.

• ControlLogix, 1756-IF8. The module's channel configuration carries its own scaling: set low signal 4 mA to low engineering 0.0 and high signal 20 mA to high engineering 100.0, and the input tag reads PSI directly with no ladder math. If the module is left scaled in milliamps: PSI = (mA - 4.0) x 6.25. At 12.00 mA, 50.0 PSI.
• Siemens S7, channel configured 4 to 20 mA: PSI = raw x 100.0 / 27648. Or NORM_X with 0 and 27648, then SCALE_X with 0.0 and 100.0.
• Siemens S7, channel configured 0 to 20 mA: 4 mA sits at 5530 counts, so PSI = (raw - 5530) x 100.0 / 22118.
• Allen-Bradley 1769-IF4, Raw/Proportional, 4-20 mA range: PSI = (raw - 6241) x 100.0 / 24965.
• MicroLogix and SLC 500 have SCP, Scale with Parameters, on the Advanced Math tab in RSLogix 500: enter input min and max, scaled min and max, and let the instruction do it. SCP is not universal across the family. It requires an SLC 5/03 with OS301 or later and is not available on the SLC 5/01 or 5/02. Check the processor before you plan around it.

Do the arithmetic in floating point. Integer division of (raw - 5530) x 100 by 22118 truncates and costs resolution. The multiply overflows a signed 16-bit integer long before it reaches the divide.

For a transmitter that is not zero-based, change the engineering endpoints and leave the current endpoints alone. A device ranged 20 to 120 PSI has EU_low of 20 and EU_high of 120, still measured against 4.00 and 20.00 mA.

Verify with three points, not one. Inject 4.00, 12.00, and 20.00 mA and confirm 0.0, 50.0, and 100.0 PSI. Correct at both ends and 70.7 at midspan means square-root extraction is enabled somewhere you did not expect. On DP flow it is often enabled in the transmitter and the PLC both, by mistake.

How do I simulate a 4-20mA signal to test a PLC analog input?

Use a loop calibrator in simulate mode, wired in place of the transmitter. The PLC or field supply drives the loop, and the channel gets exercised exactly the way the real device will exercise it.

Simulate and source are not the same mode. Confusing them is the usual mistake. In simulate mode the calibrator behaves like a 2-wire transmitter: it regulates the current but supplies no driving voltage, and loop power comes from the PLC or field supply. In source mode it supplies both. Connect a sourcing calibrator to a channel that provides its own loop power and you have two supplies pushing against each other, and you can damage the channel.

• Disconnect the transmitter at the field terminals or in the marshalling cabinet. Note polarity before pulling anything.
• Set the calibrator to simulate, connect in place of the transmitter, and step 4.00, 8.00, 12.00, 16.00, 20.00 mA.
• Record both the raw count and the scaled tag at each step. The raw count proves the card; the scaled tag proves your math.
• Then go to 3.6 mA and 21.5 mA and confirm the alarm and fault logic responds. Skip it and the fault handling stays untested until a real fault arrives.

Without a calibrator, a 24 V supply in series with a 1 kΩ fixed resistor and a 5 kΩ potentiometer will sweep roughly 3.8 to 19.2 mA through a 250 Ω input, with a meter in series to read the actual current. That proves the channel responds. It is not accurate enough to calibrate against. Do not use it for anything you sign.

Why is my control valve not responding to the 4-20mA signal from the DCS?

Measure the current at the positioner terminals while the DCS output is stroked from 4 to 20 mA. Correct current arriving and no movement means air or mechanical, not signal.

Three outcomes, three different problems:

• No current, or current that will not change. The output card, the wiring, or loop power. Confirm the channel is not in manual, override, or simulate. Check the card's load capability: the 1756-OF8 is rated to drive 0 to 750 Ω on a current output. A Fisher DVC6200 needs a minimum of 9.5 V at the instrument terminals for analog control and 10 V for HART communication, and Emerson gives an equivalent input impedance of 500 Ω for loop-load calculations, corresponding to 10 V at 20 mA. That is already two thirds of the 750 Ω budget. Add a safety barrier and a long cable run and the current clips below 20 mA.
• Correct current, no movement. Air. Check supply pressure at the positioner gauge, not at the header, and check the filter regulator. Fisher's own rule for the DVC6200 is a minimum recommended supply pressure 0.3 bar (5 psig) higher than maximum actuator requirements, with a maximum of 10.0 bar (145 psig) or the actuator's rating, whichever is lower. A 3-15 psig I/P feeding a 3-15 psig bench-range actuator needs at least 20 psig of clean, dry supply. Below that it moves nothing. Fisher 67C series instrument supply regulators are offered in outlet ranges that include 0 to 20, 0 to 35 and 0 to 60 psig, which is where the common airset settings come from.
• Correct current, sticky or partial movement. Mechanical. Packing friction, a bent stem, a disconnected feedback arm, or a failed actuator spring. Stroke it with a calibrator in source mode and watch the stem directly.

Two things that fool people. A positioner left in a setup or calibration routine holds position while the current changes correctly, and it reads normal on every meter you put on it. A valve with the handwheel wound in, or a travel stop set, sits at a limit regardless of what the loop does. Check the handwheel first. It costs ten seconds.

What causes a ground loop on a 4-20mA signal and how do I get rid of it?

A ground loop is the current loop referenced to earth at more than one point, with those earth points sitting at different potentials. The difference drives current through the signal path.

The loop needs exactly one ground reference. A second one arrives easily: a transmitter case bonded to grounded pipe while the loop negative is also bonded at the DCS, a shield landed at both ends, a non-isolated input card sharing a common with another grounded device. Differences of 1 to 10 V across a plant are ordinary. Across a 250 Ω sense resistor, 0.5 V driving through a low-impedance fault path is 2 mA of error. That is 12.5% of span.

Symptoms that point at grounding rather than the instrument: a constant offset that the transmitter's own display does not show, 60 Hz ripple on the reading, a shift when a nearby motor starts, or a change when you lift one conductor. Diagnose it directly. With the loop powered, measure DC and AC volts between the loop negative at the DCS and the transmitter case or local earth. More than a few tens of millivolts is a second reference you did not intend.

• Remove the extra reference. Ground the loop at one point only, normally the negative of the supply in the control room.
• Ground the shield at one end only and insulate the field end so it cannot contact the housing.
• Use an isolated input card or add a loop isolator. This is the answer when the transmitter case must be bonded for safety and the second reference cannot be removed.
• Check whether the transmitter internally bonds loop negative to the case. Some 2-wire devices do, and the datasheet will say so.

Do not fix a ground loop by lifting a safety ground. Equipment grounding conductors carry fault current and are not optional. Isolate the signal instead.

Should I ground the shield on a shielded analog cable at one end or both ends?

One end for 4-20 mA and other low-frequency analog signals, at the control room end; both ends for Ethernet, VFD motor cable, and anything else carrying high frequency.

The reason is frequency and cable length, not preference. Below roughly 100 kHz, and while the cable is short compared with a wavelength, a shield grounded at one end still works electrostatically. Ground both ends and it becomes a conductor between two earth points at different potentials. Power-frequency current then runs along the shield and couples back into the pair it is supposed to protect. Once the cable is longer than about 1/20 of a wavelength at the frequency of concern, a single-ended shield starts behaving like an antenna, and both-end bonding is what actually performs. Above about 1 MHz, skin effect keeps the noise current on the outside of the shield and the signal return on the inside. That is why both-end bonding stops being a liability up there.

• Land the 4-20 mA shield at the control room or marshalling end, on the instrument ground bus.
• Cut back and insulate the field end. Do not leave it loose in the head where it can touch the case.
• Carry the shield through junction boxes unbroken. Splice drain to drain and do not ground it at the box.
• One shield, one ground, per cable. On multipair, every drain lands at the same end.

For industrial Ethernet the rule inverts. The PROFINET assembly guideline is explicit that to keep the shield effective at high frequencies it must be earthed at both ends of the cable, normally through the connector, and that the devices themselves must be adequately earthed. Shielded EtherNet/IP practice is the same. Where a link runs between separate buildings or parts of buildings, an equipotential bonding conductor is laid parallel to the data cable; the minimum cross-sections come from IEC 60364-5-54, and Siemens EMC installation guidance for PROFIBUS and PROFINET quotes them as 6 mm² copper, 16 mm² aluminium, or 50 mm² steel. A shielded cable with a floating shield performs worse than plain unshielded cable. If you are not going to bond it, do not buy it.

Can I use an NPN sensor on a PNP (sourcing) input card, or do I need a relay?

It depends on which way current flows at the input terminal, not on the label. A card whose input common ties to DC common (0 V) is a sinking input and needs a PNP sensor. A card whose input common ties to +24 V is a sourcing input and needs an NPN sensor.

The naming causes most of the confusion. An NPN sensor switches its output to 0 V and sinks current. A PNP sensor switches its output to +24 V and sources current. A card described as a sourcing input sources current out of its terminal, so it wants an NPN sensor to pull that terminal down. A card described as a sinking input, an Allen-Bradley 1756-IB16 for instance, sinks current into common and wants a PNP sensor. Find the card's common terminal on the print. See whether it lands on 0 V or +24 V. That settles the question in ten seconds, and it answers both phrasings of it.

If you have an NPN sensor and a card that wants PNP:

• Change the sensor. Cheapest and cleanest. Most photoelectric and inductive proximity sensors are stocked in both variants with the same part number family.
• Check the card first. Some modules are dual-mode: the Siemens SM 1221 24 V DC input modules for the S7-1200 are catalogued as sink/source and are selected by how the common is wired. Others are fixed, and the catalog number decides it. On POINT I/O, 1734-IB8 is a sinking input that expects a PNP sensor and 1734-IV8 is a sourcing input that expects an NPN sensor, so you order the one you need rather than reconfiguring it.
• Interposing relay. Drive a 24 VDC coil from the NPN output to 0 V and feed the card from a dry contact. It costs a relay and panel space, and it adds contact life as a wear item. It works reliably.
• Pull-up resistor. A resistor from the sensor output to +24 V, typically 2.2 kΩ to 3.3 kΩ at 0.5 W, makes the signal readable. It also inverts the logic and holds the input near its threshold instead of driving it hard. Size it from the sensor's output saturation voltage and the card's ON-state current, not from the range alone. In an electrically noisy panel it is a bad trade.

For a safety-related or interlock input, change the sensor or use the relay. Do not use the pull-up resistor.

Why does my sensor false trigger whenever the VFD runs, and how do I stop the electrical noise?

The drive is coupling common-mode noise into the sensor circuit. The fix belongs at the drive and the cabling, not at the sensor.

A PWM drive switches at a carrier frequency that is typically adjustable over roughly 2 to 16 kHz, with IGBT output transitions on the order of 50 to 200 ns. Rockwell's own drives application paper puts IGBT output transitions at 0.05 to 0.1 µs, and newer wide-bandgap devices are faster still. Treat these as typical rather than specified. Those edges push high-frequency common-mode current through motor cable capacitance and motor frame capacitance, and that current returns by whatever path exists. Often building steel. Often your signal wiring. The sensor is not defective. It is responding to a real electrical event.

Work the source, then the path, then the victim:

• Motor cable. VFD-rated shielded cable with the shield terminated 360 degrees at both ends, drive and motor. This is the highest-value single fix. Unshielded conductors in conduit are not equivalent.
• Drive grounding. Bond the drive to the panel ground bus with a short, wide strap or braid rather than a long round wire. High-frequency impedance is dominated by inductance. Length matters more than gauge.
• Separation. Get the sensor cable out of the wireway, tray, and conduit carrying motor leads. Cross at 90 degrees where crossing is unavoidable.
• Carrier frequency. Dropping from 8 kHz to 4 kHz or 2 kHz reduces noise energy. It also makes the motor audibly louder and may require derating the drive. It is a legitimate move and it is a trade.
• Victim side. Shielded sensor cable grounded at the panel end, plus input filter time. The available filter settings are narrow and module-specific. Read the spec rather than assuming. A 1756-IB16 offers a user-selectable filter of 0, 1 or 2 ms off-to-on and 0, 1, 2, 9 or 18 ms on-to-off, on top of a hardware delay of a few hundred microseconds. A Siemens SM 1221 24 V DC input module is configurable from about 0.2 ms to 12.8 ms, in groups of channels. Filter time hides the symptom. Use it after the source work, not instead of it.

If the false trip happens only at one specific drive speed, check mechanically as well. Resonance at that frequency can move a target or a bracket far enough to trip a proximity sensor for real.

How far apart do 480V power cables and 4-20mA signal cables need to be in a cable tray?

Use a separate tray or a fixed barrier. The NEC generally does not permit Class 2 or Class 3 signal cable to share a raceway or tray with light and power conductors at all, and for noise rather than code, plan on 12 inches of air from 480 V.

Two different questions get mixed together here, and the code one does not answer the engineering one.

The code question. NEC 725.136 starts from a prohibition in 725.136(A): cables and conductors of Class 2 and Class 3 circuits shall not be placed in any cable, cable tray, compartment, enclosure, manhole, outlet box, device box, raceway or similar fitting with conductors of electric light, power, Class 1, non-power-limited fire alarm, and medium-power network-powered broadband communications circuits, unless permitted by the subsections that follow. Those subsections are the exceptions. They turn on a permanent barrier, or on how each set of conductors is enclosed, not simply on inches of air. The 2 in. (50 mm) figure people quote comes from the "other applications" subsection, and it applies only where neither set of conductors is in a raceway or in metal-sheathed, metal-clad, non-metallic-sheathed or Type UF cable. It is not a general permission to share a tray. Medium-voltage cable does not share a tray with signal cable.

Confirm the edition your jurisdiction has adopted before you rely on any of this. The 2023 NEC reorganized Chapter 7: Class 1 circuits moved out of Article 725 into a new Article 724, and common cabling rules moved into a new Article 722. Confirm too whether your instrument cable is Class 2/3, Class 1, or Type ITC. They are treated differently, and in the 2023 edition Type ITC moved out of Article 727 and into Article 335 in Chapter 3.

The engineering question is separate, and it is the one that decides whether the loop works. Coupling scales with parallel run length and falls off with distance. 10 feet at 4-inch separation is a different problem from 400 feet at the same spacing. Common instrument design practice:

• Under 120 V control wiring: 6 inches, or a separate tray.
• 120 to 480 V power: 12 inches minimum, more as the parallel run gets longer.
• VFD output cable: its own tray. Do not share at any spacing.
• Crossings: at 90 degrees, which are far less of a problem than parallel runs.

Those numbers are design practice, not code, and they are worth naming as such. The separation tables they descend from are in IEEE 518-1982, a guide IEEE withdrew in 2002, and published copies of its table do not agree with each other on the exact figures. For a number you have to defend, use your project's electrical design specification.

A grounded steel barrier or a separate grounded conduit buys more than several extra inches of air. And a twisted shielded pair with the shield grounded at one end tolerates far more than an unshielded pair or a 0-10 V signal would.

0-10V or 4-20mA - which should I use for a sensor 300 feet from the panel?

Use 4-20 mA, but not for the reason usually given. 0-10 V does not fail at 300 feet from voltage drop in the wire. It fails because it has no defense against ground potential difference and coupled noise.

Run the wire resistance numbers first. The common explanation is wrong. 300 feet of 18 AWG pair is 600 feet of conductor at 6.385 Ω per 1000 feet for solid copper at 20 °C, about 3.8 Ω round trip, and a few percent more for stranded. In a current loop that costs 0.077 V of supply headroom at 20 mA and no accuracy at all. The current is the signal. In a 0-10 V circuit feeding a 10 kΩ input, the same 3.8 Ω costs about 0.04%. Neither number is the problem.

What actually breaks 0-10 V over distance:

• Ground potential difference. A 0-10 V signal is referenced to ground at the sensor and read against ground at the panel. 0.5 V between those two grounds is 5% of span of error, and it moves whenever plant loads change.
• Coupled noise. A current loop is driven by a current source and terminated in about 250 Ω. A voltage induced in the loop barely changes the current. A voltage input sits at 10 kΩ to 1 MΩ and reads induced noise almost in full.
• No live zero. With 0-10 V, 0 V means both "sensor reads zero" and "wire is broken." A live 4 mA is 0%, not a dead loop, and 0.0 mA is a broken wire. That distinction comes free.

0-10 V is right in places. Inside a panel. Short runs in a clean electrical environment, roughly under 50 feet. Into a device that accepts voltage only. On a fast signal where loop bandwidth is the limiting factor. If the sensor is 0-10 V only, put a signal conditioner at the sensor and run 4-20 mA back to the panel, rather than running 300 feet of voltage.

Should I use shielded or unshielded Ethernet cable in a control panel with VFDs?

Shielded, but only if the shield is going to be bonded properly at both ends. A shielded cable with a floating shield performs worse than plain unshielded cable.

A panel containing VFDs is not the controlled environment TIA and ISO cabling standards assume. The noise source is inside the same enclosure as the cable. Use S/FTP or SF/UTP Cat 5e or Cat 6 with shielded RJ45 or M12 connectors, and bond the shield 360 degrees at both ends through the connector shell to panel ground. High-frequency common-mode current from the drive needs a defined low-impedance return path. The shield is that path.

The failure to avoid is shielded cable installed with unshielded connectors, or landed at one end only. A shield left floating gives up the high-frequency benefit it was bought for and can measure worse than equivalent unshielded cable. The money is spent and the performance is lost.

• Keep Ethernet out of the wireway carrying motor leads. Eight inches of separation inside a panel is cheap, and crossing wireways at 90 degrees is free.
• Cat 5e is sufficient for 100 Mbps device-level networks. Cat 6 buys bandwidth headroom, not noise immunity.
• Use factory-made patch cords where the length allows. Field-terminated shielded RJ45 is the most common place shield continuity is lost.
• Short jumpers between adjacent DIN-rail devices away from the drive section are usually fine unshielded.

If the panel is already built with unshielded cable and comms are intermittent, replace the cable last. Check drive grounding, motor cable shield termination, and physical routing first. Those are more often the actual cause, and they are cheaper to correct.

When this page runs out

This page covers reading and troubleshooting the loop itself. It does not cover accredited calibration, hazardous-area certification, or safety instrumented function verification. An accredited calibration certificate with traceable uncertainty comes from an accredited calibration laboratory. If the device is under OEM warranty, or the fault is inside a sealed sensor module, the manufacturer is the right choice, and Bridges Industrial will say so. If the loop is part of a safety instrumented function, the proof test interval and the SIL calculation belong with whoever holds the safety requirements specification, not with a troubleshooting page. And if the transmitter checks out at 4.00 and 20.00 mA against a calibrator, the instrument is not the problem. No amount of instrument work will make it one.