Some process variables swing both positive and negative: torque that can drive or brake, flow that can run forward or reverse, a position that moves either side of a datum. A signal that only goes one direction cannot represent these, so a bipolar input, typically plus-or-minus 10 volts or plus-or-minus 20 milliamps, is used, and its raw value becomes signed. This guide explains why bipolar inputs exist, how the raw counts turn negative, why zero sits in the middle of the range rather than at the bottom, and how the scaling equation spans a symmetric range.
Bipolar signal scaling in one line: Bipolar analog signal scaling handles inputs that swing both positive and negative, such as plus-or-minus 10 V or plus-or-minus 20 mA, used for quantities like torque, bidirectional flow, or two-sided position. The raw count becomes signed, so it goes negative for negative inputs, and the zero of the measurement sits at mid-range, at 0 V or 0 mA and a raw count near zero, rather than at the bottom. The scaling line spans the symmetric range, mapping the most negative input to the low engineering value and the most positive to the high.
A unipolar input, such as 0-to-10 V or 4-to-20 mA, only represents values from a minimum up to a maximum in one direction; its whole range lies on one side of zero. That is fine for pressure, level, or temperature, which do not go negative in normal use. But some quantities genuinely have two directions. Torque on a shaft can be positive when driving and negative when braking or being driven. Flow in a line can run forward or reverse. A control surface or actuator can move either side of a center position. To carry these, the input itself must be able to go negative, which is what a bipolar range does.
A bipolar input is symmetric about zero: plus-or-minus 10 V runs from minus 10 through zero to plus 10, and plus-or-minus 20 mA from minus 20 through zero to plus 20. The measurement's zero, the point where torque is neither driving nor braking or where flow is neither forward nor reverse, corresponds to the middle of the electrical range, at 0 V or 0 mA, not at the bottom. This mid-range zero is the defining feature of bipolar scaling and the thing that most distinguishes it from the unipolar case, where zero engineering value sits at the low end of the signal.
This changes how the range is read. On a unipolar 4-to-20 mA level, the bottom of the signal is empty and the top is full. On a bipolar plus-or-minus 10 V flow signal, the bottom is full reverse, the middle is no flow, and the top is full forward, so half the electrical range represents each direction. The signal's midpoint is the operating center of the measurement, and the two halves are mirror images, which is exactly why the scaling has to be set up symmetrically rather than as an offset from a low end.
Because the input goes negative, the analog-to-digital converter reports a signed count. Where a unipolar channel might report an unsigned count from zero up to its maximum, a bipolar channel reports a signed count that runs from a large negative value through zero to a large positive value, commonly stored in two's complement so the sign is handled naturally by the arithmetic. A negative input produces a negative count, 0 V or 0 mA produces a count near zero, and a full positive input produces the maximum positive count. The mental model is a number line centered on zero rather than starting at zero.
Scaling a signed count is still the same two-point line, just anchored across the symmetric range. You anchor the most negative count to the most negative engineering value and the most positive count to the most positive engineering value, then interpolate, and because the arithmetic is done in signed numbers, a count near zero lands near the engineering zero automatically. For a bidirectional flow scaled to plus-or-minus five hundred units, minus full scale of the input maps to minus five hundred, the electrical zero maps to zero flow, and plus full scale maps to plus five hundred, with the sign carried through cleanly.
The two things to get right are that the count really is being treated as signed and that the endpoints are entered with their signs. If code treats a two's complement negative count as a large unsigned positive number, a small negative input reads as a huge positive value, which is a classic bipolar scaling bug that produces wildly wrong readings when the signal crosses zero. And if the engineering endpoints are entered as if the range were unipolar, the mid-range zero is lost and the whole measurement is offset. Enter the symmetric endpoints honestly, keep the count signed, and the same line equation that handles a unipolar channel handles the bipolar one.
Bipolar measurements are less common in general field monitoring than unipolar pressures and levels, which is exactly why they deserve care: they are the channels a routine setup is most likely to misconfigure by assuming everything starts at zero. Bidirectional flow on a line that can run either way, torque or load on rotating equipment, and position feedback that swings either side of center are the typical cases in oil and gas and other industries. Each of them carries meaningful information in its sign, forward or reverse, driving or braking, above or below datum, so losing the sign or mishandling the mid-range zero throws away the very thing the measurement exists to capture.
The failure signature is distinctive. A bipolar channel mis-scaled as unipolar tends to read fine on one side and jump or fold on the other, because the negative half of the range has no valid mapping. A signed count mistaken for unsigned reads sensibly while the signal is positive and then leaps to a large wrong value the instant it crosses zero into negative. Because these symptoms appear only when the process actually goes negative, they can lie hidden through commissioning if the equipment was only ever exercised in one direction, and then surprise operators the first time flow reverses or torque changes sign.
In a cloud SCADA architecture such as Merobix, the platform displays whatever signed engineering value the edge device produces, so the signed-count handling and the symmetric endpoints are configured at the RTU or card. What the platform contributes is clarity that a value is legitimately two-sided: carrying the tag's full range including its negative end as metadata, and keeping the raw signed value visible alongside the engineering value, lets an analyst confirm the zero sits at mid-range and that negative readings map correctly. A bidirectional flow that only ever trends positive on the dashboard, or one that spikes to an implausible value whenever it should go slightly negative, is a signal that the bipolar handling needs checking at the source, and having both the raw and scaled values on hand makes that diagnosis direct.
A unipolar input represents values in one direction only, with its whole range on one side of zero, so the engineering zero sits at the bottom of the signal, as with 4-20 mA or 0-10 V. A bipolar input is symmetric about zero, such as plus-or-minus 10 V, so the measurement's zero sits at mid-range and the signal can represent negative values. Bipolar scaling therefore uses signed raw counts and anchors a symmetric range rather than offsetting from a low end.
Because the input voltage or current genuinely goes negative, the analog-to-digital converter must report negative numbers, so it produces a signed count, commonly in two's complement. A negative input gives a negative count, the electrical zero gives a count near zero, and a full positive input gives the maximum positive count. If code mistakenly treats that signed count as unsigned, a small negative input reads as a huge positive value, a classic bipolar scaling error that shows up when the signal crosses zero.
For process variables that swing both positive and negative: torque that can drive or brake, flow that can run forward or reverse, and position or actuator feedback that moves either side of a center datum. In each case the sign carries real meaning, so the measurement needs an input that can go negative and a scaling that puts zero at mid-range. These channels are less common than unipolar ones, which is why they are the most likely to be misconfigured.
Merobix reads your field devices into a cloud SCADA - the real thing behind these terms, live in days from any browser.