The decision that drives DP transmitter accuracy is not the sensor's reference accuracy figure (commonly 0.1–0.5% of span [S2]) but how the upper range limit (URL) of the cell relates to the calibrated span you actually configure, and where the normal operating point falls inside that span [S1][S3].
Two DP cells with identical 0.1% reference-accuracy spec sheets can deliver very different field error if one is ranged 0–100 mbar on a 0–100 mbar URL while the other is ranged 0–10 mbar on a 0–100 mbar URL: the second unit carries the same absolute zero uncertainty but a 10× larger percentage of reading [S2][S3].
URL, Calibrated Span, and Turndown in Plain Terms
URL is the maximum differential pressure the cell can measure; the calibrated span is the range you actually configure (LRV to URV) using the lower- and upper-range value trim [S4]. Turndown is the ratio URL/span, and most industrial DP cells support turndowns of 4:1 to 100:1 depending on diaphragm and electronics [S1][S2].
Sizing rule of thumb from working engineers: pick a URL roughly 4× to 10× the maximum process dP you expect, so the normal operating point sits in the 25–75% band of the calibrated span [S1][S3]. Going past 10:1 turndown starts to multiply the percentage of reading, because zero drift and quantisation are fixed in absolute units [S2][S3].
Where the Error Actually Lives: % of Span vs % of URL
Datasheets express accuracy two ways, and the choice changes the math: "% of calibrated span" means the error scales with your configured range (set 0–5 bar on a 0–10 bar sensor, 0.1% of span equals ±0.005 bar), while "% of max range / URL" ties the error to the full sensor capability regardless of how you range it [S2].
High-accuracy families like the PT600 specify 0.1% of max range reference accuracy, with a 0.075% high-accuracy option and total temperature drift within 0.1% of max range over −10 to +70 °C; the PT600RSH differential version reaches 0.075% on its DP ranges, while the compact PT03 sits at 0.35% (0.15% option) for general-purpose use [S2]. The LT100 level transmitter is specified at 0.1% for hydrostatic level service [S2].
Practical consequence: a 0–10 bar URL cell ranged to 0–1 bar (10:1 turndown) still carries the same absolute error in bar as a 0–10 bar configuration, so the % of reading climbs 10×. The differential pressure transmitter cell is accurate in absolute terms, but only the pressure transmitter spec phrasing tells you whether the % figure is friendly or hostile to large turndowns [S2].
Operating Point: Why the Middle Third Wins
Engineers converge on a "middle third" rule for a reason: avoid the bottom third of the calibrated range because zero drift becomes a large share of the reading, and avoid the top 10–20% because small process upsets push you into over-range, especially on DP flow where dP scales with the square of flow [S3].
For a DP flow loop, the relationship is brutal: at 10% of maximum flow, dP is only 1% of the maximum pressure drop, so pressure fluctuation and electronic noise dominate, and the square-root extractor may even switch to a less accurate segment of its curve [S3]. A common design target is around 75% of full scale at the normal flow, where calibration can minimise third-order non-linearity, but only if the actual measurement band is reasonably stable [S3].
The same logic applies to hydrostatic level service: range the level transmitter so the normal level sits at roughly 50–70% of the calibrated span, leaving headroom for level swell, pump-down transients, and density-driven zero shifts without clipping the LRV or URV [S1][S2].
Re-Ranging an Existing DP Flow Loop
When a DP flow loop is operating at 40–60% of the current span, re-ranging the cell so the normal flow lands at 80–90% of span can lift accuracy, because dP at 50% of max flow is only 25% of the cell's full-scale dP, and re-calibrating the dP at 50% of the URL roughly doubles resolution while still allowing flow to climb to 70.7% of the original before over-range [S3].
Re-spanning a transmitter that already operates in the middle third of its span gives only marginal gain, and a 10% flow increase on a transmitter ranged to 80% will push it over-range, so confirm the maximum credible flow before committing to a tighter span [S3]. For a comparable walk-through of the safe valve sequence used to take a DP flow transmitter out of service for the re-range, see the field calibration procedure at calibrating a DP flow transmitter [S6].
A common mistake is re-ranging the analog output alone: that may only sharpen the display, not the underlying cell resolution, and it does nothing for the dP-squared compression at low flow [S3][S4].
Static Pressure, Temperature, and the Total Error Band
Reference accuracy is only one slice of the error pie; datasheets add static-pressure effect (especially for low DP ranges on a high-line-pressure cell), temperature effect on zero and span, and long-term stability per year, and the sum is often called total error band [S2]. The PT60-type remote-seal level transmitter, for example, quantifies extra error from temperature differences along the capillary and from static-pressure changes, and the manufacturer recommends insulating the capillary to keep that contribution inside spec [S2].
For low ranges (e.g. below 300 mm water column) or interface service with similar densities on both sides, pay particular attention to capillary-fill fluid temperature/density variation, nozzle resolution loss on 2″ or 3″ connections, air or liquid pockets in the hookup, and the uncertainty that appears when the cell's maximum possible calibration range is much greater than the actual calibrated range [S5]. Static-pressure zero and full-scale drift on a high-line-pressure DP can be compensated, but only if the compensation pressure is actively measured [S5].
A pragmatic guardrail: if you cannot keep the calibrated span above 4× the expected operating dP, consider a capacitance level transmitter or a remote-seal DP with matched capillaries, both of which handle small spans and dirty, foaming, or fouling services better than a bare DP cell [S5]. For absolute-pressure service in vacuum or sealed-reference loops, an absolute pressure transmitter with a tighter URL avoids the same turndown penalty.
Decision Matrix: URL / Span Options Against Real Criteria
Three common sizing choices for a 0–60 mbar normal dP on a steam flow loop, with a 0.1% of span reference spec: [S4]
Option A, 0–100 mbar URL ranged 0–100 mbar (1:1 turndown): accuracy at 60 mbar is about ±0.1 mbar (0.17% of reading); zero drift is negligible share of reading; no turndown penalty, but any process upset near 100 mbar clips the URV. Best for stable, well-understood flows with tight over-range control [S1][S2].
Option B, 0–500 mbar URL ranged 0–100 mbar (5:1 turndown): if the spec is % of span, accuracy is still ±0.1 mbar at 60 mbar; if the spec is % of URL, the absolute error stays the same but % of reading at 60 mbar is roughly 0.83%, almost 5× worse than Option A; this is the usual "great spec sheet, poor field accuracy" trap [S2][S3]. Best only when % of span is contractually guaranteed.
Option C, 0–250 mbar URL ranged 0–100 mbar (2.5:1 turndown): balanced choice; healthy headroom for process upsets to about 158 mbar before over-range, and a 0.1% of span spec gives ±0.1 mbar absolute, or 0.17% of reading at 60 mbar; this is the recommended range for most steam, air, and water DP flow services [S1][S3].
The "75% of full scale at normal flow" rule is fine when the actual measurement band is stable, but the rule is brittle when the operating point is highly variable: in that case, prefer Option A's wider calibrated span to keep turndown low and protect against over-range on flow upsets [S3].
Standards, Documentation, and Failure Modes to Watch
For hazardous-area service, dual-certified ATEX category 2 and IECEx cells are the default on European chemical-plant builds, and the diaphragm, O-ring, and fill-fluid materials must be checked against ISO 15156 / NACE MR0175 for sour-service compatibility; impulse tubing and seal-pot layout on dry-leg and wet-leg level schemes are governed by long-standing instrumentation practice, including the wet-leg labelling on the three-valve manifold and the drainage pot at the low point of a dry leg [S5].
Failure modes that show up only after the wrong URL/span pairing: clipped URV on pump-start transients, square-root extractor non-linearity at very low dP, zero shift after a thermal cycle on long capillary runs, and quantisation steps in the HART or analog output becoming a visible share of the reading at extreme turndowns [S2][S3]. The right pairing removes all four at once.
Track for the next planning cycle: tighten the contractual accuracy statement to "% of calibrated span, not % of URL" for any cell beyond 4:1 turndown, and verify the long-term stability figure (typically given as ±X% of URL per year) rather than assuming the reference-accuracy number holds for the full 5-year calibration interval [S1][S2].