Control Valve Sizing & Valve Authority — Why a Bigger Cv Controls Worse

The valve that hunts at turndown

Night shift calls. The cooling-water temperature controller on E-2101 has been swinging ±3 °C for an hour. The trend shows the valve output sawing between 6 % and 14 %, the temperature chasing it a minute later. Day shift's note says: "CV-2101 replaced last shutdown — upsized from 4″ to 6″ because it was suspected of restricting flow."

Guess first: the valve is now twice the size it was. Did the loop get better, worse, or stay the same?

It got worse, and not marginally. The old valve sat at 81 % travel at design and 23 % at minimum flow. The new one sits at 31 % and 8 % — barely off the seat, in the region where the plug contour has stopped being a contour and flow is set by clearance, burrs and seat geometry. Every 1 % the positioner moves is now a large, unrepeatable step in flow. The loop cannot find a resting place, so it hunts. Meanwhile the throttling is happening across a few millimetres of annulus at high velocity — the seat will be wire-drawn within a year.

The wrong belief, stated plainly: a bigger Cv gives you more capacity and therefore more control; if in doubt, go one size up. Capacity, yes. Control, no. Control is not capacity — control is the ability of the valve's motion to change the flow, and oversizing attacks that from two independent directions at once.

Cv, Kv and the sizing equation

Cv is a measured capacity, not a geometric property: the number of US gallons per minute of 60 °F water that pass through the fully open valve at a 1 psi drop. Kv is the same idea in metric clothing — m³/h of water at 1 bar drop.

Kv = 0.865 × Cv            Cv = 1.156 × Kv

For a non-flashing, non-choked liquid the sizing equation is just that definition scaled:

Q = Kv · √(ΔP_valve / SG)          Q in m³/h, ΔP in bar
Kv_required = Q / √(ΔP_valve / SG)

Two things go wrong before the datasheet is even issued:

• The flow is padded. Process gives a "design" flow already carrying 10–20 % margin; the instrument engineer adds another factor "for safety"; the result is a Cv 30–60 % above duty.
• The ΔP is starved. Every other item in the circuit is sized with its own margin, so the hydraulic calculation hands the valve whatever head is left over — often a token 0.3–0.5 bar. That single decision, made by nobody in particular, sets the valve's authority for the life of the plant.

Then the body size is rounded up to match the line size, full-size trim is fitted because it is the standard build, and the valve arrives three times too big. Nothing in that chain was an error; every step was conservative. Conservatism on Cv is not conservative on control.

ΔP_choked = F_L² · (P₁ − F_F · P_v)          F_F ≈ 0.96 − 0.28·√(P_v / P_c)

Symbol key — every symbol on this sheet

The convention is that the subscript names where the pressure drop is measured: ΔP_valve is across the valve alone, ΔP_system is everything else in series, ΔP_total is the pair of them.

• Cv — flow coefficient: US gpm of 60 °F water at 1 psi drop · gpm/√psi
• Kv — metric flow coefficient: m³/h of water at 1 bar drop · m³/h/√bar
• Q — volumetric flow through the valve · m³/h
• SG — specific gravity (density relative to water) · dimensionless
• ΔP_valve — drop across the valve at the stated flow · bar
• ΔP_system — drop across everything else in series (pipe, exchanger, orifice) · bar
• ΔP_total — the differential available across the whole controlled circuit · bar
• N — valve authority = ΔP_valve / ΔP_total at design flow · dimensionless
• h — valve travel (lift) as a fraction of rated stroke · – or %
• R — rangeability = Kv at max controllable travel ÷ Kv at min · e.g. 50:1
• gain — installed loop gain = Δflow % ÷ Δtravel % · dimensionless
• F_L — liquid pressure-recovery factor of the valve body · dimensionless
• F_F — liquid critical-pressure-ratio factor · dimensionless
• P₁, P_v, P_c — inlet, vapour and thermodynamic critical pressure · bar a

Valve authority — the number that decides controllability

Authority is the share of the circuit's pressure drop that belongs to the valve, at design flow:

N = ΔP_valve / ΔP_total = ΔP_valve / (ΔP_valve + ΔP_system)

Why it matters: the rest of the circuit is a resistance that obeys ΔP_system ∝ Q². When the valve opens, flow rises, the system's share of the drop grows as the square, and the valve's own share collapses. The valve is therefore fighting a moving target — and the more of the total ΔP the system owns to begin with, the less the valve's motion can achieve.

• Inherent characteristic — flow versus travel on a test bench at constant ΔP. A property of the trim contour, printed in the catalogue.
• Installed characteristic — flow versus travel in your circuit, where ΔP_valve sags as flow rises. A property of the system, and the only one the controller ever sees.

At N ≥ 0.5 the two curves are close and a linear trim stays roughly linear. At N ≈ 0.25 the distortion is real but equal-% trim compensates. Below N ≈ 0.25 the installed curve collapses towards a quick-opening shape: most of the achievable flow change happens in the first fifth of travel, and above that the valve can open all it likes and the flow barely moves. The loop gain then swings by an order of magnitude across the operating range — tuned for one end, it is unstable at the other.

Watch it collapse — set the ΔP split, the Cv and the trim, and the explorer plots the inherent and installed curves on the same axes with the design and minimum-flow points marked, plus the installed gain against the controllable 0.5–2.0 band: ▶ open the interactive: instrumentation control valve authority calc

One subtlety worth holding on to: authority and oversizing are different faults. Authority is fixed by how the hydraulic designer split the pressure drop; it does not change when you buy a bigger valve. Oversizing decides where on the stroke the valve has to sit. A valve can have excellent authority and still be useless because it lives at 8 % travel — or be correctly sized and still uncontrollable because it was only given 5 % of the circuit ΔP. The explorer reports both, separately, and so should your datasheet review.

Trim, rangeability and a worked example

Trim shape is the one cheap correction available, because it can be chosen to bow the opposite way to the distortion the system imposes:

Linear trim:      Kv(h) = Kv_rated · [1/R + (1−1/R)·h]  equal travel → equal flow step
Equal-% trim:     Kv(h) = Kv_rated · R^(h−1)           equal travel → equal PERCENTAGE step
Quick-opening:    most of the Kv in the first third of travel (on/off duty, not control)

• Use linear when the valve already owns most of the drop (N ≳ 0.5) — level loops on a gravity drain, bypass circuits, anything where ΔP_valve is nearly constant.
• Use equal-% when the system takes the bigger share (N ≈ 0.2–0.5) — the usual case with a heat exchanger, a long run of pipe or a flow orifice in series. The inherent curve's downward bow cancels the system's upward bow and the installed result comes out close to linear.
• Neither saves you below N ≈ 0.1. That is a hydraulics problem; give the valve more ΔP, or accept a loop that only works over a narrow band.

Rangeability R is the trim's own claim (30:1 or 50:1 for a contoured globe plug; ~15:1 for a standard butterfly). The installed turndown is always less, because at minimum flow the system's resistance has almost vanished, ΔP_valve is near the full pump head, and the valve must close hard down to hold the flow back.

See the trim itself — the cut-away model animates the plug lifting out of the seat, and the toggle swaps a linear parabolic taper for the slim equal-% needle so you can see why the same travel gives 51 % of Kv on one and 14 % on the other: ▶ open the interactive: instrumentation control valve authority 3d

Worked example — cooling water to E-2101. Design 100 m³/h water (SG 1.0). Pump differential across the control station 4.0 bar; exchanger + piping + isolation at design flow 2.0 bar.

ΔP_valve   = 4.0 − 2.0 = 2.0 bar        N = 2.0 / 4.0 = 0.50      ← healthy
Kv_required = 100 / √(2.0/1.0) = 70.7    → Cv_required = 81.7
Selected: 4" globe, Cv 100 (Kv 86.5), linear trim, R = 50
Travel at design  = (70.7/86.5 − 1/50) / (1 − 1/50) = 0.81  → 81 %
Travel at 40 m³/h : ΔP_sys = 2.0×0.4² = 0.32, ΔP_v = 3.68,
                    Kv = 40/√3.68 = 20.9 → travel 23 %
Installed gain at design = 0.55          ← inside the 0.5–2.0 band

Now the margins: process quotes 120 m³/h "design", the instrument engineer applies ×1.3 on Cv, the vendor's standard 6″ body with full trim gives Cv 250.

Oversize factor = 86.5-equivalent → Kv 216 / Kv_required 70.7 = 3.06 ×
Travel at 100 m³/h design flow = 31 %
Travel at 40 m³/h minimum flow =  8 %      ← the hunting valve of the opening story

Authority never changed — it is still 0.50. The valve was ruined purely by sitting on its seat.

And the other failure mode, same plant, different sin: a pump uprate leaves ΔP_total = 6.0 bar with ΔP_system = 5.7 bar, so ΔP_valve = 0.3 bar and N = 0.05. Travel at design is a respectable 84 %, but the installed characteristic has collapsed: the valve reaches 40 % of maximum flow at 6 % travel, and the installed gain at design is 0.06. Perfectly sized, completely uncontrollable.

Common pitfalls

• Sizing on the padded flow instead of the real duty, then rounding the body up to line size.
• Quoting a flow coefficient without saying whether it is Cv or Kv (a 16 % error, silently).
• Checking only the design point. Minimum flow is where oversizing shows up — always compute travel at both ends.
• Treating authority as a valve property. It is a hydraulic-design property; the valve inherits it.
• Specifying equal-% "because it is standard" on a loop with N ≈ 0.8 — you have then created a non-linearity that was not there.
• Reading rangeability off the catalogue and calling it turndown.
• Curing hunting by detuning the controller. It hides the symptom at the cost of response, and the seat still erodes.
• Ignoring ΔP_choked: on a high-pressure letdown the valve may be choked at every opening you care about, and the liquid sizing equation is simply the wrong equation.

Outcome

• Cv/Kv are measured capacities, not sizes: Kv = 0.865·Cv, and Q = Kv·√(ΔP/SG) for clean non-choked liquid.
• Stacked margins (flow padding × Cv factor × body rounding) routinely produce a 3× oversized valve that lives near its seat at 8–30 % travel — hunting, poor resolution, seat wear.
• Valve authority N = ΔP_valve/ΔP_total is set by the hydraulic split, not by the valve. Above 0.5 the installed curve ≈ the inherent curve; below ~0.25 it collapses toward quick-opening and the loop gain swings wildly.
• Equal-% trim bows the opposite way to the system distortion and linearises the installed curve for N ≈ 0.2–0.5; linear trim belongs where the valve already owns the drop; nothing rescues N < 0.1.
• Rangeability is a trim claim; installed turndown is what you get, and it is always smaller.
• Interactive: ▶ open the interactive: instrumentation control valve authority calc — ΔP split, inherent vs installed, travel at design and turndown, gain band, verdict.
• 3D: ▶ open the interactive: instrumentation control valve authority 3d — cut-away globe valve, travel slider, linear vs equal-% plug contours side by side.

Open items

• Add gas/steam sizing (expansion factor Y, x_T, critical-pressure ratio) — the current tool is liquid-only.
• Add a real pump curve so ΔP_total falls as flow rises; the constant-head assumption slightly flatters authority at high flow.
• Add noise prediction (IEC 60534-8-3/4) and an anti-cavitation trim selector to the choked-flow branch.
• Worked example against a real vendor Cv table and F_L value rather than generic figures.

Educational teaching material — simplified and illustrative. Not for engineering design use.
© EPiC Circle · provided by PipeAxis