Why bars need a running start
Reinforcing steel only works where it’s anchored. A bar can’t deliver its yield strength at a cross-section unless enough embedded length lies beyond that section for bond — the grip between deformed ribs and concrete — to transfer the force. That required embedment is the development length, ld, and it explains a detailing rule that puzzles newcomers: bars always run past the point where the calculation stops needing them. End a bar exactly where the moment diagram says its force goes to zero, and for the whole region before that point the bar was never fully anchored at all.
ACI 318-19 covers this in §25.4, and the logic of its equations is worth internalizing before touching the arithmetic: development length grows with the force to be anchored (the yield strength fy) and shrinks with the concrete’s gripping ability (which scales with √f′c, not f′c). Everything else is a correction for circumstances — where the bar sits in the pour, what’s coating it, how big it is, and how well the surrounding concrete is confined. This guide follows one bar through the whole chapter: a 16 mm bottom bar (a hair over a US #5) with fy = 420 MPa — the metric Grade 60 — in 28 MPa (about 4,000 psi) normalweight concrete.
The general equation, factor by factor
The refined tension equation, Eq. 25.4.2.4a in SI form, reads ld = [fy·ψt·ψe·ψs·ψg / (1.1·λ·√f′c·((cb+Ktr)/db))] · db. The ψ factors each price one circumstance. ψt = 1.3 for top bars — those cast with more than 300 mm of fresh concrete below them, where settlement and bleed water weaken the bond underneath the bar — and 1.0 otherwise. ψe = 1.5 for epoxy-coated bars, whose slick coating cuts the grip, with the product ψt·ψe capped at 1.7 so the two penalties don’t fully compound. ψs = 0.8 rewards small bars (19 mm and under); ψg charges for high-strength steel at 1.0 up to fy = 420 MPa, 1.15 to 550, and 1.3 beyond; and λ = 0.75 penalizes lightweight concrete.
The denominator’s confinement term (cb + Ktr)/db is the interesting one. cb is the smaller of the side cover and half the clear bar spacing; Ktr indexes the transverse reinforcement crossing the potential splitting plane. Together they measure how hard it is for the concrete around the bar to split away — the actual failure mode of an under-developed bar. The code caps the term at 2.5: beyond that, splitting no longer governs and extra cover or stirrups buy nothing shorter. Two absolute floors close the section: tension development never falls below 300 mm, compression never below 200 mm.
The 16 mm bar, worked through
Assemble the example bar’s factors: bottom bar, so ψt = 1.0; uncoated, ψe = 1.0; 16 mm ≤ 19 mm, so ψs = 0.8; fy = 420 MPa sits exactly at the ψg = 1.0 boundary; normalweight, λ = 1.0; and generous cover and stirrups put the confinement term at its 2.5 cap. The equation: ld = [420 × 1.0 × 1.0 × 0.8 × 1.0 / (1.1 × 1.0 × √28 × 2.5)] × 16 = (336 / 14.55) × 16 = 369 mm — about 14.5 inches, or 23 bar diameters. Every input is auditable, and changing any single circumstance moves the answer by exactly its factor.
Try the two most common what-ifs. Make it a top bar and ψt = 1.3 lifts the length to 480 mm. Make it epoxy-coated as well and ψt·ψe = 1.3 × 1.5 = 1.95 would apply — except the 1.7 cap intervenes, holding the length to 1.7 times the base case, 628 mm… which is the point where hand-tracking factors starts inviting errors. The development length calculator runs the full ACI 318-19 §25.4 chain — both tension equations, all five factors with their caps, the lap-splice classes, and compression development — from the bar diameter, strengths, and condition toggles, so the factor bookkeeping can’t silently go wrong.
General vs simplified: what the shortcut costs
ACI offers a simplified alternative, Table 25.4.2.3, which for small bars reads ld = [fy·ψt·ψe / (2.1·λ·√f′c)] · db (divisor 1.7 for 22 mm and larger). For the example bar it gives (420 / (2.1 × √28)) × 16 = 605 mm — 64% longer than the general equation’s 369 mm. Nothing changed about the bar; the shortcut simply refuses credit for two things the general equation counts: the small-bar factor ψs = 0.8 and the actual confinement, which the table conservatively fixes rather than measures.
That gap is a design decision in disguise. On a drawing with a handful of bars, the simplified length costs a few hundred millimetres of steel and saves an hour of confinement bookkeeping — a fine trade. On a heavily reinforced job where every splice repeats thousands of times, the general equation’s shorter answers pay for their own paperwork many times over. Both are code-legal; the choice is economics. What isn’t optional is consistency: the confinement credit belongs only to bars that genuinely have the cover, spacing, and stirrups it assumes.
Lap splices: Class B is the default, not the exception
Bars come in stock lengths, so bars get spliced, and a tension lap splice is just overlapped development: each bar must hand its force to the concrete and back to its partner. ACI 318-19 §25.5.2 defines two classes — Class A at 1.0·ld and Class B at 1.3·ld — and the structure of the rule surprises people: Class B is the default. Class A is permitted only when the reinforcement provided is at least twice what’s required over the splice region and no more than half the bars are spliced within one lap length — that is, when there’s slack capacity and the splices are staggered.
For the example bar, the Class B lap is 1.3 × 369 = 480 mm, call it 19 inches of overlap for a 16 mm bar. Note what the 1.3 multiplies: the full development length with all its factors already applied, so a top-bar epoxy-coated splice compounds quickly, and note that the multiplier is not a strength increase — a Class B splice develops the same fy; the extra length covers the uncertainty of force transfer where many bars terminate at one section. Staggering splices isn’t just good practice for constructability; it’s literally what buys the shorter class.
Compression is easier — and the whole bar in one view
Bars developed in compression get a shorter assignment, because end bearing helps and there’s no flexural cracking working against the bond: §25.4.9 gives ldc = max(0.24·fy/(λ·√f′c), 0.043·fy) · db, floored at 200 mm. For the example bar: max(0.24 × 420/√28, 0.043 × 420) = max(19.05, 18.06) = 19.05 diameters, so ldc = 305 mm — almost exactly 12 inches, and notably less than the 369 mm tension length under identical conditions. The second term in the max() is a high-strength-concrete backstop: past roughly f′c = 31 MPa, √f′c stops helping and 0.043·fy governs.
The full ledger for one 16 mm Grade-420 bottom bar in 28 MPa concrete: tension development 369 mm by the general equation (605 mm by the simplified table — the cost of skipping the confinement math), Class A lap 369 mm, Class B lap 480 mm, compression development 305 mm, with floors of 300 mm and 200 mm underneath it all. One caveat belongs in every recap: these are straight-bar lengths. Standard hooks (§25.4.3) and headed bars (§25.4.4) solve the same anchorage problem in less space with their own equations — a different chapter of the same story.