Where Railing Sway Really Comes From: 0.51 mm in the Top Rail vs 3.89 mm in the Post — Splitting the Deflection Three Ways

A balustrade passes inspection, and then somebody leans on it and it moves. The reflex response on site is almost always the same: the top rail tube is too light, specify a heavier one. So the 50.8 x 1.5 mm tube becomes 63.5 x 2.0 mm, sometimes 76 x 2.5 mm. The railing is rebuilt, somebody leans again — and it still moves. Material cost is up by a third and the complaint has not gone away.

The reason is straightforward: the horizontal movement at the top of a railing is not produced by the top rail. Split one lean into its parts and the rail tube itself contributes 0.510 mm while the post contributes 3.89 mm at its tip — a factor of 7.6, with the post alone accounting for 88.4% of what the hand actually feels. Re-specifying the rail means spending money on a component that carries roughly one eighth of the problem.

This article separates the horizontal deflection at railing top into three parts — top-rail bending, post cantilever bending, and base/joint slip — and gives section properties, formulas and numeric results for each. It then does the two things that are actually useful on a live project: a compliance matrix for four common square post sections at 900 and 1200 mm spacing, and a per-metre steel weight comparison between the two remediation routes (thicken the post, or close up the spacing). One result runs against intuition: for the same verified performance, closing the spacing from 1200 mm to 900 mm uses about 5% less steel than thickening the post wall from 2.0 mm to 3.0 mm.

Scope and assumptions. The horizontal load at railing top is taken as the characteristic value q_k = 1.0 kN/m from GB 50009-2012 Table 5.5.2 (Load code for the design of building structures), applicable to residential, dormitory, office, hotel, hospital, nursery and kindergarten occupancies. Crowded public venues are commonly raised by the design documents; results for a raised 1.5 kN/m are noted where relevant. Deflection is verified under the characteristic combination (q = q_k); strength is verified under the fundamental combination (q_d = gamma_Q x q_k with gamma_Q = 1.5). Projects still working to the older partial factor of 1.4 can multiply the stress figures here by 0.933. Material is 304 / 06Cr19Ni10 stainless: modulus of elasticity E = 193 GPa, design bending strength f = 205 MPa (CECS 410:2015, Technical specification for stainless steel structures), density 7930 kg/m3. Carbon steel Q235 (f = 215 MPa, E = 206 GPa) is not mixed in. Two deliberately conservative simplifications are used: the top rail is calculated as a single simply supported span (in reality it is a multi-span continuous beam, with mid-span deflection roughly 0.2 to 0.6 of the simply supported value), and the post is calculated as a pure cantilever with a fully fixed base (in reality a lower rail, glass infill or vertical bars provide composite action and reduce movement). Results are therefore upper-bound estimates, which is the safe direction for a pass/fail judgement. Post-tip drift limits come in two tiers and the two must not be mixed: the code limit answers compliance, the 11 mm perception threshold answers whether it feels firm. Code limit: JGJ/T 470-2019 (Technical standard for building guardrails) clause 4.3.4 sets the displacement limit at the top of a guardrail post under design loads at 30 mm (clause commentary: 30 mm is the overall maximum deformation of the guardrail), with the top-rail deflection limit at 1/250 of rail length; clause 4.1.5 of the same standard requires a minimum 2.0 mm wall for stainless tubular posts and 1.5 mm for stainless tubular top rails. Perception threshold: on top of the code limit, this article also uses the engineering convention h/100 = 11 mm as a stricter visual-acceptance threshold, which is convention, not a code requirement, and must be confirmed by the designer of record. Excluded: base plate and anchor deformation, initial geometric imperfections of the post, weld flexibility, composite stiffness of glass infill, and wind or seismic combinations. Steel weights use 304 net section at 7930 kg/m3 and exclude weld allowance, processing waste and glass weight.

1. Separate the movement: one deflection, three sources

What the hand feels at the top of a balustrade is the sum of three flexibilities acting in series:

1. Top-rail bending deflection, f_rail — the rail is pushed sideways between two posts; governed by the rail's second moment of area I and the post spacing L;
2. Post cantilever drift, delta_post — the force arrives at the post tip and the post leans; governed by the post's I and the cantilever height h;
3. Base and joint slip, delta_base — anchor clearance, gasket compression, weld deformation between the base plate and the structure. There is no code formula for this one; it is controlled by detailing, not calculation.

These three add in series: with the hand at mid-span, the felt movement is approximately f_rail + delta_post + delta_base. Improving the feel therefore requires knowing which term dominates — spending money on the term that carries 5% of the movement will not be visible to anybody.

2. Choosing the load: what 1.0 kN/m is, and when 1.5 applies

The horizontal load on a balustrade is not "how hard one person pushes". It is a codified line load applied along the rail top. GB 50009-2012 Table 5.5.2 gives a characteristic value of 1.0 kN/m acting at the top of the railing, covering leaning and crowd pressure rather than everyday hand contact.

Two points that are routinely confused:

• Deflection checks use the characteristic value (serviceability limit state, characteristic combination), q = 1.0 kN/m = 1.0 N/mm;
• Strength checks use the design value (ultimate limit state, fundamental combination), q_d = gamma_Q x 1.0 = 1.5 kN/m = 1.5 N/mm.

Both are tabulated below: read 1.0 for movement, 1.5 for stress. For crowded occupancies (retail, stations, schools, stadia) designers frequently raise the characteristic value to 1.5 kN/m, which pushes the design value to 2.25 kN/m; multiply the stress results in this article by 1.5 in that case.

L is the post spacing and 1.1 is the interior-support reaction coefficient for a multi-span continuous beam under uniform load. This is the bridge in the whole analysis — it converts a load per metre of railing into a point force at one post tip.

3. Contribution one: deflection of the top rail itself

Two of the most widely used stainless round tubes:

Round tube properties follow I = pi(D4 - d4)/64 and W = pi(D4 - d4)/(32D), with d = D - 2t. Taking the rail as a simply supported single span under uniform horizontal load:

With q = 1.0 N/mm (deflection), q_d = 1.5 N/mm (strength) and E = 193 000 MPa:

The conclusion is blunt: the top rail is not the constraint. Even the lightest tube, 50.8 x 1.5 mm, at a generous 1500 mm spacing uses only 74% of the 205 MPa design strength and deflects a mere 4.84 mm. There is essentially no "not enough" case at this level, and any improvement from a heavier tube lands below the threshold a human hand can detect.

4. Contribution two: post cantilever drift — the actual driver

Now the post. Once the force reaches the tip, the post behaves as a cantilever fixed at its base:

Note the h3 term. Raising the rail from 900 mm to 1100 mm multiplies drift by (1100/900)3 = 1.83. This is what makes balustrades unusual: the lever arm is the railing height, and the height is fixed by code (residential balconies 1.05 m minimum clear, public building edges 1.10 m), so drift cannot be bought back by lowering the rail.

Taking h = 1100 mm and R = 1.1 · q · L:

Verdict basis, two tiers: stress at or below 205 MPa (304 design strength, CECS 410:2015); post-tip drift against the JGJ/T 470-2019 clause 4.3.4 code limit of 30 mm, with this article additional 11 mm perception threshold also listed. The fails-both flag in the table is judged on the 11 mm perception threshold; against the 30 mm code limit every post in this table satisfies drift and the only failing criterion is strength - 50 x 50 x 1.5 and 50 x 50 x 2.0 at any spacing, plus 50 x 50 x 3.0 at 1200 mm and 60 x 60 x 2.0 at 1200 mm, all exceed 205 MPa at the base, while 60 x 60 x 2.0 at 900 mm and both 60 x 60 x 3.0 cases pass. Separately, JGJ/T 470-2019 clause 4.1.5 requires a minimum 2.0 mm wall for stainless tubular posts, so the 50 x 50 x 1.5 post also breaches the wall-thickness floor outright, independent of the stress check.

The single most useful result in this table: the two most common posts on the market, 50 x 50 x 1.5 and 50 x 50 x 2.0, both fail a full 1.0 kN/m strength check at 1100 mm height and 1200 mm spacing. For 50 x 50 x 2.0 the stress is 368.6 MPa, 1.80 times the design strength, and the drift is 20.54 mm - inside the JGJ/T 470-2019 30 mm code limit, but 1.87 times the 11 mm perception threshold. That is not "slightly springy" — it is a genuine shortfall in load-bearing capacity. Real installations rarely fail in practice because lower rails and infill contribute composite action and because 1.0 kN/m is a deliberately conservative crowd-condition value; but relying on composite action as your safety margin is a bet, not a design.

5. Combining them: 88% of the sway you feel comes from the post

The code load is an extreme condition. What occupants actually notice happens under a much lighter one. Reproduce a single-person lean: a horizontal point force P = 0.5 kN at rail mid-span, post spacing L = 1200 mm, rail 63.5 x 2.0 mm, post 50 x 50 x 2.0 mm.

Substituting P = 500 N, L = 1200 mm, E = 193 000 MPa, I_rail = 182 883 mm4, I_post = 147 712 mm4, h = 1100 mm:

Here is the quantitative answer to "why does a heavier rail not help". Moving from 50.8 x 1.5 to 63.5 x 2.0 raises I from 70 647 to 182 883 mm4 — a factor of 2.6 — and cuts rail deflection from 1.32 mm to 0.51 mm, saving 0.81 mm. The 3.89 mm term does not move at all. Total drops from 5.21 to 4.40 mm if the rail were the only change, or from 4.40 to 3.59 mm on the 63.5 mm baseline: an 18% improvement, below the level occupants perceive.

Apply the same material increase to the post instead — 50 x 50 x 2.0 to 60 x 60 x 2.0 raises I from 147 712 to 260 459 mm4, only a factor of 1.76 — and tip drift falls from 3.89 mm to 2.21 mm. Total movement becomes 2.72 mm, a 38% reduction from roughly the same material addition. Twice the effect for the same spend, because the money went where the movement is.

6. Contribution three: base slip — no formula in the code, and the one that bites on site

The first two can be calculated; the third cannot, because it depends on workmanship. Typical sources:

• Gap between base plate and structure — uneven embedment or uncompressed shims leave a visible gap that closes under load, measured in millimetres;
• Anchor hole clearance — an M10 anchor in a 12 mm hole gives 1 mm of clearance per side, so 2 mm of base plate slip, the same order as the post's 3.89 mm;
• Expansion anchor loosening under repeated load — small at first but cumulative, and the usual cause of "it was fine when installed and wobbles six months later";
• Weld flexibility and local distortion at the post foot — thin-wall tube fillet-welded to a base plate will show plate warping under moment.

The remedy here is detailing, not arithmetic: base plate at least 8 mm thick, at least four anchors per plate, chemical or undercut anchors in concrete rather than plain expansion bolts, non-shrink grout under every plate, and a 0.5 kN hand-push retest on every post at handover. Holding this term under 1 mm is worth more per dollar than any section change.

7. Remediation compared: closing the spacing beats thickening the wall

Back to the practical question: the post has been shown inadequate, so where does the money go? Two routes — thicken the post (60 x 60 x 2.0 to 60 x 60 x 3.0 at unchanged spacing) or close the spacing (1200 to 900 mm at unchanged wall). Computed per metre of railing, including a 63.5 x 2.0 mm top rail, one 38.1 x 1.2 mm lower rail and 1100 mm post height:

Performance is effectively identical — 8.74 versus 8.17 mm of drift, 188.1 versus 175.9 MPa, both comfortably under 205 — yet the closer-spacing option saves 0.48 kg per metre, about 5.3%. Closing spacing from 1200 to 900 mm adds 33% more posts at unchanged section; thickening the wall from 2.0 to 3.0 mm raises mass per post from 3.680 to 5.424 kg/m, up 47%. The marginal cost of wall thickness is higher than the marginal cost of spacing.

Scale the view up: from the non-compliant 50 x 50 x 2.0 at 1200 mm (6.96 kg/m) to the lightest compliant option, 60 x 60 x 2.0 at 900 mm (8.66 kg/m), is 1.70 kg per metre, a 24.4% increase. Compliance costs about a quarter more material; remediation costs several times that in labour — replacing posts after handover means stripping the rail, the glass and the embedment, and labour runs well above the material delta.

One caveat: closer spacing is not always available. Glass panel module widths, the visual rhythm of the facade, and the position of cast-in embeds all constrain spacing. Where the architect has locked 1200 mm, the options are a thicker wall or a switch to a rectangular post oriented with its depth in the load direction (for example 80 x 40 x 3.0), which delivers a larger second moment of area for the same mass. Putting material further from the neutral axis is always more efficient than simply making the wall thicker.

8. Specification and handover checklist

In one line: railing sway is roughly eight-tenths a post problem, not a rail problem — under a 0.5 kN single-person lean a 63.5 x 2.0 mm top rail deflects just 0.510 mm while a 50 x 50 x 2.0 mm post moves 3.89 mm at its tip, an 88.4% share, so a heavier rail buys only an 18% improvement in feel; checked against GB 50009's 1.0 kN/m characteristic load, 50 x 50 x 1.5 and 50 x 50 x 2.0 posts at h = 1100 mm and L = 1200 mm reach 476.8 and 368.6 MPa at the base with 26.57 and 20.54 mm of tip drift, failing both criteria; and on remediation, closing spacing to 900 mm with a 60 x 60 x 2.0 post (8.74 mm, 188.1 MPa, 8.66 kg/m) uses about 5.3% less steel than thickening to 60 x 60 x 3.0 at 1200 mm (8.17 mm, 175.9 MPa, 9.14 kg/m) — while the gap between the non-compliant 6.96 kg/m baseline and the lightest compliant option is only 24% more material, against several times that in labour if it has to be redone.

Frequently asked questions

The railing feels loose — will a larger top rail tube fix it?

Largely no, and it is the most expensive way to achieve nothing. Split the movement and the reason is clear. Under a 0.5 kN lean at mid-span with L = 1200 mm, a 63.5 x 2.0 mm rail deflects f = PL3/(48EI) = 500 x 12003 / (48 x 193000 x 182883) = 0.510 mm, while a 50 x 50 x 2.0 mm post moves delta = (P/2)h3/(3EI) = 250 x 11003 / (3 x 193000 x 147712) = 3.89 mm at its tip — the post carries 88.4%. Upgrading the rail from 50.8 x 1.5 to 63.5 x 2.0 multiplies I by 2.6 and saves only 0.81 mm of rail deflection; total movement falls from 4.40 to 3.59 mm, about 18%, which occupants do not notice. Put the same material into the post instead — 50 x 50 x 2.0 to 60 x 60 x 2.0 multiplies I by only 1.76 — and tip drift drops from 3.89 to 2.21 mm, cutting total movement by 38%. Spend where the movement is.

Is a 50 x 50 x 2.0 mm stainless post adequate at 1.1 m height and 1.2 m spacing?

Against GB 50009-2012's characteristic horizontal load of 1.0 kN/m it is not adequate - though the failure is in strength rather than drift: drift stays inside the JGJ/T 470-2019 30 mm code limit while the base stress does not. The interior post reaction is R = 1.1 x 1.0 x 1.2 = 1.32 kN, giving a base moment M = 1.32 x 1.1 = 1.452 kN·m; under the fundamental combination (gamma_Q = 1.5) that becomes 2.178 kN·m, and with W = 5 909 mm3 the stress is sigma = 2 178 000 / 5 909 = 368.6 MPa, 1.80 times the 205 MPa design strength for 304. Drift under the characteristic combination is delta = 1320 x 11003 / (3 x 193000 x 147712) = 20.54 mm - inside the JGJ/T 470-2019 30 mm code limit, but 1.87 times the 11 mm perception threshold, which is why it reads as loose in the hand. Workable configurations are 60 x 60 x 2.0 at 900 mm spacing (188.1 MPa, 8.74 mm) or 60 x 60 x 3.0 at 1200 mm (175.9 MPa, 8.17 mm). Two qualifications: modelling the post as a pure cantilever is conservative, and composite action from a lower rail or glass and bar infill reduces real movement — which is why many existing railings that fail this check nonetheless perform acceptably — but treating composite action as the safety margin is not a defensible design position. Deflection and stress are checked under different combinations, and where a crowded occupancy raises the characteristic load to 1.5 kN/m the stress figures here must be multiplied by 1.5.

On remediation, should I thicken the posts or add more of them?

Add more posts, unless the architect has locked the spacing. Per metre of railing: 60 x 60 x 2.0 at 900 mm needs 4.50 kg/m of post and 8.66 kg/m in total including a 63.5 x 2.0 top rail and a 38.1 x 1.2 lower rail, with 8.74 mm drift and 188.1 MPa. 60 x 60 x 3.0 at 1200 mm needs 4.97 kg/m of post and 9.14 kg/m in total, with 8.17 mm and 175.9 MPa. Performance is effectively the same — 0.57 mm apart on drift — but the closer-spacing option saves 0.48 kg per metre, about 5.3%. Closing spacing from 1200 to 900 mm adds 33% more posts at unchanged section, whereas thickening from 2.0 to 3.0 mm raises mass per post from 3.680 to 5.424 kg/m, up 47%; the marginal cost of wall thickness is simply higher. If glass module widths or facade rhythm make spacing untouchable, the more efficient alternative is a rectangular post placed with its depth along the load direction, such as 80 x 40 x 3.0: the same mass delivers a larger second moment of area, because moving material away from the neutral axis always beats making the wall thicker.