Sash Windows You Are Not Allowed to Replace

A box sash counted as six sealing lines, not four

Take a traditional box sash apart in your head before measuring anything for it. Two sashes slide vertically in a box built up from a pulley stile, an inner lining and an outer lining, with a hollow pocket down each side carrying a cast iron or lead counterweight on a cord over a pulley wheel. The outer sash runs between the outer lining and the parting bead; the inner sash runs between the parting bead and the staff bead. Where they cross, the two meeting rails are usually splayed so that closing the catch pulls them together. Every one of those parts is a separate face that air crosses, and the sash itself is not where any seal goes — the beads carry it, which is why the retrofit is a bead replacement rather than a strip stuck to a window.

That anatomy wrecks the perimeter arithmetic that works on a door. On an opening of height H and width W, a door takes seal around one leaf: 2H + 2W. A box sash takes seal around two leaves, each about half the height but the full width, so the verticals are counted twice — 4 x (H/2) of stile, plus the head, the cill, and the meeting rail where the two sashes cross. Work it as two sashes and you get 2H + 4W. On a 1.5 m wide by 2.0 m tall opening, the door arithmetic says 7.0 m and the sash arithmetic says 10.0 m: forty per cent more seal on the same hole in the wall. That single error is why sash draught-proofing jobs run out of carrier two windows from the end.

The counting is generous by one window width where the system seals only one face of the meeting rail rather than both, and generous is the correct side to be wrong on for something sold in whole rolls and carrier lengths. What no amount of counting will fix is the pulley: it is a slot in the stile with a moving cord through it, opening into a box void that is often connected to the masonry behind, and no brush profile closes it. Blanking covers exist, and so does converting to spiral balances so the box can be closed and insulated, but that second one removes the weights and is a change to how the window works — a conservation decision, not a draught-proofing one.

Three interventions, three different physics

Draught-proofing changes air leakage and does not change the U-factor of the window by any amount worth measuring. Secondary glazing changes the U-factor and, incidentally, the air leakage. Reglazing changes the U-factor and nothing else. Occupants routinely expect the first to fix the second, and a room that has been perfectly sealed can still be unpleasant to sit in because the glass is running at a temperature that starts a convection loop down its face. Deciding what to buy means deciding which complaint you are answering.

The numbers, in the units the European literature uses: four millimetre single glazing calculates to roughly 5.7 W/m2K by the method in BS EN 673; single glazing with secondary glazing behind it is commonly reported in the region of 2.5 to 3.0 W/m2K depending on the cavity and on whether the secondary pane carries a low-emissivity coating; a slim double-glazed unit in the sash lands somewhere around 1.6 to 2.2 W/m2K depending on gas fill and coating. Those are ranges from published calculation methods and maker literature, not a specification, and the figure that goes into a submission is the one on the tested product.

The heat loss calculator below is written in the North American convention, which is a different unit for the same quantity: U-factor in Btu per hour per square foot per degree Fahrenheit, capped at 1.3. Divide a W/m2K figure by 5.678 before typing it. That puts bare single glazing at about 1.00, secondary glazing at about 0.48, and a slim unit at about 0.32 — all inside the range. Type 5.7 into the field and it will be rejected, which is the correct behaviour and a confusing one if you have never seen the imperial convention. Four sashes at 1.0 by 1.7 m is 6.8 m2 of window; going 1.00 to 0.48 across a 50 degF design difference comes out near 1,900 BTU per hour, about 0.56 kW off the design heat loss of that room. That is a rate on the coldest day, not an annual bill, and it says nothing at all about the draught.

Why the Georgian sash beats the Victorian one, until it does not

Here is the result that changes where the money goes. Timber is a better insulator than single glass — BS EN ISO 10077-1 tabulates frame transmittances by material and thickness, and a softwood sash frame sits nearer 2 W/m2K than the 3.5 the calculator opens at, which is a metal-frame default. Single glass is 5.7. So on an unmodified sash window, the glazing bars are the best-performing part of the assembly, and a six-over-six Georgian sash with roughly a third of its area in timber runs an area-weighted composite around 4.5 W/m2K, while a two-over-two Victorian plate sash of exactly the same size, with maybe fifteen per cent timber, runs closer to 5.2. The window with more joinery in it is the warmer window.

That ranking narrows the moment secondary glazing goes in. Bring the glass component to about 2.7 W/m2K and the same two windows land at roughly 2.5 and 2.6 — the bars have almost stopped mattering, because the glass is no longer the outlier. Which tells you where a glazing upgrade is worth most: on a plate-glass sash, where glass is eighty-five per cent of the area, and least on a heavily barred one, where a third of the opening was never the problem. Run the comparison twice, once with bare single glazing at 5.7 W/m2K and once in the improved state: the glass field takes both, and the arithmetic is the same area-weighted sum either way.

Secondary glazing is joinery with a gap in it

Four decisions make the difference between a secondary glazing job that disappears and one that ruins a room. Where the subframe lands — against the sash box, back in the reveal, or on the face of the architrave. How it opens. How deep the cavity is. And what the panel is made of. Only the last of those is a shopping decision; the other three are setting-out, and they are settled by the reveal you have, the shutter boxes that may still be folded into it, and the curtain track somebody screwed into the soffit in 1970.

Cavity depth is where thermal and acoustic ambitions pull in opposite directions, and it is worth being blunt that you cannot have both from one drawing. Thermally, the resistance of a still air layer stops improving once the gap is deep enough for convection to start circulating in it — the calculation method in BS EN 673 shows the plateau, and it arrives in the teens of millimetres, so a deeper cavity buys nothing more. Acoustically the opposite holds: the mass-air-mass resonance of the double leaf falls as the cavity deepens, taking it below the frequencies traffic occupies, and maker literature and the tested figures — measured to BS EN ISO 10140-2 and rated to BS EN ISO 717-1 — put the useful range at a hundred millimetres and upward. A panel set tight against the sash is a heat answer. A frame set well back in the reveal is a noise answer. Choose which complaint you are being paid to fix.

It has to open, and for three separate reasons. The cavity faces attract dust and will need cleaning, which nobody does through a lift-out panel more than once. Summer ventilation through the sash has to survive. And where the window serves as an emergency escape opening — Approved Document B, Volume 1 in England, Section R310 of the International Residential Code in North America — a secondary panel that cannot be opened from inside without a tool has closed an escape route, which is the one failure on this page that is not merely expensive. Glazing in a critical location takes safety glass or an equivalent; Approved Document K covers impact in England, BS 6262-4 is the code of practice, Section R308 of the IRC is the North American counterpart, and BS 8213-1 covers safety while cleaning. On an upper floor a light acrylic panel on a hinged frame is often the better answer than glass on a lift-out one.

The subframe seal is the part that gets skimped and the part the whole assembly depends on. Bed the frame to the reveal on all four sides, sealed continuously, before the beads or stops go on. Old reveals are the awkward case here: friable lime plaster and chalky distemper give a sealant nothing to bond to, so rake back to something sound first. Watch the coverage assumption too — the tube figure in the calculator below is for a quarter-inch bead, and coverage falls with the square of the bead width, so a three-eighths bead pressed into a ragged reveal gets nearer eleven feet from a tube than twenty-five. That is not a rounding error; it is a doubling of the tube count on a house full of windows.

Ask the house how much of the leak the sashes actually own

Before any of this is bought, there is a question worth putting to the building rather than to a supplier. A house with twelve sashes may well have most of its air leakage in a suspended timber floor, three open flues, a loft hatch and the boxing round a soil stack. The windows get done first because they are visible and because they are what the occupant points at, and a great deal of money can buy an improvement in the low tens of a per cent while the floor goes on doing what it always did.

The measurement is a subtraction and any tester with a blower door can do it in an extra half hour. Test the house as found, to whichever procedure your tester works to — BS EN ISO 9972 and ASTM E1827 both describe the fan pressurisation method. Then tape polythene over every window opening and retest. Tape to the plaster line, not to the sash: the box void behind the pulley stile is part of what the window costs the house, and taping to the sash alone credits the window with a leak that no brush seal will ever close, which flatters the case for the work you are trying to justify. The difference between the two readings is the windows' share.

Put both readings through the same normalisation so they are comparable. A 110 m2 house with 2.6 m ceilings reading 2,400 CFM50 as found comes out at about 14.3 air changes per hour at fifty pascals; the same house taped, reading 2,050, comes out at about 12.2. Two air changes, roughly fifteen per cent of the total leakage, is what the windows own. Fifteen per cent is well worth having if the windows are also the comfort complaint and the noise complaint, and is poor value on its own if the floor has never been touched. What replaces the ventilation you remove — trickle vents kept open, extract that actually runs, and the requirements in Approved Document F or ANSI/ASHRAE Standard 62.2 — is the other half of that conversation, and the specific hazard of open-flued appliances in a house you have just tightened sits in the general draught-proofing guide, along with setting a target for the retest before the work begins.