Ordering Insulated Glass Units: Working Back From the Pocket

Nobody gets to choose the overall thickness

The enquiry arrives as one bad window and turns into fourteen. A unit greys over at the bottom of the cavity, someone stands back and looks along the elevation, and the same fabricator's work is failing across the whole block on roughly the same timetable. The frames are sound, the hardware is sound, the sashes are staying. What is being bought is glass to fit pockets built years ago by a company that may no longer exist, to a makeup nobody wrote down.

That is the constraint the entire order hangs from. Rebate depth, platform width and the bead profile were all cut for one overall thickness, and none of them is adjustable without new beads or new sashes — at which point the job has stopped being a glass order and become a window replacement with a different price and a different set of approvals. So overall thickness is not a figure you specify. It is a figure you measure, and every improvement you would like to make to the glass has to happen inside it.

Inside it there are three things and no more: an outboard pane, an inboard pane, and the sealed cavity between them. Fix the panes and the cavity is the remainder. Fix the cavity and the panes are the remainder. There is no fourth variable to spend, which is why the argument about going laminated on the inboard leaf for acoustics is never really about the laminate — it is an argument about what the spacer is allowed to be once the laminate has taken its share.

Getting the numbers off a unit that is still in the frame

Two dimensions govern the order and neither is the one people quote on the phone. The first is the overall makeup — outboard pane, cavity, inboard pane, added up. The second is the tight size: the actual glass rectangle, not the aperture you can see. Sight size is smaller than tight size by the edge cover on all four sides, and a batch ordered to sight size arrives short in both directions, fits nothing, and cannot be recut because it is a sealed unit.

The makeup can be read without breaking anything. A glass thickness gauge works off the reflections from each surface: a target held against the outer face produces a set of images offset from one another, and the spacing between them scales with the glass and the cavity. The comb version reads pane thickness alone; a digital meter reads both panes and the gap in one pass, which on a survey of forty windows pays for itself in an afternoon. Where a spacer is printed, the stamp on the bar gives the width and the manufacturer directly, and a torch held at an angle usually finds it.

Do not survey the whole elevation from the outside. Take the bead off one sash of each type, measure the rebate depth and the platform width, and record how much glass the bead actually covers. That single opening tells you what the pocket will accept, what edge cover the old unit was set with, and whether the drainage slots in the rebate are clear — and it is the difference between a schedule that fits and a schedule that gets remade at your cost.

1. Take one bead off per window type and measure the rebate depth and platform width before anything is ordered.
2. Read the overall makeup with a glass thickness gauge at the centre of the pane, away from the edge seal, and record all three figures separately rather than just the total.
3. Look for a printed spacer stamp with a torch raked along the cavity; it names the bar width and often the manufacturer.
4. Measure the sight size at both ends and the middle in each direction, take the smallest, then add the edge cover for both sides to reach tight size.
5. Check every opening for square across both diagonals — old timber sashes rack, and a unit cut square to a racked opening will bind on one corner.
6. Pass a low-emissivity detector over each face to establish which surface carries the coating and confirm it matches on every unit you are copying.
7. Photograph one spacer corner, one seal edge, and any capillary tube crimp, so the fabricator is quoting against evidence rather than a description.

Reading the dead unit before it goes in the skip

Glass surfaces are numbered from outside in: surface 1 faces the weather, surface 2 is the cavity face of the outboard pane, surface 3 the cavity face of the inboard pane, surface 4 the room. The coating lives on 2 or 3 and the choice is climatic, not arbitrary. Where the load is heating, the coating goes on surface 3 to hold long-wave heat inside; where cooling dominates, it goes on surface 2 to reject solar gain before it enters the cavity. Copy the position of the unit you are replacing unless you have a reason to change it, and if you change it, change it on every unit in the elevation — mixed coating positions on one facade read as a colour difference from the street.

Coating type matters for a different reason. A pyrolytic hard coat is fired into the glass and survives handling and cutting. A sputtered soft coat is not durable in air and has to be removed around the perimeter — edge deletion — so the sealant bonds to clean glass instead of to a film that will corrode inward from the edge. Edge deletion is a fabrication operation, but it is your problem when a unit arrives with a visible deleted band wider than the bead covers, and the width the fabricator uses is worth asking about before the batch is cut rather than after.

The edge seal tells you what the unit was built for. The standard build is dual seal: a thin polyisobutylene primary bead between the spacer sides and each glass face, which is what actually holds gas in and moisture out, backed by a structural secondary seal filling the channel behind it. Polysulphide and polyurethane secondaries hold gas well; silicone is far more permeable and is chosen when the edge is exposed to ultraviolet light or when the unit has to carry structural sealant glazing loads, which is the situation ASTM C1249 exists to guide. A silicone-sealed unit will lose argon faster than a polysulphide one of the same geometry, and that is a specification decision rather than a defect.

Two other things are worth a look before the old unit is broken out. A crimped capillary tube at the edge means the unit was made to travel over an altitude change and was breathed and sealed on arrival — and a unit with a breather tube was not gas filled, because you cannot do both. And an interlayer visible as a fine line at the exposed edge means one leaf is laminated, which changes the acoustic result, the safety classification and, critically for your arithmetic, the thickness left for the spacer.

Wider is not better once you are past the middle of the teens

Heat crosses a sealed cavity three ways: conduction through the gas, convection within it, and radiation between the two glass surfaces. Widening the cavity cuts the conduction, because there is more gas to cross. It also makes convection easier, because a taller, deeper cell of gas circulates more freely — warm gas rising at the room-side pane and falling at the cold one. The two effects pull opposite ways, so the total conductance falls with width, reaches a minimum, and then climbs again.

Where the minimum sits depends on the gas, on the emissivity of the coated surface and on how the unit is oriented, and EN 673 is the calculation method that puts it in the right place for a specific makeup rather than in general. For a vertical, argon-filled, low-emissivity double unit the minimum falls in the middle of the millimetre teens; for air it sits somewhat wider, because air conducts more and convects less readily than argon at the same width; for krypton it is much narrower, which is why krypton turns up in slim triple units and almost never in a 20 mm cavity. Sloped and roof glazing behave worse than vertical, because gravity and the temperature gradient are no longer at right angles and the convection cell has an easier time of it.

The practical consequence for a replacement job is blunt. If the subtraction leaves you at 20 mm, you have not bought a better unit than the 16 mm one next door — you have bought a wider one that performs about the same or slightly worse, and paid for the extra argon to do it. Spend the pocket on glass instead: a thicker or laminated leaf, or a better coating, both of which return more per millimetre than the last four millimetres of cavity ever will.

The bar is a desiccant container and a thermal bridge

A spacer does three jobs and only one of them is spacing. It holds the two panes apart at a controlled dimension, it carries the desiccant that scavenges the residual moisture sealed in at manufacture, and it conducts heat around the edge of the unit from the warm pane to the cold one. Traditional aluminium bar does the first two well and the third far too well, which is why the coldest strip of a well-specified window is the two inches nearest its frame.

That edge conduction is quantified as a linear thermal transmittance — a psi value, in watts per metre kelvin — and it is calculated by the numerical method in the EN ISO 10077 series rather than measured on site. Warm-edge bars replace most of the metal section with polymer or thin stainless, cut the psi value, and lift the interior glass edge temperature by a few degrees. On a whole-window U-value the improvement is modest and real; on condensation risk at the edge it is the difference between a sill that is wet every January morning and one that is not, which is what the customer will actually judge the job by.

The desiccant is molecular sieve, either loose beads blown into a hollow aluminium or stainless bar, or held in a matrix in a warm-edge profile. Its capacity has to cover the moisture sealed in during fabrication plus whatever crosses the seal over the unit's life, and it is sized by the fabricator against the cavity, not by you. What is yours to worry about is the corner detail: a bar with bent corners has one joint per unit instead of four, and every joint is a place the desiccant can leak and the primary seal has to be made good by hand.

What the safety rules do to the glass you were going to order

A replacement unit inherits the location of the one it replaces, and locations carry duties. Glazing in and beside doors, below a defined height in walls, around baths and showers, and at the bottom of stairs is treated as a critical location almost everywhere, and the glass in it has to be a safety material regardless of what was in the frame before. In the UK that route runs through BS 6262-4 and the impact classification in EN 12600; in the United States through the critical-location provisions of the IBC and IRC and the material requirements of CPSC 16 CFR 1201 and ANSI Z97.1; in Australia through AS 1288 and in New Zealand through NZS 4223, with the glazing material itself specified to AS/NZS 2208 in both. Finding an unmarked annealed pane in a door sidelight is common on older stock and it is not an option to copy it.

Which leaf becomes toughened or laminated is a decision with a thickness consequence, and that lands straight back on the spacer. A 4 mm annealed inboard leaf swapped for a 6.4 mm laminate to satisfy a critical location has taken 2.4 mm out of the cavity, and the makeup has to be re-run rather than assumed. Toughened glass in the same nominal thickness keeps the arithmetic intact but brings its own consequences: it cannot be cut or drilled after processing, it has a small residual risk of nickel sulphide inclusion failure, and heat soaking is the accepted mitigation where a spontaneous breakage would be unacceptable.

Thickness is also a structural question independent of safety marking. Wind load, unit size, aspect ratio and the load sharing between two panes of an insulated unit are what decide whether 4 mm is adequate, and they are handled by ASTM E1300 in North America and EN 16612 in Europe rather than by a table of spans. Large units, exposed elevations and anything above the second floor deserve that check before the panes are fixed, because increasing a pane thickness afterwards changes the spacer, the weight, the lifting method and possibly the frame.

Where a replacement batch actually goes wrong

The expensive failure is a schedule ordered to sight size. Sealed units cannot be trimmed, so every unit in the batch is scrap and the frames stay open behind boarding until the remake arrives. The second most expensive is a unit ordered to the right size with the coating on the wrong surface, which is invisible on delivery, obvious once the elevation is glazed in daylight, and equally unfixable.

Then there is the unit that fits and is installed the wrong way round. On an asymmetric makeup — a laminated or thicker leaf on one side, or a coating on 2 rather than 3 — flipping the unit reverses the performance and, where the laminate was there to satisfy a critical location, moves the safety glass to the wrong face. Mark the outboard face on the interlayer or the packaging at the fabricator and glaze to the mark, because nobody can tell by looking at a wrapped unit on a stillage.

Handling and storage account for most of the rest. Units are stored on edge on timber bearers, never flat and never leaning against the top edge of another; a sealed unit left face-loaded in the sun can deflect enough to stress the seal. Blocks go under the glass at the quarter points so the load runs into the frame and not through the edge seal, and rebate drainage has to be left clear — an edge seal sitting in standing water fails on a predictable schedule.

Finally, order the batch as a schedule and not as a sequence of phone calls. Fabricators cut in runs, and one late addition of two units in a different makeup can add a fortnight to a lead time that was fine. Get every makeup, every tight size, every coating position and every safety requirement onto one sheet, have the fabricator confirm the spacer widths against what they actually stock, and only then release it.