http://shannon.engineeringcommunity.net/

Engineering Community, Scottsdale, AZ (2026)

09/08/2026

In the spring of 1978 Diane Hartley was a 21-year-old senior at Princeton, finishing her thesis on a new Manhattan tower. She got hold of its design calculations, and her adviser told her to check a quartering wind, one that hits a corner. A head-on wind presses one wall. A corner wind presses two. She worked the figures and thought a corner wind pushed harder, and the steel should have been sized for it.

So she rang. A junior engineer told her it could handle it: the code asked only for the head-on case. On the code he was exactly right.

The tower was Citicorp's, on Lexington Avenue. As it went up, Bethlehem Steel proposed bolting the brace joints instead of welding them. Cheaper, and routine. William LeMessurier had designed that bracing. His own New York office approved it in 1974. He heard of it in May 1978, four years later, and it did not worry him.

In late July, he ran the corner case himself and the number surprised him. His office had checked the head-on case and nothing else. He took his figures to Alan Davenport, whose lab had run the tower's wind tunnel tests. By Joe Morgenstern's account for The New Yorker, a sixteen-year storm could bring the weakest joint down, with the counterweight on the roof out of action.

He told Citicorp. By August, welders came in once the tenants had emptied out of the offices. They covered more than two hundred bolted joints with two-inch steel plates, from eight at night until four.

On 8 August, Citicorp put out a release. The bracing would be strengthened, on new wind tunnel data, and its engineers had assured the bank there was no danger. Two papers ran it. Neither asked what the new data was. That afternoon LeMessurier told the city's building commissioner everything.

The public got none of it. Hurricane Ella came up the coast while the welding went on. Plans were drawn with the police, the city's emergency office and the Red Cross, to evacuate the tower and a wide area around it. New York's major papers were on strike, and the work finished weeks before they came back. Morgenstern printed the story in May 1995, seventeen years later.

In 2019, the National Institute of Standards and Technology, a US government lab that studies wind on buildings, went back to it. Dat Duthinh and colleagues compared the two cases on a generic tower, and the corner wind loaded it less, not more. Each of the two walls has a worst push. The 1978 work added the two, as if they came at the same moment. Wind comes in gusts, Duthinh writes, so they do not. The corner number came out too big. The Citicorp tests are proprietary, so a generic tower is what they had.

No one here behaved badly. Hartley did the work and asked the question. The office answered her correctly. LeMessurier checked anyway, and had the tower strengthened before anyone made him. Hartley found out before Morgenstern printed it, in the early nineties, from a television report, holding her baby. Nobody had told her. Forty-one years later, Duthinh says the answer she was given was right, and no outsider can check that against the building.

This story joins Petroski, our line on engineering failures, as the one where nothing broke, with the sources behind every claim and the seven plates before it: shannon.engineeringcommunity.net.

Image: Norman McGrath, 1978. Library of Congress, LC-DIG-ppem-02530.

09/05/2026

We get corrected a lot. Readers in the United States tell us that windscreen should be windshield, and that toughened should have been tempered. That second one came up on our post about chemically strengthened glass. Readers in Britain tell us the reverse. Same thing with metres and meters. So we went to the standards to see who was right.

At first the two rulebooks do look like they have taken sides. UN Regulation 43 covers safety glazing on vehicles, and its Annex 4 is titled "Toughened-glass windscreens". The words "windshield" and "tempered" do not appear anywhere in the English text published in the Official Journal on 31 August 2010.

The American standard, FMVSS 205, uses the other pair. Its 2002 edition covered glazing made before September 2006. In that edition, one item is titled "Tempered Glass-Plastic for Use in All Positions In a Vehicle Except the Windshield".

Then you notice that the same 2002 American text also says "Motorcycle windscreens". Three times, and each one is a rule about the part of a motorcycle screen below a line 380 millimetres above the seat. We cannot tell you why it uses that word there and not the other one, because the document does not say. But both words are in the same American standard, in different places.

Meter and metre work the same way. NIST, the American government's measurement laboratory, published a guide to using the International System of Units. Its preface says which spelling it uses: "meter," not "metre."

Going forward, here is our approach. We will name the document we are quoting and its edition when it matters. If it uses a word in a particular way, we keep that word, even when the other one reads more naturally. And where two names really are two names for one thing, we pick one and give the other alongside it. If we have used the wrong one by your standard, tell us. You are probably right, and we would rather hear it.

09/04/2026

At Mach 3 the SR-71 was a bigger aircraft than the one that had left the ramp.

Its skin ran hot. A NASA history of the programme puts average surface temperatures between 462 and 622 degrees Fahrenheit. Friction and the engines took some parts of it "to temperatures as high as 1,050 °F".

Metal grows when it is heated. Titanium keeps its strength at those temperatures, which is part of why the aircraft is made of it, but it changes size like other metals.

So the aircraft had to be built to change size. A lot of what looks strange about it follows from that, including the thing it is famous for.

Let's start with the skin. The account says the outer surfaces of the wing panels were "beaded and corrugated to permit the skin and structure to expand and contract".

That is only half of it. The skin and the structure under it do not heat at the same rate, so they grow by different amounts. Something has to give, and it is the join: the panels are held on by fasteners "capable of sliding as the panels expanded and contracted".

Then the fuel, and this is where it gets expensive.

The SR-71 does not carry separate tanks. Its structure is the tank: the fuel sits directly against the airframe. So the joints that had to move were ones that had to hold fuel.

They were sealed with ten thousand linear feet of sealant, nearly two miles of it. It had to work across a range the account gives as minus sixty to more than six hundred degrees. That sealant "leaked a considerable amount of fuel", because of the very provisions that let the tanks expand and contract.

The leak came out of the very thing that let the tanks move.

There is a stranger consequence, and it depends on how much fuel is left. Down each side of the forward fuselage runs a long blended edge, called a chine. As fuel is used, the upper fuselage and "the cooler lower surfaces where fuel remained" end up at different temperatures. The two grow by different amounts, and that "caused the chines to be deflected downward", changing their aerodynamics slightly.

The shape it flew with was not quite the shape it was built as.

That is the history's account. NASA's own measurements come from a different aircraft, and they were after something else. It took a YF-12A, an interceptor from the same family with a different nose.

They wired it with strain gauges, which read how much a piece of metal is stretching, and heated it. The structure had been designed to minimise thermal stress. Even so, in some of those gauges "the thermal effects were the significant, and in some cases, the dominant, stress".

In places, the heat was doing more to the structure than the air loads were.

Which brings us to the part of the story you already know.

The usual ending is that once it was up to speed the metal expanded, the gaps closed, and the leaks stopped. We published that ending a year ago this week, and we would not publish it now.

Neither NASA document says the leaks stopped. Neither gives a figure for how much it grew, though "several inches" turns up in a lot of retellings without a source under it. Colonel Richard Graham flew the aircraft and wrote a book about it. Graham is quoted in the present tense: "Yes, the plane does leak fuel, but not enough to require refueling after takeoff."

We have not found a leak rate measured at temperature, in anything we read. It may have sealed. We cannot show you that it did.

There is a second account we cannot settle either, and it is worth having anyway. The SR-71 took on fuel again shortly after take-off, and the reason usually given is the leaks. Graham's explanation, reaching us through an article quoting his book, is that it was about the tanks. The fuel is a low-volatility blend called JP-7, and at cruise it gets hot enough to give off vapour. Vapour and air in the same tank is a fire risk. Filling each tank right up pushed the air out, and nitrogen then held it out. On that account, without the step the aircraft was held to about Mach 2.6.

We have not read his book. We are telling you what we hold.

The documents are AIAA 2009-1522, by NASA Dryden's historian, and NASA TM-104317 on the YF-12A thermal loads programme. Both are public. Every claim above is tagged in our open archive, next to the plate on the A-12 the shape came from: shannon.engineeringcommunity.net

Image: NASA, EC98-44817-2.

09/03/2026

In February 1968, a Corning engineer named John Howitt filed a patent for a car windscreen. The inner sheet was chemically strengthened glass, hardened in a hot salt bath rather than by heat, which makes it very hard to break. He put a deliberate flaw in it. The patent calls it a stress raiser. Its job is to make the strong sheet fail on purpose, at a chosen load.

A windscreen has two jobs and they pull against each other. It has to stop a head going through it, and it has to stop that head without injuring it. A sheet that simply resists does the second job badly. The patent's answer is to let the windscreen bend first and break second. The flaw acts, in Howitt's words, "after the glass has been deflected enough to slow down the impacting body".

He was not guessing. The patent says the idea was already borne out by "tests run at Wayne State University in August 1967".

Both halves of it were read out in May 1969, at one SAE meeting in Chicago, one paper after the other. Corning's described the design. Wayne State's gave the numbers for a windscreen built the same way, without saying who had made it.

The numbers are worth having. The standard windscreen of the day let go at 26 miles an hour. This one held to 35, and the peak head accelerations it produced were lower than any other type tested. On impact it pushed out into a bulge, which the authors say "decelerates the head over a long distance with low accelerations". The pocket that bulge made was lined with fragments less cutting than ordinary glass.

Then it reached a car, and here the record thins. Corning called the glass Chemcor. Two enthusiast histories of the 1970 American Motors line say a Chemcor windscreen was standard on the Javelin and the AMX, and neither of them cites a source. Corning's own history lists car windscreens among the things the glass was meant for, and never says one was built. That is as far as we could get.

Tested, patented, presented, fitted. And then, in the early 1970s, Corning shelved it.

Five Corning engineers, writing in 2017, give the reason. Other solutions were reaching the market, and the one they name is Pilkington's float glass, which had "dramatically reduced the cost of sheet glass".

We have written about float glass here before. It is the process that made ordinary window glass cheap and flat, and it is the one Corning names. Their own history says only that Chemcor did not ultimately "find its way into the intended applications as a fully-commercialized product line".

The chemistry outlived the project, and to see why you have to look at how the glass was hardened.

Our last plate in our archive covered the older way, with heat. Warm the sheet until it is soft, then blow cold air at both faces. The skin goes rigid while the middle is still hot, and once the middle cools the skin ends up squeezed. Squeezed glass is hard to break, because a crack has to beat the compression before it can start.

But that depends on the outside going cold while the middle is still hot. A thin sheet has too little depth between face and middle to hold much of a difference, and below about three millimetres the ordinary process becomes very difficult.

Chemical strengthening reaches the same squeezed surface another way, and it does not need that temperature difference. The sheet sits in a bath of molten potassium salt, hot, but far below the temperature at which glass softens. Sodium ions near the surface are replaced by larger potassium ions. The surface ends up squeezed because it is now carrying ions that do not fit the space they are in.

That works on sheets far thinner than heat can manage. Thin enough for a phone screen.

Corning is plain about what came next. Chemcor, it says, "was not the right product for mobile consumer electronics", and what the company sells for phones now is a different composition and a different product. But Chemcor and the chemistry behind it "offered scientists an entry-point". The windscreen that did not sell is where the phone glass started.

In 2017 the arrangement in the patent came back, on the Ford GT: an outer pane, a plastic film, and a thin chemically strengthened sheet behind them. C&EN reports the construction could cut about thirty per cent from a windscreen's weight, nearly seven kilograms on some SUVs. In tests it took more than twice the impact energy to form star cracks. It has been a factory option on the Jeep Wrangler and Gladiator since the 2021 model year.

It is not the same glass, though, and it does not carry Howitt's flaw. No modern account mentions a stress raiser, and his preferred inner ply was about 1.8 millimetres where the Ford's is 0.7. The AMC records do not describe the individual plies either, so we cannot say what the 1970 car actually had. Between it and the Ford we found nothing at all.

Howitt filed his patent in 1968. By 1970, a Chemcor windscreen was on a Javelin. By the early 1970s the programme was gone, beaten by glass that cost less. The chemistry went on without it, into a pocket nobody had thought of. Half a century later a version of the same three layers came back to the job it was built for.

The patent is US3473997A. The papers are SAE 690484 and 690485, and the 1970 follow-up is 700428. All of it is tagged in our open archive, next to the plate on float glass, which is the one Corning names: shannon.engineeringcommunity.net

Images: NHTSA, DOT HS 813 753; and US Patent 3,473,997, John S. Howitt, Corning Glass Works, 1968. Both public domain. Cropped, text added

09/01/2026

A car carries two different glasses, and it is worth knowing which is which. The windscreen is laminated. The side window, on most cars, is toughened, which is the same thing American makers call tempered. Both start as ordinary glass. They are finished in different ways, because they are not being asked to do the same job.

Toughening is done with heat. Take the sheet, warm it until it is soft, then blow cold air hard at both faces. The outsides cool and go rigid while the middle is still hot and loose. Then the middle cools, and tries to shrink, and the rigid outsides will not let it. The middle ends up stretched, and the skin ends up squeezed.

That squeeze is the whole trick. Glass usually breaks the same way: a flaw at the surface gets pulled open, and a crack runs from there. In toughened glass those flaws are being pressed closed, so anything trying to start a crack has to beat the squeeze first. Ordinary window glass carries 3 to 8 megapascals of compression in its surface. In toughened glass that surface compression is 80 to 150.

The stretch in the middle is stored energy, and it is why the window breaks up the way it does. Once a crack gets down to it, the crack branches, and branches again, and the whole sheet turns to small pieces.

A windscreen is built the other way about. Two thin sheets of ordinary glass bonded around a plastic film. Hit it and the glass cracks, and the film holds the pieces together. Nothing is squeezed and nothing is stretched, and all the strength is in the sandwich.

American regulators wrote the difference down in two sentences. Tempered glass, they say, is made to take rough treatment but is "not resistant to pe*******on". Laminated glass is not as tough, cracks more easily, and is "very resistant to pe*******on".

So the windscreen is the weaker of the two, and that is the point of it. The regulator credits it with stopping things, not with surviving them.

Toughened glass does the opposite, and it does one more thing. It does not crack and stay in the frame. It leaves. The regulator has a phrase for it, out of the procedure for breaking a pane before a crash test: tempered glass will "completely evacuate the opening".

Whether that was ever a reason anyone put toughened glass in a side window, we could not find out. The records we read give two reasons for it, and neither one is escape. The first is that it takes rough treatment. The second is how it breaks. There is a fracture test it has to pass.

The regulator gives that test a purpose in plain words: "to minimize the risk of injury caused by fragments of fractured glazing material". So the glass is there to survive a door slam and to come apart into pieces that will not cut you. Getting out through the opening is a third thing the same glass happens to allow, and none of the records we read claims it as a reason.

It matters now, because side glass is changing. Most people killed by being thrown out of a car go out through a side window rather than through any other opening. American regulators counted 6,412 such deaths a year, working from crashes between 1997 and 2008. The rule written to bring that number down is expected to save 373 lives a year.

The agency expected curtain airbags to do most of that work. For the faster of the rule's two tests, a maker may put laminated glazing behind the curtain as well. Carmakers have a reason of their own to like it, because laminated glass is harder to break into.

Among American 2018 models, about a third had laminated glass in the front side windows. Only four per cent had it front and back, so on most cars that have it, the rear side windows are still the toughened kind.

In 2019 the AAA tested six escape tools on both kinds of window. Four of the six broke the toughened glass. None of the six broke the laminated glass, which cracked and stayed exactly where it was. Rescue crews hit the same wall much earlier. In 2001 a trade journal reported that the spring-loaded punch a firefighter carries "caused only a spider pattern and pierced the laminate". It has to be cut.

You can settle which kind you have without leaving the driveway. Look at the bottom corner of the side window for the word TEMPERED or LAMINATED, or a maker's own name for the same thing. The AS code stamped there will not settle it, because AS2 covers both.

We keep these on an open archive, next to our plate on how a sheet of glass gets made flat in the first place. Every claim is tagged with where it came from, and the ones we could not settle are marked as open: https://shannon.engineeringcommunity.net/p/automotive-safety-glass

Image: hail damage to a windscreen, Moore, Oklahoma, 2010. FEMA photo by Win Henderson, public domain.

08/30/2026

Pan Am wanted a double-decker. Inside Boeing the argument went both ways, and plenty of Sutter's own people started out liking the idea. The objection that kept coming back was evacuation. A second deck sits a long way up, and emptying one in a hurry is hard.

So the aeroplane went wide rather than tall. A Boeing engineer named Milt Heinemann made the case in Pan Am's boardroom by stretching a twenty foot clothesline across it, so the airline could see what that much cabin actually looked like. Full size mock-ups settled it later.

Now the bump, which is an interesting decision. Juan Trippe of Pan Am did not expect the 747 to be carrying people for very long. Supersonic airliners were coming, and they would take the passengers. He told Boeing's chairman that when its time was done the aeroplane could go back to hauling freight, a Mack truck of the air.

So Boeing wanted it able to load through the nose. Freight does not require that and side doors exist, but a long load goes straight in through a nose and has to be turned through a right angle to go in through a side. They wanted the straight line, and what that costs you is the front of the aeroplane. The loading path runs where the flight deck would normally sit. So the flight deck had to move, and it went up. The first shape they put around it made too much drag, and it was drawn out backwards into the teardrop everybody can picture.

Then the lounge. Walking the mock-up, Trippe found the leftover space behind the cockpit and asked for passengers in it rather than crew bunks.

Boeing borrowed against all of it. Debt topped two billion dollars, and the one point two billion owed to its banks set a record for all corporate borrowing, not only in aviation. Bill Allen's verdict was that it was really too large a project for them. John Steiner, who ran product development, put it another way: we have never revealed how close we got to the edge.

In the middle of that, Allen proposed saving money by taking a thousand engineers off the programme. Joe Sutter, who was running the design, told him they needed eight hundred more. He went home certain he had lost his job. Nobody mentioned it the next morning, so he kept working.

Getting people out came back. There is a rule about how long it can take to empty an airliner, and while the 747 was being built it got harder: a hundred and twenty seconds became ninety. The new limit applied to this aeroplane. So in February 1970 they put four hundred and ninety nine people inside one, in a darkened hangar at Boeing Field, and emptied it in ninety seconds.

None of it happened the way Trippe expected. Boeing's own supersonic airliner was cancelled by Congress in 1971, before it flew. Concorde flew, and ended commercial service in 2003. The last 747 was handed over at the start of 2023, and it was a freighter.

We keep these on an open archive. This one lands next to our plate on that ninety second rule, and next to Concorde, which we published this week. Every claim is tied to the document it came from, including the two we got wrong on the way here: shannon.engineeringcommunity.net

Image: the nose cargo door of a Boeing 747-8F, open during static display at the 2011 Paris Air Show. Photograph by Olivier Cleynen, CC BY-SA 3.0, via Wikimedia Commons.

08/29/2026

Every second Concorde was at cruise, the air arriving at its intakes was doing about 1,350 miles an hour. And every second, that air had to be brought below Mach 1 before it reached the engine face. Subsonic air is the only kind the compressor could use.

There is a crude way to do that, and it costs you. Put a shock wave square across the duct and let the air hit it head on. A shock is a sudden change rather than a gradual one, and sudden is the expensive kind. At Mach 2, one shock set square to the flow destroys about 28 per cent of the total pressure. Total pressure is the measure of what the intake has left to hand over. So about a quarter of what it had just captured would be gone at the door.

So Concorde's intakes did it in stages. Two ramps in the roof of each one made a series of gentle squeezes instead of one violent one. One of them was shaped as a smooth curve rather than another sharp shock, and NASA's study of this kind of intake gives the reason. It was kinder to the thin, slow layer of air dragging along the wall.

Then they threw air away on purpose. Six per cent of everything the intake had just caught left again through a slot in its own throat, so that the rest would flow cleanly.

At what the study calls the critical operating points, meaning with the shock sitting exactly where it should, ninety six per cent of the total pressure survived.

There is nothing to see in any of it. Two ramps turning in the roof of a duct, and a slot letting a little air go. On the other side of that wall, a cabin full of people having lunch at twice the speed of sound.

You will read that Concorde's intakes made most of its thrust. It might well be right. We went looking for the document that settles it and did not come back with it, so it is not in this story.

We keep these on an open archive. This one is the 1976 entry on the commercial aircraft line, alongside the Comet, the 747 and United 232. Every claim there is tied to the document it came from, and the ones we could not settle are published as open questions: shannon.engineeringcommunity.net

Image: looking up into one intake pair on Concorde F-BVFA, on display at the Steven F. Udvar-Hazy Center. The two pale panels sloping down into the ducts are the ramps. Photograph by Dockurt2k, CC BY-SA 2.5, via Wikimedia Commons.

Engineering Community

09/08/2026

09/05/2026

09/04/2026

09/03/2026

09/01/2026

08/30/2026

08/29/2026

Address

Website

Alerts

Shortcuts

Share

Listing takedown request