One thing I didn't understand was the author's comment that "I bought the part off eBay, not from a reputable supplier, so it could have come from anywhere."
A 5-pack of quality 7805's can be found on Amazon for $5, including Prime shipping, (e.g. http://amzn.com/B00H7KTRO6), so what's the incentive to buy parts of unknown provenance on eBay or the like?
I ask not being a hardware guy myself, so genuinely curious, as I've heard stories like this before.
Again, fantastic overview of the chip, though. I learned a lot.
Now we had a big problem. Was counterfeit product getting into the mix? Our distributor claimed they were clean, the manufacturer claimed they were clean too.
Turns out that the manufacturer had temporarily moved production to a different plant during the Fukushima tsunami. There was some...confusion...during the move.
Tantalum prices being the way they were, they had the workers out there along the tracks picking up the chunks of metal from the weeds, collecting the pieces in plastic bins.
It was sent back to the smelter to be reprocessed, but the temporary shortage caused real problems with their delivery schedules.
And thanks again for the article itself. Very much enjoyed it.
On to your question about why: There are hundreds of electronics parts that are fairly standard and if people paid that much for them they would need a lot of money. Many /most people buying on ebay do it as a hobby.
A quick browse on digikey shows that <40 cents is easily found for a 7805 (http://www.digikey.com/product-detail/en/NCP7805TG/NCP7805TG...). Ebay would be even cheaper.
There is a huge problem in the marketplace right now with LM323K series regs (basically a 3 amp die in a TO3 "big transistor" mounting case). The problem is old video game systems (not home systems, but video arcade) and some other old appliances use them, so you could modify the heatsink and stick a modern LM323T (aka TO-220 package) or hot wire in a COTS switching supply, but it wouldn't look "stock" anymore. So the market is getting flooded.
Its "well known" among CPU collectors that when you get a (insert obscure CPU here) from China, you're not really getting that CPU, you're getting some random 40 pin DIP with repainted markings. The only honest and reputable seller I'm personally aware of in China who doesn't do this is utsource, the stuff they ship actually works.
Its bad when they take a random chip like a 16550 or 8255 and remark it as a Z80 or 8086 for the collectors, but its actually worse when you really do need a 10 MHz rated Z80 or 6502 and you're sold a remarked 2 MHz part which kinda sorta sometimes works. Or if you want a real PITA some of the repainters are fairly ignorant and will ship repainted 6809 for 6809E and vice versa. Or another hilarious one, not all 6502 are pin compatible so you get a WDC product repainted to be sold as a rockwell product, which again doesn't work a lot of the time.
No, ebay in 2014 is not really fun at all for a retrocomputing enthusiast or whatever you call it.
I guess the closest HN sports analogy would be ebay is flooded with collectors items claiming to be the 1790 world series collectors plate or the 1925 football superbowl.
If you're just trying to regulate 600 mA for your rasp pi or whatever its no big deal but its a hassle for repair/restoration of old equipment.
As for how the LM323 supports 3 amps, the datasheet attributes it to "new circuit design and processing techniques". I'm tempted to open one up and see what's different from the 7805 (I'm guessing a bigger output transistor), but I've already got too many chips I want to look at.
A logic chip like a microprocessor is designed for a particular supply voltage, if this voltage drops too much the logic circuitry will switch falsely. Say we had only a capacitor and we tried to power the logic chip with it. As the chip draws current the capacitor discharges - this is because current is movement of charge, so the charge (and energy) can come only by draining the capacitor. For an ideal capacitor the voltage is directly proportional to the charge across it, so as the charge drains the voltage falls. To hold the voltage constant we need to keep 'topping up' the capacitor with charge. This is what a voltage regulator does - it uses a negative feedback loop to sense the capacitor voltage and when that voltage falls the circuit provides just the right amount of charge 'juice' for the top-up.
As we take the foot off the clutch pedal in a car, the load gets engaged to the engine and if we sense a stall we press the gas pedal a bit. That's the imagery of a voltage regulator in action.
The capacitor plays a key role because the regulator feedback loop isn't very fast - one trouble with fast feedback circuits is chatter, or responding to every blip. Negative feedback circuits are designed to be more like ship wheels - they like to steer sedately and not respond to every excited cry from the mast. But what happens if a current blip arises because a logic circuit block turns on all at once (in response to some block of code)? That local current blip is provided by the capacitor, it acts like an ATM to provide local draws - but it still depends on the regulator to top it up.
In fact you can think of a battery as a capacitor that tops itself up via electrochemistry, it works as long as there are ions in the electrolyte. If instead of 'bandgap energy' we used the chemist's terminology of 'electrochemical potential difference', then the system similarity becomes evident.
Between different manufacturers there is easily an 10x difference in price of 7805, so counterfeiting is certainly worthwhile, even more so when 7805 is more of a description of function than of implementation. 7805 means three-pin linear regulator with dropout <= 2V, such and such accuracy, line/load regulation and noise and typical design uses these parameters, not actual parameters guaranteed by given name-brand manufacturer (which are sometimes significantly better than for original LM7805), so mostly no-one will notice if you take random Chinese 7805 and repackage it into name-brand package.
What I know, is that about 5 years ago I have found part marked NEC uPD2805 (or maybe 2905) in power supply for Canon fax machine and somehow found out that it is 7805-esque voltage regulator and used it in my bench power supply as such.
Also, these parts have a brand logo on them, but are not made by that manufacturer. If Hyundai sold cars with the word Honda and a Honda logo on them, without permission from Honda, they would indeed be counterfeits.
But of course, this happens a lot in experimental circuits.
Also check out the temperature derating curves.
That series has a thermal shutdown ckt so "operating nominally" above boiling pt is kind of hard to define.
There are also non-regulator thermal issues, like most cases are not designed to operate continually at 125C internal temps, and the electrolytic caps nearby the reg have their lifetimes drop with some polynomial of temp above room temp so smoking hot reg isn't good for everything nearby even if isolated by itself it would be OK.
Finally the general idea is still true, that you can get a nice 2nd degree burn or think something is "really hot" at a temp well below rated.
There are plenty of high power transistors rated to 150C junction temp. Not impossible with a really good package and heatsink (low thermal resistance) and good assembly technique to be able to boil water on one of those. As a stunt in the lab everyone tries this sooner or later to see if their anti-thermal runaway ckt is good enough and their assembly technique is good enough.
Starting in the 80s Low Drop Out LDO regs started hitting the market, using mosfets as the pass transistor instead of bipolar trans (and a few other changes) means you can run with much lower voltage drops.
However, legendarily, the bipolar voltage drop automagically provided some oscillation dampening and the LDO regs are notorious for being unstable, almost as much of a PITA as using a switching reg, almost. So sometimes with reactive (usually inductive) load at certain currents and voltages, they turn into little RF oscillators. Sometimes the oscillation is too high of a freq to interest the overheating and other self protection ckts so they literally catch fire and so on.
These 80s stereotypes of LDO regs have been continually improved since the 80s so they're somewhat more stable now.
Part of the shocker of both the bad 7805s in the story and the oscillating 80s era LDOs is the 7805 series used to be legendarily bulletproof because they have a lot of self protection circuitry built in. Blowing one up is non-trivial compared to most other chips. Usually decent reverse current protection, some reverse polarity protection, overheating protection, short ckt protection, they're pretty touch chips, so its surprising when they aren't.
I guess BJTs at that power level are slow, as opposed to mosfets, which are much faster, so there's that as well.
The non-rectangular resistors were inconvenient to represent. Also, the 7805 has a lot of overlapping transistors (e.g. sharing the collector), which messed up my original system that assumed everything inside a box belonged to the box, so I ended up tagging each component with a different color. Maybe this is more than you wanted to know about the system :-)
Fuses are a different alloy as well http://en.wikipedia.org/wiki/Fuse_%28electrical%29#Construct...
Bond wire diameters are (for historical reasons) usually specified in mils. A common bond wire diameter is 1 mil, which, depending on length, can reliably handle a bit more than an ampere without fusing. (Shorter lengths and lower temperatures help this.) Note that this number is quite conservative, since this is the fusing current at 125 degrees Celcius, and there is a quasi-exponential dependence on temperature.
(Also note that multiple bond wires can be used in parallel to increase the fusing current; in one image from the linked article  it appears that there are two bond wires in parallel, but kens points out below that this is a force-sense arrangement and not a double bond.)
As a result, when running a chip at room temperature, very high currents will almost always blow up the circuit rather than the bond wire. In over ten years I've only once managed to melt bond wires without also causing the chip to crater, and that was on a power converter with gigantic transistors (like, "see them with the naked eye" big) that was designed to be quadruple bonded and was only double bonded in an engineering sample.
More recently, copper bonding has become increasingly common. This is driven by higher conductivity of copper (and thus lower bond wire resistance and higher fusing current) and substantially lower cost compared to gold. However, copper bonding requires special care. For example, the bond pad structure must be built to withstand much more bonding force without cracking.
For a technology as old as the one being used to build this 7805---a mid-80s bipolar or BiCMOS fab, 1 or 2 layers of aluminum interconnect---it would probably take substantial effort to redesign the bond system for copper bonds (and there may not be enough metal layers to provide cushioning, so it might just be impossible to reliably bond with copper). Beyond that, there are only four bond wires (see the aforementioned image), so the cost is minimal (and at most bonding houses, packages come with a number of "free" bonds built into the price of the package). In addition, redesigning the chip would involve substantial engineering effort (redesign the layout, redesign the package, redo all the reliability qualifications), the cost of which would probably not be recuperated via the price difference between copper and gold.
EDIT: kens, good catch on the Kelvin connection. Now that I look at the die photo again, it's clear the left one is for sensing.
TL;DR: one bond wire to the output carries the current and the second bond wire on the output senses the voltage.
Very, very nice explanation of the 7805 though. This is the bread and butter of linear regulators.
ST's datasheet also has this standard disclaimer, so it's also perfectly acceptable for them to change it from a legal perspective:
STMicroelectronics NV and its subsidiaries (“ST”) reserve the right to make changes, corrections, modifications or improvements, to this document, and the products and services described herein at any time, without notice.
Maybe 90% or even 99% of designers would only be interested in taking a regulator, connecting input, output and ground, and perhaps some adding bypass caps, and stopping right there. Good enough to make it work in most situations.
But that final 1% are the really good engineers. They want to truly understand or grok what all the components of the circuit are doing. That's what lets them achieve so much more than the typical engineer.
From what I've seen (but I haven't really studied his designs in detail) Steve Wozniak was (is?) the epitome of a great design engineer. His designs were magical for their time. The article also mentions Bob Widlar, who was also truly one of the greats.