IR drop: The 'gift' that keeps on taking
There's no getting away from it: one of the oldest manifestations of Ohm's Law still affects us daily. There's voltage loss (drop) when current flows through resistance, your basic V = IR.
This phenomenon is so pervasive that I am still amazed that an internal-combustion car can even start. You've got a nominal 12-V supply and a cranking current of around 100 A, while you need 6 to 8 V to get that starter to turn over. Do the math and you see that even a mere 0.5 Ω resistance in the battery, connectors, and leads will take away five precious volts.
Yes, there are lots of low-power battery-powered devices in use (you can make your own list), but high-current devices are also part of our world. You've got home systems which are drawing tens of amps, all the way to servers with 100 to 200 amps/board and kilowatts per rack. That means you are fighting both IR drop and subsequent thermal dissipation.
The solution in many cases, of course, is to trade current for voltage. It's an approach that was known to early electric-power pioneers such as Edison, Tesla, Steinmetz, and Westinghouse, among others. It's also one of the few cases where the engineering tradeoff is not painful, and you get "something for almost nothing."
The IR drop is complicated by products that have wide ratios between their peak, average consumption, and quiescent consumption. In most cases, you have to design for the peak draw, which means you'll need heavy-gauge wires and connectors to minimize resistance, yet most of the time that virtue isn't critical, as there's a huge difference in drop between a 10A peak-current mode and the 0.1A quiescent state. So it's design overkill, except when it's not; under-designing leads a poorly regulated supply rail, which in turn leads to erratic performance even if you are not concerned about the dissipation and losses themselves.
I see many otherwise good designs which overlook this consideration. The schematic says "here's your power" but the physical cabling say "not quite." It's easy to inadvertently overlook a fundamental factor such as IR drop when you are worried about hardware issues such as nanosecond signal timing, signal integrity, EMI/RFI, PC board layout, drive currents, I/O, switching, and the many other issues which affect product design.
Have you ever seen a sophisticated design that missed an obvious fundamental like IR drop? Have you ever committed such a mistake yourself?