One of my recent re-emerging hobbies involves basic analog and digital electronics. The process of ramping up into the hobby has resurrected an understanding of perhaps an older school of thought around “engineering” and provided a few interesting examples of the evolution of technology.
A few weeks ago, I ordered twenty NE555 timer chips from a vendor. Each 555 chip is an 8-pin, dual-inline package with 42 total components on the chip.
The 555 is used to create free-running square wave signals at arbitrary frequencies to provide a clock for timing digital systems or as part of the process of generating triangle and sine waves useful in audio applications. The chip dates from 1971 and currently costs about $0.50/chip even in quantities of 1.
The shipment arrived and I launched my tinkering successfully but received an email two days ago from the seller relaying a notice they received from Texas Instruments notifying distributors that TI was closing the remaining fabrication plant still making that original 1971-vintage part. They will still make the part in the same 50-year old DIP packaging format but will be using a newer fabrication plant which uses newer / smaller wafer-making / lithograph technologies. TI was giving a heads up that because there would be SLIGHT changes in MINOR characteristics of the chip’s performance, solutions depending on the old version of the chip better buy up what’s left and/or begin altering designs to accomodate any slight changes (improvements) in the specs of the chips from the newer plant.
That was a fascinating reminder of a difference between what I would call “real engineering” and “seat of the pants engineering” that is so prevalent in software driven markets today. In real engineering, planning for the life cycle of a PRODUCT has to incorporate planning for the life cycle of COMPONENTS within the product to avoid subtle changes in a component rippling through a larger system and leading functions out of designed boundaries. That reflects a world where once a product is physically made, physically recalling it to fix a minor defect in a $0.50 part is physically and economically impossible. It simply HAS to work across a WIDE margin of error.
In contrast, so much system functionality today is driven by software, engineers are prone to assume quality doesn’t have to be perfect because there are opportunities to overcome design flaws via weekly / monthly patches in perpetuity after the product is foisted on the public.
This isn’t just a philosophical exaggeration. This happens in software driving cars and it was the root cause of the Boeing MAX-777 failure with its MCAS system that resulted in two plane crashes and the loss of hundreds of lives.
Not only is Texas Instruments still operating with these long-term lifecycle concepts in mnd every day, so is this vendor. After they got the notice, they queried their database for all customer orders for that part and sent them notification as well. I had no idea they made that effort. I’m not building communications systems for Mars rovers, I’m just experimenting with guitar amps and related audio gadgetry.
A second example of the evolution of technology builds upon the first. That 555 chip has 42 elements on it (25 transistors, 15 resistors and 2 diodes). A state of the art Intel i9Core CPU might have somewhere around 25 billion transistors on its chip (arguably larger than a 8-pin 555 but still…) But a Core i9 chip is still $500 to $600 dollars. However, another area of evolution in electronics involves what are now termed microcontrollers or “SoCs” (System on a Chip) which contain a CPU, FLASH memory, Ethernet connectivity, USB connectivity and even Wifi and BlueTooth connectivity on a single board the size of a large USB thumb drive. Arduino was the first popular product in this space but now Espressive seems to be taking over with its ESP32 platform.
When today’s laptops have the equivalent of supercomputer power of 15 years ago, that level of functionality and miniturization isn’t surprising. What’s suprising is these microcontrollers cost about $9 a piece, fully assembled. For that you get a CPU running at 240 megahertz, 320kB of RAM and an open-source library allowing software to be compiled for the CPU in standard C/C++. I have two of these on order for experiments. (smile)
It’s staggering to contemplate the amount of computing power being distributed in throwaway electronics. It’s humbling to contemplate I may only be two or three years away from being unable to keep current on the cutting edge by mere dabbling in the technology. It becomes hard to gain and maintain a foothold of functional comprehension with these technologies with so many layers of abstraction piling up.