Designing for an Unstable Supply Chain

[seancsnm]

So I had the fun experience of getting my first job out of college less than a year before the COVID madness began. At the end of 2020 I had to take some time off for personal reasons. When I came back to electronics design, the design environment completely changed. First design I did used a somewhat demanding spec DC-DC converter and by the time I finished the design several components were  no longer in stock with 1+ year lead times for restocking. Eventually I managed to scrape something together but I feel it will be several years before I'll be able to replicate that design again. I'm working on several designs now and am realizing just how much the supply chain has been stripped bare. I'm looking for a microcontroller and thought I'd just grab the Atmega328 since it's pretty much ubiquitous thanks to the Arduino's popularity. But it's not being stocked at all right now at Digikey. Same story for a handful of higher performance op amps used in some popular app notes. I'm sure most here can relate.

From what I can tell, the best case scenario (for getting component stocks back up) is that a global recession brings demand down so that in a few years supplies will be brought back up. The worst case scenario is that global energy shortages continue that limit factory productivity and the conflict between China and Taiwan escalates, disrupting the Taiwan Semiconductor factories and shipments out of Taiwan. It takes years to build new factories so although they may help the situation, it won't be for 5+ years that they make a meaningful impact on the supply side. It's possible I'm being too pessimistic (and I hope that's the case), but the more I see the more I think the world is in for a very tough decade ahead.

So that brings me to the main topic of discussion: What are some strategies for designing around an extremely fragile supply chain? My generation and the past few generations of engineers have been spoiled with so many specialized components that we've taken for granted the analog skills that enabled us to advance so far in the first place. So what happens when a major design factor is whether you'll be able to get the components you need by the time you go into production rather than what IC best meets your design requirements? And even then, how do you ensure that you'll get the quantity you need or that the part will be available in a year if you need to make another production run? You can pre-order the "jelly bean" components and use them wherever possible I suppose, but that is a large up-front investment and doesn't guarantee you'll be able to meet design requirements down the road. You can build your design around popular parts with a lot of drop-in alternatives, but how do you find those common parts without a lot of experience? It seems like the more seasoned engineers just know the super popular parts. My approach has been to look up a general P/N and see if several manufacturers make it and stock it in high quantities. Looking at old texts such as application notes or The Art of Electronics also lists a bunch of the popular parts that probably have a lot of compatible parts given not too stringent design tolerances.

Have you all found things that have helped ensure you'll be able to get the parts you need for your designs? A lot of this ramble is just trying to spark ideas and discussion that will hopefully help me (and others?) improve my process for the coming years.

[DC1MC]

Here is the new agile development plan, three simple steps:
- First phase of the design is looking and see what is available now on the suppliers.
- Immediately do an initial BOM and have purchasing dept buy the WHOLE BOM for the whole estimated production of your gadget.
- Once all of the BOM items is safely in storage, start designing the actual product based of what you actually have in stock.

PRO: The wankery of having endless forum discussion about this LT or that TI DC-DC converter that has 96.83% on some specific loads vs 96.77% efficiency on all loads types is gone for ever, also the centicents price difference and having superior discussion and exasperated eye rollings and whining on that "overnighting" on some distributor didn't work properly and you had to wait until next day, the horror

PRO: Because once you have the parts in the initial BOM, is most of the time all that you have, one have to really estimate properly the actual implementation, no oopsies, or be responsible of a very expensive pile of useless stuff (that can be resold later, usually at a profit, but you'll be long fired by then).

PRO: Get rid of "independent consultants" and contractors, the companies need to have stuff in house, not buying the same designs again and again. That will clean the stables, forcing most of them in bankruptcy, early retirement and a life of collecting old TE and generally making place for a less entitled and humble generation of electronics designers, that will have to do whit what is available now for purchase, not what they think is is the next cool toy to play with and force it on the customers.

PRO: Allow for the software designers to take also the supremacy (and better pay), because your replaced MCU may be feature compatible, but the product needs a new firmware redesigned in short time.
Also this will get rid of "idiot savants", that do some crazy optimizations that only work on a specific series/model/silicone mask version of a MCU and it's totally useless and non-portable otherwise. You have a portable framework, that can be transposed to another MCU or GTFO. This in itself will make for better development process and getting rid of obsolete crap that is only alive because of this special optimizations.

So far, besides a moderate price increase, this chips shortage actually seems to only have advantages, this is why the industry maintains it, is a nice and profitable way to he wheat from the chaff, you can't cut it, you're out.

Cheers,
DC1MC


_{for highly sarcasm impaired morans, this whole post is a (almost) a joke}

[PlainName]

What, no cons?

Surely you risk ending up with loadsaparts you don't want because your initial fag packet spec turned out to be wrong but you didn't know until you did the design after buying all the parts.

And you need to know in advance how many you will sell. Sometimes you will struggle to shift any and other times your entire inventory will be gone in a flash (ho ho), and that last can be worse than the former for your reputation.

Also, you need to cash up front rather than being able to use sales to fund more sales.

[tszaboo]

So that brings me to the main topic of discussion: What are some strategies for designing around an extremely fragile supply chain?

If you build thousands of things, you buy the components for it before finishing the schematic.
If you build millions, you don't really have a supply problem, you have a lead time problem. That's also called "someone else's problem"
Also, good engineers are suddenly invaluable. Become one.