Not normal is now normalized: the last few years have been unpredictable at best. It feels time to share some experiences and observations of the state and future of small manufacturing.
Most people experience supply chain issues as simply not being able to purchase something, or only being able to do so at a greatly inflated markup. But another situation emerged that is even more frustrating: the potentially great difficulty of repairing existing things as a result of replacement parts either not being made or being unobtainable.
To lead with an actual case: we have not been able to get the STM32 microcontroller used in the Crow (and Grid) for well over two years now, and all estimates are “unknown” (the words from various sales reps I’ve spoken to, ie those at Mouser). Since we’ve sold out and also exhausted our backup-repair-stock, if someone contacts us with a hardware-related break-- for example, somehow the STM32 is fried-- there’s basically nothing we can do (that was previously standard procedure).
One resort is to salvage: when a PCB is sufficiently busted in more than one way, it enters the scrap yard (a particular cardboard box in our basement) where it can have further parts pulled from it. This was previously not economically worthwhile: the success rate of part swapping isn’t 100%. Frequently after substantial time has been spent lifting and installing, a “new” part would also be malfunctioning (likely when a PCB was subjected to “bad conditions”). So previously the parts would just go to e-waste since new components were widely available and irrationally cheap.
We are no longer in that era. So I’d like to instill this message, as strongly as possible: your machines are not easily repairable. They may be incredibly difficult (interpret: expensive) or impossible to repair. Please let this inform your treatment and expectations of the machines in your life.
It’s also potentially a good time to learn some repair skills. Soldering a DIY kit is a good start, but doing careful repair and salvaging is a fine craft. There’s a long tradition of musicians also being caretakers of their instruments (in school, as an oboe player, I was taught to make my own reeds, identify and fix leaks, tighten and lubricate various action… this is the same for basically most acoustic instruments)— consider this my call for a greater number of people to bring this attitude more widely to electronics.
Why not just put in more engineering time and use a different part?
This sometimes applies to a somewhat generic component (like a passive) but the industry now is mostly constructed using highly diverse, specialized, and non-pin-compatible semiconductors. (My frustration with this hyper-capitalist situation is the subject of another essay). For repairs, this is generally not an option.
For the production of an existing design it’s possible to rework the section of the specific part in shortage. This might mean changing a package size, moving signal connections around, adding/changing support circuitry, and/or changing firmware code. This invariably requires a full prototyping for confirmation which is expensive and long despite a large portion of the design staying the same.
And there’s no guarantee the new part won’t also go out of production, so the cycle begins again.
But of even more consequence: a single project with multiple designs adds a level of complexity which creates further problems for documentation, support, and maintenance. Recall this gem of a user experience:
We did this because the global supply of the pi 3 dried up (partially due to manufacturing crisis, partially due to basic planned obsolescence) and because the audio codec similarly was gone (due to a factory fire prior to the pandemic, which doubly made restock lag behind after the real crisis hit.
In addition to being confusing, this transition took a ton of engineering time— some of which was volunteer efforts by talented contributors who could’ve been working on more creatively impactful aspects. Support in service of the new confusion is likewise diverted from time supporting scripting education. This is the new cost of doing business.
Finally, for some parts (like the Crow’s microcontroller) there simply isn’t a well-matched similar chip, so a redesign isn’t really an effective option.
Won’t this all work itself out soon?
It’s hard to say, and most of what I’ve read and experienced suggests no, not really. Despite using predominantly domestic US manufacturing, we’re still hugely impacted by shutdowns in China due to the pandemic (which is not over, despite potential appearances). Fossil fuel prices are incredibly volatile (war/etc) and have a ripple effect out to every other commodity. Fires and floods and ice storms and hurricanes are all more frequent due to climate change, taking factories offline for potentially long stretches while aging infrastructure needs to be repaired (as happened recently in Texas). Political battles waged with embargo and import taxes (variations for each administration).
Skilled labor shortages also impact our small operation. We use a local circuit assembler. They’ve been experiencing a substantial amount of turnover during the pandemic and are frequently short-staffed. Despite their best efforts and the comparatively high costs we pay (versus an overseas assembler), quality has suffered. Which has created a load of in-house debugging and repair and costly send-back support.
As supply issues deepen it’s only more likely that small manufacturing becomes marginalized. Large corporations don’t buy capacitors from Digikey or Mouser: they go direct to the OEM. In many cases, they just buy the OEM, and have exclusive access.
Companies making expensive machinery (ie, heavy large tractors) have a pretty strong financial incentive to engage in edge procurement techniques. Such as, purchasing large quantities of new stock consumer goods (think, appliances) which contain a specific needed semiconductor, extracting that part for use, and discarding the rest. Acquiring a single part puts the remaining unused components into higher demand: a sort of avalanche effect (not to mention the huge amount of waste, since generalized recycling of e-waste still doesn’t “make sense” short of harvesting a small range of known high-value parts.)
It’s hard to see a return. More likely I’m expecting a sort of paradigm shift.
I don’t mean for this to be a drag. I think I simply need to share some of the invisible weirdness that deeply impacts this small universe. Written somewhat quickly without time to do annotations with references. I’ll attempt to edit this for clarity as certainly there are some issues/errors as written. Any suggestions or experiences or links to reading would be greatly appreciated.