Scarcity and Precarity in Small Manufacturing

Thanks for sharing this insight. I work in brewing, but I’m seeing so many echos of my own experience here as well.

On the equipment side, lead times on parts and repairs of things have been wild. We recently had a printer out for ~12 months while waiting on a replacement board for it.

Lead times in general are wild, and across the board quality has become a lot more variable even in standard soft goods (think hoses/gaskets/etc).

Volatility due to climate change on the agricultural side of things has been an issue for a while, but this past years malting barley harvest was big wake up call for a lot of people. Significantly smaller with a lower overall quality has required a lot of creativity on the side of maltsters and has required us as brewers to be more dynamic about how we are handling certain processes. We’re seeing situations like this for almost all ingredients.

Preventative maintenance and extra care / thoughtfulness are more critical than ever.

Very much this. With the changes/crumbling of the supply chain I don’t see how things can return to “normal”.

15 Likes

I kept wondering about this kind of thing during the absolute onslaught of new synth announcements at Superbooth. How is everyone going to make these things in the current environment? Or are most places designing new products around the current reality?

13 Likes

To the point of fabricating low-level controller devices like the STM32?

I agree with the above post suggesting we look towards a paradigm shift, but what to when basic parts are unobtanium with no estimates provided because corporations are amassing ten years stock at a time.

6 Likes

i believe that we’ll keep making instruments no matter how hard it gets to salvage parts. it’s in our nature - art never ends

9 Likes

No - I dont think hobbyists are able to fabricate stm32s at home - maybe you dont appreciate how much is involved in designing mcus? Even after you’ve designed it the cost to build a fabrication plant and fund a production run is into the billions of pounds so this is beyond the means of most hobbyists. You could google what is involved to gain some perspective on what you are suggesting.

A very easy alternative for the hobbyist (who has access to the schematics, firmware and board designs) is to substitute an alternative mcu - either another stm32, esp32 or even other arm-based mcus. (I have done this and it is pretty easy).

1 Like

That’s what i was questioning when it was stated that “if components are unobtainable the community can redesign relatively quickly”.

I don’t need convincing that we aren’t running home fabs anytime soon :slight_smile:

I just don’t where the base components are supposed to come from in that case!

“another stm32”

It’s not like we’re just talking a single implementation or form factor that’s been bought up for years and years of stock.

1 Like

If you have any questions on designing and fabricating stm32 I’m happy to try to answer - I used to work fot ST many years ago on the stm32 toolchain (before I left to study medicine).

It’s quite common in electronic design to redesign for alternative parts when some are unobtainable - usually the designer of the consumer device doesnt need to have a microprocessor custom-designed and fabricated as this is a hugely expensive andtime-consuming undertaking. They usually look to see what is available and redesign their product. This is a lot easier if the firmware, schematics etc are available. With modern tools (eagle, kicad, open source toolchains and access to cheap pcb manufacture) this is all easily within reach of the casual hobbyist….

esp32 are available as are other arm-core mcus. You can find small numbers of stm32 as they become available - I’ve picked up quite a few over the last 18 months - I’d be very concerned if I had to rely on supply for a commercial product but for a diy hobbyist there are small numbers availablr here and there.

I’ve also been playing with the new raspberry pi mcu (rp2040 and dev board) that is dirt cheap and easily available - TiNRS are redesigning their wobbler to use it…

5 Likes

There are still deep shortages – like zero available anywhere on the planet – of the STM32G4A1KEU6 or the STM32G491KEU6 which is the otherwise identical version with no hardware encryption support. This is the main brain of the EMpick. One of the driver chips is also currently unavailable but at least with that one they promise to have stock in a couple of months. There is no information about when the STM parts will be available. This isn’t a problem right at this moment. I do have the development hardware so I’m continuing to work on the code — all things that will have to be done before the first EMpick wakes up and says hello.

If this shortage continues I may release the build without the STM and check as much of the rest of the circuit as I can. However, the STM is still needed before anything can be shipped.

I’m still digesting this situation, still thinking it through. And still looking for at least enough parts to build one or two EMpicks so I can get to testing everything. It’s just insane. I can’t even find one chip, not one single STM32G491K. It’s not like this is some obscure chip. It’s mainstream. It ought to be easily sourced.

This is the situation. It’s not just this one part number, it’s all kinds of such chips. Many small companies are in the same boat. Larger companies with deeper pockets acted earlier and bought up everything, but that stock will be running out too, in 2022. So this could get worse. It is certainly one of the drivers of the economic slowdown, one they aren’t talking about much.”

^^
From a few blogposts of Paul Vo, who worked on the Moog guitar and is working on Wond II, on account of persons completely unable to find what they need.

Sure, a small number over 18 months for a single DIY hobbyist, I wasn’t saying that would be impossible, if i needed one i’d probably cannibalize something :slight_smile:

5 Likes

Yeah - it’s been pretty rubbish!
I was only considering the situation from the hobbyist’s point of view in keeping modules alive (or having access to them). For a manufacturer it’s crazy bad - the parts for the redesigned device are likely to have supply issues as well - again, for the hobbyist who only wants small amounts and can wait it’s more reasonable!
(But I have a number of orders with mouser that are over a year old and am awaiting estimated delivery next year….! For a hobbyist, I can manage this but not commercially!)
I think we agree though (I think!) - hobbyists could redesign stuff to use available alternative parts but commercially (for a manufacturer) this is a complete financial nightmare as it’s difficult to predict any future supplies of anything! (The hobbyist would need access to firmware, schematics, etc though in order to do this…)

I can see why my post about alternative stm32 would have touched a sore spot as you are developing a product with them! Sorry! I get how annoying and frustrating it must be - good luck with supply and development!

Ps - just been reading about the EMPick - it looks really cool!

5 Likes

Apologies if I phrased that poorly (appreciate your knowledge and perspective! thanks for taking my perspective in good faith) I am not he, I am frustrated for a number of small developers and companies who are providing a real-time depiction to me through a variety of updates how bleak things are, alongside friends in consumer products.

I am somewhat familiar with constant BOM substitutions / errata and trying to change packaging within similar substitutions, it just all sounds like a nightmare without easy individual, non-systemic fixes.

2 Likes

i realize that one reading of my post could be the subtext that monome is having a hard time, or concerned about the future. while i used our work as an example, i was speaking more broadly about the industry.

but since some people are wondering: we’re good. seriously. we may have a more difficult time building things and fixing things, but really we’re doing fine.

(i truly believe that part of why we’re fine is due to staying small and resisting the standard story. this requires more elaboration, perhaps for later).

120 Likes

this is good to hear

5 Likes

I’m sure I speak for everyone here when I say I’m eager to hear more about this. It’s so easy to think of these things in terms of a tradeoff between “ethics” and “practicality,” and I find it really refreshing to learn about situations where the ethical choice turns out to be more practical than the alternative.

Mutable Instruments ceasing production is another obvious example – her choice to open source everything from the very start also allowed her to make a clean exit without handing the company over to some larger and less benevolent entity. Component shortages notwithstanding, if you really want one of the newly discontinued MI modules, the resources are out there for you to have it made yourself. Both the producer and the consumers win.

21 Likes

Of course you are probably way too busy, but I feel like more stories from the whole Monome project would be so awesome!

Since you decided to small and deep instead of wide and shallow you ended up with some people who feel very close to the project (as of course this forum is a result of), so there are probably a lot of us here that would be super interested in all kind of real behind the scenes stuff of what it takes to do a project like this. Just a thought!

14 Likes

Some of those announcements were, I believe, probably meant to be earlier than Summer 2022.

2 Likes

I miss making things! My 25 years making printed circuit boards taught me a lot about troubleshooting and self-reliance.

Silver and gold price spikes, Arab oil embargo, OSHA, EPA, high interest rates (15.5 % for my first home mortgage) were all challenges. The IBM PC sucking up production capacity created tremendous demand and profits.

We closed in 1998 and there are few PCB shops left in my area. I moved on to IT for 12 years until that job moved offshore too.

Making stuff is hard, but very satisfying.

15 Likes

(post deleted by author)

2 Likes

I actually have an ST related question you can maybe shed some light on - why do they (still) offer hundreds of variants per core?

There seems to be every comination of 4 LQFP and BGA footprint, flash size etc, yet with the chip shortage, as the market governs the price, they all cost about the same (when they are available) - so wouldn’t it, at this point, make much more sense to keep say 3 kinds of STM32F4 LQFP and BGA available, and crank out an insane amount of them?

I have no idea other than it must make sense to them and make them more money! I dont know if silicon yield plays a part, legacy support, peripheral support (adcs, etc etc) plays a part, but they are highly tailored to different applications and there are real differences in the models in each family but at the end of the day I dont really know!

2 Likes

i’ve often thought about this as well.
i cant seem to find it right now but i remember coming across a txt file that someone on a german-language forum dug up from some ST files that listed a bunch of stm32 parts and correlated them with their unique internal die names. interestingly there were much fewer unique dies (dice?) than i expected… with many variations/families of parts being literally identical silicon presumably with sections of the chip fused off…
i vaguely remember someone doing some tests and being able to access extra ‘hidden’ flash and maybe peripherals on some parts. though i would imagine either these parts havent been tested/validated or have been tested and found defective, and potentially disabled peripherals wouldn’t have been bonded out?
actually now that i say that… i think only some of the G0 small package parts have multiply bonded pins hmm.

one surprise i remember was that a nontrivial amount of the H7 parts shared a die with the dual-core version but were nominally single-core parts… so there was a whole secret cortex M4 on-die sleeping in the shadows that would never get the chance to execute a single instruction… poor thing… interesting to think about

this could be a yield thing as forestcaver said, i also wonder if saving time on full-chip testing has any meaningful economic benefit?
i know at the low end packaging can be a significant part of the total cost per unit but again i’m struggling to understand if that plays a part…
i wouldn’t be surprised if there was only a small technical motivation behind all of this and was instead mostly driven by marketing targets.

i guess my take away from all this is that there are economic forces at play here that i really don’t understand… the more i learn the more i realise that chip fabrication is like some kind of parallel-universe-black-magic that has it’s own set of primitives and laws but it brings me so much joy that we can take part in it so easily and cheaply [current situation aside…]

i was looking at an i2c eeprom recently that listed a minimum memory retention of 200 years on the first page of the datasheet. i wonder what electronics will look like in 200 years. and i wonder what the accelerated-aging test apparatus looks like for them to be confident in the prediction…

4 Likes