When will we return to normal microcontroller stock availability?

[luiHS]

We have been out of stock of microcontrollers and other active components for a long time, such as accelerometers, CAN FD controllers, I2S sound amplifiers and others.

Last year it was said that this year the stock would return, but it has not been like that, it is worse, there are still fewer components than last year. The distributors put some expected dates, I think theoretical, of availability, in the best case for December of this year, and in the worst case for the first half of 2023.

Are these dates credible, really by the end of the year or early 2023 we will have stock of microcontrollers again?

I was farsighted and spent money to have microcontrollers and other components with which to work, although seeing now the total scarcity of everything, I think that I should have even spent more money to buy microcontrollers and other components in quantity (CAN FD controllers, amplifiers sound I2S, accelerometers).

[T3sl4co1l]

Probably next year, as factories get back online.  Remember it's not just the disruption of supply chains and labor, but also a fab was knocked out by natural disaster.  So, spotty availability for quite a while until that gets back into operation.  Or even longer as other proposed new fabs come online.

Tim

[cfbsoftware]

"These 5 Charts Help Demystify the Global Chip Shortage" IEEE Spectrum, 14 Feb 2022:

https://spectrum.ieee.org/global-chip-shortage-charts

[woofy]

It's going to be well in to next year at least. I'm seeing no signs of any immediate recovery with most quotes coming back for 2023 deliveries.
The longest lead time so far is 2027 for a TI part (via distribution).
One distributor for an ST sensor chip was unable to give any delivery date but said we should place the order now and pay up front, to secure the delivery. Yeah, right!

[SiliconWizard]

Of course the question is for us - small or even medium businesses.
The large corporations don't seem to have much problem with availability.
https://hypebeast.com/2021/11/apple-iphone-13-reportedly-sold-more-units-than-iphone-12

[brucehoult]

Apple is a bit different.

While they get 3rd parties to manufacture components for them, in many cases Apple pays for or supplies the equipment (or even whole factory) needed to make the components, and gets exclusive access to the items produced.

This of course means they need to predict demand years in advance.

[rpiloverbd]

I don't think this problem will end soon.

[SiliconWizard]

Apple is a bit different.

Pretty much all large corporations are a bit different, which was my point. Can you tell me one example of a *large corporation* making eletronic devices that is truly affected by this shortage to the point of having seen both their sales and value plummet? In comparison, many smaller ones are in this case.

While they get 3rd parties to manufacture components for them, in many cases Apple pays for or supplies the equipment (or even whole factory) needed to make the components, and gets exclusive access to the items produced.

For their own processors, sure. But they are also using a lot of third-party components. And they still do not own the means of production (foundries).
But of course, one reason of the above is just that those companies have enough cash, power and influence to go through the crisis relatively unharmed. Or even, take advantage of it to get even further ahead of their smaller competitors.

This of course means they need to predict demand years in advance.

Sure. Did they predict any of what happened during the last 2 years?

[coppice]

Apple is a bit different.

Pretty much all large corporations are a bit different, which was my point. Can you tell me one example of a *large corporation* making eletronic devices that is truly affected by this shortage to the point of having seen both their sales and value plummet? In comparison, many smaller ones are in this case.

It has less to do with size than with planning. Nobody wants to upset a really big customer, but that hasn't stopped the major car makers getting screwed, and they are a group that work very very closely with their suppliers - both the module makers, and the key component makers those module makers use - all the time. It seems most car makers cut their ongoing orders too much, so their module makers cut their ongoing orders for components too much.

[Sal Ammoniac]

Apple are the absolute masters at managing a supply chain.

[brucehoult]

This of course means they need to predict demand years in advance.

Sure. Did they predict any of what happened during the last 2 years?

Apple's revenues are very smooth and predictable.

https://twitter.com/asymco/status/1486870094697402368/photo/1

The "last 2 years" i.e. COVID doesn't seem to have affected that.

Many car makers apparently thought sales would fall, and made huge bets based on that, cancelling component orders. And were dead wrong.

Note that Apple's revenue increase in 2021 can probably be attributed to the introduction of the ARM-based Macs, not COVID i.e. the revenue and the associated production capacity was probably planned for.

Other PC manufacturers have had quite large sales increases, but Apple tends to be already running factories at full capacity, and selling everything they can make, so they can't take advantage of unexpected events that increase demand. Their production capacity is locked in one or two years in advance.

Horace Dediu ("asymco") sometimes publishes charts of Apple's capital expenditure (including buying equipment for their suppliers to use) vs revenue. They very clearly are in virtual lock-step, with a delay.

https://twitter.com/asymco/status/894913508034449408

[hans]

Apple can relatively easily predict how many phones and notebooks they must produce for a given period. If they have a 1 or 2 year refresh cycle, you just look at the previous (couple of) product release cycles and add any growth and/or inventory margin on top. They have the advantage that it's a consumer item and people buy into the hype of release, the hype around presents for the holidays, or whatever yearly trends they can have in sales.

I haven't worked at a large company design team, but I imagine that a chip is not even getting anywhere close to the schematic or PCB layout if the logistics (and legal) team haven't got the supplier on contract to supply e.g. 100M units with specified delivery dates. There are only so many crucial parts that absolutely must be available to e.g. make an iPhone, like the RF chipset from Qualcomm, or their own silicon from TSMC.

But if we look at a voltage regulator. Even though it's probably not a jellybean LM317 with a classic circuit, you can still view many DC/DCs, transistors, glue logic, connectors, etc. as general purpose building blocks. Unless you're really pushing a part up to it's absolute maximum, then a substitute must be atleast "as good". But there are so many alternatives that can fulfill a general purpose task.

I've looked at some motherboard datasheets in the past, and something I've noticed is the absence of boutique chips on their designs. I bet they rather use a dozen passives and discretes to turn on a power rail with some delay time, rather than the "integrated solution" of a high-side load switch that can be enabled using some RC time constant.

[rfclown]

Apple is a bit different.

Pretty much all large corporations are a bit different, which was my point. Can you tell me one example of a *large corporation* making eletronic devices that is truly affected by this shortage to the point of having seen both their sales and value plummet? In comparison, many smaller ones are in this case.
...

Not all "large corporations" are like Apple that sell very large volume consumer products and can make long term deals with fabs to secure parts (for which they better have figured correctly). I was at Harris (now L3/Harris) in 2018 and obsolescence/part shortages were a huge problem. In the one manufacturing facility I visited, they had Arrow using space in their building. There were two very large parts inventories in the building, their own and Arrow's; it was very impressive. They had resources to try to deal with the problem... but the problem is big. I don't know if their sales or value have "plummeted" because of the part shortage problem, but it certainly has caused them to spend a significant amount of time and money trying to deal with the situation.

[peter-h]

I think the problem continues because bigger companies continue to hoard stocks.

With near-zero interest rates, hoarding is a cheap insurance.

Eventually the whole thing will blow up, and when it starts it will blow up big-time, with panic everywhere.

Re Arrow, I find that quite funny because they are among the most useless distributors I have ever used here in the UK

[nigelwright7557]

The shortage feeds on itself.
There is a shortage so people panic buy making it even worse.

[DavidAlfa]

My company (Not my own, but where I work at) has finally come to a dead end, no way to supply lots of parts, even simple passives.
They're halting till end of the year, except very small batches.

[Karel]

Nobody wants to upset a really big customer, but that hasn't stopped the major car makers getting screwed, and they are a group that work very very closely with their suppliers - both the module makers, and the key component makers those module makers use - all the time. It seems most car makers cut their ongoing orders too much, so their module makers cut their ongoing orders for components too much.

(Electric) Car manufacturers aren't big customers for semiconductor manufacturers.


"Nope, nobody's interested in dealing with that low a volume. And with good reason. 10K/yr on 28nm for even large chips
isn't a boat of wafers. You never start less than 2 boats, and a big fab will run thousands of boats/day even when idling.
So we have to keep several sets of masks around, calibrate the machines for each set, then run them once a year, deal
with special probe cards and tester programs etc. You're a really expensive customer to keep around, and having to switch
to run just your product will slow down the rest of their production. Then top it off with all the extra paperwork aerospace
chips usually need. You think your chips are special, but to a fab, they're usually nothing but annoying noise in the overall
scheme of things.

This is the same problem we have with cars. For some reason they think they can just waltz in and demand premium treatment.
Sorry, folks, you run on the same old lines as most power electronics, and frankly even a low volume power supply run is bigger
than what a car company needs. If we have to prioritize, we'll make more money on the power supplies than we will on what they
car companies want to pay. And besides, car companies are in love with JIT delivery and as I mentioned above, JIT isn't something
you do on those very low volume runs. It's why when car companies cancelled all their orders at the beginning of the CCP flu,
then decided to reorder, we had to tell them they'd have to wait because the power guys had bought all the slots for over a year.
It's only now they're starting to get appreciable supply.

The economics of a chip fab are nasty, and they're especially bad for small run parts like aerospace or cars. It's what's driving the
incredible consolidation in the electronics industry today. Yes, my company is one of the larger car/industrial chip suppliers,
and in the overall scheme of things, that segment is about 1% of the company."


https://www.theregister.com/2022/02/04/nxp_chip_deals/

[SiliconWizard]

I think the problem continues because bigger companies continue to hoard stocks.

Yes, that's a large part of it. But apparently this is just because they are planning so much better than us poor chaps, not because they have loads of cash and influence. =)

With near-zero interest rates, hoarding is a cheap insurance.

This *was* also a significant part of it, but interest rates are going to ramp up...

[peter-h]

Sadly, interest rates (below about 5-10%) are irrelevant in this, because if you can't ship your product, you are totally buggered

Hoarding is thus a pretty "safe" option. It just screws everybody else, because in modern times the whole supply pipeline is operated tightly.

[coppice]

Nobody wants to upset a really big customer, but that hasn't stopped the major car makers getting screwed, and they are a group that work very very closely with their suppliers - both the module makers, and the key component makers those module makers use - all the time. It seems most car makers cut their ongoing orders too much, so their module makers cut their ongoing orders for components too much.

(Electric) Car manufacturers aren't big customers for semiconductor manufacturers.


"Nope, nobody's interested in dealing with that low a volume. And with good reason. 10K/yr on 28nm for even large chips
isn't a boat of wafers. You never start less than 2 boats, and a big fab will run thousands of boats/day even when idling.
So we have to keep several sets of masks around, calibrate the machines for each set, then run them once a year, deal
with special probe cards and tester programs etc. You're a really expensive customer to keep around, and having to switch
to run just your product will slow down the rest of their production. Then top it off with all the extra paperwork aerospace
chips usually need. You think your chips are special, but to a fab, they're usually nothing but annoying noise in the overall
scheme of things.

This is the same problem we have with cars. For some reason they think they can just waltz in and demand premium treatment.
Sorry, folks, you run on the same old lines as most power electronics, and frankly even a low volume power supply run is bigger
than what a car company needs. If we have to prioritize, we'll make more money on the power supplies than we will on what they
car companies want to pay. And besides, car companies are in love with JIT delivery and as I mentioned above, JIT isn't something
you do on those very low volume runs. It's why when car companies cancelled all their orders at the beginning of the CCP flu,
then decided to reorder, we had to tell them they'd have to wait because the power guys had bought all the slots for over a year.
It's only now they're starting to get appreciable supply.

The economics of a chip fab are nasty, and they're especially bad for small run parts like aerospace or cars. It's what's driving the
incredible consolidation in the electronics industry today. Yes, my company is one of the larger car/industrial chip suppliers,
and in the overall scheme of things, that segment is about 1% of the company."


https://www.theregister.com/2022/02/04/nxp_chip_deals/

Interesting numbers. NXP has about 11% of the automotive market, which is about 4.5 billion dollars of over 40 billion dollars per annum. When did NXPs revenues expand so much that 4.5 billion was 1% of their business?

Automotive semiconductors can be a bit quirky. With 50 to 100 MCUs per vehicle. and nearly 100 millions vehicles per annum, cars consume a heck of a lot of MCUs. A single application, like radar, which is not even fitted to all cars right now, is something like a billion dollar business for the semiconductor makers. On the other hand, some part for a special feature that most cars don't have might be made in quite small numbers.

Car makers don't usually buy chips, although they have a lot of say about which ones are used. They usually buy from module makers, like Denso. There are only a handful of module makers globally, so each module maker is a huge customer for a semiconductor company. In turn, those module makers have only a few options from which to source the key semiconductors they use. Automotive customers have a laundry list of specialised requirements, especially for quality control. Only a few semiconductor vendors are prepared to engage in the discipline needed to service those requirements. Perhaps, with people like TSMC developing specific automotive grade processes, more will enter the market.

[SiliconWizard]

Car makers don't usually buy chips, although they have a lot of say about which ones are used. They usually buy from module makers, like Denso. There are only a handful of module makers globally, so each module maker is a huge customer for a semiconductor company. In turn, those module makers have only a few options from which to source the key semiconductors they use. Automotive customers have a laundry list of specialised requirements, especially for quality control. Only a few semiconductors are prepared to engage in the discipline needed to service those requirements.

Absolutely.

[brucehoult]

A single application, like radar, which is not even fitted to all cars right now

It's fitted to the vast majority.

Pretty sure AEB is going to be mandatory in the USA and Australia from 2023 and in the EU from May this year, and the vast majority of cars sold already have it.

nhtsa.gov in the USA says Audi, BMW, Hyundai, Mazda, Mercedes-Benz, Subaru, Tesla, Toyota, Volkswagen, and Volvo started fitting it to ALL new vehicles three years early, and Ford, Honda, Kia, and Nissan had it in at least 75% of new vehicles sold by August 2020.

During calendar 2020, the only manufacturers with less than 90% of vehicles fitted with AEB in the USA were Kia (75%), Porsche (55%), Maserati (48%), GM (47%), Mitsubishi (39%), Fiat Chrysler (14%), and Jaguar Land Rover (0%).

Admittedly, that's not all "radar". Subaru and Tesla, for example, both use camera-based systems.

Not that this reduces the number of CPUs needed :-)

My 14 year old 2008 Subaru, by the way has camera-based dynamic cruise control, lane departure warning, automatic braking when in cruise control and a "you better brake now!" alarm when not in cruise control.

[coppice]

A single application, like radar, which is not even fitted to all cars right now

It's fitted to the vast majority.

Globally? Its becoming widespread in some countries, but I doubt it has gone global.

[brucehoult]

A single application, like radar, which is not even fitted to all cars right now

It's fitted to the vast majority.

Globally? Its becoming widespread in some countries, but I doubt it has gone global.

Mandatory in North America, Australasia, Europe. India has been considering making it mandatory from 2023, but I don't know if that has been enacted.

China is letting it grow organically. They are expecting 36% of vehicles to have AEB by 2025, and it's already compulsory for commercial vehicles since 2021.

At some point it costs more for manufacturers to leave it off.

[VK3DRB]

2024.

[luiHS]

The problem is whether now, with the Russian invasion of the Ukraine and the crazy Putin's threats to the rest of the world, this can get even worse.

It seems that we are going from bad to worse, new problems are being added that threaten the current crisis even more.

[PartialDischarge]

Even LMC555 are out of stock. Can't even design a led blinker !!!

[Karel]

The problem is whether now, with the Russian invasion of the Ukraine and the crazy Putin's threats to the rest of the world, this can get even worse.

It seems that we are going from bad to worse, new problems are being added that threaten the current crisis even more.

“In addition, the semiconductor industry has a diverse set of suppliers of key materials and gases,
so we do not believe there are immediate supply disruption risks related to Russia and Ukraine.”

https://www.semiconductors.org/sia-statement-on-sanctions-on-russia/

"Chipmakers Downplay Fears Ukraine Crisis Will Worsen Shortages
Ukraine isn’t a threat to the supply chain, industry says
Intel and GlobalFoundries see no impact on materials"

https://www.bloomberg.com/news/articles/2022-02-24/chipmakers-downplay-fears-ukraine-crisis-will-worsen-shortages

"ASML seeking alternative sources for neon gas amid Ukraine crisis
Although Ukraine is the world's biggest producer of neon, ASML sources less than 20%
of the gas from it uses from those countries, a spokesperson said on Wednesday."

https://www.reuters.com/technology/asml-seeking-alternative-sources-neon-gas-amid-ukraine-crisis-2022-02-23/

[leon_heller]

Neon is used in the lasers that are in all chip fabs.
Uh

[luiHS]

Even LMC555 are out of stock. Can't even design a led blinker !!!

The situation is not that extreme, the 555 is available from many manufacturers, most from Texas Instruments.

https://www.mouser.es/c/semiconductors/clock-timer-ics/timers-support-products/?q=555&mounting%20style=SMD%2FSMT&instock=y&sort=pricing

It is other chips that are really scarce, microcontrollers, sensors, accelerometers, I2S chips, CAN FD controllers, etc... What for now does not seem to be in short supply are memory chips, RAM, Eprom, Flash, SDRAM. Most 32bit microcontrollers have disappeared from vendor stock, this is the worst.

[SiliconWizard]

Even LMC555 are out of stock. Can't even design a led blinker !!!

If you can't design a LED blinker without a 555, you need to go back to the drawing board. =)

[PartialDischarge]

Even LMC555 are out of stock. Can't even design a led blinker !!!

If you can't design a LED blinker without a 555, you need to go back to the drawing board. =)

It was trying to be a funny overstatement

[floobydust]

Even 32.768kHz crystals that go with 32-bit MCU's, are in short supply. Some of 12.5pF but 6-9pF is months away. I know quartz crystals takes time to grow I guess it's demand related that shortage?

I am also watching LM555 stock for amusement, not that anyone designs these in - but it's been around for 50 years and imagine the dwindling stock running out.

[rfclown]

Even 32.768kHz crystals that go with 32-bit MCU's, are in short supply. Some of 12.5pF but 6-9pF is months away. I know quartz crystals takes time to grow I guess it's demand related that shortage?

I am also watching LM555 stock for amusement, not that anyone designs these in - but it's been around for 50 years and imagine the dwindling stock running out.

I actually put one in a design at work last year. We had the 1PPS coming out of a ublox GPS module, which I think has about 100ms pulse width at 3.3v. In the former design it went to the input of a uC, no problem. The requirement changed in that it needed to mimic a military GPS unit for which the 1PPS is specified (if I'm remembering correctly) 5V, 20 usec into 50 ohms. Using a 555 as a one shot, I was able to get a pulse narrower, and have enough umph to drive 50 ohms (picking the right 555). Pulled a DIP 555 out of my at home stash, tried it out on a breadboard, and bingo, it worked. Used an SO8 of the same part for the design. 555 is still useful, but there are simpler ways to blink an LED.

[brucehoult]

Even LMC555 are out of stock. Can't even design a led blinker !!!

If you can't design a LED blinker without a 555, you need to go back to the drawing board. =)

A low end PIC or AVR is several times cheaper than a 555 and with a couple of ADC inputs can do everything a 555 can do, at similar maximum frequency/latency. The only difference is most of them can't source/sink 200 mA.

[peter-h]

So, where is all the production going now?

ST etc must be running flat out, making hay while the sun shines (while prices are high) because they know that eventually this bubble will collapse and then nobody will be making money.

[Kjelt]

Probably to those chips that earn the most $/layer.
The newest processors and flash chips have smaller and more layers.
The first smallest layers are processed on the expensive euv machines but the layers above are done on the same duv machines where the jelly bean chips are also made. So the global throughput of silicon wafers per day is decreased due to the latest chips from Apple, NVidia etc. But those companies probably pay the most to the foundries. Also the demand for chips has increased with over 50% due to everyone working at home.
So to say how long this takes really depend on the exact chip you are waiting for and in what dateslot the production of that chip has been secured by the manufacturer. Some chips are still not slotted so there is no new production scheduled which makes the date of availability unknown.

[PlainName]

I am also watching LM555 stock for amusement, not that anyone designs these in

I've just used one (albeit not an LM) as a missing pulse detector.

[floobydust]

I am also watching LM555 stock for amusement, not that anyone designs these in

I've just used one (albeit not an LM) as a missing pulse detector.

I'm not against using a 555, simplicity has a place; but it's way overpriced - triple the price of small 8-bit MCU's.
Yes, missing is the wide VCC range and high output drive current but I kind of expect a 50 year old (~25 transistor) 555 to cost less than an MCU, not way more.

[floobydust]

So, where is all the production going now?

ST etc must be running flat out, making hay while the sun shines (while prices are high) because they know that eventually this bubble will collapse and then nobody will be making money.

ST raking in the cash: "STMicroelectronics has reported solid financial results for 2021, with fourth quarter net revenues reaching $3.56 billion, up 9.9% year-over-year and 11.2% sequentially. For the full year, ST achieved net revenues of $12.76 billion, an increase of 24.9%, while net income increased by 80.8% to $2 billion."
"ST expects the global chip shortage to gradually improve in 2022. “But it might not return to ‘normal’ until the first half of 2023,” said {President and CEO} Chery."
source https://www.eetasia.com/stmicroelectronics-expects-to-sustain-strong-growth-in-2022/
The CEO seems to be OD'ing on kool-aid, sky high prices and no STM32 stock, union pissed off strikes at plants, look I saw a unicorn and a rainbow, and made $12.76 billion 

Record semi sales, double digit across the board YOY. ~26%.
It seems to be smartphones eating up most of the silicon, sales surged 31% in 2021 record high $35.0B in sales.
"Embedded microprocessor sales... are projected to grow 9% in 2022 to about $21.5 billion after increasing 11% in both 2021 and 2020."

[descartes]

ST raking in the cash

Not so much a rake as a fleet of JCB bulldozers.

Last 20+ days I've lost hours each day playing hunt the MCU trying to figure out a game plan. Indeed hours looking for simple stuff like electrolytic caps. Last week 2 pole CO PCB mount relays, still not found anything of reasonable provenance in a qty of 1,000+. By the time I've found some, I'll need 2,000.

But the MCU's, that's just madness. ST seem to be 3 times the price, when you can find something, than anything similar to Atmel SAM.

In the UK the main distributors, Farnell & RS, may as well de-list the usual ranges to save us the bother of using their search. RS just adds to the misery by not showing stock levels in the list pages.

However the question I'd really like answered is how we arrived at so many variants. I'm fully aware of yields, but if you have a part that can be 32, 64, 128, 196 or 256KB of flash, that's not yield, that's turning off chunks of the die. And then you multiply that out with the packaging options.

Right now I'd be happy to go with a 256KB flash 64 pin in a small shop friendly format like TQFP (as opposed to any BGA), even if I only need 64KB flash on 32 pins for the actual project. But I have no clue what the strategy is, so anytime I'm looking at KiCAD PCB on screen, I just come out in a cold sweat that I can't find the part / package I'm guessing on a design for.

With the lead times we've ended up with, Gigadevice & Nuvotron ARM look increasingly attractive for a short soak test, their websites sell direct & appear to have stock on hand, ST & Microchip could find them selves with a whole generation of engineers who leave them to their corporate sales pipelines whilst the rest of us move designs over to smaller but more dynamic players.

So if any of the big boys are reading this, how about stop trying to make some of everything, just build lots of the upper end of each range in more common formats and don't ship 100,000 to a customer that you know will only use a third of that in the coming year.

[coppice]

Even LMC555 are out of stock. Can't even design a led blinker !!!

If you can't design a LED blinker without a 555, you need to go back to the drawing board. =)

A low end PIC or AVR is several times cheaper than a 555 and with a couple of ADC inputs can do everything a 555 can do, at similar maximum frequency/latency. The only difference is most of them can't source/sink 200 mA.

Where are you buying low end PICs and AVRs for a fraction of the cost of a 5 cent 555 timer? Are the undercutting Paduak where you live?

[Kjelt]

But I have no clue what the strategy is, so anytime I'm looking at KiCAD PCB on screen, I just come out in a cold sweat that I can't find the part / package I'm guessing on a design for.

Your strategy is faulty, you should buy the components first in a quantity that will keep you afloat till end of next year, when it arrives and you have it on the shelves only then do a redesign of create the product. That is why you can not find anything, if something comes on the market it will be bought right away.

With the lead times we've ended up with, Gigadevice & Nuvotron ARM look increasingly attractive for a short soak test, their websites sell direct & appear to have stock on hand, ST & Microchip could find them selves with a whole generation of engineers who leave them to their corporate sales pipelines whilst the rest of us move designs over to smaller but more dynamic players.

If you know that they can produce the same amount of chips as ST/Microchip with their own fabs why not. If they outsource to foundries then you can expect the same logistic reliability problems just after their stock is depleted.

don't ship 100,000 to a customer that you know will only use a third of that in the coming year.

Those are long term solid valued customers that have been buying those quantities for the last ten+ years from them. No one wants to loose their valued customers they just sell them for higher prices as usual. And no they also don't get everything they ordered over two years ago. It is a big mess for everyone.

[brucehoult]

Even LMC555 are out of stock. Can't even design a led blinker !!!

If you can't design a LED blinker without a 555, you need to go back to the drawing board. =)

A low end PIC or AVR is several times cheaper than a 555 and with a couple of ADC inputs can do everything a 555 can do, at similar maximum frequency/latency. The only difference is most of them can't source/sink 200 mA.

Where are you buying low end PICs and AVRs for a fraction of the cost of a 5 cent 555 timer? Are the undercutting Paduak where you live?

555s aren't 5c anywhere I can buy them.

Interesting I'm now seeing slow 100 kHz ones (i.e. original spec) for 40c. I haven't found them for less than $1.50 before, at which price an ATTiny85 is cheaper (and I always have a bunch on hand), let alone an ATTiny25.

https://www.digikey.com/en/products/detail/texas-instruments/NE555P/277057

[PlainName]

LCSC C6986 - TLC555 38p 7828 in stock (allegedly).

[coppice]

555s aren't 5c anywhere I can buy them.

Interesting I'm now seeing slow 100 kHz ones (i.e. original spec) for 40c. I haven't found them for less than $1.50 before, at which price an ATTiny85 is cheaper (and I always have a bunch on hand), let alone an ATTiny25.

https://www.digikey.com/en/products/detail/texas-instruments/NE555P/277057

With some parts you need to buy big numbers to get a reasonable price. With a 555, there are so many sources they are just jelly bean prices, even in modest quantities. Digiikey is a very expensive place to buy semiconductors, but even they only want you to buy 2500 to drop the price to 11 cents on the page you referenced. You can buy them by the reel for half that.

[descartes]

But I have no clue what the strategy is, so anytime I'm looking at KiCAD PCB on screen, I just come out in a cold sweat that I can't find the part / package I'm guessing on a design for.

Your strategy is faulty, you should buy the components first in a quantity that will keep you afloat till end of next year, when it arrives and you have it on the shelves only then do a redesign of create the product. That is why you can not find anything, if something comes on the market it will be bought right away.

Certainly a strategy for anyone making product. Although I was actually referring to what the mainstream manufacturer's strategy was for replenishing the supply chain.

Sadly I don't have product on my side of the office, so not a suggestion that's a great fit, I create small runs of proof of concept through to initial batches of custom IoT, frequently as co-development on the firmware side which means the client often has a preference for the MCU based on their staff skills. Sure I have a certain amount of standardisation and I am buying when I can and I'm explaining it real slow to prospects that there isn't much availability so if they want to do a project, they need to bring their own supply to the party or give me free rein to find what I can.

If you have it all under control, I'd be happy to pick up a loan from you to fund the cashflow to buy 22 months worth of parts. 

With the lead times we've ended up with, Gigadevice & Nuvotron ARM look increasingly attractive for a short soak test ....

If you know that they can produce the same amount of chips as ST/Microchip with their own fabs why not. If they outsource to foundries then you can expect the same logistic reliability problems just after their stock is depleted.

I don't know, otherwise I'd not have raised it as a discussion point. Do you?

don't ship 100,000 to a customer that you know will only use a third of that in the coming year.

Those are long term solid valued customers that have been buying those quantities for the last ten+ years from them. No one wants to loose their valued customers they just sell them for higher prices as usual. And no they also don't get everything they ordered over two years ago. It is a big mess for everyone.

For our other manufacturing / product based range we have allocated proportionally based on prior purchases so the smaller businesses that we support now will be more inclined to stay with us in the years ahead. None of the larger ones seem that bothered as they are clearly putting shipments on the shelf and by asking them what they actually need, we can plan accordingly.

[brucehoult]

555s aren't 5c anywhere I can buy them.

Interesting I'm now seeing slow 100 kHz ones (i.e. original spec) for 40c. I haven't found them for less than $1.50 before, at which price an ATTiny85 is cheaper (and I always have a bunch on hand), let alone an ATTiny25.

https://www.digikey.com/en/products/detail/texas-instruments/NE555P/277057

With some parts you need to buy big numbers to get a reasonable price. With a 555, there are so many sources they are just jelly bean prices, even in modest quantities. Digiikey is a very expensive place to buy semiconductors, but even they only want you to buy 2500 to drop the price to 11 cents on the page you referenced. You can buy them by the reel for half that.

I don't buy things by the reel. I buy one, or maybe ten maximum.

My working assumption, not having any actual data to the contrary, is that people who buy in volume get similar percentage discounts on everything relative to the Digikey / Mouser / Element14 qty-1 price and get AVRs and PICs much much cheaper than I can too.

So you can get a 555 for 5c. Cool. What can you get an ATTiny25 or the equivalent 8 pin PIC for?

[coppice]

So you can get a 555 for 5c. Cool. What can you get an ATTiny25 or the equivalent 8 pin PIC for?

Its a long time since I looked, but the reason people were amazed by the 3 cent Padauk MCUs is they a fraction of the price of anything else. I think PICs probably start around 10 cents for those really simple ones in tiny packages, in high volume. As you add more pins the A&T costs rapidly rise.

[T3sl4co1l]

Where are you buying low end PICs and AVRs for a fraction of the cost of a 5 cent 555 timer? Are the undercutting Paduak where you live?

You keep bringing up low prices but curiously, you haven't opened a storefront extending those savings to the small buyer.  Why is that?  Surely even a small markup on top of $0.05/pc would justify such a venture.

Tim

[brucehoult]

So you can get a 555 for 5c. Cool. What can you get an ATTiny25 or the equivalent 8 pin PIC for?

Its a long time since I looked, but the reason people were amazed by the 3 cent Padauk MCUs is they a fraction of the price of anything else. I think PICs probably start around 10 cents for those really simple ones in tiny packages, in high volume. As you add more pins the A&T costs rapidly rise.

One factor in the 3c Padauks is that they are OTP. They have ones with flash program memory too, but they cost quite a bit more.

[coppice]

So you can get a 555 for 5c. Cool. What can you get an ATTiny25 or the equivalent 8 pin PIC for?

Its a long time since I looked, but the reason people were amazed by the 3 cent Padauk MCUs is they a fraction of the price of anything else. I think PICs probably start around 10 cents for those really simple ones in tiny packages, in high volume. As you add more pins the A&T costs rapidly rise.

One factor in the 3c Padauks is that they are OTP. They have ones with flash program memory too, but they cost quite a bit more.

That OTP issue intrigues me. It really adds to the die area to make a small flash MCU self-programmable, because of the die area of the charge pump required. However, if you accept that the flash needs to be programmed by an external programmer, and you put the voltage generator in there, the use of reprogrammable flash doesn't really have an overhead.

[SiliconWizard]

Yes it does have an overhead however you see it.
But the main point is elsewhere: as I think we already discussed in earlier threads, CMOS processes that offer embedded Flash are fewer and more expensive, so the cost of dies is significantly higher, whether you actually embed Flash or not, or even if it's just a small area of it. While OTP is usually available on a wider range of CMOS processes at a lower price point.

[coppice]

Yes it does have an overhead however you see it.
But the main point is elsewhere: as I think we already discussed in earlier threads, CMOS processes that offer embedded Flash are fewer and more expensive, so the cost of dies is significantly higher, whether you actually embed Flash or not, or even if it's just a small area of it. While OTP is usually available on a wider range of CMOS processes at a lower price point.

Most modern processes are only set up for flash. Perhaps these tiny MCUs are pad limited, despite their low pin count, so an old large geometry process with EPROM support is still the cost effective solution. Does anyone know who fabs for Padauk?

[PlainName]

Don't those tiny MCUs have a reprogrammable development version? Doesn't matter if it costs £20 or something - you use that single device to develop and then roll out a billion OTP parts knowing you've got the code right.

[coppice]

Don't those tiny MCUs have a reprogrammable development version? Doesn't matter if it costs £20 or something - you use that single device to develop and then roll out a billion OTP parts knowing you've got the code right.

As you said, it not a big drawback. I just found it odd that what appears to be recently developed silicon didn't naturally gravitate to a reprogrammable approach.

[SiliconWizard]

Yes it does have an overhead however you see it.
But the main point is elsewhere: as I think we already discussed in earlier threads, CMOS processes that offer embedded Flash are fewer and more expensive, so the cost of dies is significantly higher, whether you actually embed Flash or not, or even if it's just a small area of it. While OTP is usually available on a wider range of CMOS processes at a lower price point.

Most modern processes are only set up for flash.

Nope. But just have a look at the offering at TSMC, GlobalFoundries and whatever else you like.

[coppice]

Yes it does have an overhead however you see it.
But the main point is elsewhere: as I think we already discussed in earlier threads, CMOS processes that offer embedded Flash are fewer and more expensive, so the cost of dies is significantly higher, whether you actually embed Flash or not, or even if it's just a small area of it. While OTP is usually available on a wider range of CMOS processes at a lower price point.

Most modern processes are only set up for flash.

Nope. But just have a look at the offering at TSMC, GlobalFoundries and whatever else you like.

Specify some actual processes that are not flash, or flash without a charge pump acting like some kind of fake EPROM, that isn't ancient.

[brucehoult]

Yes it does have an overhead however you see it.
But the main point is elsewhere: as I think we already discussed in earlier threads, CMOS processes that offer embedded Flash are fewer and more expensive, so the cost of dies is significantly higher, whether you actually embed Flash or not, or even if it's just a small area of it. While OTP is usually available on a wider range of CMOS processes at a lower price point.

Most modern processes are only set up for flash.

Nope. But just have a look at the offering at TSMC, GlobalFoundries and whatever else you like.

Specify some actual processes that are not flash, or flash without a charge pump acting like some kind of fake EPROM, that isn't ancient.

I don't know the answer to that.

I do know, from my own experience that the SiFive FE310 (TSMC 180nm) and FU540 (TSMC 28nm) both provide a few kb (8 and 16 respectively) of OTP, and no flash. They use external SPI flash for firmware.

Incidentally, the code in OTP (as delivered from the factory on the HiFive1 and HiFive Unleashed boards) starts with a RISC-V FENCE instruction 0x0000000F, which can be easily burned later to an up to ±1 MB relative jump 0xnnnnn06F.

I'd be interested to know the similar tricks that work on other ISAs.

Having all 0s be a NOP would be the simplest thing, but the RISC-V designers deliberately made both all 0s and all 1s be illegal instructions to try to detect execution off into the weeds ASAP.

[passedpawn]

Even 32.768kHz crystals that go with 32-bit MCU's, are in short supply. Some of 12.5pF but 6-9pF is months away. I know quartz crystals takes time to grow I guess it's demand related that shortage?

I am also watching LM555 stock for amusement, not that anyone designs these in - but it's been around for 50 years and imagine the dwindling stock running out.

I actually put one in a design at work last year. We had the 1PPS coming out of a ublox GPS module, which I think has about 100ms pulse width at 3.3v. In the former design it went to the input of a uC, no problem. The requirement changed in that it needed to mimic a military GPS unit for which the 1PPS is specified (if I'm remembering correctly) 5V, 20 usec into 50 ohms. Using a 555 as a one shot, I was able to get a pulse narrower, and have enough umph to drive 50 ohms (picking the right 555). Pulled a DIP 555 out of my at home stash, tried it out on a breadboard, and bingo, it worked. Used an SO8 of the same part for the design. 555 is still useful, but there are simpler ways to blink an LED.

The pps signal on ublox chips is programmable.  Not sure about pulse width, but certainly the pps. 

[Deni]

37 years lead time...

[SiliconWizard]

At least they're not trying to lead you on.
(The situation is not funny though.)

[Monkeh]

"Hey guys, when can we get these parts, our customers are asking"
"2059, now fuck off!"

[Doctorandus_P]

And what is the Lead free time? 

[tepalia02]

Wanna know the answer too.....

[hans]

Atleast it's not on a Friday the 13th. Then atleast, you don't have that bad luck working against you.

[anda3243]

This crazy lockdown in China will not help eather. I think a second big supply chain disruption is coming. The big port in Shanghai has been closed for almost a month. Most businesses in and around Shanghai is closed, including the few IC manufacturers. Have been closed for a month. Beijing will shut down soon I think.
So far the Shenzhen area has had very few shutdowns, but it might come as well. Alot of ICs are made in Shenzen area.

Overall China might not be a huge exporter of chips. But these stupid covid shutdowns will cause more chip related shortages. More than the Ukraine war for sure.

[SiliconWizard]

Noticed that when something I ordered a while ago now was supposed to arrive more than a month ago, and is still blocked there.

[nigelwright7557]

RS Components is limiting semiconductor supplies.
I can get one off PIC32mx230 for £4-ish
Their 30 off much cheaper option is out of stock.
There are plenty of one off's to cover the 30 offs too.

The prices are going up too. A PIC that was around £2 is now nearer £4.

[SiliconWizard]

FPGA prices have gone through the roof as well.