Have you noticed any easing of the chip shortage?

[Nortek-Chris]

Two quotes back over the last few weeks from UK Brokers

VNB920B5-E - £92.00 EACH!

ATMEGA328-AU - £16.11 EACH!

Had similar prices back from a couple of Trusted sources in China, so I guess we are all chasing the same stock, and its not the middlemen doing the profiteering.

[AndyC_772]

Might not have reached hobbyists but there are signs that inventory is growing: https://www-theregister-com.cdn.ampproject.org/c/s/www.theregister.com/AMP/2022/05/26/sharp_chip_inventory_correction_looms/

Geez, it's as though these analysts write their reports without even speaking to anyone in the industry.

Of couse demand will fall, if critical components are simply unavailable for an extended period of time. Why would I even bother to order components Y and Z, if component X (which is also on the BoM for the same product) has been vapourware for the last 12 months solid? What am I going to do with them? Stick them on Ebay? It might be more profitable than building them into products that - surprise! - people can't afford because the cost of living is rising at the fastest rate in living memory. But then we'd be making the problem worse for everyone else, and that's not something that anyone with an actual conscience would choose to do.

Of course products have been discontinued; we simply can't make them. If we could, for a while, it's been by scraping the bottom of the supply chain barrel and paying ridiculously inflated, predatory prices for the last few parts available for sale by anyone, anywhere. That's not sustainable. D'uh.

[dietert1]

Take that publication with a grain of salt. Nowadays Britains are expected to speak bad about Europeans. Like calling others hobbyist.

Regards, Dieter

[SilverSolder]

In order to maintain price levels as demand falls, sellers have to constrain supply - it's as simple as that.  Otherwise prices would fall, and "they" really hate that.

Seems to me their game is to constrain supply and hope the economy improves...  at which point, the supply issues will magically disappear and we will be able to buy as much as we want...  at the new, higher price levels that have now become the norm!

[AndyC_772]

I wonder whether those offering 'analysis' really understand that semiconductors aren't fungible.

If there's a shortage in supply of one model of car, I can just choose to buy a different, competing model. For 'car' read dishwasher, cellphone, television, webcam, or any of the hundred other things a consumer might purchase that contains electronics. Personal brand loyalties aside, switching to an alternative product is easy.

This is, of course, the general public's experience of electronics in a commercial sense. If you can't buy one thing that you wanted, you probably can "just" buy something else instead, and it'll do much the same job.

Can't buy an STM32H723? Very different problem.

[SilverSolder]

I wonder whether those offering 'analysis' really understand that semiconductors aren't fungible.

If there's a shortage in supply of one model of car, I can just choose to buy a different, competing model. For 'car' read dishwasher, cellphone, television, webcam, or any of the hundred other things a consumer might purchase that contains electronics. Personal brand loyalties aside, switching to an alternative product is easy.

This is, of course, the general public's experience of electronics in a commercial sense. If you can't buy one thing that you wanted, you probably can "just" buy something else instead, and it'll do much the same job.

Can't buy an STM32H723? Very different problem.

Some semiconductors are more fungible than others, though?  -  obviously redesigning a product for a different family of CPU is a major endeavour, but we can probably live with a different source of resistors etc. 

[FrankBuss]

Can't buy an STM32H723? Very different problem.

Not much. There are many competing microcontrollers. And I write my programs for bigger projects in a way that the hardware layer is abstracted away (and even write emulations for it to test it on a PC), so that I can change to another vendor easily. But right, it is not exactly the same as buying a different car. And redesigning the circuit is annoying.

Would be nice if every microcontroller would be a FPGA, with standardized analog features, a set of standard footprints, and same pin configurations. Then different manufacturers could build compatible versions, and just add additional features. If you don't use these features, you would have many second sources. Price would be probably cheaper than normal FPGAs, because of the competition. Only problem might be the power consumption, which is usually higher with FPGAs than hard IPs (for the same performance). But might be not a problem for many projects. You could just use "one computing chip, model x", as you would now buy "one car, for 4 persons".

[AndyC_772]

we can probably live with a different source of resistors etc.

Yes, of course, but it's the most difficult component that dictates whether or not a given product can be manufactured, not the least. I've answered a great many emails this year along the lines of "we can't get <passive>, can we use <other passive> instead?", and the answer is usually 'yes'. Sadly the same can't be said when the component in question is a semiconductor, though.

I write my programs for bigger projects in a way that the hardware layer is abstracted away (and even write emulations for it to test it on a PC), so that I can change to another vendor easily. But right, it is not exactly the same as buying a different car. And redesigning the circuit is annoying.

Are you doing this commercially, or just for fun?

I'd love to say it's possible to switch "easily" because of how the software is written, but you can't side step the fact that you still have to research a new CPU that's more readily available, design a new PCB, prototype it, update the code, then test functionality, redo EMC testing, keep track of the new part number and its (incompatible) firmware, update production tools and processes... it's never going to be "easy" to introduce a new product that has no benefit over its predecessor beyond "possible to buy (for now)".

Would be nice if every microcontroller would be a FPGA, with standardized analog features, a set of standard footprints, and same pin configurations.

I think the closest we'll see to that is the ARM core and its associated, standard peripherals (NVIC etc).

[SilverSolder]

[...]  it's never going to be "easy" to introduce a new product that has no benefit over its predecessor beyond "possible to buy (for now)". [...]

Maybe that's an opportunity...  make some product improvements as well, so the redesign is "worth the effort"?  Also makes it easier to sell the inevitable inflation-driven price increases...

[FrankBuss]

I'd love to say it's possible to switch "easily" because of how the software is written, but you can't side step the fact that you still have to research a new CPU that's more readily available, design a new PCB, prototype it, update the code, then test functionality, redo EMC testing, keep track of the new part number and its (incompatible) firmware, update production tools and processes... it's never going to be "easy" to introduce a new product that has no benefit over its predecessor beyond "possible to buy (for now)".

Yes, this is true. Only the core firmware can be reused, still a lot of work and expensive, but not too difficult.

Would be nice if every microcontroller would be a FPGA, with standardized analog features, a set of standard footprints, and same pin configurations.

I think the closest we'll see to that is the ARM core and its associated, standard peripherals (NVIC etc).

Probably in future more RISC-V cores. This would make it even possible for low power devices to have many different second source parts. Just define the pin positions for a crystal for all vendors, and make all internal peripherals configurable to all pins, and then provide some basic functions for all chips, like DAC, ADC, and GPIO features like pull-up/down, open drain etc. and it should be possible.

Problem would be features like timers, PWM etc., they are pretty different between microcontrollers (has the PWM dead-time or other required features?). For this, maybe a small FPGA part would be useful.

The manufacturers could define such a generic microcontroller in a group, same like for example OpenGL was defined for a standardized graphics standard.

[dietert1]

It is rather odd that there's stock out there of brand new devices ...

ST has multiple manufacturing sites, TSMC, Samsung, but also smaller ones in the EU, that they own. I think they don't get the capacity in the far east. At least someone with insider info was hinting me something like this. For example their wireless microcontrollers are all made here, and they are supposed to be better stocked.

Another interpretation: I'd assume for STM the appearance of STM32 clones (GD32) was a hostile act. Certainly they took measures to avoid that in the first place and they have to take other measures now to avoid cloning of their complete STM32 lineup. The world has become a strange place. In former times second-source used to be a requirement.

Regards, Dieter

[Keith956]

Take that publication with a grain of salt. Nowadays Britains are expected to speak bad about Europeans. Like calling others hobbyist.

Well I'm a hobbyist, and also a European, but don't care much for racist remarks.

[cvriv]

Its not going to get better.

[jmelson]

Not chips, but connectors.  I use a lot of On-Shore 2-piece screw-terminal plugs and headers.  For some of them, I buy from a local distributor/factory rep that gives me 50% off the Digi-Key price.  On some other parts, I don't get much discount, so I just buy from Digi-Key.  Now, Digi-Key seems to be reducing stock and requiring massive minimum orders.  I just got caught running low on a 2-pin, 3.81mm screw-terminal plug, and couldn't find any stock in the US.  I finally found a compatible part at Arrow for over $2 each, where I used to pay $0.77 from Digi-Key.  (Digi-Key part # ED2894, the tricky bit is that the keying tabs face opposite the wire entry side.)
Jon

[MR]

Several FPGAs still have a very poor availability, and those which are available have a 2-3x higher price than usual.
I'm still glad that I have ordered various EVBs years ago and I'm happy to do prototyping with them and finally try to find a silicon which will be suitable (and available) for my application.

[abquke]

The ethernet PHY I needed last year has lead times down to 52 weeks!

...oh wait. It was 52 weeks a couple of months ago.

[AndyC_772]

"52 weeks" doesn't mean 52 weeks, it means 'a year or more'. It's open-ended, ambiguous, just a placeholder for an actual lead time that the supplier simply isn't able to give you.

[jc101]

I ordered some PIC18 from Microchip with a date of December 2022 a few weeks ago, I've had two order updates since then, the latest was shipping mid-June, and that was 2022 too.

Some Analog devices power ICs I ordered from DigiKey in June 2021 are due to be delivered next week, so almost a year later.  They originally had a shipping date of August 2021.  They did honour the original price though, they have increased 25% since then.

[Pineapple Dan]

I notice the DSPIC33EP64GS502 is back in stock some places but not the 504/506. Prices much higher than before though

[JPortici]

Yeah dsPIC/PIC24 had their price increased by about 25-30%
PIC18 are starting to coming through.. we've almost replenished our yearly supply. Delivery dates were correct.

[uer166]

In low volumes I actually found that my assembly house was able to source pretty much every component on a power conversion board, I was surprised. The prices are a little eye-watering on some parts but at least they're able to get and assemble them:

STM32G474VET6: $28.2
UF3C065030B3 FET: $27 a pop
UCC5350SBDR 4-channel isolator: $22

The rest of the 48 BOM lines are actually pretty reasonable, incl. passives. I'm not sure if it means that they have some back-channels (most of these parts have no availability in any disty), or the shortage is easing up, or is it simply the free market speaking: if you can afford to pay 2-10x the usual price for a part, you can probably get it somehow.

[josuah]

Trying to figure things out: we end-up with compromises like below?

Option #1 contains:

* Same complex components, waiting that they become available again
* Uncertainty about when the next component will be available
* Alternative compatible passives

Option #2 contains:

* Another complex component slightly incompatible
* Extra engineering time to port the project to the new component
* Alternative compatible passives

Option #3 contains:

* Stop production of that product altogether, focus on the next iteration
* Extra engineering time for the new product, eventually planned already
* Picking components among those available

Option #4 contains:

* In case of availability at a higher price, pay the extra price to still be able to produce without a redesign
* Uncertainty about how long the priucier component will be available
* Alternative compatible passives

Option #5 contains:

* No reaction, letting the doubt settle about everything
  (Not a real "option", but would probably happen in spite of any option chosen.)

Option #6 contains:

* Alternative compatible passives
* Using adapter boards for different footprint or packages
* Alternative incompatible active, now made compmatible
* Moderate amount of engineering for integrating the adapter board in the process.

Option #7 contains:

* Alternative compatible passives
* Change the footprints and find chips whose signal are compatible,
* Alternative incompatible active, now made compmatible
* Moderate amount of engineering for integrating the new footprints.
* Moderate amount of engineering for making the new footprint flexible to more chips in the future.

[EDIT: added Option #6 from Mr. schmitt trigger]
[EDIT: added Option #7 from Mr. SMdude]

[schmitt trigger]

I would add a 6th option: using adapter boards for different footprint or packages.

[Infraviolet]

"52 weeks" doesn't mean 52 weeks...
The nasty part is when you see a date to backorder for and see it is in the next month, so you order, a month's not too long to wait, then you realise it was actually for August 2023.

[dietert1]

Recently i noticed some cheap stm32l01.. mcus were on stock and for sale on the stm website. Anybody tried that?

Regards, Dieter