General > General Technical Chat

Resilience of semiconductor supply chains

(1/7) > >>

profdc9:
As most of you probably know, we are in the midst of a semiconductor shortage due to a number of factors, only some of which relate to the worldwide epidemic.

https://www.washingtonpost.com/technology/2021/03/01/semiconductor-shortage-halts-auto-factories/

In this article it says that:


--- Quote ---Car manufacturers began using electronics to control automobiles in the 1970s, replacing older mechanical controls. Gradually, the number of tiny chips known as microcontrollers increased inside cars, powering a wide array of functions, from lights to engine cooling systems.

The 38 microcontrollers in an Audi Q7 come from eight companies, highlighting the complexity of auto supply chains, according to research firm IHS Markit.

Yet because TSMC manufactures nearly three-quarters of all auto microcontrollers, any capacity crunch at the company has ripple effects through the entire auto industry.

--- End quote ---

Cars are now awesomely complex machines.  That there are 38 microcontrollers is not surprising, and the article doesn't say how many of these are the same or difference microcontrollers, only that eight companies provide the microcontrollers necessary to complete an Audi Q7.  Furthermore, in the USA auto assembly plants are idle because the silicon is not available to complete the cars.

I guess the question is: do we really need so many specialized parts from so many manufacturers?  Of course, this was easier when there were fewer choices available, for example the 8051 was the workhorse of the 1980s.   I am not suggesting that we all return to using the 8051 as the lowest common denominator.  But aren't microcontrollers supposed to be more general purpose devices with lots of peripherals so that they can be customized for different applications?  There doesn't seem to be any attention paid to engineering for resilience or longevity.  Perhaps that is because the consumer market can or will not bear those costs. 

Like for example, consider a boost converter like the MT3608.   There are many similar devices with the same pinout such as HM1549. XR3403, etc.  This would suggest that a design with a MT3608 could be robustly supported with substitutes.  I realize that the requirements for the automotive industry are much more stringent than a consumer-level Chinese generic boost converter.  But is every boost converter really so different that there can't be a few generic types that are commonly used?  Just at Texas Instruments alone they list 185 boost converter ICs with integrated switches!

https://www.ti.com/power-management/non-isolated-dc-dc-switching-regulators/step-up-boost/boost-converters-integrated-switch/products.html

It may be a bit unfair to target TI as they also have many legacy products from National Semiconductor and others, and I think it is great that they support so many products.  And yes, power efficiency requirements keep increasing, boost converters have to adapt to new kinds of battery chemistries, etc.  Still, if the one of the 185 boost converters TI sells is out of stock, and one didn't anticipate and stockpile, you could be SOL.

I don't have an answer for this but I think part of it could be fewer generic, configurable components that have, if possible, generic manufacturers.  There used to be 74 TTL/CMOS chips, Z80 processors, op amps like the LM741/NE5532, power regulators such as the LM78xx, LM79xx, TL494, UC384x, standard transistors like the 2N2222, 2N3904, 2N5551, 2N7000, TIP41, TIP42,  etc.  If you used these parts you can find many sources.  Now everyone has their own walled garden of proprietary components.  It looks like it is only worth the Chinese semiconductor companies time to make compatible components (assuming they are sufficiently compatible) because they don't have to research new designs.

This problem has been ongoing for quite a while.  Perhaps these manufacturers can get together and realize it is in no one's interest if demand falls because factories are idle waiting for parts.

ataradov:
The reason we have so many parts that are doing basically the same thing is the need to maintain the parts that are already designed into product, while still innovating.

Chip vendors would be excited to drop 90% of the parts they supply if customers were to move to the new and shiny thing the same year it is introduced. This is not happening though, some commercial products still use 20 year old parts, and manufactures have to ensure continuous supply.

In the old days you had many of fixed function devices. There is not a lot you can innovate on the 74xx logic. 8051 is an architecture, there are a ton of devices using it. Most of them are not compatible on a peripheral level.

In modern days manufacturers competing on shaving uA is current consumption. There is no way they can all come together and design a common architecture. That would be too rigid and hard to innovate, so it would be immediately followed by deviations.

floobydust:
I know people in the automotive industry furious at NXP because they cannot fulfill orders for their 32bit MCU's - that were signed contracts well beyond a year in advance. They did know the parts needed to be made but no decent explanation as to what happened. I wonder if they were worthy of buying Freescale in the first place. Now the MCU's are all zero stock at any distributor and 52+week lead time.
It even was on Jim Cramer, Wall Street is getting annoyed their car stocks are sinking.
"NXP Is Causing 'the Biggest Problem' in Chip Shortage" and he mentioned trying to get the NXP CEO on his show for an interview but no success. I would say there's so many lawyers ready to sue them into oblivion, he would never say a peep.
People are buying semi's from the scalpers even paying 10X price just to keep car production going. The shortage seems to be getting worse from people panic buying.

I suspect geopolitics more than anything. Taiwan is constantly getting buzzed by the chinese military and may be letting the world know how important they are. Now it's water shortages for the fabs, in the news.

Berni:
ST has the same issue with STM32 micro controllers. Lots of them are out of stock and the lead times are skyrocketing.

The whole covid thing probably shaken things up enough to cause a wildly unstable demand for chips, the chip vendors tried to rapidly adjust the production to meet it and got it really wrong. Until eventually warehouse stock of chips started running dry, costumers started to panic about chip availability, started buying up extra chips to make sure they don't run out, making the problem even worse. Ending up with a situation where fabs are still coming back up to speed but demand is at an all time high.

We also had to resort to buying overpriced chips from weird sources. Hopefully the fabs get up to speed soon and sort this out, but my guess is that this will take at least a few months to sort out.

daqq:
How long does it take to switch a fab line to make an existing design? I mean, 52 weeks, that's insane!

Navigation

[0] Message Index

[#] Next page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod