How is Chipageddon affecting you?

[TimFox]

The "CK" prefix was used by Raytheon on their receiving and transmitting tubes.  The first transistor I ever bought, back in Cub Scouts (ca. 1960) was a CK722 germanium PNP device marketed to amateurs.
One important sub-mini tube made by Raytheon (and others) was the CK587 pentode, rated for < 0.1 pA grid current in electrometer service.

When still working in the private sector, long before the current chip shortage, my engineer friends would gripe about a useful device (especially Maxim) that had been discontinued after they had designed it into their system.  My response was "well, I can still buy 6L6s" (which RCA introduced before 1940).

[bdunham7]

My response was "well, I can still buy 6L6s" (which RCA introduced before 1940).

And you still can.

https://www.ebay.com/itm/122010201786?epid=1601260339&hash=item1c685feeba:g:~wEAAOSwtN9eE7Xb

[IDEngineer]

I recall there were a handful of [production] TV sets still using tubes [other than the CRT] til the... late 70s at least, not sure if early 80s.

CRT based computer monitors were being newly manufactured at least as late as 1996-1997. I remember getting my first "big screen" monitor back then, it was a CRT that consumed half of my workbench. The computer monitor market was always minute compared to retail TV's so I presume they weren't keeping the CRT factories running solely for monitors.

[T3sl4co1l]

I mean, for a long time I had a Trinitron made in 1999, and even televisions were made in that type (including HD resolution) for some years after.  But that's a very special application; I meant tubes other than the CRT.  (Magnetrons also a very special application, still surprisingly common today.  Power RF GaN may finally displace them in the next decade or two; or maybe not, we'll see!)

I think some of the longest-lived non-CRTs were line output, or high voltage rectifier?  Think I've also seen video output (cathode/grid driver) done with a triode cascoded on top of a BJT, but that's probably not a late-model thing, they certainly had video output transistors by then.

Tim

[IDEngineer]

WRT shortages, this isn't helping either:

https://www.cnn.com/2021/06/30/economy/china-power-shortage-intl-hnk/index.html

A lot of things important to the electronics industry comes out of that region.

[EEVblog]

I've been hit big time.
After my next back in July, new stock of BM786's aren't due until Feb 2022 due to a shortage of the main IC.
Also, BM235's are delayed until late Nov this year.

[rgarito]

It's been affecting me considerably.  In my EE work (I actually have 2 jobs!), I design STM32-based hardware mostly.  We are currently discussing availability with ST and for instance, one of our main designs uses the STM32F429.  Can't get them ANYWHERE.  Had a long discussion with ST and our suppliers and basically we're out of luck.  (one of our suppliers sent us an obviously canned response listing everything as a cause practically including solar flares! Some points were things I had not considered and others were a real stretch of the imagination). 

ST told us that the H7 series (which we do use in another newer product) is safe, but most of the rest of their parts are going to be a crap shoot to try to obtain for the foreseeable future (52+ weeks).

We have discussed switching to someone else, but I'm not sure anyone else is much better, and it would take considerable redesign (both hardware and firmware, as we are a bare metal design) so that realistically, by the time we are ready to ship redesigned products, things may have stabilized.  And that does not address the orders we estimate on receiving for the existing designs over the next 12 months...

Interestingly, another part that we've had a LOT of trouble with is crystals.  ST's chips are pretty picky on crystals (especially RTC 32khz crystals) and going from their certified component lists, several times in the last 2 months, when I locate one that has parts in stock, between the time I find it and the time I have our purchaser do the ordering--BOOM--out of stock w/ 52+ week lead time!  I had one board I even had to at the last minute update the gerbers for a different footprint and between when I placed the PCB order and when they arrived, the crystals were out of stock!  (we got parts for this board run, but we're gonna probably have to redesign for the next one... that is, if we can even get the STM32F429 that the board requires at all, by then (not looking good)).


We do have a few older designs that I am not involved with, that use PIC32 and I'm told those, we are having no problem sourcing...

[Siwastaja]

CRT based computer monitors were being newly manufactured at least as late as 1996-1997.

You have that timeline off by a decade. In late 1990's, no replacement was even near.

LCD displays started replacing CRTs in non-critical consumer market somewhere around 2003 while being expensive and quite crappy image quality wise, and in more demanding applications, maybe around 2006 when they started to offer acceptable image quality, still at quite a cost. I think I bought my last new CRT monitor in 2002 or 2003 and there was no consideration back then, it was just the normal choice. After that, like many, I ran used CRT monitors for almost a decade.

So CRT monitors were definitely made in not insignificant numbers at least as late as 2006-2007. Specialized products likely longer than that.

[Ian.M]

CRT based computer monitors were being newly manufactured at least as late as 1996-1997.

You have that timeline off by a decade. In late 1990's, no replacement was even near.

LCD displays started replacing CRTs in non-critical consumer market somewhere around 2003 while being expensive and quite crappy image quality wise, .....

LCD monitors were around a bit earlier than that but they were pretty crappy, <double-expletive> expensive and really only popular for niche applications where the bulk of a CRT was difficult to accommodate.

e.g. my oldest LCD is a Sceptre LC12W SVGA (800x600) monitor, which was manufactured circa 1997-1998 (to a 1996 design).  It uses an IBM ITSV50D 6 bit/color RGB TFT LCD panel, 4:3 aspect ratio, 12" (visible) diagonal.  Apart from the lowish color depth, it also doesn't do scaling, so all lower resolution video modes that it actually supports are letterboxed.  Its good enough for noodling around CMOS setup screens and the like but you certainly wouldn't want to use it for CAD!   It still gets dragged out and used occasionally if I've FUBARed remote access to a usually headless box.

[EEVblog]

CRT based computer monitors were being newly manufactured at least as late as 1996-1997.

You have that timeline off by a decade. In late 1990's, no replacement was even near.

LCD displays started replacing CRTs in non-critical consumer market somewhere around 2003 while being expensive and quite crappy image quality wise, .....

LCD monitors were around a bit earlier than that but they were pretty crappy, <double-expletive> expensive and really only popular for niche applications where the bulk of a CRT was difficult to accommodate.

e.g. my oldest LCD is a Sceptre LC12W SVGA (800x600) monitor, which was manufactured circa 1997-1998 (to a 1996 design).  It uses an IBM ITSV50D 6 bit/color RGB TFT LCD panel, 4:3 aspect ratio, 12" (visible) diagonal.  Apart from the lowish color depth, it also doesn't do scaling, so all lower resolution video modes that it actually supports are letterboxed.  Its good enough for noodling around CMOS setup screens and the like but you certainly wouldn't want to use it for CAD!   It still gets dragged out and used occasionally if I've FUBARed remote access to a usually headless box.

I worked on LCD monitors at Keycorp in 1994. They were 640x480, and we were just prototyping a new fangled 800x600 jobbie.
Back then the LCD's were not guaranteed to be perfect, you had to accept what a certain number of dead pixels were possible!

[peter-h]

Haven't read the whole thread, but...

Been in mfg & design since 1978. This happens regularly. Prices naturally fall, eventually the mfgs start spreading rumours, using the "a"-word (allocation) which stirs up mortal fear in buyers (whoops I meant to say accredited supply chain managers ) and then everybody with money buys up the supply pipeline for a year or so, so lead times go to 1 year.

6 months later all this stuff is delivered but most of the buyers don't need it - because they aren't making any special quantities of their end product. So they have huge stocks and stop buying for a year or so.

So you have a bloodbath.

Give it a year and you will see.

It will always happen because the sales people are mostly on commissions. Nobody (below the level of car manufacturers and their close relationships with chip makers) looks at long term.

The only solution is buy up strategic stocks of single sourced parts.

Also a lot of parts are so cheap that buying say 10k is fine even if it would last 100 years. Passives... buy enough for a few years.

Just In Time is no more than a business practice whereby a big company forces a small company (the supplier) into keeping the stuff in stock and despatch it when the big company wants it It it not ethical but is widespread. And nobody learns from these crises, because screwing suppliers is the best way for a corporate climber to demonstrate his virility (screwing customers is not good, and screwing employees needs to be done very carefully).

I have just secured a supply of some 32F4 parts and other bits, but forgot about one little item on the BOM: an Adesto SPI FLASH! Very very nearly got burnt on it too, but the one version we are using is ex stock in an SMT package which probably nobody wants, but which just fits on the SOP pads. Had to buy it from Mouser so paid almost 2x the right price.

A lot of stuff is totally unaffected e.g. just bought a load of TS391 comparators at about 10p. The skill (or luck) here is to pick a cheap and multi sourced item. So if you can use an LM358, use that, every time. If you use chips from Maxim, LT, etc, it is only a matter of time before you get burnt. These firms made money for a good reason: premium specs and no second source.

[VK3DRB]

It's been affecting me considerably.  In my EE work (I actually have 2 jobs!), I design STM32-based hardware mostly.  We are currently discussing availability with ST and for instance, one of our main designs uses the STM32F429.  Can't get them ANYWHERE. ...

We have discussed switching to someone else, but I'm not sure anyone else is much better, and it would take considerable redesign (both hardware and firmware, as we are a bare metal design) so that realistically, by the time we are ready to ship redesigned products, things may have stabilized....

Interestingly, another part that we've had a LOT of trouble with is crystals.

I switched mid-term in a design from an STM chip to an MSP430 variant that was fit for purpose. It was an excellent decision. No problems getting the MSP430's. Like Bosch, STM will lose market share long term as engineers find alternative vendors who are capable.

Once a design is locked down, engineers won't be changing back in a hurry, if at all. Furthermore, there will be a "once bitten, twice shy" approach by engineers in choosing vendors for new designs. In fact, those manufacturers who can supply components to engineers for R & D purposes during Chipageddon have a golden opportunity to capitalise on this because when Chipageddon ends, their components will be designed in.

Yes, I have also had problems with crystals and oscillators. Best thing to do it use a common footprint so alternatives can be used. Recently, I had to use an ECS oscillator instead of an AVX in one circuit, and a Kyocera instead of an ECS in another. Footprints all compatible, specs almost the same, price is high for the Kyocera. In fact today for one client I am adding extra fields in an Altium library to handle alternative parts.

[peter-h]

If you buy no-name chinese xtals you can buy thousands for very little money. Qualify the part in the lab over temperature and just keep using it.

The alternative is AVX or Kyocera at 10x the cost, and the supply dries up because all the big users have it in their parts list and just keep buying it (and get amazing prices which you will never see).

[EEVblog]

If you buy no-name chinese xtals you can buy thousands for very little money. Qualify the part in the lab over temperature and just keep using it.
The alternative is AVX or Kyocera at 10x the cost, and the supply dries up because all the big users have it in their parts list and just keep buying it (and get amazing prices which you will never see).

Make sure you buy 10x the volume you actually need, just in case, because it's likely you won't get the same part again.

[EEVblog]

Once a design is locked down, engineers won't be changing back in a hurry, if at all. Furthermore, there will be a "once bitten, twice shy" approach by engineers in choosing vendors for new designs.

Maxim leanred this the hard way. Best sample service, databooks, and parts in the business, but volume? That'll be 40 weeks lead time Sir.
Everyone knew to avoid Maxim parts because they were unobtainium in volume.

[peter-h]

"Make sure you buy 10x the volume you actually need, just in case, because it's likely you won't get the same part again."

Yes indeed, but often this doesn't matter. I still buy cables from China (custom parts) but I buy enough for ~ 2 years, so given that the vendor will be gone within that time frame, it doesn't matter much. That is the only way to deal with China - the place is like a scene from Mad Max these days, and everybody with >2 braincells is trying to make a fast $ and most don't care about long-term. I pulled out all product assembly from there ~ 2 years ago (lost stock and test gear too many times, and narrowly avoided a disaster when one assembly house went bust right as the batch was loaded onto the ship!).

"Maxim leanred this the hard way"

There are very few Maxim parts which one needs. I still use MAX232 MAX3232 MAX489 MAX3089 and MAX705 and they seem to be ok, but they are old parts, and some are second sourced e.g. TI do the MAX3232 very cheaply. Most driver etc chips can be got from TI.

Here in the UK, we used to buy from Maxim in Ireland, but they behaved incredibly arrogantly by refusing to communicate by phone or email. To have any contact e.g. to get a quote you have to open a ticket on a ticket system on their website. I eventually complained to their CEO in the US (who probably had no idea Maxim Ireland were cocking up his company so well) but while they responded interestingly, nothing actually changed.

[nctnico]

"Make sure you buy 10x the volume you actually need, just in case, because it's likely you won't get the same part again."

Yes indeed, but often this doesn't matter. I still buy cables from China (custom parts) but I buy enough for ~ 2 years, so given that the vendor will be gone within that time frame, it doesn't matter much. That is the only way to deal with China - the place is like a scene from Mad Max these days, and everybody with >2 braincells is trying to make a fast $ and most don't care about long-term. I pulled out all product assembly from there ~ 2 years ago (lost stock and test gear too many times, and narrowly avoided a disaster when one assembly house went bust right as the batch was loaded onto the ship!).

Or how about this one: a casing manufacturer stopped because the Chinese government confiscated the land the factory is on. They where not going to re-open business in a different location. 

[Bassman59]

Once a design is locked down, engineers won't be changing back in a hurry, if at all. Furthermore, there will be a "once bitten, twice shy" approach by engineers in choosing vendors for new designs.

Maxim leanred this the hard way. Best sample service, databooks, and parts in the business, but volume? That'll be 40 weeks lead time Sir.
Everyone knew to avoid Maxim parts because they were unobtainium in volume.

I don't think Maxim ever learned. They just kept up with databooks full of interesting parts that you couldn't buy.

Last year it was announced that Analog Devices entered into an agreement to buy Maxim. That deal was supposed to close this summer, so likely imminently. One wonders if ADI will simply prune the catalog of parts Maxim never actually made while also guaranteeing availability for everything else.

[peter-h]

Today I have spent a few hours making sure we can build the first 100 of a new product, later this year.

Passives are no problem, except brand name crystals (Kyocera, AVX) which have gone crazy.
Multi sourced chips e.g. TL431 are easy. Well, the older versions of the TL431 are gone, suggesting this is big slow-acting companies buying them up.
Old chips are generally ok.
Single sourced chips from last few years have disappeared.
Trendy chips e.g. Adesto FLASH AT45DB321, Silicon Labs SI8641, all gone.

The lack of demand for passives confirms there is no underlying demand in terms of a greater end product volume. This is purely a purchasing bubble which will explode.

[tom66]

The lack of demand for passives confirms there is no underlying demand in terms of a greater end product volume. This is purely a purchasing bubble which will explode.

This is rational self-interest which leads to an irrational outcome.  Classic game theory!

We have pre-purchased 500 pcs of FPGAs (spot value £50 ea.)  Holding £25k of stock on site makes the boss nervous.    Not to mention it's expensive to do so.   But if we didn't buy up these parts then we wouldn't be able to get them when we need them - and a check of Octopart confirms my fear here, nobody has stock of this part and lead times are 50+ weeks.  So if we didn't do this, then come later this year when we want to manufacture this device, there would be no parts.

The problem really is, the market has got itself into a panic, and just like toilet roll ran out at the first wave of the Covid pandemic,  chips are running out because everyone is buying parts to cover themselves against a shortage that they are creating.

But as long as other people do this the most rational thing for an individual actor to do is to do the same thing -- the overall behaviour becoming irrational, of course.  Chip makers are wary to add too much capacity, because they need to be careful not to over-supply the market. This will cost them more in inventory costs, and will inflate their assets which will concern shareholders. 

It will probably be over by late 2022,  but right now it's a Tulip Bubble, just with supply instead of price.

[EEVblog]

The lack of demand for passives confirms there is no underlying demand in terms of a greater end product volume. This is purely a purchasing bubble which will explode.

This is rational self-interest which leads to an irrational outcome.  Classic game theory!

We have pre-purchased 500 pcs of FPGAs (spot value £50 ea.)  Holding £25k of stock on site makes the boss nervous.    Not to mention it's expensive to do so.   But if we didn't buy up these parts then we wouldn't be able to get them when we need them - and a check of Octopart confirms my fear here, nobody has stock of this part and lead times are 50+ weeks.  So if we didn't do this, then come later this year when we want to manufacture this device, there would be no parts.

The problem really is, the market has got itself into a panic, and just like toilet roll ran out at the first wave of the Covid pandemic,  chips are running out because everyone is buying parts to cover themselves against a shortage that they are creating.

But as long as other people do this the most rational thing for an individual actor to do is to do the same thing -- the overall behaviour becoming irrational, of course.  Chip makers are wary to add too much capacity, because they need to be careful not to over-supply the market. This will cost them more in inventory costs, and will inflate their assets which will concern shareholders. 

The same psychology is at play in other markets too. Shares, crypto, NFT's, commodities, and it's happening like crazy in housing now too.

[VK3DRB]

The ATSAMA5D27C-LD2G-CU MPU from Microchip is now no longer available anywhere, until April 2022 . I need qty 350 now for an engineering pre-production build.

Microchip has suggested alternative parts which are as useless as tits on a bull. Microchip used to pride itself on parts supply. No longer. Hello McFly.

Why do I waste my time in this crap industry .

[EEVblog]

Not sure if I have replried to this thread before on this, but new orders for my BM786's are delayed until end Nov 2021, and the BM235's until mid Feb 2022 due to component shortage of the main chip.
Just got in my last stock today, and I must have cleared them out of chip inventory because I said I'll buy as many as they can make and they could only deliver a few hundred of each.

[EEVblog]

The ATSAMA5D27C-LD2G-CU MPU from Microchip is now no longer available anywhere, until April 2022 . I need qty 350 now for an engineering pre-production build.
Microchip has suggested alternative parts which are as useless as tits on a bull. Microchip used to pride itself on parts supply. No longer. Hello McFly.

Any alternative packages available and use of a bodge board?

[IDEngineer]

I just got a quote on connectors which usually cost $2-3.

100 pieces were quoted at $14 each.

That is not a typo.

It's not just semiconductors....