How is Chipageddon affecting you?

[Tomorokoshi]

I bet we'll see a lot of parts get abandoned as part of this. They'll want to focus on more profitable parts and this gives them an excuse to orphan parts that might otherwise have been sustained as "support" for a few more years. Not being cynical, they just have limited fab resources and need to allocate fab time where it's most profitable.

A datapoint from Digi-Key:

PMIC - Voltage Regulators - DC DC Switching Regulators, Active, Tape & Reel: 23,280

I expect a lot of these will get obsoleted.

[bdunham7]

The key is not just drop the cherries but getting something else to replace them and sell this as an improvement. I'd prefer raspberries or blackberries over cherries any day.

Word from the kitchen is that I'm getting a raspberry-strawberry keto black forest cake. 

It depends a lot on what you are redesigning.  If my 40 year old design (that I ship 500 units a year of) uses a PDIP-40 Z8 processor, an 'update' may be quite an issue. 

[IDEngineer]

We have quite a few in-production designs where we use "older" parts simply because they are known, reliable for their purpose, cheap, and multiply sourced. I'm thinking of things like modest opamps and positive DC linear regulators. I'm quite concerned some of those are going to be dropped, and while pin-compatible replacements are available, changing the BOM requires qualification steps that are a PITA.

[floobydust]

Part A cannot be found, so the engineer does a lot of work to come up with a new discrete circuit  to have the same functionality as Part A. Firmware is kept compatible, cost is not much higher. Redo PCB layout, BOM etc.
Manager and Supply Chain come into office, "um uh we can't get Part B. Can you workaround that?"
"You told me you could get all necessary parts except A". "Um yes but not now".

Engineer flips out and refuses to do anymore workarounds or redesigns until all the semi's are sitting on his desk. Seem reasonable?
What's happening is the usual - people don't understand how complicated it is to change semiconductors in a design, pcb layout, then test, regulatory etc. etc. is so much work.

Does the use of Agile methods in hardware design with its two-week "Sprint" schedule hide long-term issues?

Hardware is not software and using "agile methods" or other software engineering strategies on hardware is for complete morons.
Software can be changed in seconds. As I've said, hardware is literally pouring the cement and changing it requires getting out the jackhammer.

[nctnico]

Part A cannot be found, so the engineer does a lot of work to come up with a new discrete circuit  to have the same functionality as Part A. Firmware is kept compatible, cost is not much higher. Redo PCB layout, BOM etc.
Manager and Supply Chain come into office, "um uh we can't get Part B. Can you workaround that?"
"You told me you could get all necessary parts except A". "Um yes but not now".

Nowadays I make sure the parts get reserved / bought before finalising the PCB design.

[floobydust]

It takes a long time to get a Purchase Order issued, even $5,000-$10,000 worth of IC's that will get used, a no-brainer - it has to go through approvals and get signed off by the executive at head office. He must consult his astrologer or something. In the many days lost, the parts get scooped up by time the P.O. is good to go 
I'm not going by people's word: "we're ordering those parts" translation: "we have yet to order them but probably think we should, it will take a week or so".

Mouser had parts i.e. ferrite chip beads, then no stock and 23 week lead-time, and I checked and magically they appeared in stock after 1 week. I can't even keep up with what parts are available.

[IDEngineer]

WRT the crypto subthread: "El Salvador looks to become the world’s first country to adopt bitcoin as legal tender"

https://www.msn.com/en-us/money/markets/el-salvador-becomes-the-world-s-first-country-to-adopt-bitcoin-as-legal-tender/ar-AAKK3CW

Granted it's "just" El Salvador, but big things often start with small steps. Crypto is known the world over, and once nations start treating it as legal tender it's probably here to stay.

[tszaboo]

I think hobbyists and very small startups aside, no one use "agile" or quick hardware design cycles. Now would be the time to do that despite all naysaying, especially from those making the argument about regulations. Do quick cycles using what is available, or die.

You know it damn well that not everyone can. I have 2-4 months of certification for a product before I can legally place something on the market. And only that much because we target the European market with the product, I hear otherwise its a 12-18 months.

[Siwastaja]

changing the BOM requires qualification steps that are a PITA.

And this is what needs to be changed. Regulation and qualification processes are a luxury we were able to afford. Now they have become a threat to the survival of the companies.

[Siwastaja]

I think hobbyists and very small startups aside, no one use "agile" or quick hardware design cycles. Now would be the time to do that despite all naysaying, especially from those making the argument about regulations. Do quick cycles using what is available, or die.

You know it damn well that not everyone can. I have 2-4 months of certification for a product before I can legally place something on the market.

Don't get me wrong, I understand the struggle and I understand well enough how the industry operates - or operated before the crisis.

So law prevents you from doing quicker design cycles requiring long certification cycles. But firstly, have you actually looked at the law? It tends to be very roundabout. The whole industry works based on interpretations of law and assumptions that following certain standards strictly is "the easiest way" to fulfil the requirements of the law. But if it stops being easy due to external situations, the whole argument falls down.

Secondly, large players especially have absolutely no need nor desire to follow any regulations or laws if they so wish, the VW (and others!) emissions scandal being a typical and recent example. Companies routinely break the law, often severely, often causing actual damage to the others just for the sake of a tiny bit of competitive edge, some money saved, or a slightly better product. Now we are in an exceptional situation and it's actually about the survival of the companies and you can't even take a risk of interpreting the law with a risk of accidentally breaking it, give me a break.

I know all this sounds very radical to some and I'm not advocating such change but we are just some half a year into the crisis, and many expert opinions say this may go on as long as two years. Come back to this post a year from now and see if it still sounds radical.

And you don't need to drop all regulation and certification and start building death traps; engineers already have a lot of responsibility and experience building safe products, and quicker certification cycles (like 1 month instead of 4 months) can still catch the worst offenders.

In exceptional circumstances, exceptional measures are sometimes needed, you can't stick to the old and safe habits. If you are starving to death, you will just kill an animal to feed yourself and your family even if that legally required a permission process to do so.

Of course, shortening certification processes to allow quicker redesign cycles isn't a silver bullet that solves the chip crisis (because there are shortages in almost every chip!), but it might end up being a necessary tool for survival of the companies, I don't know. We'll see.

Finally, this isn't the problem of the engineers, but the problem of management. As an engineer, all you can do is to communicate so that management can't say they "didn't know".

[T3sl4co1l]

In exceptional circumstances, exceptional measures are sometimes needed, you can't stick to the old and safe habits. If you are starving to death, you will just kill an animal to feed yourself and your family even if that legally required a permission process to do so.

Heh... with a very large caveat:

It's morally acceptable for an individual to violate laws (or certain ones, anyway) for their survival.  (Legally acceptable, even.  And note that laws are themselves amoral, at best being parallel but different.)

Whereas it may be morally right -- or at least inconsequential -- for a corporation to "die" in such a case.

Corporations will still fight for their existence, as all living things inevitably do (for the alternative is the nullity of nonexistence), but the humans who run them mustn't lose their perspective.

Tim

[VK3DRB]

It takes a long time to get a Purchase Order issued, even $5,000-$10,000 worth of IC's that will get used, a no-brainer - it has to go through approvals and get signed off by the executive at head office. He must consult his astrologer or something. In the many days lost, the parts get scooped up by time the P.O. is good to go 
I'm not going by people's word: "we're ordering those parts" translation: "we have yet to order them but probably think we should, it will take a week or so".

Mouser had parts i.e. ferrite chip beads, then no stock and 23 week lead-time, and I checked and magically they appeared in stock after 1 week. I can't even keep up with what parts are available.

I worked for a company like that named IBM. A two year financial planning cycle. Once I wanted to buy a novel piece of test equipment a Toneohm shorts locator.

Q: "Why did you not plan for this?!"
A: "It did not exist when the planning was done."

I had present a full business case on spending $3.5K at the time that was not planned for. I worked out the return on investment would be a week at an estimate. I was told that the ROI is too optimistic and I had to fiddle the figures to make it a month, because no-one will believe me. Three re-spins of the business case. Had to go through all sorts of approvals and by bureaucratic management. After some weeks I got approval, with the money being creatively accounted for out of another bucket. It turns out that the return on investment was ONE HOUR. Within half a day we had recovered $70K worth of bones pile RS/6000 processor boards, with many shorts being a bent pin on one of the many soldered PGAs.

Lessons learnt:

1. Bureaucracy costs money, costs time, costs morale  .
2. Toneohm shorts locators are terrific 

[Siwastaja]

Often the problem is in internal bureaucracy like VK3DRB described, not due to law.

Blaming the law is the easy answer to shoot down ideas that require changes. My rule of thumb is, whenever someone says it's the law that requires us to do something: let's see if it really does, show me the law.

[nctnico]

I think hobbyists and very small startups aside, no one use "agile" or quick hardware design cycles. Now would be the time to do that despite all naysaying, especially from those making the argument about regulations. Do quick cycles using what is available, or die.

You know it damn well that not everyone can. I have 2-4 months of certification for a product before I can legally place something on the market. And only that much because we target the European market with the product, I hear otherwise its a 12-18 months.

But that doesn't have to stop you from stockpiling / ordering parts for a production run after the qualification process. Sure the qualification process can halt the production for a while but a 1-2 month delay is better than having nothing to sell for the next 12 months. You can start buying parts as soon as the level of certainty versus monetary risk is positive for a part. There is no need to wait for the entire qualification process to end.

[Howardlong]

Does the use of Agile methods in hardware design with its two-week "Sprint" schedule hide long-term issues?

Hardware is not software and using "agile methods" or other software engineering strategies on hardware is for complete morons.
Software can be changed in seconds. As I've said, hardware is literally pouring the cement and changing it requires getting out the jackhammer.

The key points here are risk, mitigation, cost and scalability.

Certain facets of agile can be used, but the fundamental strategy of agile (and devops, CICD etc) is that in exchange for frequent small incremental changes there will be added risk as full regression testing is too expensive and takes too long. Risk is mitigated by also being able to reverse out changes quickly, although often the damage has already been done.

The problem of small incremental changes on hardware is that the cost of spinning a new board every week or two at production levels is hugely expensive, not just in setup costs, but also for when your production run fails, especially once shipped. Rework is hugely expensive as unlike software, it just doesn’t scale well.

[NiHaoMike]

If only there was some system in place so that safety critical equipment like certain kinds of medical and automotive electronics get the first pick of the parts since any change means a lot of work to verify that the relevant safety requirements are still met, while other products get to fast track redesign to use what parts are more readily available.

[floobydust]

Does the use of Agile methods in hardware design with its two-week "Sprint" schedule hide long-term issues?

Hardware is not software and using "agile methods" or other software engineering strategies on hardware is for complete morons.
Software can be changed in seconds. As I've said, hardware is literally pouring the cement and changing it requires getting out the jackhammer.

The key points here are risk, mitigation, cost and scalability.

Certain facets of agile can be used, but the fundamental strategy of agile (and devops, CICD etc) is that in exchange for frequent small incremental changes there will be added risk as full regression testing is too expensive and takes too long. Risk is mitigated by also being able to reverse out changes quickly, although often the damage has already been done.

The problem of small incremental changes on hardware is that the cost of spinning a new board every week or two at production levels is hugely expensive, not just in setup costs, but also for when your production run fails, especially once shipped. Rework is hugely expensive as unlike software, it just doesn’t scale well.

You're talking as if hardware can take "frequent, small incremental changes"- which I am saying is a myth.
Hardware has little agility, you're dealing with physical items, not software.

Risk is not mitigated by going in reverse, let's ask Boeing about that approach, the concrete has hardened and reverse is not a option. Hardware is not malleable and this is why I'm saying agile methods are of no use, particularly with chip shortages.
The cost of doing a board spin is low, it's the regulatory approvals that require months and $$$$$. Does not fit into "agile" at all.

Products with functional safety, it would be great if they had priority but the fabs went where the money is. Automotive semi's are low margin, high volume and the automakers thought they are King of the Hill and could throttle back orders without consequence, and that the fab industry's excess capacity would just sit their idling.

[tszaboo]

Secondly, large players especially have absolutely no need nor desire to follow any regulations or laws if they so wish, the VW (and others!) emissions scandal being a typical and recent example. Companies routinely break the law, often severely, often causing actual damage to the others just for the sake of a tiny bit of competitive edge, some money saved, or a slightly better product. Now we are in an exceptional situation and it's actually about the survival of the companies and you can't even take a risk of interpreting the law with a risk of accidentally breaking it, give me a break.

Yes, I have actually read the law, not just the standard. Its quite strict.
If I make a mistake, here is the worst case scenario: An oil refinery burns down, with dozens of people in it. I saw a video of such an industrial accident. A tank containing some flammable material just flew out of the site, and landed some hundreds of meters away.
If the chip that is designed in and safe and verified and checked and triple checked isn't available, then I wait.

[Howardlong]

Does the use of Agile methods in hardware design with its two-week "Sprint" schedule hide long-term issues?

Hardware is not software and using "agile methods" or other software engineering strategies on hardware is for complete morons.
Software can be changed in seconds. As I've said, hardware is literally pouring the cement and changing it requires getting out the jackhammer.

The key points here are risk, mitigation, cost and scalability.

Certain facets of agile can be used, but the fundamental strategy of agile (and devops, CICD etc) is that in exchange for frequent small incremental changes there will be added risk as full regression testing is too expensive and takes too long. Risk is mitigated by also being able to reverse out changes quickly, although often the damage has already been done.

The problem of small incremental changes on hardware is that the cost of spinning a new board every week or two at production levels is hugely expensive, not just in setup costs, but also for when your production run fails, especially once shipped. Rework is hugely expensive as unlike software, it just doesn’t scale well.

You're talking as if hardware can take "frequent, small incremental changes"- which I am saying is a myth.
Hardware has little agility, you're dealing with physical items, not software.

Risk is not mitigated by going in reverse, let's ask Boeing about that approach, the concrete has hardened and reverse is not a option. Hardware is not malleable and this is why I'm saying agile methods are of no use, particularly with chip shortages.
The cost of doing a board spin is low, it's the regulatory approvals that require months and $$$$$. Does not fit into "agile" at all.

Products with functional safety, it would be great if they had priority but the fabs went where the money is. Automotive semi's are low margin, high volume and the automakers thought they are King of the Hill and could throttle back orders without consequence, and that the fab industry's excess capacity would just sit their idling.

I'm agreeing with you! :-)

[tszaboo]

So I hear from our purchaser, that the missing parts are available. Someone in China bought them, entire reels, and they are selling it for 20 USD each, instead of the usual 0.6 USD price. We see similar issues on 4-5 parts. They of course cannot give invoice, no info on storage conditions, or anything else.

[Evan.Cornell]

A part that normally costs $2 at moderate volume was just quoted by a similar type of vendor for $16 a pop. And $70 per chip at single unit quantities. The gouging is incredible.

[SilverSolder]

Sadly, the only way prices will come down, is if customers (us) refuse to buy...   

Certainly, my own customers are not going to accept price increases in the hundreds of percent, so why should I? 

[madires]

Ferengi Rule of Acquisition #9: Opportunity plus instinct equals profit.

[harerod]

So I hear from our purchaser, that the missing parts are available. Someone in China bought them, entire reels, and they are selling it for 20 USD each, instead of the usual 0.6 USD price. We see similar issues on 4-5 parts. They of course cannot give invoice, no info on storage conditions, or anything else.

If everybody could stick by the rules regarding sensitive equipment, all those components must be considered worthless for anything requiring controlled sourcing.

I could think of some suggestions as to where those gamblers may put their treasures.

We have a very rough stretch ahead of us and not every company is going make it. I just hope that the rest will have learned several expensive lessons. Fat chance, though.

My personal experience when preaching the same mantra regarding procurement since last Christmas:
- some have had more important stuff to take care of than procurement
- some switch to Chinese parts (with a different development team)
- some hope their storage will last till early 2023
- some are in denial
- some go on the dole

[Calaverasgrande]

Haven't got an ideological polemic to hang my hat on today.
Just wanted to chime in with my experience as a hobbyist.
I'm working on a replica of a famous/infamous Solid State Logic audio program compressor.
 I have a few other audio related kits and projects in the works.
One guitar pedal is short a half dozen 2N5457 JFETs, which everyone is out of, showing obsolete in many places now.
An EQ and a Limiter I'm working on are both going to gather dust until the specific audio transformers they require are back in stock.
An OEM just dropped out of the audio transformer biz leaving a lot of 2nd tier boutique suppliers like Ed Anderson, CAPI and Hairball audio in the lurch.
But the SSL Compressor replica is waiting on THAT VCAs which nobody has. And it's one of those things with one legit supplier per region.
So long story short I'm paying shipping and some markup to get JFETs and THAT VCAs from Germany to the US so I can wrap these things up.
There aren't normally hard to come by at all. But hey, supply chains aren't just snaring the 7nm silicon.