How is Chipageddon affecting you?

[DC1MC]

You reap what you sow mofos, do another 300% increase 

https://slashdot.org/story/22/12/22/2238222/micron-to-cut-10-of-workforce-as-demand-for-computer-chips-slumps

[peter-h]

Anyway, the shortage seems to be easing indeed. A number of parts are starting to reappear in reasonable quantities.
Whether it's just temporary or not, I don't know.
And, whether it's a good sign or not - unfortunately, I strongly suspect it is just recession rearing its ugly head.
But fact is, I can actually get parts at the moment.

I came back to this long thread because I am seeing the same.

But we aren't there yet. Look at e.g. Mouser and FT232BL. 8k+ in stock.  Bearing in mind the recent history of this chip ("the factory is not accepting orders") this is totally and utterly amazing. Only ~ 6 months ago their UK disti said that if we want some "allocated" by FTDI we will need to pay £1 on top of the then 4k+ price of £2.20 i.e. £3.20. And I had to beg for this, emailing their Taiwanese HQ and everybody else. So they got £2k extra "extortion fee" out of me for a 2k production run. Now this is not an expensive product I am selling so I was well pissed off with this blatent extortion especially as the FT232BL became ex stock more or less when the last of the 2k chips were shipped to us I don't mind posting this because I designed out the FT232BL and went to another chip which is a) available (I bought 2k from Digikey, ex stock) and b) is much cheaper at about £1.50.

I still have 4k bare boards which can be used only with an FT232BL but that's OK because I am sure it will be down to £2.20 before too long There is a man in Italy selling the FT232BL for just over £2.20 but
- he will sell only 5k
- he wants cash in advance
- no discussion entered into beyond a middle finger (I will save you my view on that but let's say it involves credit card disputes with Italian ski shuttle companies)
- no chance of testing one first for counterfeit reasons
- as I know very well, there is absolutely zero chance in hell of getting a penny back from anybody in Italy (except via a credit card, and even that is very very hard because they fight till death)!

On boxing day I did a PCB which carries both the FT232 and the new chip and JLC are now making some; just shipped them, they say

So what is happening? Well, what has always happened in the 45 years I have been in this business. You get a hoarding run on chips, prices go sky high, often 10x, then it collapses into a bloodbath. But in every case the distis tell you nothing until the total and complete collapse. They are all crooks, and want to book orders at inflated prices until the last moment.

And that is where we are now, with Mouser:

1:   £5.90
10:   £5.10
50:   £4.75
100:   £4.45
1,000: £3.85

The 3.85 should be ~2.50 so we are at the "keep the imminent collapse quiet for another few weeks/months" stage. Mouser's 8k stock is amazing but is not moving! I was there a week ago. And they say "43,117 Expected 12/01/2023" so that will be a total collapse of the "FT232 extortion game".

So early 2023 is gonna be interesting

As the old saying goes: be careful who you screw on your way up, because you will have to kiss their bum on your way back down. And we will see a lot of companies on their way down, whose parts were designed-out in the past year. Maxim, FTDI, Microchip, and about half the parts list of my new product.

The Micron story above is hilarious...

[tom66]

But we aren't there yet. Look at e.g. Mouser and FT232BL. 8k+ in stock.  Bearing in mind the recent history of this chip ("the factory is not accepting orders") this is totally and utterly amazing. Only ~ 6 months ago their UK disti said that if we want some "allocated" by FTDI we will need to pay £1 on top of the then 4k+ price of £2.20 i.e. £3.20. And I had to beg for this, emailing their Taiwanese HQ and everybody else. So they got £2k extra "extortion fee" out of me for a 2k production run. Now this is not an expensive product I am selling so I was well pissed off with this blatent extortion especially as the FT232BL became ex stock more or less when the last of the 2k chips were shipped to us I don't mind posting this because I designed out the FT232BL and went to another chip which is a) available (I bought 2k from Digikey, ex stock) and b) is much cheaper at about £1.50.

It's great things are getting back into stock, but there's nothing extortionate about charging market rate for a product.  Look at it this way:  if you take those 2k FTDI chips and put them into widgets you sell, but are lucky to have customer demand for 3k products, are you...

a) going to sell 2,000 products then tell the other 1,000 customers to wait X years for new product
b) offer a lottery where customers can 'win' the chance to buy a product, 2 in 3 chance of winning
c) increase your price until the demand falls to meet supply

In (a) or (b) you see your product (if sufficiently popular) on sale on eBay for the real market price.

That's the reality of limited supply, the price must rise, which causes demand to fall.  We will see semiconductor prices cooling down now supply is coming back, no doubt about that.  You can already see early days of this with the broker prices dropping.  Bosch IMUs were $120 a piece in mid 2020, now they are $15 from brokers.

In your case, you found a way to eliminate the need for that particular chip, so that's efficient too, as you were able to increase supply without increasing your costs too much.  You basically decided it was not worth paying a premium for it, and spent your time on an alternative solution.  It's frustrating to have to do that, but you clearly felt it was worth the cost/certainty.  We've been in the same situation, with many TI parts still difficult to get, they have been designed out.

[peter-h]

I did have to pay the extortion fee on 2k of them.

I call this "opportunism". I bet you anything there is no extra actual demand for these chips. The "demand" is created (and always has been in the past) by quoting crazy lead times. Normally the "a-word" (allocation) strikes fear into buyers and they go crazy ordering everywhere, multiples of their actual needs.

The way this normally works is that prices naturally fall over time. Eventually they fall too far, and a "signal gets generated" to reverse the trend, and the "a-word" is where it starts. This has always worked.

After a bit of time - and this shortage has run for a lot longer than previous ones, driven by covid-induced hoarding of everything from chips to mountain bikes - it all collapses, but the supply industry suppresses the public admission until they can't. So we have nutty stuff like Mouser saying FTDI are not accepting orders and they have 8k in stock (which is not moving) and they will have another 40k next month. Does this add up? And do you think FTDI will drop the FT232BL when it is selling (probably) millions?

This is also interesting and will work in our favour



The chinese have been able to supply chips which the western firms didn't supply. For example while you could not get STM 32F4 you could get the ESP32 stuff. So Espressif made a packet (although of course STM made a huge packet too, selling a lot of chips at inflated prices, to large customers, while not shipping to the disti channel) but there is a political risk... if china gets too aligned with Russia, a lot will change fast. The chinese will retain the bare PCB etc business (because the western PCB industry basically collapsed 20 years ago, and it was junk-QA before then) but CPU choices are made long-term.

[SiliconWizard]

Don't get your hopes up too high anyway, nothing that is bound to happen in the next few years is going to be pretty IMHO. Any glimmer of hope is likely going to be only temporary.
Cutting ties with China while they handle most of our production these days? Sure. Right. No problem. Nobody is going to suffer.

We can just take advantage of the temporary improvement and not put all our eggs in the same basket.

[peter-h]

I think moving some business out of China will reduce their arrogance a bit, which is a good thing. And right now there is an undeniable political risk. It's totally crazy to be trading with a country which wants to invade the place where most of the more complex chips we use are made. This "trade will ensure stability" didn't work with Russia but Germany (and others) completely disregarded the fact that Russia keeps war and trade totally separate. China is a bit smarter about this stuff but they are still apparently hoping to have a go.

[tom66]

Russia's economy is much less developed than China's though, and it's highly dependent upon oil and gas revenues.  The cost to Russia of the invasion is to lose the ability to sell most natural gas to Europe, but Putin's gamble was apparent to all.  Capture Kyiv quickly (that's why VDV - Russian airborne forces - was at Hostomel day 1 of the invasion), install a puppet government, and hope Europe is so desperate for gas during a time of already heightened gas prices and low storage levels that they are too afraid to do anything.  This appeared to work for some member states,  Germany was very reluctant to intervene, holding back on all sorts of sanctions, and effectively had to be bullied into them.  But, ultimately the gamble failed, because Europe supported Ukraine, and thanks to Russian incompetence, the military failed to capture the capital.   Putin's objectives, once the puppet state was installed, would be to export the large amount of gas in Donbas to the mainland to sell on to Europe (the existing Siberian fields are running dry) and to retain a secondary buffer state against NATO, which he sees as a threat.  Russia had already invaded Ukraine, in 2014, with limited response from the EU, so what was another invasion anyway?

It's not impossible that China would turn against the West for Taiwan, but if their aim is to recapture Taiwan for technology, what is the point in doing so if so much of that technology is dependent upon the West?  I think China knows this all too well.  Not to mention, Taiwan has limited wealth outside of technology, no major natural resources, and Taiwan is a small island.  55% of their exports are technology related.  The economy of Taiwan is relatively small (~5% of China's).  The GDP of Taiwan is comparable to around 1-2 years worth of Chinese exports.  The population of Taiwan would not integrate well with the mainland, there would be too much political opposition, protests, riots, sabotage etc.  So the only reason for China to invade would be for political aims, to fulfill the objective of returning Taiwan to China, based on a corrupted ideology.  It is not impossible to believe that could happen, of course, but it does seem unlikely. 

I do think it is essential that semiconductor manufacturing not be concentrated in a few countries though, if only for long term economic stability.

[VK3DRB]

...I do think it is essential that semiconductor manufacturing not be concentrated in a few countries though, if only for long term economic stability.

I have been saying that for the since the Sumitomo semiconductor plastics fire in the 90's. All eggs in one basket is risky, especially in dodgy countries run by communist dictators. In the electronics industry, JIT was never a great idea because it relies on everything in the supply chain working like clockwork at all times. At the moment, a hybrid approach seems a lot safer - stockpile those critical single-source chips that are at risk, buy do JIT on the jelly bean parts.

Microchip and Texas Instruments have some chips starting to appear on the market again and some lead times are reducing a little. The big question is can we ever trust these companies again? Maybe after their CEOs resign and they get new blood in who are competent and customer focused. I did a hardware design and embedded coding for a client using a TI MSP430 variant. These chips became unobtainable for SME's like pretty much everything from TI, but when 2000 appeared at TI Direct, I strongly urged my client to buy 1000 chips immediately for his first small production run in a few months because there is no guarantee those chips will be available again in the medium term when he needs them. He bought them plus some 1000 battery charger chips, and he can now sleep at night.

[peter-h]

In the electronics industry, JIT was never a great idea because it relies on everything in the supply chain working like clockwork at all times.

JIT never really existed. It is a euphemism for a big company (customer) shafting a small company (supplier) into keeping stock free of charge. In electronics especially, it can't work.

[Karel]

Maybe after their CEOs resign and they get new blood in who are competent and customer focused.

With a few exceptions, CEO's now are all shareholder focused. I don't see that changing soon.

[peter-h]

Most things will not change.

Screwing suppliers is an integrap part of the culture of "big" companies. If you want to climb up the corporate ladder, you have some choices
- screw customers (a very bad idea)
- screw employees (they tend to sue so you have to be very careful)
- screw suppliers (always a good one)

JIT works all the time you have willing suppliers.

[HwAoRrDk]

- screw customers (a very bad idea)
- screw employees (they tend to sue so you have to be very careful)
- screw suppliers (always a good one)

Unless you are Amazon, in which case you can do all three with impunity!

[peter-h]

I know how Amazon do it (I have a shop there also): nearly everything they sell is either chinese crap (which can be resold with huge margins) or well known branded products (which also have huge margins, and whose owners needs to watch Amazon constantly for counterfeits). Normal western-made products are difficult to sell via Amazon, mainly because right under your product they have the "Products related to this item" banner, which is a pile of chinese stuff; mostly crap. The company is successful because they sell stuff at prices people like (chinese, mostly), they deliver fast, and it is usually easy to get a refund. For a supplier with "normal" margins, Amazon is hard work, expensive in a number of areas, and they are almost impossible to communicate with. AWS is a similarly shit service; OK when it is running and when something goes wrong you are dealing with complete morons.

Off topic though

Mouser is still showing 8,741 In Stock of FT232BL

[peter-h]

Now 51,758 FT232BL in stock



Not if but when...

[Perkele]

Now 51,758 FT232BL in stock



Not if but when...

I was monitoring FT4232 availability for several months now. It is getting better, but the lead times are still pulled from /dev/random.

[peter-h]

Indeed; my point is that to maximise profit, the lead times are faked until the very collapse.

[Kasper]

A 'coworker' recommended an IC supervisor for a button circuit.  I rejected it due to low stock and made a more generic design with easily replaceable parts but signed up for notifications anyways.  Now I'm getting notifications: part went obsolete.

[VK3DRB]

A 'coworker' recommended an IC supervisor for a button circuit.  I rejected it due to low stock and made a more generic design with easily replaceable parts but signed up for notifications anyways.  Now I'm getting notifications: part went obsolete.

Well done, but somewhat alarming when a nice chip goes off the radar for good. This sometimes happens when a smaller company is swallowed up by a larger one and then the bean counters decide to scrap a unique chip because volumes are now too low. But COVID might have taken some small companies under.

I used a nice chip (ST or TI, I can't remember) in a design several years ago that was a momentary button control chip, where a short press turned provided a power-on signal, another short press provided a power-off signal, and a long 10s press caused a full power-on-reset if case of CPU lock-up. Similar to an on-CPU watchdog timer but it wasn't a watchdog timer. The CPU was in ultra low power sleep mode when it was off. I used that little button control chip it in a product where PCB real estate was very tight. Using a few of MOSFETs and and RC circuit could do the same thing but take up more space and add more BOM line items. It would be nice to remember what it was, but I no longer have the schematics.

[Kasper]

That does sound nice.  There are a variety of them with different options.  Some sound useful.  Just not ideal for me at a small startup.

I wanted button, charger or programmer to turn it on.  MCU to hold it on (when desired) and long press to turn it off even with frozen MCU. 

Also needed fast turn on from charger, to notify users quickly, low on-resistance, and level shifting between MCU and button circuit.

That narrowed the selection down to few options for 'all in one' IC solutions and those solutions required almost as many components as my jellybean version.  They also required a fair bit of time digging through datasheets to find them.

I used a 4 pin load switch, 2 FETs in 1 package, 3 diodes in a common cathode package a few resistors and a cap.  More BOM lines and not quite as small as the IC based solution but seemed safer and worked first try which the original recommendation from 'coworker' (before he recommended a low stock part) did not.

[peter-h]

add more BOM line items

Do some companies have such a rule? It would be utterly bizzare. Resistors and caps and common transistors are always available, and are always cheap. Using commodity parts to replace a unique chip, especially a weird thing like that, is always a good idea.

[T3sl4co1l]

add more BOM line items

Do some companies have such a rule? It would be utterly bizzare. Resistors and caps and common transistors are always available, and are always cheap. Using commodity parts to replace a unique chip, especially a weird thing like that, is always a good idea.

Yes? Most of them? More line items means more reel changes on the PnP, more labor required, more production cost.

Tim

[peter-h]

Not if you design everything with 10k resistors etc (you get my drift)

You can always tell a novice working without supervision. The circuit is full of 9.1k resistors A lot of Honeywell avionics are like that e.g. the KFC225 autopilot - hundreds of weird component values which are so obviously pointless.

[Kasper]

add more BOM line items

Do some companies have such a rule? It would be utterly bizzare. Resistors and caps and common transistors are always available, and are always cheap. Using commodity parts to replace a unique chip, especially a weird thing like that, is always a good idea.

Yes? Most of them? More line items means more reel changes on the PnP, more labor required, more production cost.

Tim

I'm not really familiar with PnP but I've heard removing an item can have extra impact if for example their PnP has 20 reel slots and your BOM has 21 or 41 line items.

I'm not sure how many slots my assemblers have or how much that difference actually costs.

I just know that's more items for me and everyone else to deal with and I prefer to have less.

[Kasper]

Not if you design everything with 10k resistors etc (you get my drift)

You can always tell a novice working without supervision. The circuit is full of 9.1k resistors A lot of Honeywell avionics are like that e.g. the KFC225 autopilot - hundreds of weird component values which are so obviously pointless.

My guess is they had 10k everywhere but were fine with 7k to 13k.  Needed a precise 9.1k somewhere so they changed all their 10k to 9.1k.

[peter-h]

No; they manage to use up much of the E91 range

But the point is that being able to manufacture something is a lot better than using some weird chip which you can't get.

I am now reworking most boards to use multiple alternative chips where possible. For example the LM2936M-5.0 became unobtainable but ST make a similar part, not pin compatible, rather obscure, and half the price! Then there is the MIC5201. MAX3089 can be replaced with a MAX489 which can be replaced with various other chips...

Getting rid of anything from Maxim is a great idea.