Thoughts about uC shortage

Thread Starter

BobTPH

Joined Jun 5, 2013
6,080
I, like many of you, have experienced the shortage in microcontroller chips. I am using a pic24 in 44-pin TSOP. I chose (before the shortage) the most capable member of the chosen family, just because I like to leave room for expansion.

So, which one becomes unavailable? You guessed it, the one I chose. Ones with smaller memory are still available, which is okay for me because I can fit in the smaller memory. But for a commercial product that needed the max memory it would be a disaster.

So why would Micrichip produce the less capable chips and not the more capable one that can replace all of the smaller ones? It makes no sense to me. They would even benefit from the higher selling price if they sold only the larger one to customers who didn’t need it. And fewer customers would be stuck with no chip they could use.

Bob
 

nsaspook

Joined Aug 27, 2009
10,698
IMO

Who says they produce the less capable chips and not the more capable ones? Inventory levels are very customer driven in the current production climate. There are usage based inventory targets for each product type per batch run and there are very few that get to jump to the front of the line. The queue for every popular chip is a mile long.
 

Thread Starter

BobTPH

Joined Jun 5, 2013
6,080
When the lead time I is over a year, as it is for these chips, I have to think they are giving priority to other chips.

Bob
 

nsaspook

Joined Aug 27, 2009
10,698
https://www.jabil.com/blog/global-chip-shortages.html

Why The Chips Are Down: Explaining the Global Chip Shortage
Beyond wafer foundries, wire bonding, substrates, materials and testing are all seeing shortages or delays. In Southeast Asia, COVID-19 outbreaks have impacted the supply of lead frames and assembly and testing. Suppliers have been forced to temporarily shut down facilities to curb the spread of the virus, further cutting overall semiconductor capacity and output.

This continued mismatch between demand and supply is pushing lead times ever longer. As of December 2021, lead times for most semiconductors — no matter the type — are running 40 weeks or more. Essentially no waits are shorter than 28 weeks, but most are far longer. For certain microcontroller and FPGA families, customers are reporting average lead times of 52 weeks.

Analog chip suppliers are seeing an average book-to-bill ratio (the number of orders placed vs. the number of orders filled) of 1.5:1, indicating backlogs are significant. High-end semiconductors like microcontrollers (MCUs) and chipsets are even more constrained, with average book-to-bill ratios of 1.5-2.5:1. Most of these components are in allocation, with most lead times running 52 weeks or more.
 

nsaspook

Joined Aug 27, 2009
10,698
When the lead time I is over a year, as it is for these chips, I have to think they are giving priority to other chips.

Bob
Exactly what are these other priority chips you're talking about that are preventing others? Chip are going out per allocation as they are made as negotiated by contracts. The last, hold the line, super hot chip run I remember was for some vaccine related refrigeration controller that was health critical.
 

MrSalts

Joined Apr 2, 2020
2,617
When the lead time I is over a year, as it is for these chips, I have to think they are giving priority to other chips.

Bob
They may make the most profitable chips. The chips that can be made most quickly with solid demand in applications that can support the prices they want. Some "high end" chips may have longer cycle time or more cycles though the deposition equipment. Why make them if they block up the system and lower the total number of chips made each month. The high end chips cost more because they take longer to make, require more resources.
 

nsaspook

Joined Aug 27, 2009
10,698
So I guess they take large customer orders and put them in a queue.

Bob
Playing favors in the current shortage environment could get you into court in a second. Everybody has lawyers looking for 'fair' allocations per contracts and binding agreements.

I'm not minimizing the personal frustration of the shortage but it's worldwide, on everything.
 

MrSalts

Joined Apr 2, 2020
2,617
Playing favors in the current shortage environment could get you into court in a second. Everybody has lawyers looking for 'fair' allocations per contracts and binding agreements.

I'm not minimizing the personal frustration of the shortage but it's worldwide, on everything.
I am waiting 20 weeks on a bicycle and hopefully the couch I ordered in Late January will be here in July. Everything is right. A local startup company trying to make a wearable device was just told they could get a mold built for their device until Feb next year. Their investors are not happy. All of the electronic components for 10,000 units had been purchased per investor's requirements and nobody looked into the "easy parts" of the device.

The design team has been laid off and founder/CEO is taking the summer off - essentially pushing the pause button. The CEO is still looking for alternative tool builders but they are having trouble getting mold base steel and retaining employees. Tool/die makers are enjoying life right now, huge demand for their skills as manufacturing is "on-shored" and nobody realized what skills are needed in the US. Salaries are way up.
 

Thread Starter

BobTPH

Joined Jun 5, 2013
6,080
I think you are still missing my point.

If I am am selling 3 chips, A, B, and C in my XYZ family, with the only difference being A has 32K of memory, B has 64K, and C has 128K, the C chip could be used by customers who need any of the 3, but the A chip can satisfy only the A customers. It makes sense to me to only make the C chip.

And @MrSalts, do you really think takes longer or costs more ti make the C chip? Often, the different versions actually all have the same die but parts of the chip are disabled for the cheaper versions.

Bob
 

Externet

Joined Nov 29, 2005
1,960
Smells more like a planned manufactured global shortage to pump prices up to be in the same speculative opportunistic trend as everyone else :(
 

nsaspook

Joined Aug 27, 2009
10,698
I think you are still missing my point.

If I am am selling 3 chips, A, B, and C in my XYZ family, with the only difference being A has 32K of memory, B has 64K, and C has 128K, the C chip could be used by customers who need any of the 3, but the A chip can satisfy only the A customers. It makes sense to me to only make the C chip.

And @MrSalts, do you really think takes longer or costs more ti make the C chip? Often, the different versions actually all have the same die but parts of the chip are disabled for the cheaper versions.

Bob
Not missing it, your point is mainly mistaken on the only difference and same die.
 

Thread Starter

BobTPH

Joined Jun 5, 2013
6,080
Not missing it, your point is mainly mistaken on the only difference and same die.
So you have certain knowledge that the chips that differ in memory size are not the same die with memory disabled? Intel actually sold 4 core processors with 2 cores disabled because it was cheaper than having two dies. I don’t see why it would not be the same for Microchip.

Edited to add: There are 32 members of the family I am using which differ in memory sized and numbers of a few peripherals.

Bob
 
Last edited:

Lo_volt

Joined Apr 3, 2014
280
Microchip has always held limited stock of their processors. That's the most frustrating part about using their parts in a product. In the past, when preparing to use any Microchip parts, I learned to give the assembly house a heads up plenty of time before they'd need to receive the parts. 3 to 6 months lead time on larger quantities was not unusual. For smaller quantities the distributors typically held on to some stock of Microchip parts.

I will say that a 1 year lead time is very unusual, even for Microchip.
 

nsaspook

Joined Aug 27, 2009
10,698
So you have certain knowledge that the chips that differ in memory size are not the same die with memory disabled? Intel actually sold 4 core processors with 2 cores disabled because it was cheaper than having two dies. I don’t see why it would not be the same for Microchip.

Bob
Because they operate in different environments (limited range of large expensive processor vs wide range of small cheap controllers) things like same size die optimization don't work (added testing and production costs per die) as well for things like memory (memory takes up a lot of silicon area) features on controllers.
 
Last edited:

MrSalts

Joined Apr 2, 2020
2,617
So you have certain knowledge that the chips that differ in memory size are not the same die with memory disabled? Intel actually sold 4 core processors with 2 cores disabled because it was cheaper than having two dies. I don’t see why it would not be the same for Microchip.

Edited to add: There are 32 members of the family I am using which differ in memory sized and numbers of a few peripherals.

Bob
Did they "disable" 2 of the cores or did only two of the 2 cores pass inspection so they are sold as two-core unit and you got your money's worth.

And I do feel that a core with more peripherals snd more memory may take longer to make or there may be more "fall out" (off spec units per batch) so they know the sweet spot. Or, they simply have more demand for the mid or low end parts. Most people want a microcontroller to very specific things snd only very specific things. That's why there are so many microcontrollers on a car. Why would a car company pay for a bunch of high end micros if it only gets used to control a mirror, or only the hvac fan and thermostat or only gets used to monitor window positions (possibly a single window's position) snd over current in case some kid has their hand stuck in the window. Few people would fill up a high-end micro with code. The mid or bottom range is typically enough.
 

Thread Starter

BobTPH

Joined Jun 5, 2013
6,080
I give up. I am talking about 1 specific family, of which there are probably hundreds at Microchip. I am not suggesting one could or should use a 16-bit 512K memory uC for something that would be best be done by an 8-bit controller with 1K of memory. I just don't believe that have a different die for each uC model they sell, this would run in the 1000s. If they did that the lead time would likely be multiple years, as it would not make sense to do many different runs at the fab in one day.

Bob
 

nsaspook

Joined Aug 27, 2009
10,698
I give up. I am talking about 1 specific family, of which there are probably hundreds at Microchip. I am not suggesting one could or should use a 16-bit 512K memory uC for something that would be best be done by an 8-bit controller with 1K of memory. I just don't believe that have a different die for each uC model they sell, this would run in the 1000s. If they did that the lead time would likely be multiple years, as it would not make sense to do many different runs at the fab in one day.

Bob
Your idea is a logical one for some scales but not on others. Even if one die per family were true, the yields per die would leave very little fractional memory defect controller product to sell knowing the defect profile of most devices and yields.

Why do you think it takes so long for one specific controller? Maybe because there is a huge amount of runs for different devices even in the same family. You can can't compare this tier of semi manufacturing to a Intel process flow.
 
Last edited:

metermannd

Joined Oct 25, 2020
334
Even 'common' parts are getting hard to find.
I just tried to order some ULN2803A Darlington arrays from any of the big suppliers.
ZERO stock and not even a "we expect more on (date)..." status!
Had to go on Ebay to get some - hope they're legit.

ETA: They seem to be the real McCoy - whew!
 
Last edited:
Top