Interrupts: can be lost?

I think people put too much into processor longevity. Code can be compiled for another core at much lower cost (including tooling cost) than people are willing to pay for using sub optimal hardware today in the hope it will pay off decades later. About 20 years ago I had something like 10 megabytes of sources written in 68k assembly. At the time - early 90-s - when I picked the 68k (a cpu32 really) processor this was just the best part I could buy. At some point early this century the best part I could buy was power (PPC as they had it back then) based; it did cost me a year to manage what now is vpa (virtual processor assembler) and move on, having practically all my code working again. Was by far the better deal for me than if I had wasted myself into compromising using this or that, looking for toolchains etc. No chance I could have been 10% as efficient as I have been just because I had complete control over my tools.

IOW get the best deal you can get at the moment where other people except you are involved (e.g. chip makers) and do things you have complete control of really well so you don't have to do them again.

Dimiter

====================================================== Dimiter Popoff, TGI

If your goal is to create a product that needs to be constructed and sold f or many years, being able to use programmable logic is not of much utility. Unless the board was originally designed with the same part available tod ay, a board respin will be required which will mean a re-certification effo rt which is often more expensive than designing the board in the first plac e.

If you are just building something with no critical requirements, then sure , use whatever chip you want. Back when ARMs were just starting to be wide spread I predicted they would take over from the other MCUs because people would want a common platform available from many makers. Someone explained very clearly that the CPU is the least important part of the CPU. Compile rs manage the details of 99% of the code and the startup code is often prov ided. But the peripherals is the part that varies between makers hugely an d that can be the source of significant code changes when porting to new pr ocessors.

So RISC-V, ARM, 8051... not really so important unless you need to design a nd certify a product once for the next 20 years or longer. Then the 8051 s tands out from the crowd.

Seems to me that companies that box themselves into a corner over processors and second sources get what they deserve and is generally a sign of poor strategy and project management. I've worked on a couple of projects using essentially obsolete 8051 hardware, recommended a redesign using a more modern architecture, but they were not listening. Result was a greatly extended development time, costs, difficult ongoing software maintenance and compromises everywhere due to banked architecture for code and data. Having said that, have used the silabs 80F... series for a couple of small projects which worked really well. You wouldn't want to look at the asm op from the compiler, but more than adequate for the task and all written in C...

Chris

ries

ioned bugs that are being worked out. Not encouraging. Is Gigidevice a co mpany that makes FPGAs? Or am I mixing them up with someone else? Ah, I s ee I have downloaded their data sheets, so I guess I was looking at a low c ost RISC-V board and also have info on their ARM ST clone chip. Wait, it's the same chip! No sign of FPGAs. That was one of the other companies lik e AGM, Anlogic or maybe Gowin. AGM has an interesting data sheet, but it's two years old and no sign of the device.

If a product requirement is to use only second sourced devices, that's the requirement. How do you propose getting around that?

Fire the manager involved ?.

As I said, if you box yourself into a corner in the design, then choices become limited. I prefer to take a higher level view and try to abstract the hardware into functions that are common to most micros. For example comms, timers and other core functions. Define a set of call interfaces into that required functionality, then all that's needed is an abstraction layer between that and the underlying hardware. There are great benefits in terms of code reuse, reliablity, cost and project times. Use a subset of processor capabilities and deal with special cases as needed. There is some work, but develop over a few projects, to end up with a set of known good library code that can be added to and reused indefinately. I think it was Win NT that first publicised the idea of a hw abstraction layer, but with the speed of modern micros, you can apply that idea to just about any embedded project.

The whole idea is to design assuming that the hardware will change over time, new requirements, performance, cost, whatever, as new micros come to market. The exact opposite of s design frozen in time, with all the hassle that involves...

Chris

series

ntioned bugs that are being worked out. Not encouraging. Is Gigidevice a company that makes FPGAs? Or am I mixing them up with someone else? Ah, I see I have downloaded their data sheets, so I guess I was looking at a low cost RISC-V board and also have info on their ARM ST clone chip. Wait, it 's the same chip! No sign of FPGAs. That was one of the other companies l ike AGM, Anlogic or maybe Gowin. AGM has an interesting data sheet, but it 's two years old and no sign of the device.

the requirement. How do you propose getting around that?

Ok, I get it. You don't understand requirements. There are perfectly vali d reasons for wanting second sources on parts that obviously have never app lied on projects you've worked on. So you don't understand why anyone will need that requirement.

Whether you understand or not, that's a valid requirement and firing the ma nager to change the requirement will only result in a failure of the projec t.

If you want to discuss the issue, I'm happy to do that, but you are going t o need to change a lot of your assumptions for us to communicate.

Even with a strict second-source policy, actually validating the device after with the second source will require some additional work. Do product managers require validation of correct operation with any second source *before* they need to switch to it? Because if not, they should expect the possibility of additional work (including minor code changes) when they do decide to switch. All of which means that in many or most cases, it's entirely possible to consider the GD32 chips as a second source for the STM32 ones.

CH.

???...

Oh, I understand requirements, but while second sourcing is always a desirable thing for any manufactured item, it gets more and more difficult to ensure with the ever increasing complexity and feature set of modern micros, all of which differ. All I was trying to do was suggest a different way of looking at things, but if you have already made your mind up, not much point really. Find a different solution, it's that simple. Just because you demand something won't make it magically happen, or that it's even possible.

The only people who have a financial muscle to ensure second source and decades supply guarantee are probably defence and avionics, but most of industry doesn't work that way, far too expensive. Best you can hope to do is choose a vendor agnostic architecture, like arm, for example, where there's chance you will be able to make at least some of the code common to several vendor's product...

Chris

valid reasons for wanting second sources on parts that obviously have never applied on projects you've worked on. So you don't understand why anyone will need that requirement.

Of course they would verify the second source worked. I designed a board I expected to be making for some time. The FPGA came in two flavors, 3.3 vo lt only or 1.8 volt core. I designed the board to use either with an optio nal step down regulator. I had to build the board with two of the 1.8 volt chips to verify they worked ok. Good thing too... the FPGA subsequently w ent EOL and while Arrow bought up a bunch of them, the price of the 3.3 vol t only version has doubled and tripled. I bought 2000 of the 1.8 volt vers ions a year ago for about $2 each in contrast to the $30 price the 3.3 volt versions are getting now. Just waiting for the next order.

I also used a hybrid footprint to work with either an ADI or Maxim analog s witch. The ADI part has gone up in price a bit, but still cheaper than the maxim part. I also tested that. Never had to use it though.

Having choices in parts purchasing can extend the life of a produce or keep it affordable.

I think it was 2008 when I designed that board.

F series

mentioned bugs that are being worked out. Not encouraging. Is Gigidevice a company that makes FPGAs? Or am I mixing them up with someone else? Ah, I see I have downloaded their data sheets, so I guess I was looking at a l ow cost RISC-V board and also have info on their ARM ST clone chip. Wait, it's the same chip! No sign of FPGAs. That was one of the other companies like AGM, Anlogic or maybe Gowin. AGM has an interesting data sheet, but it's two years old and no sign of the device.

s the requirement. How do you propose getting around that?

valid reasons for wanting second sources on parts that obviously have never applied on projects you've worked on. So you don't understand why anyone will need that requirement.

e manager to change the requirement will only result in a failure of the pr oject.

ng to need to change a lot of your assumptions for us to communicate.

You do not understand, which shows in your use of the term "desirable".

This is getting absurd. You want to make the issue about the whim of a man ager or myself and you don't understand that redesigning a produce at times is literally not an option because of the huge expense. You must not have worked on anything that needed more than just testing and it was out the d oor.

My current product had to go through various levels of certification before it could be used. A part on the board is getting expensive to buy and eve ntually will be unavailable. I might be willing to design a new board with a different part, but the company who buys them may not be willing to pay for the certifications. At that point the product will be dead.

There are apps where the certs are very expensive, like space or various sa fety related uses. Nuke plants are an example.

I'm not going to continue to bat this around with you when you clearly are not getting the concept.

Nice work.

And so the GD32 parts could likewise be validated as a second source for STM32, contrary to your earlier claim.

CH.

y valid reasons for wanting second sources on parts that obviously have nev er applied on projects you've worked on. So you don't understand why anyon e will need that requirement.

nd

s)

rd I expected to be making for some time.

keep it affordable.

Hardly. From what I believe you have written, they have had several issues of non-compatibility and have fixed those issues. For my use, sure, that would not be a bad thing. For many uses, no changes are allowed because po tentially no longer meeting some spec.

If you have a serious need for a component that will meet all specification s for an extended period of time Chinese knockoffs are not a second source. I'm pretty sure I would not even consider them.

The part I use can be bought on Aliexpress and other venues for much cheape r than through approved vendors. I ain't doin' that because there is too m uch chance of counterfeits and crap.

One should also consider the political risk with some politically unstable countries like USA, Russia or China. For this reason. better use components in which the primary manufacturer is outside these political risky countries to avoid licensing issues.

I don't know that. But it has been an issue with chips that have been cloned in a similar way, for example the AVR-like chips that aren't from Atmel. Yet the majority of Arduino sketches (yuk, child's play, admittedly) still "just work". The differences are known and can be worked around, to write code that works on both authentic and cloned chips.

I also don't know (and doubt) that the clone chips have been revised, but as long as you buy the correct version, there's no problem.

If you validate the design on two different chips, and can continue to acquire the same versions, then there should be no problem.

Wake up sunshine! Almost all the "US" chips are now made in China. One of the last hold-outs is Intel, and they're falling behind. You'll need to take early retirement if you refuse to use Chinese chips.

CH

Well, if you have read the news over the past couple of days, that may rule out intel processors. Unless, that is, intel can actually get its act together with its 7nm process.

Usually, this Log Term Support requirements come from an industry (like Defense) where the 'customer' is going to spend a LOT of effort evaluating, proving and certify the design. And because of this they want to be sure that the design (and exactly the design) that was certified will be usable for what they see as the effective life of the program.

f you used a part that becomes no longer available, and re-spin the design to use a new part, they need to evaluate how much now needs to be re-certified, which may cost YEARS of effort and a LOT of money. Now, if you are willing to take PERSONAL responsibility, that the new item behaves exactly as the old, and if some grunt doesn't notice an anomalous result and bombs a village, YOU pay the reparations, go ahead. Or for a different segment, a medical device that has gone though rigorous testing to prove safety and some small change you make happens to be able to be linked to a possibility that it killed a patient.

Many systems don't need the newest and shiniest system parts, but want proven and reliable parts. Parts with long term supply promises, which often include things like second source agreements, tend to provide those promises better.

This market isn't like the consumer market where you actually often want a soft of short lifetime of a given product so you can come out regularly with new and improved versions, and perhaps convince people to upgrade.

Oh, I understand requirements, but while second sourcing is always a desirable thing for any manufactured item, it gets more and more difficult to ensure with the ever increasing complexity and feature set of modern micros, all of which differ. All I was trying to do was suggest a different way of looking at things, but if you have already made your mind up, not much point really. Find a different solution, it's that simple. Just because you demand something won't make it magically happen, or that it's possible.

The only people who have a financial muscle to ensure second source and decades supply guarantee are probably defence and avionics, but most of industry doesn't work that way, far too expensive. Best you can hope to do is choose a vendor agnostic architecture, like arm, for example, where there's chance you will be able to make at least some of the code common to several vendor's product...

Chris

I think I get the concept, but if there is no second source for a part, what's your solution ?. Only thing I can see to ensure continuity is to buy a product lifetime set of parts...

Chris

Basically lifetime buy is the only way indeed, second source or not; second sources are not immortal just like first ones. While Rick's point about certification is valid it is hard to imagine how a customer will be willing to bear the cost of that certification and not the (probably minor in comparison) cost of a lifetime part buy. But people do all sorts of things when planning ahead, you never know what is behind the corner. We all cope with the times which "are a changin'" , sometimes we get it right sometimes we don't.

Dimiter

====================================================== Dimiter Popoff, TGI

Look "sunshine", if you are going to talk nonsense, there is no point in discussing the issue. Chips made in China are not the same as "knockoffs" and most chips are not made in the PRC. Most chips are made in Taiwan and other Asian countries.

Get your facts straight before your resort to insults. It makes you look fatuous.

I don't think I'm going to continue to discuss this issue with you since you have repeatedly failed to understand the significant issues.

Enjoy.