(1) Which STM32F2 and (2) purchased where?

[HForr.1]

(1)

Please tell me if I'm on the correct track here. I have an STM32F1 design (STM32F105RCT6TR) and a STM32CubeMX .ioc file that I used for pin-out assignment in the 64-LQFP package. This morning I created a new .ioc file for an STM32F2 (more specifically STM32F205RETx), also using the 64-LQFP package. The two pin-outs seem effectively identical. So, if both .ioc files seem to get done what I need to get done with regard to pin-out and library function setups, then might it really be true that I could simply insert a well-chosen F2 part onto my board originally intended for an F1?

Beyond that, what I also would like to know is, can I make my next redesign of the board intended for the F2 simply support one F2 in the 64-LQFP package -- and then actually get a hold of ALMOST ANY F2 and, as long as the Cube thinks it has all the pin facilities that I need, be successful in putting that F2 on this board?

Maybe I should rephrase. Is it true that all the STM32F1's and all the STM32F2's in the same package are for the most part pin-compatible? If they have the same resource, they'll be found on the same pin, including hard locations like power, crystal and JTAG?

(2)

However, when I search for any SMT32F2's in stock, all the "regular" places show me a big goose egg (that's "0"). So where in the world, literally, am I supposed to be able to get a hold of low 100's quantities of any of these puppies? (And also, therefore, a large part of the reason for my first question!)

Thanks!

[Tesla DeLorean]

It's an old part, got to be a decade since I migrated from F1 to F2/F4 parts

Watch power pins, and new VCAP pins

CubeMX should have a Part Fitting / Will-It-Blend tools to indicate compatible parts. Basically does a sieve on the Data Sheet pin.

Then print out the Data Sheet pin list and do your own due-diligence of pin number vs function.

The F2/F4 have a much better pin-mux strategy than the F1, so might be more accommodation of actually moving/swapping pins.

Again its an old part, expect much demand on it, suggest you pick something newer and less loved.

Perhaps in the G4 range. Perhaps in a different package. More pins, less size even?

Again consider what you're actually using, and how much better those functions might map on something else.

Try places like OCTOPART

Most anything directly ordered will be into mid-2023 at this point

[gbm]

In most cases, L4 series is pin-compatible with F1, so no PCB redesign would be required. F2 series is not futureproof.

[HForr.1]

Thanks @Community member​ and @gbm​ . Actually, I was going to keep the F1 and add a second MCU, as an F2, beside it for particular reasons. @gbm​ implied F2 not good for new designs. @Community member​ assumed I was getting rid of my F1 and thus implied the same about F1 not being good for new designs, or continued designs. I need to UNDERSTAND CLEARLY, please. I take it that I should move *BOTH* of my new-need MCUs to at least an F4 if not a G4.

I don't know the difference in F & G yet, with customer breathing down my neck to get new boards into production. The greater the change, the greater the risk, plus that Good/Fast/Cheap-Pick2 quandary. mid-2023 is of course a non-starter. The customer has already been ordering through gray market, successfully. I thought it was because of general COVID-related chip shortage. Maybe it's both that and obsolescence, although the F1 used isn't marked obsolete. I need to look closer at lifetime info from ST. In 30+ years, I've never seen it this bad. (I've only got 2 years on STM32, that being just one product with F1.)

Yes, I've been using Octopart for years. I do see more hits on F4 at traditional sources.

NEW QUESTION: What's liable to go wrong with footprint and firmware if I jump from F to G? What's the general difference in F & G? I will go try to research on my own, of course, but I'm time limited for this project as well as three other unavoidable irons in the fire competing for my time.

[waclawek.jan]

'F2xx and 'F4xx with the same xx are virtually twins, you should be OK with footprint and maybe even with code down to binary (maybe some minor differences, didn't look extremely closely).

The older STM32 are fairly pin compatible, see the Nucleo64 board for example. There are also "pin compatibility" diagrams in the datasheets, and there are migration appnotes.

'G0 and 'G4 deliberately break pinout compatibility compared to other STM32 in the name of less power pins. They are also the newest amongst the "normal" STM32, and there's some 8 or so years between 'G4 and 'F4, which shows on the differences in the individual peripherals.

There's no OTG USB in 'G4. 'G4 is intended to replace 'F3 not the 'F4. You may also want to read this and this.

Availability is poor across all families, I don't think 'G4 are better off than 'F2/'F4 at the moment.

JW

[Tesla DeLorean]

I don't see anything alarmingly different. I port a lot of material back-n-forth between platforms, and maintain many projects. My portfolio extends to ATMEL, MICROCHIP, SILABS, MAXIM, and a host of others.

I'm removing STM32's from a number of designs due to repeated failure-to-deliver issues.

My overriding recommendation is to find parts you can actually source, buy them and use part foot-prints that can be multi-sourced. And then port the software

[HForr.1]

@Community member​ it sounds that your "overriding recommendation" is to leave STM32 altogether and go with a different processor manufacturer. I've used Microchip plenty of times before. Am I understanding you correctly? My biggest problem with that is I have a moderately complicated code base already running on STM32 on top of HAL by CubeMX. That big a change seems way too much for right now. Instead, I was hoping to have a wide range of footprint compatible parts, as well as a first build quantity in hand before going to even SCH/PCB update.

@gbm​ (Please read this, @Community member​ and @Community member​ ), you said "F2 series is not futureproof.". But at this ST webpage I see lots of 10-year commitment beginning in 2022. I see that for the F2 at that link. I see the same for the F4 at this link. I'm a little nervous at going to G, especially given @Community member​ 's comment.

Please correct me where I may be wrong.

Right now, my plan is to find almost any F-ing F2 or F4 (was that a pun, LOL?) with sufficient quantity at regular sources, get it ordered TODAY (maybe tomorrow) for a one-year supply. Overnight shipping. Get the Mo'Fo's in hand. Then kick off the SCH/PCB revision work... Me, cus? Never!

Thanks EXTREMELY MUCH for the quick reply from all you folks. I'm on a daily routine of pulling my hair out right now...

[HForr.1]

...meanwhile, what about the STM32F466 line. It seems to want slightly different power connections or something like that. I guess I could build jumpers into the board for that. Comments? THANKS!

[waclawek.jan]

You mean 'F446. Yes, that's a good option if you can get them. It does not have ETH but if you don't need it, it's OK it does have two OTGs.

JW

[Tesla DeLorean]

I'm saying if the only citrus at the grocery store is Lemons, I'm going to fill my cart with those, and we're going to make lemonade on the line, not orange squash, because making nothing isn't going to pay anyone's bills

10-year commitments are worth next to nothing. At least Micron has the wisdom to commit to pin-outs and form-fit-and-function level commitments. Why are they worthless? Well they are entirely situational, if the fab burns to the ground, or a piece of irreplaceable equipment breaks, all bets are off. If the next CEO decides it's stupid, or impractical, it's gone, if they sell or break-up the company it's gone, if the company/government that owns the fab decides to go in a different direction, it's gone. ST's got a geometric / exponential problem.

Very few commercial/consumer usages of significant volume are on 10+ year product cycles. And those companies that are, a) own enough of the supply chain to have such expectations, or b) buy enough up-front to cover whatever tail they are willing to commit their own money / resources too. Banks have no interest in inventory, stuff is mostly considered a liability, not an asset, and definitely not a revenue stream. Although the hedge funds are currently making bank on it.

The record for true pin-to-pin level compatibility is extremely poor. Clone manufactures at least understand what the actual goal is.