Opinion on 32F401, UFQFPN48 package

[peter-h]

Somewhat bizzarely I am looking at replacing an Atmel 90S1200 (yes, last time buy ~ 10 year ago, been using up old stock) with a 32F401. Atmel offer the 8313 but on their past record they will drop it by the time I get around to doing the PCB

I have spent last few years of my life on a 32F417 product and the 32F401 DS specifically says the 32F401 is "totally compatible" with the 32F417. I could just strip out MbedTLS and save 200k and strip out other junk and that would be it. Re-use the entire development set-up.

I would run it from a 4MHz ceramic resonator. It just reads a dipswitch and implements a timer, plus it interrupts from an incoming GPIO edge. Really trivial.

Does anyone know any bad stuff about this chip? Data sheet is new - Feb 2021.

Anything from ST is likely to run for longer than anything from Atmel...

[NorthGuy]

Atmel offer the 8313 but on their past record they will drop it by the time I get around to doing the PCB

There's no Atmel any more. It's Microchip now. They still sell stuff which is even 20 years old.

[SiliconWizard]

I'd use something much lower power, lower cost and simpler as a replacement (like some cheap Cortex-M0 stuff), but obviously if you're familiar with the STM32F4 series, using a 32F401 will reduce development costs significantly, so I guess you already made up your mind. There's nothing particularly wrong with this MCU. If the STM32F4 series fits your bill, go for it.

[DavidAlfa]

I did lots of stuff with the 411 "Black Pill", with 128/512KB ram/flash and 100MHz it's a great chip.
The 401 is basically a stripped version, slighly slower and 96/512KB.
Consider using tqfp instead, unless the qfn is considerably cheaper or your design is space constrained, soldering is always a bit tricky with QFNs.

[wek]

Just FYI, ATTiny2313/ATTiny2313A is a mostly compatible successor to AT90S1200, and is actively manufactured. There's a migration appnote, AVR093.

IMO a great way to start with a new STM32 is to get a cheap Nucleo board and exercise the features you'd expect from it.

JW

[peter-h]

Thanks wek yes I saw the 8313 still exists. My worry is how long. This is a very old family now. Maybe I should just go for it.

I would not need to get a dev board for the 401; I could just use my existing target and wire up a few of the existing unused GPIO pins (I have about 8 spare).

Agreed about the package; a TQFP can be hand soldered, although the QFN also works OK with the ex-JLCPCB stencils which is how I normally do prototyping. BGA I totally avoid.

[hans]

I like soldering QFNs more than QFPs, less bridging. I now use hot air, but before that I used this method: https://youtu.be/d-f-SBC0GrU

I also used the F411 before. Its a great little chip. Its only a slightly trimmed down version of a F407. I used it in an USB DAC, still works great to this date.

[peter-h]

One can use hot air from above even if it has a thermal pad. Just don't set the hot air gun to +500C

[NorthGuy]

Stencils became cheap. So, I now use stencils and ovens for everything. Much faster than a heat gun.

[SiliconWizard]

Stencils became cheap. So, I now use stencils and ovens for everything. Much faster than a heat gun.

Yes. Even if you don't own a oven, hot plates are usually much easier to use than hot air. You will never have a very focused heating area though (although there are these mini hot plates), but that rarely matters for prototyping.

[langwadt]

I did lots of stuff with the 411 "Black Pill", with 128/512KB ram/flash and 100MHz it's a great chip.
The 401 is basically a stripped version, slighly slower and 96/512KB.
Consider using tqfp instead, unless the qfn is considerably cheaper or your design is space constrained, soldering is always a bit tricky with QFNs.

ebay et.al. listings for "Black Pills" often have both 411 and 401 versions

[peter-h]

What I find attractive about a chip which is a subset of the 32F417 is that the same Cube IDE devkit can be used straight off. I just need to put the 10-pin SWD connector on the PCB and can load code straight in. In fact I probably don't even need to change the CPU type in Cube (which is exactly how I handle a 32F437; the code auto detects the CPU ID at runtime and adjusts a few RAM locations to use the extra 64k).

Otherwise, what most people do is they buy a dev board and use the Cube MX code generator feature to knock up some startup code and maybe some C code for the functionality, which obviously won't work properly so you need to hack it around, while posting a million questions on the ST forum like "my LED won't flash - please help me", and in the end you waste a chunk of your life doing that - as I know only too well from the 32F417 project!

[NorthGuy]

Otherwise, what most people do is they buy a dev board and use the Cube MX code generator feature to knock up some startup code and maybe some C code for the functionality, which obviously won't work properly so you need to hack it around, while posting a million questions on the ST forum like "my LED won't flash - please help me", and in the end you waste a chunk of your life doing that ...

To be fair, people using Cube post quite a bit of questions as well

[DavidAlfa]

Cube needs its own learning curves, know some common stupid bugs, then it's pretty straighforward.

[peter-h]

That's true guys but nothing is as tricky as MX and the half-baked code it generates.

[zapta]

I did lots of stuff with the 411 "Black Pill", with 128/512KB ram/flash and 100MHz it's a great chip.

The F411 is a great choice and is its black pills are reasonably priced. For example from Weact Studio which open source their designs and provide product longevity (I am just a customer, no other interests).  https://www.aliexpress.us/item/3256801269871873.html  https://github.com/WeActStudio/WeActStudio.MiniSTM32F4x1

[Scrts]

Last time I've looked, the 48 pin packages are pin-to-pin compatible between all STM32F series. You can even go with Cortex-M0+ STM32F0 if your STM32F4 is not available...

[Doctorandus_P]

Names of IC packages are getting just as ridiculous as the names for monitor resolutions.   

[SiliconWizard]

Names of IC packages are getting just as ridiculous as the names for monitor resolutions.   

I'm sure you can come up with something better. Please share.

[ataradov]

I don't see how it is different from a normal QFN48 package. ST has blog post where they describe it as some sort of a miracle. It is literally just a standard QFN. I guess compared to LQFP48, it is small. But writing a whole post about a new package, really?

At 0.55 mm, it does look thinner than others. So, it is something, I guess. I doubt 0.3 mm compared to a regular package would be an issue for a lot of designs. May save some money on epoxy though.

[langwadt]

I don't see how it is different from a normal QFN48 package. ST has blog post where they describe it as some sort of a miracle. It is literally just a standard QFN. I guess compared to LQFP48, it is small. But writing a whole post about a new package, really?

At 0.55 mm, it does look thinner than others. So, it is something, I guess. I doubt 0.3 mm compared to a regular package would be an issue for a lot of designs. May save some money on epoxy though.

a couple of decades ago when I was doing Bluetooth, a big hurdle to get designed into a cellphone was whether the package was 1mm or 1.5mm tall

[ataradov]

I understand that there are use cases where size matters, but in that case you are not likely to use QFN in a first place.

It is fine, there is no reason to want thicker package, it is just strange to be so excited about it.

[peter-h]

It seems unusual which is why I asked for opinions, mainly re long term availability. A lot of packages get phased out over time.