Cheap Bluepill, very likely it has fake STM32 right ?

[OwO]

Yes, people just need to understand that "stm32" on ebay/aliexpress doesn't mean "stmicroelectronics stm32", it means "stm32 compatible" which generally adheres to the datasheet and reference manual. Code that depend on undocumented behavior may not work across all manufacturers, but in all my tests of the peripherals including ADC, DMA, timers, USB, SPI, and flash writing I've not found anything that deviates from the spec sheet. Maybe if you use a broken USB library (the one supplied by ST is broken!) it won't work on a CKS32, but so far with libopencm3 all peripherals including USB worked perfectly on all compatibles I've tried.

For someone new to STM32, I don't think the compatibles are a bad choice, and I'd even recommend testing your code on all compatibles including a genuine ST STM32 from mouser so that you can be sure your code is portable and doesn't depend on unspecified behavior. Just like you should test your web app on several browsers, and not just a Genuine Microsoft (R) internet explorer (TM).

[panoss]

Yes, people just need to understand

Peolple don't need to understand anything.
When somebody orders an STM32 and receives an "stm32 compatible" it's simply a fraud.
And he will have to make the seller 'understand' this.

[BravoV]

Part numbers are not generally considered trademarks here. If I produce a transistor that adheres to all 2n3904 specifications, I can sell it as a 2n3904, even if some characteristics not on the datasheet are a bit different.

In the olde days, among "big" manufacturers used to practice that among them selves.

Your example on the popular transistor, or chips like LM317, LM324, 78xx regulator and etc.

Its just in those days the electronics communities didn't use ugly words like "clone" , "copy" and etc, they used the nice word called ... "cross referenced" . 

[langwadt]

Part numbers are not generally considered trademarks here. If I produce a transistor that adheres to all 2n3904 specifications, I can sell it as a 2n3904, even if some characteristics not on the datasheet are a bit different.

In the olde days, among "big" manufacturers used to practice that among them selves.

Your example on the popular transistor, or chips like LM317, LM324, 78xx regulator and etc.

Its just in those days the electronics communities didn't use ugly words like "clone" , "copy" and etc, they used the nice word called ... "cross referenced" . 

and each put their logo on them so if you wanted a specific type you could tell them apart.  And an MCU is many orders of magnitude more complicated than a voltage regulator/opamp etc.

[BravoV]

Its just in those days the electronics communities didn't use ugly words like "clone" , "copy" and etc, they used the nice word called ... "cross referenced" . 

I prefer "second source".

BTW, I don't appreciate if someone put a second source chip in a package with the original's logo on it.

I'd be totally fine if someone rolls out an STM32 board with clear photo showing the chip is not from ST, and a firm disclaimer in item description.

Agree, that is simply a crime.

[techman-001]

Yes, people just need to understand that "stm32" on ebay/aliexpress doesn't mean "stmicroelectronics stm32", it means "stm32 compatible" which generally adheres to the datasheet and reference manual. Code that depend on undocumented behavior may not work across all manufacturers, but in all my tests of the peripherals including ADC, DMA, timers, USB, SPI, and flash writing I've not found anything that deviates from the spec sheet. Maybe if you use a broken USB library (the one supplied by ST is broken!) it won't work on a CKS32, but so far with libopencm3 all peripherals including USB worked perfectly on all compatibles I've tried.

For someone new to STM32, I don't think the compatibles are a bad choice, and I'd even recommend testing your code on all compatibles including a genuine ST STM32 from mouser so that you can be sure your code is portable and doesn't depend on unspecified behavior. Just like you should test your web app on several browsers, and not just a Genuine Microsoft (R) internet explorer (TM).

" I'd even recommend testing your code on all compatibles including a genuine ST STM32"

I think your advice is for Chinese users of the home grown 'compatibles' but I wouldn't do that because I'm only using genuine ST STM32 chips, and their documents tell me exactly what to expect, there is no ambiguity here for me.

My situation here in the West is the reverse of yours, I can buy the genuine STM32 chips at a good price, there is no need to try and buy them from China which is now far too risky given that I may receive up to 3 different 'compatible' non genuine chips, marked as STM32F103x instead. My motivation to buy 'compatibles' is exactly zero.

" but in all my tests of the peripherals ... I've not found anything that deviates from the spec sheet "
There is no single "spec sheet", there are many STM32F1x documents, including ones from ARM. It takes months just to read them all.

Furthermore the STM32F1xx has 37 peripherals, 413 registers and 3044 bitfields, so I don't believe you could have tried more than a few combinations.

We have a old saying here in the West, "lack of proof is not the same as proof of lack".

I don't want to give the impression I'm anti Chinese, because I'm not. I'm only anti counterfeit and fakes, which are illegal in the West.

I think the Chinese people have made a terrible mistake with the STM32F103 counterfeit chips, at least from a perspective of selling them here in the West. I believe those chips will mainly be sold in China now as no one wants them here.

Contrast the 'compatibles' to the GD32VF103, it has the open source RISC-V core which anyone is free to use, and a lot of the STM32F103 peripherals which have their own GD32VF103 names now. It's not a 'compatible' of anything, it's a new chip, all Chinese. There is a GD datasheet that's ok, but has some terrible errors.

I have no complaints about the chip, I received a couple of GD32VF103's in different boards and they seem to work ok, tho it's early days yet. GD may possibly sell a *lot* of the GD32VF103 here when they are available as chips only, and if the price is good.



Look at the excitement the Padauk PMS150C "3 cent micro controller" has caused since it came out, it's not a copy of anything, the documentation is great, the IDE and programmer software (windows only) is free to download and they have hardware simulators to develop applications so buckets of OTP PMS150C's aren't required for development. I think it's a excellent, honest product with first class documentation and support.

I believe that honest, reliable, innovative and low priced Chinese MCU's may sell well in the West, but we have long memories and don't like being ripped off with counterfeits and fakes.

I believe you may see the impact of this in lower chip sales to the West eventually because the buyers here know they can't trust they will get what they paid for when buying electronics from China.

Earning back buyer trust may take China decades.

[techman-001]

I believe that honest, reliable, innovative and low priced Chinese MCU's may sell well in the West, but we have long memories and don't like being ripped off with counterfeits and fakes.

They didn't rip you off if they made it clear that those are CKS32Fxxx or GS32Fxxx, and they told you those are legally legit STM32 clones. Those chips are properly licensed by ARM, and you can't protect a peripheral set by any IP laws nor a naming schema rest of STM as a trademark.

ST did try to made them illegal by claiming that they used ST's header, libraries and linker script files, but since then they wrote their own.

Running ST firmware on those chips is still illegal as ST header and library license prohibits as such, but that's the liability of the user, not the chip manufacturer.

Also, GD doesn't even care how their chips sell in the West. They care to sell the chips to large quantity appliance manufacturers, and they will trickle to the West in those final products.

I don't think Chinese fablesses even care to sell to the West directly. Rest of ESP series, I can't even recall a single manufacturer that actually supported low volume manufacturers. Put it simple, they don't care.

Chips don't grow on trees in China. They are manufactured at the same cost, if not more, due to the higher tax and embargo of high end chip making tools, and they are sold cheap due to lower service cost, which a great part of it is customer support cost.

That's also why ESP will never be a super leader in the industry. Some of their chips are more expensive than Nordic and TI parts, and how do you expect them to hit mass market?

You like it or not doesn't matter. What matters is the market likes AllWinner and MTK, not TI, at least not in Asia and most households in even the West. Qualcomm and Samsung do have similar products, and Amazon even used them in early Echo series. Guess what, those got replaced with MTK.

TI DaVinci had the same success in Chinese IP camera market, and guess what, AllWinner and HiSilicon completely crushed the market despite the lack of good quality documentation and zero support availability.

And before you ask, I believe OwO's point was remarking a GD to STM is bad, but that's not done by GD. You get fake chip makers in China, and we hate them too. It doesn't make GD clones of STM any bad.

And to your point, why do sellers market them as STM, well, if they went far enough to remark those chips, do you think they have integrity? Again, that makes the sellers bad, not GD.

And as a disclaimer, I don't work for GD, nor AllWinner or HiSilicon. I do work in research industry looking into second sourcing the latest Western technology, but we do clean IP from scratch, and build our own IP portfolio licensing to mainly Chinese mainland and Taiwanese IC manufacturers. I work with Chinese second source chips as a hobby, like GD/CKS vs STM, AllWinner vs TI/NXP, Gowin vs Lattice and more, but I don't get a single dime from them.

It should be fairly obvious I only have problems with counterfeits. You must have seen the decapping pics by Noopy in the recent posts which included a CS32 which was clearly relabeled as a STM32F103 and  that's *exactly* the kind of ripoff I mean.

A legal clone is what we call a 'second source', and that's perfectly acceptable, no problems there. I wish the legit Chinese semiconductor industry all the best, innovation and competition is good for tech consumers like me.

[SiliconWizard]

It should be fairly obvious I only have problems with counterfeits. You must have seen the decapping pics by Noopy in the recent posts which included a CS32 which was clearly relabeled as a STM32F103 and  that's *exactly* the kind of ripoff I mean.

A legal clone is what we call a 'second source', and that's perfectly acceptable, no problems there. I wish the legit Chinese semiconductor industry all the best, innovation and competition is good for tech consumers like me.

Agreed. (Also - let's not forget the original thread was about the Bluepill, which is a cheap dev board, which is why I personally focused on the hobbyist use case.)

Counterfeits are just illegal, and the additional downside for the user is that they can't be guaranteed in any way usually.

Now China (in particular) has a lot of regular clones (not counterfeits), and even a lot of original and very low-cost MCUs these days, so the point of choosing a counterfeit over those if you're targetting low-cost and chinese vendors is completely dubious. There's just no reason to do it, except to effectively rip-off the original vendor (leveraging their popularity, dev tools, code base, etc.)

[HwAoRrDk]

The mention of part numbers and trademarks sparked something in my mind...

ST have a trademark on the term "STM32". It's a stupid thought, but what if the reason why these CKS/GD chips re-marked to be ST are marked with "STM32F" is because they are being careful not to tread on the exact trademark term? Although, perhaps not, given that most IP protections only seem to warrant a passing thought in China.

If they ever tried to pursue legal action, I'm guessing ST would dearly hope that a court would not decide that the mark "STM32" does not go the way of Intel and be deemed to have been 'genericised'. Back in the '90s Intel famously lost the case against AMD using the term "386", and had their trademark invalidated. This then set the course for Intel to start using names like "Pentium" for their CPU range.

[SiliconWizard]

I don't know whether ST has ever tried to take action on those counterfeits or not.

Since ST seems relatively popular in China (and probably has an interesting market there), they may just kindly ignore the problem in order not to irritate the chinese. Dunno.

[techman-001]

It should be fairly obvious I only have problems with counterfeits. You must have seen the decapping pics by Noopy in the recent posts which included a CS32 which was clearly relabeled as a STM32F103 and  that's *exactly* the kind of ripoff I mean.

A legal clone is what we call a 'second source', and that's perfectly acceptable, no problems there. I wish the legit Chinese semiconductor industry all the best, innovation and competition is good for tech consumers like me.

Agreed. (Also - let's not forget the original thread was about the Bluepill, which is a cheap dev board, which is why I personally focused on the hobbyist use case.)

Counterfeits are just illegal, and the additional downside for the user is that they can't be guaranteed in any way usually.

Now China (in particular) has a lot of regular clones (not counterfeits), and even a lot of original and very low-cost MCUs these days, so the point of choosing a counterfeit over those if you're targetting low-cost and chinese vendors is completely dubious. There's just no reason to do it, except to effectively rip-off the original vendor (leveraging their popularity, dev tools, code base, etc.)

Exactly.

I personally ranted against the Bluepill because I thought the STM32F103 was too old. Even tho it's still very capable there are literally a ton of far better models for hobbyists to spend time learning. As a Forth user I operate close to the metal and I see the STM32F103 issues intimately.

Compared to the STM32F0xx that old chip is a pig. A very fast pig 

Then the Bluepills started coming out with the counterfeit chips and customer confidence plummeted. It was never that high to start with due to issues with the USB port breaking off, no bootloader, bad USB resistors etc.

The clone situation and resultant confusion here prompted me to make my bootable USB Forth based STM32F103-diagnostics program, and along the way I came to appreciate the STM32F103 a bit more than I had before. At the end of the project I had become so used to the response of a 75MHz STM32F103 that my STM32F051 development system is now (over)clocked at 75 MHz also.

$2 for a Bluepill was insanely cheap and hobbyists couldn't resist them, as you say it's a new age of "cheap at any cost" insanity. Kids have grown up thinking a $20 MCU is expensive, but those of us who paid $175USD for a Motorola MC6800 think otherwise.

But cheap doesn't have to mean crap, look at the Padauk PMS150 ?

I did because I thought it would be utter rubbish but after looking at the website I changed my mind totally, it's a very classy product ... great job Padauk!

In summary I think the Bluepill's days are over, no one in the West trusts them any more because of the counterfeits. Greed killed off old grandfather Bluepill in so many ways.

Perhaps it's time for honest homegrown $2 Chinese GD32Pills or a CS32Pills (if anyone can find a CKS32 technical reference) ?

If they come with the MCUs promised there may still be room for them in the ultracheap hobbyist parts boxes?
 

[mskeete]

The mention of part numbers and trademarks sparked something in my mind...

ST have a trademark on the term "STM32". It's a stupid thought, but what if the reason why these CKS/GD chips re-marked to be ST are marked with "STM32F" is because they are being careful not to tread on the exact trademark term? Although, perhaps not, given that most IP protections only seem to warrant a passing thought in China.

If they ever tried to pursue legal action, I'm guessing ST would dearly hope that a court would not decide that the mark "STM32" does not go the way of Intel and be deemed to have been 'genericised'. Back in the '90s Intel famously lost the case against AMD using the term "386", and had their trademark invalidated. This then set the course for Intel to start using names like "Pentium" for their CPU range.

Intel lost because the courts ruled that you cant trademark a number. 386 / 486 are just numbers and no-one has an exclusive right to them.

[GromBeestje]

ST did try to made them illegal by claiming that they used ST's header, libraries and linker script files, but since then they wrote their own.

And now ST put their libraries on GitHub. The Drivers are under a 3 Clause BSD license so can legally be used by any compatible devices. There are some parts still under SLA0044 license, such as the USB Middleware. A high level USB implementation is easy enough to replace. (I wrote some stuff to replace it)

Perhaps it's time for honest homegrown $2 Chinese GD32Pills or a CS32Pills (if anyone can find a CKS32 technical reference) ?

Well... that's the problem with the C(K)S32. The lack of any documentation. Not even mentioned on the CKS website, so is it even made by them?

For the GD32, there are datasheets, technical reference manuals, errata, libraries, everything is available. There are distributors in the West. I would trust making a design using GigaDevice parts.

Maybe some Pills with a MM32F103CBT6 by MindMotion. There are datasheets, reference manuals and libraries at their site. This beast can run at 168 MHz, but is not STM32F103 compatible. To achieve that speed the layout of the RCC registers is changed. I'd have to go through the documentation to see what else has changed.

I just checked LCSC, and a new *32F103 variant popped up. There is an HK32F103CBT6.  From what I can tell it is made by http://www.hsxp-hk.com/ but their website is down (Error 522 from a Cloudflare-like service)

[thm_w]

Maybe some Pills with a MM32F103CBT6 by MindMotion. There are datasheets, reference manuals and libraries at their site. This beast can run at 168 MHz, but is not STM32F103 compatible. To achieve that speed the layout of the RCC registers is changed. I'd have to go through the documentation to see what else has changed.

I just checked LCSC, and a new *32F103 variant popped up. There is an HK32F103CBT6.  From what I can tell it is made by http://www.hsxp-hk.com/ but their website is down (Error 522 from a Cloudflare-like service)

It says CPU = 96MHz, is 168MHz guaranteed or is it just added to the PLL table?
https://lcsc.com/product-detail/mindmotion_MindMotion-Microelectronics-MM32F103CBT6_C316700.html

and now its even more confusing, they have a STM32L373, which doesn't exist from ST:
https://lcsc.com/product-detail/Microprocessor-Microcontroller_MindMotion-Microelectronics-MM32L373PS_C212250.html

HK32F103 basic datasheet attached.

[GromBeestje]

It seems they've pulled that from their datasheets. It used to say, speed 96 MHz, Turbo mode 168 MHz. Unfortunately my hard disk crashed and I've lost the version of their datasheet making that claim.

I've found the reference: http://www.mindmotion.com.cn/userfiles/images/MM32F103XiLieWenDang/DS_MM32F103xx_o_V1.13_EN.pdf
Under "1.2Product Feature" it says:

Standard operating frequency is up to 96MHZ
Maximum operating frequency is up to 168MHZ

This one is missing from http://www.mindmotion.com.cn/userfiles/images/MM32F103XiLieWenDang/DS_MM32F103xx_n_V1.05_EN.pdf

It is not clearly stated what the difference between the "n" and "o" is, but one thing I noticed in the datasheets is that the "n" version has 64 or 128 KiB flash and the "o" has 256 or 512 KiB flash.

and now its even more confusing, they have a STM32L373, which doesn't exist from ST:
https://lcsc.com/product-detail/Microprocessor-Microcontroller_MindMotion-Microelectronics-MM32L373PS_C212250.html

The most confusing part is that the MM32F103CBT6 is not compatible with an STM32F103CBT6. But at least there is documentation.

HK32F103 basic datasheet attached.

Thanks! It looks like another 5,5 Volt capable one.  Just like the BLM32 and MM32. Those two are not STM32 compatible, and I haven't been able to locate any manuals for the BLM32. Oh well. I've ordered a few to see what they are, and add them to my comparison.

Edit: found the 168 MHz reference

[Noopy]

Excellent work, I love your pics.

Thanks!

If anyone has an other "strange looking" STM32 I can do more investigations.


Well I´m no expert regarding these STM32-clones.
I wondered why it was labeled CKS32. A lot of websites talk about "CS32*" and the die marking shows also CS32...

I have some chips here, another "STM32F103 compatible" part the APM32F103. I could send you a sample if you like. Also, I have the cs32f103c8t6 and cs32f103cbt6. I wonder if they are the same as the cks32 labelled part.

Thanks to GromBeestje I have some more parts to look inside. 

The first two are the least interesting ones (now I have looked inside): cs32f103c8t6 and cs32f103cbt6

https://www.richis-lab.de/STM32_03.htm

I didn´t even upload the dies because they are equal to the csk32f103c8t6  I already had (viewed from a high level).

=> csk32f103c8t6 == cs32f103c8t6 == cs32f103cbt6

 

[Noopy]

NEWS


Today I have a APM32F103C8T6 for you (Thanks to GromBeestje again):

https://www.richis-lab.de/STM32_05.htm







It seems that SEC-CHIP has done a lot of the design.
I found only little information about SEC-CHIP. They seem to cooperate a lot with APEXMIC.


There is kind of a small similarity to the CKS32/CS32:
https://www.richis-lab.de/STM32_03.htm
These two testpads inside the logic block...


Overview here:
https://www.richis-lab.de/STM32.htm

[profdc9]

Despite the fact that STM32F1 is old, it is good enough for some projects.  When designing hobbyist or open source projects (which I know is not the main concern here), there isn't the quantity to ensure part availability long into the future.  I have made open source projects for NXP parts and many of them are still available, but hard to get, and could be end-of-life soon.  The STM32F1 seems like it is used widely enough that it should be available long into the future.  If, for example, STM32F4 or STM32F7 had similar widespread adoption, I would use it.   However, it is not necessarily the cost that makes it attractive for projects, but the low cost makes it more likely to be used, which raises demand and therefore increase the likelihood that it will be available in the future.  There are definitely frustrations and limitations of the STM32F1, but because of its ubiquity there is a strong incentive to use it if it is sufficient for the purpose.

[Noopy]

Hi all!


Thanks to GromBeestje I was able to add a GD32 to my collection: A GD32F103CBT6 and a GD32F101CBT6. They both look the same although the 101 has less RAM and no USB interface. Probably the 101 and the 103 are the same parts…


https://www.richis-lab.de/STM32_06.htm


It´s the interesting die-stack zeptobars already has decapped.




Very interesting: Since zeptobars has decapped the GD32F103CBT6 they apparently have reduced the die size of the flash memory!




Some dies get really dirty while glueing the dies together.


 

[GromBeestje]

The GD101 checks though. When I flash a binary compiled for 20 KiB of RAM, it immediately hardfaults. Trying to enable USB doesn't trigger a hardfault but it just doesn't work.  So I guess, there is some hidden area in the flash containing the configuration. Setting the amount of memory, and selecting the enabled peripherals.  Kinda like the wounding Cypress uses in their PSoC4 series.

[Noopy]

I´m pretty sure the manufacturer can somehow configure the GD32. 
It´s possible that the configuration is placed in the flash but it´s also possible that there are some fuse bits on the big die. Perhaps there are real fuses or some other one time programmable memory bits.

[GromBeestje]

Some articles how this is handled on the PSoC

http://www.dmitry.gr/?r=05.Projects&proj=23.%20PSoC4
http://www.dmitry.gr/?r=05.Projects&proj=24.%20PSoC4%20confidential

It has been speculated the GD32 copies the content from flash to some designated RAM area and start executing from there. So I'd imagine they've some bootrom that performs this task, as well as configure the available features. In some article, when the GD32 was hot, was wondered whether one could "unlock" this memory area to have some more RAM. However, after the initial burst of interest, there has been little written about this GD32.

[iMo]

Btw, there is the APM32F103 with simple memory mapped FPU inside (most probably add/sub/mult/div only). Their SDK shows the FPU is indeed real.
Mind it is CM3..

https://www.apexmic.com/en/newproduct/apm2/16

[ale500]

Btw, there is the APM32F103 with simple memory mapped FPU inside (most probably add/sub/mult/div only). Their SDK shows the FPU is indeed real.
Mind it is CM3..

If this is true, it will be a hard sell:

Running power consumption: 340mA/MHz

[iMo]

I will run it at 96mHz clock then..