Air105 / Air32F103 - New chinese arm chips

[DavidAlfa]

Found a new company making stm32-like chips on AliExpress
https://a.aliexpress.com/_ujJBWl

Specs are pretty nice:

Air32F103
216MHz, 32K RAM, 128K Flash
https://chowdera.com/2022/198/202207171843163101.html
https://wiki.luatos.com/chips/air32f103/index.html
(Actually the RAM can be unlocked to 97KB!)

Air105
204MHz, 640KB SRAM, 4MB flash
https://developpaper.com/aizhou-air105-uses-keil-mdk-dap-link-to-burn-and-debug/

[pcprogrammer]

Not cheap though for the AIR32F103 that is. ~10 euros shipped

[thm_w]

What I don't get is they claim "compatible STM32 For Arduino". But the screenshots are showing Keil, and the AIR105 is clearly a completely different beast from the STM32F103.
I guess the AIR32F103 is the only compatible one then? and AIR105 has no STM equivalent at all.

There are also more in your link, AIR101, AIR106

If the 32F103 is a drop in compatible replacement, then its a smart way to enter the market. As gigadevices did.
But all of the non-GD clones, from what I've seen, haven't been as good in terms of compatibility. Even the GD one isn't perfect.

Not cheap though for the AIR32F103 that is. ~10 euros shipped

Here you go, $2: https://www.aliexpress.com/item/1005004606391844.html

edit: the AIR103 module is not the same pinout as bluepill, or blackpill, is it compatible with any pinout? Don't know what these people are thinking, just use an existing pinout. Completely worthless board now.

[pcprogrammer]

Not cheap though for the AIR32F103 that is. ~10 euros shipped

Here you go, $2: https://www.aliexpress.com/item/1005004606391844.html

For me this one is ~4 euros shipped, which is comparable to others, but with 216MHz compared to 72MHz much faster. Is it reliable though?

But with a pin out of the board that does not match the bluepill it becomes its own thing and not a drop in replacement for something you already made. For me as a hobbyist not a problem, but already have enough development boards lying around

[DavidAlfa]

Who buys bluepills board for professional products...?
It's all for hobbists.

[thm_w]

For me this one is ~4 euros shipped, which is comparable to others, but with 216MHz compared to 72MHz much faster. Is it reliable though?

But with a pin out of the board that does not match the bluepill it becomes its own thing and not a drop in replacement for something you already made. For me as a hobbyist not a problem, but already have enough development boards lying around

Exactly, it becomes its own thing, which makes it a problem:
- You can't just drop it in your existing design and test the reliability.
- You have to spend development effort to use it anywhere

This guy is implying its drop in: https://blog.csdn.net/michaelchain/article/details/125458496
But doesn't provide much info, and no datasheet to be found. I'm assuming the 216MHz is a simply PLL register change (72MHz * 3), as the mindmotion F103 was (?)

[rooppoorali]

Are these chips available somewhere else? Except for Aliexpress?

[thm_w]

Are these chips available somewhere else? Except for Aliexpress?

Taobao.
Google is your friend, or find a re-shipper if you need the latest and greatest directly from china.

[fanoush]

But doesn't provide much info, and no datasheet to be found. I'm assuming the 216MHz is a simply PLL register change (72MHz * 3), as the mindmotion F103 was (?)

Datasheet in chinese is linked here https://wiki.luatos.com/chips/air32f103/hardware.html thanks to Chrome browser builtin translation those pages are quite readable with lot of info and the pdf can be translated by google translate too. The flash latency is interesting  - up to 108MHz the value is 0 and >108-216 is 1. Original needs 2 at 72 (and APM32F103 needs 3 at 96MHz) so I am quite interested in benchmarking code running from internal flash withthis board. With APM32F103 at 96 code from flash runs only a tiny bit faster than at 72, only when code is copied to RAM it runs *1.33 faster

I have ordered one of those $2 boards from aliexpress just to test performance at 216Mhz and compatibility. BTW, the datasheet shows power consumption at various clocks and it is much lower at 216 than STM32F103 original at 72Mhz (38 vs 50 mA).

[DavidAlfa]

Don't forget the F1 series have 15 years already...

[IOsetting]

For your info:
1. AIR32F103x is an OEM product, the original type is Mega Hunt MH32F103x, you can still find MH32F103 chips on Taobao, they are the same.
2. AIR32F103x is compatible with STM32F103x in most cases but there are quite a few difference, here is a detailed list (in Chinese) https://wiki.luatos.com/chips/air32f103/switchFromSxx.html
3. All other chips (Air101, Air103, Air105... ) are not compatible with STM32.
4. The library and example projects for Keil5 IDE https://gitee.com/openLuat/luatos-soc-air32f103/tree/master/ModuleDemo
5. The gcc arm project template (with the same library from above repository) https://gitee.com/iosetting/air32f103-template

Possibly you can find more details from the example code.

[fanoush]

Thanks, the gcc template will be helpful :-) MH32F103 gives some additional interesting results in google search.

As for example code I was interested mostly in setting up higher clock speed (216,108) with having USB still working with correct divider and found the code here https://wiki.luatos.com/chips/air32f103/enhancement.html so I am looking forward to testing it once my board arrives.

BTW, The datasheet also mentions CCT6  version with 64K RAM, 256K Flash. Board with that one would be even better than the CBT6 one. But I guess CBT6 one is cheaper and maybe even more compatible with STM32F103 (regarding flash block size?). If I had few cheap bare chips of CCT6 variant, it would be fun to replace existing chips in some devices with STMF103 I have where the 128K of flash is hitting the limit of what the firmware can do.

[IOsetting]

216MHz code example for gcc arm embedded toolchains
https://gitee.com/iosetting/air32f103-template/tree/master/Examples/NonFreeRTOS/RCC

The highest clock speed would be 256MHz when PLL=32

I found recently there are more variants of these series

[bidrohini]

Has anyone used these?

[fanoush]

I/m about to receive board with  Air32F103 any day now so will test it a bit then. Ordered AIR32F103 variant from https://www.aliexpress.com/item/1005004539387853.html

[fanoush]

So I got the board and I like the chip. I have flashed binary of my build of Espruino for STM32F103 and it runs out of box just fine.

USB works, GPIO works, I can see the unused SRAM over 20K up to 32K works. Also looks like the flash latency is much better, even with unchanged STM32 binary at 72Mhz (=with flash latency set to 2) this board is much faster when code is running from flash!

here are some results (full code https://gist.github.com/fanoush/ca350c56d0b6847e223fa9cbb877286b )

STM32F103, 72Mhz, flash latency 2
benchall(20000);
time: 3.71646881103 result= 936372485
time: 1015.77520370483 result= Infinity
time: 12676.12385749816 result= Infinity

APM32F103 96Mhz, flash latency 3
benchall(20000);
time: 2.85291671752 result= 936372485
time: 990.22006988525 result= Infinity
time: 12141.72148704528 result= Infinity

Air32F103, 72Mhz, flash latency 2 (probably ignored)
benchall(20000);
time: 3.77774238586 result= 936372485
time: 705.31296730041 result= Infinity
time: 9807.07502365112 result= Infinity

the first line is fully native code running  from SRAM
second is compiled JS = compiled code is running from SRAM but is calling into JS runtime in flash a lot
third is interpreted JS fully running from flash

It can be seen that only first line is based on raw CPU speed and 96Mh is faster than 72Mhz however when code is run mainly from flash the latency kills it for both STM32 and APM32 (so in second line even the APM32 is not much faster than STM32 at 72Mhz). Air32F103 flies here.

Will test higher clocks too later. Also will try to set lower latency at 72 as the Air32 datasheet says latency should be 0 for such low speed, not 2 as set for STM32

EDIT:
And BTW this board comes by default with CMSIS-DAP daplink firmware (with usb to serial, WinUSB, WebUSB and mass storage modes enabled too) so it is handy to use it as SWD dongle too

[bidrohini]

Thanks for then review after using.

[Scrts]

I understand the AIR103 does not have SWD or JTAG? So only UART prints and GPIO toggles for debug?

[DavidAlfa]

Of course it has SWD. Where did you took that idea from?

[Scrts]

Of course it has SWD. Where did you took that idea from?

I am guessing. Would be good if someone tries.

For example check here:
https://mecrisp-stellaris-folkdoc.sourceforge.io/bluepill-diagnostics-v1.6.html Table: Device Info.
APM32 and CKS32 do not support SWD. You can flash over SWD wires, but you cannot debug.

[GromBeestje]

APM32 and CKS32 do not support SWD. You can flash over SWD wires, but you cannot debug.

I don't recall having any problems debugging APM32 or CKS32 over SWD.

[Scrts]

APM32 and CKS32 do not support SWD. You can flash over SWD wires, but you cannot debug.

I don't recall having any problems debugging APM32 or CKS32 over SWD.

What debugger did you use? STM32 debugger doesn't work. It didn't work with J-Link EDU flashed on ST-Link debugger.

[IOsetting]

I understand the AIR103 does not have SWD or JTAG? So only UART prints and GPIO toggles for debug?

AIR103 and AIR32F103 are different, I guess you meant AIR32F103?

AIR103 is a totally different arch(C-Sky), the OEM type is W806, QFN56 pack.

[fanoush]

What debugger did you use? STM32 debugger doesn't work. It didn't work with J-Link EDU flashed on ST-Link debugger.

OpenOCD works. You can run it like this to force different CPU ID (and larger flash size)

openocd -d2  -f interface/cmsis-dap.cfg  -c "set FLASH_SIZE 131072" -c "set CPUTAPID 0x2ba01477" -f target/stm32f1x.cfg

Then you can use any IDE that supports gdb debugging or use gdb directly.

EDIT: oh sorry for jlink  it is -f interface/jlink.cfg or  for stlink -f interface/stlink.cfg
you can get recent binary for windows e.g. from https://github.com/openocd-org/openocd/releases/

[GromBeestje]

Regarding OpenOCD, please use a recent version v0.11 or the latest git

When using the ST-Link, the stlink-dap.cfg is preferrred over stlink.cfg. The stlink-dap.cfg won't need CPUTAPID,

Note when using the stlink-dap.cfg you'll need to have at least firmware version V2J24 on The ST-LINK. 

Code: [Select]

openocd -f interface/stlink-dap.cfg -f target/stm32f1x.cfg
When using J-Link

Code: [Select]

openocd -f interface/jlink.cfg -c "transport select swd" -f target/stm32f1x.cfg
Note to select the interface after including the config file for the debugger but before the config file for the chip.  This generally holds for any non-st-link type of debugger, eg, cmsis-dap. Otherwise it will default to JTAG.

Code: [Select]

openocd -f interface/cmsis-dap.cfg -c "transport select swd" -f target/stm32f1x.cfg
OpenOCD will run with these stand alone, or in an IDE like the embedded version of Eclipse, as shown in the attachment