ADC speed for Olimexino-GD32 vs. RedPill

Hi everyone,

I’ve recently got two Olimexino-GD32s from Amazon and was quite surprised to find the ADC conversion taking less than 1 microsecond (using just a simple analog.read() in Maple IDE). Somewhere else in the forum I’ve read and later confirmed STM32’s 2 us limit, so GD32’s ADC seems to go twice as fast. My question is: can anyone else confirm this for the generic GD32 using Arduino IDE? My Olimexino-GD32s seem to have some ADC-unrelated issues (basically my Mac stopped recognizing the USB port after 10 uploads or so for both boards – the last successfully uploaded sketch was running fine, so I’m guessing the MCU was okay) so I would like to find as alternative.

Thanks!

GD32 are significantly faster than STM32.

What I find a bit amusing, is that Olimex have jumped on the GD32 bandwagon, after initially writing very disparaging blog posts about GD ripping off STM etc etc

Olimexino-GD32? I cant find it in their product catalog… is this official product?

$9 on Amazon:

https://amzn.com/B01APN27ES

Anyone know what clock speed the HSE is ?

I will need to add support for this, because my other GD32 board has 12Mhz HSE Xtal and I suspect this has 8Mhz as the probably just replaced the STM32 with the GD32 equivalent

If I add support for this, you can run the board at 120Mhz and still USB Serial (As it has extra USB PLL DIV settings in the GD32)

RogerClark wrote:GD32 are significantly faster than STM32.

What I find a bit amusing, is that Olimex have jumped on the GD32 bandwagon, after initially writing very disparaging blog posts about GD ripping off STM etc etc

Perhaps they are just noticing that everyone can buy much cheaper boards from AliExpress.

Because, like you say, the price difference is definitely not just the price of the GD32 vs STM32

It’s a bit weird that they are not for sale or even listed on their website?

Olimex blog post about the GD32 Here claims they are 20% cheaper to buy than the ST.

1. All the products provided by GigaDevice were developed by GigaDevice independently, and GigaDevice has its all intellectual property rights. As the third-party intellectual property rights, GigaDevice has obtained the intellectual property rights holder’s authorization.

2. All the products provided by GigaDevice do not infringe any third-party intellectual property rights.

3. It’s sure that for the record, GigaDevice are a legitimate ARM licensee :
https://community.arm.com/community/arm … iconductor

4. And GigaDevice is officially the ARM connected community member and the ARM university programme partner.

5. GigaDevice GD32 series of MCU got the trade mark in world wide such as US, Korea& Japan

6. The right of layout-designs of integrated circuits design of GD32 MCU is protected in China.
GigaDevice also have lots of US,EUROPEAN & INT’L PATENT concerning GD32 MCUs, examples check the attached patent file.

1 PCT/CN2013/075149 PCT International EMBEDDED SYSTEM CONTROLLER

2 14/399,858 US EMBEDDED SYSTEM CONTROLLER

3 13788134.8 EU EMBEDDED SYSTEM CONTROLLER

4 PCT/CN2015/071990 PCT International MCU

5 PCT/CN2015/071983 PCT International MCU

6 PCT/CN2015/071556 PCT International MCU

7. We also provide the “Intellectual Property Rights Guarantee letter” and the “Trademark Certificate” for your reference.

8. We will continue to highlight the advantage of our own in market, especially the family roadmap.

Manny wrote:It’s a bit weird that they are not for sale or even listed on their website?

Document CD00161566 (datasheet for STM32F103C[8B]) – page 12 – note 3: “To have an ADC conversion time of 1 μs, APB2 must be at 14 MHz, 28 MHz or 56 MHz.”

So it seems also STM MCUs can achieve 1 μs conversion time..

The Pig-o-scope thread also has some ideas about this topic.
http://www.stm32duino.com/viewtopic.php … &start=230

Quick update: I’ve managed to get my Olimexino-GD32s up and running again reflashing the bootloader via serial using ST Flash Loader Demonstrator. The bootloader is for STM32: prodTest_OLIMEXINO_STM32.hex.

So playing again with the ADC (12-bit):
Again, Maple IDE: 909 ns per conversion (1.1 Msps).
New, Arduino IDE (Maple 3 board): 833 ns per conversion (1.2 Msps).

I’m impressed for a $9 board.

EventHorizon wrote:Quick update: I’ve managed to get my Olimexino-GD32s up and running again reflashing the bootloader via serial using ST Flash Loader Demonstrator. The bootloader is for STM32: prodTest_OLIMEXINO_STM32.hex.

So playing again with the ADC (12-bit):
Again, Maple IDE: 909 ns per conversion (1.1 Msps).
New, Arduino IDE (Maple 3 board): 833 ns per conversion (1.2 Msps).

I’m impressed for a $9 board.

What is the quality of the conversion at that speeds?

ahull wrote: Is that per channel, or using two channels interleaved?

Pito wrote:What is the quality of the conversion at that speeds?

Its a 12 bit adc, (max count 4096), so a rough guestimate is 2 counts in 4000 is probably +/- the LSB or maybe a little bit worse. I would imagine that the lack of a separate analog ground will also have an impact of a few parts per thousand on the accuracy too.

This gives various hints for improving things, but I would suggest that driving the ADC even faster (but reducing accuracy to say 8 bits) would be worth investigating. I suspect if we could clock it fast enough we might even get 4 mega samples per second or more, trading accuracy for speed of course. I wonder what is the maximum speed we can clock the ADC at (not the maximum stated by the datasheet, but the maximum we can actually overclock to), and how fast can we actually read and reset the ADC, without bothering to let it settle.

Now if 12 bits is three nibbles, that makes it a tribble… I wonder how much trouble we can get in to from messing with these tribbles.

The STM’s doc on ADC does not mention ENOB at all..