Feedback on ESP32-C3?

[SiliconWizard]

I'm considering using an ESP32-C3 for one project which requires Bluetooth. Specs look fine, they are low-cost and based on a RISC-V core, which is uh, cool.

Now I'd like to have some feedback from people having used it in commercial projects. Is Espressif a reliable company? Is there some support? Are those chips themselves reliable? Have you encountered any issues? Are the docs adequate?

Now for the downside I've seen so far: they have an SDK that sounds OK(ish), but it looks almost impossible to do anything serious with this chip without this SDK. Apparently all the radio part of the chip is not documented (or poorly, dunno?) so you can't write code "baremetal" for Bluetooth or WiFi. I found a project for baremetal dev on those on github, and they clearly state this. They can access pretty much everything except the radio block. While having to use the SDK is not a big showstopper, I do not like a lot to invest dev time in a solution that isn't "open" enough to develop for it baremetal if required.

So, yeah. Any experience and thoughts welcome.

[lucazader]

We have used the original esp32 chip in our products reliably for the last almost 5 years.
Originally the SDK was a bit hit or miss. but it has seen a LOT of development over that time and is pretty good overall.
Support wise they have been quite good overall. They even provide services to look over schematics and module designs if your company is rolling its own module or direct on board solution.


As a company you can ask for support and report bugs on the github repo for the IDF. However you can also email them directly with bugs etc and they will look into them and prioritize fixing them for commercial customers. At least they have done that for a bug we found int he past.

In all honesty they seem much more receptive to finding and fixing bugs than ST ever has (though I think that says more about ST than it does about espressif).

TLDR: We have used them for 5 years. Have been happy with the support we received, and happy with the IDF and improvements made to it over that time.

[westfw]

Apparently all the radio part of the chip is not documented (or poorly, dunno?) so you can't write code "baremetal" for Bluetooth or WiFi.

That seems to be very common for chips or modules that support BT or WiFi.

[SiliconWizard]

Apparently all the radio part of the chip is not documented (or poorly, dunno?) so you can't write code "baremetal" for Bluetooth or WiFi.

That seems to be very common for chips or modules that support BT or WiFi.

Yeah. Actually could be cool to access the radio even if just on a lower level - a bit like with the NRF24L01/+ (that are unfortunately abandoned and now looks like Nordic only has higher-level solutions?)

[Simon]

My colleague found these and we will use them, well the S3 version not based on RISC-V. They seem to be aimed at being programmed higher level. They don't explain hardware in detail which annoyed me particularly with the ADC. Reading between the lines the reference voltage is 950mV and higher voltage ranges up to 3V are allowed by on chip voltage dividers however the possible error is not great at 0.95V input range (no "attenuation") and at 3V range is a whopping 50mV which just makes a mockery of the 12 bit. But you can't have everything, I doubt the silicon they are made of was meant for analogue and we will simply use an external ADC (another micro).

[SiliconWizard]

That's all what I'm seeing. It's not well documented. You have to use the SDK and forget about the details. But it's very cheap for the specs. So that's what it is. Btw, this situation looks rather common for chinese MCUs in general.

Also, I downloaded the ESP-IDF SDK and uh. It's absolutely gigantic. Good lord. Takes 1.2 GB or so.

[tellurium]

The TRM for C3 (https://www.espressif.com/sites/default/files/documentation/esp32-c3_technical_reference_manual_en.pdf
) was underdeveloped in comparison with the TRM for original ESP32 (https://www.espressif.com/sites/default/files/documentation/esp32_technical_reference_manual_en.pdf). But they seem to improve the TRM over time, so things that are underdocumented now, could become documented.

Except the radio part. There were multiple attempts to reverse engineer radio registers, though no usable results are published. So apart of the radio, the rest of the peripherals are usabe. You might take a look at the rudimentary bare metal setup for C3 at https://github.com/cpq/mdk - it requires only make and RISCV GCC, nothing else.

[SiliconWizard]

The TRM for C3 (https://www.espressif.com/sites/default/files/documentation/esp32-c3_technical_reference_manual_en.pdf
) was underdeveloped in comparison with the TRM for original ESP32 (https://www.espressif.com/sites/default/files/documentation/esp32_technical_reference_manual_en.pdf). But they seem to improve the TRM over time, so things that are underdocumented now, could become documented.

Yeah. Maybe. That's again a common issue with chinese MCUs. Those get released before any serious documentation is out there, so you're supposed to entirely rely on SDKs and too bad if you need detailed info. Sometimes even just how much current the thing draws in given conditions is not documented. Espressif it not the worth there for this. But yeah.

Except the radio part. There were multiple attempts to reverse engineer radio registers, though no usable results are published. So apart of the radio, the rest of the peripherals are usabe. You might take a look at the rudimentary bare metal setup for C3 at https://github.com/cpq/mdk - it requires only make and RISCV GCC, nothing else.

Yep, that's what I mentioned earlier.
Unfortunately, I think using ESP32 MCUs if you don't use the radio doesn't make sense. There's better and not overly expensive stuff out there for that. Just my opinion here though.