⚠️ Lolin Wemos C3 mini needs hardware mod to work ⚠️

[Jason2866]

no user got Tasmota working reliable. In the official Lolin Aliexpress shop there is feedback like this too.
No proper designed ESP32 board needs to reduce wifi tx power from the standard choosen from espressif https://www.wemos.cc/en/latest/c3/c3_mini.html#about-wifi

This is NOT a Tasmota problem. Tasmota does work well on every other board we and users have tried.

Imho the board has a hardware design problem, which can not solved by software.
Stay away from this board and try to get a refund.

EDIT a hardware mod is found to get it working. See #15443 (comment) Thx @hallard

Jason2866

[agners]

That sounds very much like a power issue. Quickly checked their schematic, they are using MicroOne ME6211 LDO with 500mA peak current, the ESP32-C3 specified 335mA@3.3V peak power when RF TX is working. In theory, it should be good enough.

The NodeMCU ESP32-C3-32S-Kit uses a AMS1117-3.3 1A LDO, Espressif's ESP32-C3-DevKitM-1 uses the SGM2212-3.3XKC3G/TR, a 800mA LDO.

I wonder if bypassing the LDO and supplying 3.3V directly makes it work reliable.

[fangspeen]

For the record, it does not seem to help to supply 3.3v directly or even additionally, see my replies to TD-er

[Jason2866]

Give it a try ;-) Whatever reason it is, the board is not useable as it is sold.

[taste66]

Is it failing at the moment you are writing to flash? This might require significantly more power.

[arendst]

Update: I have three of these.

• All are able to be flashed over embedded usb once you know how to get it in upload mode (hold GPIO9 down (button) and press and release reset button, then release GPIO9). After that initial task any esptool upload and reset works flawlessly.
• One consistently fails to wifi connect after it connected only once. No matter if it is fed by the embedded usb interface, a serial to usb converter connected to Rx/Tx using 5V or 3V3 supply. I consider this one useless.
• Two others are working stable but only when wifipower 8.5 is set as recommended by Wemos.
• I managed to make it functional with an AM2301 connected but considering the tiny LDO I don't expect larger loads can be used.

Considering all this I think it's safe to say this device is not recommended for development purposes.

[agners]

Ugh, that really doesn't sound promising.

Two others are working stable but only when wifipower 8.5 is set as recommended by Wemos.

Is that also true when connecting 3V3 supply directly?

[TD-er]

Hmm the need to reduce TX power on WiFi suggests an antenna impedance mismatch, or a resonance frequency mismatch.
Both may result in TX power reflecting back into the transmitter circuit.

Does it make a difference if you put your thumb on the antenna? If so, then it is for sure an antenna mismatch.
You may also want to check if the antenna feels warm to the touch, which is also a good indication of bad design.

[arendst]

Touching the antenna parts makes no difference. No temperature differences from a "functional" device.

It might even be the cause of the issue as those parts are normally hidden by a shield but here they are unshielded and might have been broken by ESD.

[TD-er]

Or just bad soldering which adds enough resistance to cause voltage drops when drawing current.

[Jason2866]

Bad soldering affecting all devices? I knew about 10 of this modules, and none is working.
I share your opinion -> Bad HF design.
Lillygo failed too with there first attempt of there C3 board (It was a HF Antenna design issue).

[fangspeen]

I've just done this and touching the antenna does make some difference, in that I get a couple of steps further along the wifi setup before the crash.
For interest sake I tried adding 3.3V via breadboard power unit and that made no difference.

I have previously been able to get this same unit to work somewhat by reducing TX power but my current program is Rust + ESP_IDF based and I have yet to figure out how to reduce TX.

Lolin C3 mini V2.1 has a different design and antenna, and seems to work without a problem.
Lolin C3 pico v1 also differs but so does the error I get, kernel panic: "Guru Meditation Error: Core 0 panic'ed (Store access fault). Exception was unhandled."

[Jason2866]

There is a fix found by @hallard . Thx!
His post from Discord:

Hi there, since my failed (no wifi) last month test on ESP32 C3 Mini, I had postponed this one. Now with official release of 2.0.3 I tried again having in mind what @Jason2866 told me, bad RF concept on this board. But since this can't be easier (RF outpout goes to matching network then to PCB antenna trace). And since PCB trace antenna is known to work I decided to look an the matching network. First remove the capacitor (or inductor not sure), and first try suddently saw AP, connected, configured Tasmo with my own AP and reboot then connected, magic as simple as that (may be not best matching couple but works better than nothing). My guess an error on the cap/inductor on factory side or bom, anyway C3 now works as expected. Here is the component to remove.

[hallard]

Take care it's a dirty fix interesting to test distance performance.

Wemos state another fix (cleaner or not) is to change wifi Tx power as follow,

WiFi.setTxPower(WIFI_POWER_8_5dBm);

[Jason2866]

Change the TX power to 8.5dBm does work really bad. Not a solution.

[hallard]

Nice you tried, avoid me doing this stupid test. But to be honest not sure my fix will do better, for sure result will not be as good as a good matching circuit.

But following solution should give better results, just remove resistor and soldering 3.1cm wire len on the pad (or 6.2cm or 12.4cm)

[asergunov]

I've checked the design guidelines and wemos schematic. Looks like the full schematic is inductor plus two capacitors from both sides. But wemos put only one capacitor and replaces inductor with 0 ohm resistor. Looking on Reference PCB design it's clear that external antenna connector goes after the CLC circuit. Could result be better if we populate all the circuit with recommended elements?

[hallard]

As I suspected this fix is not really the solution, this afternoon I played with fixed board for a customer, and I never had success to get ESPNow working and send data, so I give a try with unmodified board with soft fix as follow

  // Set device as a Wi-Fi Station
  WiFi.mode(WIFI_STA);
  WiFi.setTxPower(WIFI_POWER_8_5dBm);

It worked immediately, so depending on your usage hard fix or soft fix is up to you, but my 2 cents, stay away from these board and use a well known one https://www.adafruit.com/product/5405 (not sure it was tested but I'm sure it's working)

[jnovonj]

Hi,

I have the same problems as reported here:

In the URL [https://www.wemos.cc/en/latest/c3/c3_mini.html] Wemos shows the Schematic V1.0.0:

C15 (1pF) y C16 (NC) Capacitor
L1 Inductor

I am unable to measure this capacitors and inductor.

The Espressif pdf says:

I am unable to debug through the serial port but this is only my problem

Best Regards.

[fluppie]

I see there is a v2.1.0 version now that doesn't mention this 7.5 dBm wifi power issue. Ordered a bunch of them and will let you know if it works out of the box.
https://www.wemos.cc/en/latest/c3/c3_mini.html

[barbudor]

They replaced the PCB antenna by a ceramic one
Good move

[Jason2866]

The sold all borked boards the had. Never made a notice nor the did an exchange program.
No apologize. Nothing... Very evil.

[hallard]

They replaced the PCB antenna by a ceramic one
Good move

If it is correctly matched of course 😂

[sfromis]

It does appear than some capacitor adjustment was done with the new revision. And they added a Neopixel as basis for saying "new and enhanced" without mentioning problems with the first iteration.

[Jason2866]

It is possible to design a cheap working C3 board. Buy this https://wiki.luatos.com/chips/esp32c3/board.html (contracted version) instead of Lolin and supporting an evil company who is fooling his buyers. It is available from different sellers on Ali under 3€.

[fluppie]

I can confirm the v2.1.0 boards work out of the box.
@Jason2866 you mean this one? https://nl.aliexpress.com/item/1005004490215444.html?spm=a2g0o.productlist.0.0.178d619dGIJQGO&algo_pvid=85be1b87-c562-44b8-88e6-ba6fcc25e7f9&aem_p4p_detail=2022092714262215302205461094970007300218&algo_exp_id=85be1b87-c562-44b8-88e6-ba6fcc25e7f9-0&pdp_ext_f=%7B%22sku_id%22%3A%2212000029340839373%22%7D&pdp_npi=2%40dis%21EUR%213.13%212.5%21%21%211.75%21%21%402101e9d016643139826951544e58e5%2112000029340839373%21sea&curPageLogUid=r2wMa3M5zFTZ&ad_pvid=2022092714262215302205461094970007300218_1

[Jason2866]

yes

[msx80]

@fluppie where did you buy v2.1.0 ? On https://lolin.aliexpress.com/ there's still v1.0.0 listed (at least now).

Still angry that i bought some of the 1.0.0 rev, i guess that's a risk for such ultra low cost boards. The rgb led is fancy tho.

[sfromis]

Probably sold out, then ...

[hallard]

waiting for mine that should arrive, hoping it's new revision, surprise in the mailbox soon...

[hallard]

Arrived today, conform, with external antenna, not tested yet

[ncsEXE]

Have you had time to test it and see if it works?

[PBrunot]

I have this board v2.1 and I still have occasional issues with Wifi. The issue was especially present when the board is mounted on a breadboard, when it's "flapping around in the breeze" with an USB cable the Wifi mostly succeeds. The txpower setting improves the situation but I would not call it reliable.

[msx80]

I had a couple myself, they don't seem to have a very good range, do they? From my work desk they have a bit of an hard time connecting while other boards (like the S2 or other brands) had less problems. Anyone noticed something similar?

[sfromis]

If your Wemos C3 Mini is the older V1.0.0, the bad Wifi is a well-known issue, the reason for even having this thread.

[msx80]

no sorry i forgot to specify they are the new 2.1.0 version with ceramic antenna

[Jason2866]

Do you have a other C3 board, which you can compare?
Every MCU has a different wifi hardware and software. So only the same MCU type can be compared.

[martin072]

Can anyone confirm v2.1.0 C3 works ok? I want to replace some tasmota esp8266 with an esp32 and it would help enormously if I can use the same footprint as the current d1 mini so I can keep the already builtin pcb's.
I do have some v1.0.0, but as mentioned the software mod is needed to get it to work, and not sure what the range is with this mod.
I also have some S2 version's but not sure if tasmota is stable on these?

[sfromis]

In my experience, Tasmota is as stable on ESP32-S2 (as in Wemos S2 Mini) as other variants. Only thing is that there may be antenna trouble with non-Wemos branded clones of that board.

I've heard nothing bad about the new revision of the C3 Mini board, but for me the old one does actually not work too badly with the workaround of WifiPower 8.5, but would not place it where signal strength matters.

[martin072]

I have the original Wemos boards. But the C3 v1.0.0 has the antenna issue. Just curious if this is fixed on the V2.1.0 (before I buy a bunch of them ;)

[sfromis]

Well, IME the Wemos C3 Mini V2.1.0 connects with {"WifiPower":"17.0"} instead of having to {"WifiPower":"8.5"}

[martin072]

Ok thanks, just to confirm that this (fix/workaround) is still needed on the v2.1.0 boards as well?

[dosipod]

Hello , I have Lolin c3 v1.0.0 and I have tried tasmota so many times but i could not get that to work . the board worked when flashed with wled using below commands.

esptool.exe -p COM10 erase_flash
esptool.exe -p COM10 write_flash 0x0 C3_bootloader.bin
esptool.exe -p COM10 write_flash 0x8000 C3_partitions.bin
esptool.exe -p COM10 write_flash 0x010000 6_may_lolin_c3_mini_T2.bin

I have only one to test and I am yet to try the same process with tasmota but if you need to test the files I used the look here https://discord.com/channels/473448917040758787/779395228624617512/1104464327068622959

[sfromis]

My Tasmota experience with that board is that the workaround of WifiPower 8.5 got wifi to work in good conditions, but I'd not really expect good range. To apply the WifiPower 8.5 setting with precompiled builds, I needed to use USB to get console access after first boot. Until then, no wifi available.

[Jason2866]

@dosipod Just dump the Lolin C3 rev 1.0 in the bin. There are other cheap working C3 boards.
Waste of time.

[dosipod]

I agree with you and we have only one of each new chip c3 , s2 and s3 to test with students and they all rev 1,0. I would not go again and buy but I cant just throw money in the bin . We have plenty of normal esp32 /8266 and we test new chips for learning as it is a challenge to make something like this work and we were about to do the HW mod listed in this post . If that is working with wled might be at least you could try to compile in tasmota and provide the bins ( including bootloader and partitions ) so others might be able to fix existing boards with tasmota . We are thinking that tasmota bin could also be good but the flashing process is the issue as it seem with wled

[barbudor]

I don't understand your point
The hardawe is bad and there is nothing that can be done except reducing the transmit power.

If that is working with wled might be at least you could try to compile in tasmota

If you have identified a software fix for that specific board, you are free to fork the source code, apply the change and compile your own build. That would surely be a very good exercise for students.

We will never add into Tasmota any code that is specific to a given board. Our build are generic per processor type and should work as-is with any board with that CPU.

[TD-er]

Just an idea if you need to flash lots of units with the same issues...
Create one with the right settings you need (maybe even use it on 1 modified version if you really can't get it to work without) and then use esptool.py to read the entire flash.
Then use that image of the entire flash to flash all other boards including the tweaked settings.

[dosipod]

I am not sure about other suggestions but best thing is for the person suggesting something is to test it by himself which we did already with wled . As the title of the post is "Lolin Wemos C3 mini needs hardware mod to work " I think it is a bit confusing if the software fix is enough or not but it seems so with wled you do not need a HW mod

[sfromis]

Having some workaround (like lowering WifiPower) to get better results is not the same as having a fix of the problem. "Working or not" is not a binary black/white, but many shades of grey. Of course, if someone knows details of what WLED is doing differently......

[TD-er]

I have no idea what WLed does and also not what Tasmota does.
Based on this thread, and for the simple fact I am currently implementing support for S3/C3 for my project (ESPEasy), I also have ordered a batch of similar boards as shown in this thread. (Since some of my users will probably also use garbage, I also try to test those known-to-be-garbage boards)
Among those are a few with an ESP32-S2 and I found one that's absolutely horrible regarding WiFi stability when sending at max. TX power.
In my project, I have a feature to dynamically adjust the TX power based on the RSSI of the AP I connect to. (based on the assumption the RSSI of the signal as received on the AP will be in the same order of magnitude)
When enabling this feature on that horrible S2 board, I got very usable WiFi performance.

So maybe WLed does something similar?

[sfromis]

I've seen indications of clones of the Wemos S2 Mini having pretty bad antenna design. I'm having no trouble with the original Wemos-branded one.

[sblantipodi]

I received the v2.1 board and I can confirm that the antenna does not work reliable even on the newer version of the board.
The only way to use it on a breadboard is to reduce the wifi power to
WiFi.setTxPower(WIFI_POWER_8_5dBm);

@TD-er you wrote this:

In my project, I have a feature to dynamically adjust the TX power based on the RSSI of the AP I connect to. (based on the assumption the RSSI of the signal as received on the AP will be in the same order of magnitude)
When enabling this feature on that horrible S2 board, I got very usable WiFi performance.

can you show us that feature? is it public? may we use it in our projects?

[TD-er]

Sure, it rougly comes down to this:
https://github.com/letscontrolit/ESPEasy/blob/f5a1e6b573955d472bebde86aeb4f79ebd0a0ccc/src/src/ESPEasyCore/ESPEasyWifi.cpp#L675-L826
Periodically call SetWiFiTXpower() which will then base the TX power on the current RSSI.
When connecting to an access point, you need to use the RSSI you got from the scan results as the parameter since you can't use WiFi.RSSI() before connecting.

The basic idea is that the RSSI is in the same order of magnitude on both ends (AP receives your data with similar strength as you receive the AP)
But if the antenna of your device is worse, you should add some dB's to the "sensitivity margin". (I use a default of 3 dBm)
The ESP does have a specified sensitivity (taken from the datasheets), which also depends on the connection mode. (802.11b/g/n)
When not yet connected, I use the worst sensitivity, which is for the 802.11n.
I adjust the TX power based on the RSSI, where I try to stay at least "sensitivity margin" dB above the specified sensitivity, but no more than a specified max TX power in my settings.
See: https://github.com/letscontrolit/ESPEasy/blob/f5a1e6b573955d472bebde86aeb4f79ebd0a0ccc/src/src/ESPEasyCore/ESPEasyWifi.cpp#L694-L717

The datasheet parameters are defined here:
https://github.com/letscontrolit/ESPEasy/blob/f5a1e6b573955d472bebde86aeb4f79ebd0a0ccc/src/src/Helpers/Hardware.h#L39-L76

More in-depth documentation on this: https://espeasy.readthedocs.io/en/latest/Tools/Tools.html#wifi-tx-power

Whenever you use it, I would appreciate it if you mention in your code where you got it from.

[sblantipodi]

many thanks for the kind answer @TD-er. I appreciate it.
I have tried your awesome code but it does not work here...

The ESP return an RSSI of -30 when it's on the air and perfectly connected to the network,
it returns an RSSI of -33 when on the metal breadboard when not connected during a scan network.

when on the metal breadboard the ESP can't connect to wifi, if I remove it from the breadboard it connects pretty instantly.

with this RSSI it's not possible to calculate the correct signal strength, I think that this board is bad designed and nothing can be done...

[TD-er]

Please also read the link to the documentation I added at the end of my post.
You have to remember that RSSI alone isn't a good indicator of the expected WiFi link quality.

Especially when you're placing the ESP's antenna close to metal, you may end up with a lot of reflections of the WiFi signal.
In such cases the ESP may act as if it has a badly tuned antenna.

First some very basic antenna theory...
A perfectly tuned antenna emits 100% of its signal and nothing gets reflected back to the transmitter.
Such an antenna has a SWR of 1.
However, since nothing is perfect, there will always be some reflection of the RF signal.

N.B. not only the length of the antenna has an effect on the SWR, but also the materials surrounding the antenna.
Even the orientation of the glas fiber strands in the PCB haven an effect on it.

Simply hold the antenna in your hand, or place it in an enclosure and the SWR may change.

Such reflections can have numerous effects on the WiFI performance.
One such effect is that the radio may become temporarily a bit "deaf" and thus less sensitive to low power WiFi signals.

The ESP processors do have some ways to automatically tune the antenna within some parameters.
But I'm not 100% sure this tuning does take any set max. TX power into account.
What I am sure about is that any metal close to the antenna will affect such RF tuning as it is essentially what this tuning is for, to minimize the amount of reflections.
So what you can try here is to make sure to limit the TX power before calling WiFi.begin(), or maybe even before turning the WiFi on.

I personally would try to start with a lower maximum TX power in such setups.

Right now, I have no idea how to 'detect' in software whether the ESP experiences a lot of RF reflections.
Just keep in mind that with 'reflections' I mean receiving your own TX signals back into the transmitter.
This is less of an issue when receiving WiFi signals.
So received RSSI values can still be very loud, but if your transmitted signal is quite distorted, you will not be able to connect to an AP.

[arendst]

Implemented in dev branch v12.5.0.3

Use command WifiPower 0 to enable it.

[TD-er]

Yay!