D1 mini Pro boards (clones) no longer work with 5V external power!? Software issue?

Message

Wookbert · #1 Post by **Wookbert** » 30 Jun 2021, 12:33

I'm seriously confused. I have a bunch of Wemos D1 mini Pro clones (those with ceramic antenna and 4 MB), which don't work with 5V supplied to the 5V pin, although I definitely know that the only board from that batch I've used so far did work (with 5V supplied to the 5V pin and some older ESPEasy release).

Details/findings:

all boards work flawlessly with 5V supplied to MicroUSB
with 5V supplied to 5V pin
- no sign of life (blue LED not blinking once)
- approx. 3.3V measured on 3V3 pin

Releases tried:

ESP_Easy_mega_20210503_normal_ESP8266_4M1M.bin
ESP_Easy_mega_20210615_normal_ESP8266_4M1M.bin (using experimental Browser flash tool)
ESP_Easy_mega_20210615_normal_ESP8266_4M1M_VCC.bin

My PSU board shown in the picture is perfectly fine, when attaching the D1 mini with printed antenna shown in the picture, that board works instantly. But no matter how many of those D1 mini Pro clones I unwrap (still have a dozen fresh in anti-static bags), they only work with power via MicroUSB.

Any ideas what's going on here? Have there been any changes in the ESPEasy code in the past months?
(No clue, if this could be a software-inflicted issues and suddenly requires me to adjust some software settings).

: IMG_0868.JPG (2.5 MiB) Viewed 37632 times

: IMG_0869.JPG (2.06 MiB) Viewed 37632 times

: IMG_0870.JPG (2.13 MiB) Viewed 37632 times

TD-er · #2 Post by **TD-er** » 30 Jun 2021, 12:36

How much older were the releases that worked?

Wookbert · #3 Post by **Wookbert** » 30 Jun 2021, 13:19

According to my post here, I began using the D1 mini Pro (clone) board on the very same PSU board on November 27, 2020. I assume I've flashed the board with the latest ESPEasy release available back then, but can't tell for sure. So if this is a Software bug, I guess it must have found its way into the code after November 2020.

Guess I'll try the same setup with a Mid-2020 release again and report accordingly.

P.S.: Perhaps you could/should add a release date selector into the Browser flash tool

TD-er · #4 Post by **TD-er** » 30 Jun 2021, 13:43

Wookbert wrote: ↑30 Jun 2021, 13:19 [...]
P.S.: Perhaps you could/should add a release date selector into the Browser flash tool

And a "wipe flash" checkbox

One of the things changed since nov. 2020 is related to WiFi TX power and periodical scan WiFi.

Please try with "Send with Max TX power" checked and periodical scan WiFi unchecked.

Wookbert · #5 Post by **Wookbert** » 30 Jun 2021, 14:35

No changes with

Send with Max TX power“ enabled and „Periodical scan Wifi“ disabled
ESP_Easy_mega_20200829_normal_ESP8266_4M1M.bin
ESP_Easy_mega_20200516_normal_ESP8266_4M1M.bin

Gotta do my taxes now, but will systematically try all releases between April and December 2020 on a couple of boards.
Still heavily confused. Should I measure on the PCB if those 3.3V I measure on the 3V3 pin actually reach the ESP8266 IC?

TD-er · #6 Post by **TD-er** » 30 Jun 2021, 14:53

Or maybe just flash a 'blink' sketch to it which will blink an LED.
This way you can see if it does anything.

It could be the USB chip is stuck in some limbo state resetting the ESP.

I do see the boards use a CP2102 chip, which does reset the ESP as soon as you connect to it using some terminal application.
The older boards using the CH340 USB to UART chip behave different.

Wookbert · #7 Post by **Wookbert** » 30 Jun 2021, 15:12

Uploaded this blink sketch (.ino file), but not sign of life. Oh... Idiot, me. Just saw, that it is ESP-01, not ESP8266. Got a sketch at hand?

TD-er · #8 Post by **TD-er** » 30 Jun 2021, 16:14

esp01 should also work at esp8266 boards with a larger flash.
If it flashes the LED when connected via USB, then it is a good test.
You may need to change the pin to GPIO-2 for your board. No idea what ESP01 used for LED.

ThomasB · #9 Post by **ThomasB** » 01 Jul 2021, 04:58

Details/findings:
all boards work flawlessly with 5V supplied to MicroUSB
with 5V supplied to 5V pin
no sign of life (blue LED not blinking once)
approx. 3.3V measured on 3V3 pin

This might be related to the low performance 3.3V Vreg chip that was used on some D1 mini clone boards. The installed Vreg may have difficulty maintaining clean/stable voltage in all situations (e.g. during boot or WiFi activity). Especially since some ESPEasy Mega firmware releases can draw a bit more current. So I suggest checking the LDO VReg's exact part number and confirm its current rating is robust enough. I think >700mA is ideal, but maybe >500mA rating would be sufficient.

FWIW, powering by USB (VBUS) versus the D1 Mini's 5V input pin, is nearly the same thing; Both provide 5VDC to the 3.3V VReg. But typically the USB VBUS voltage is several hundred mV lower than the 5V input method due to the diode protection on the VBUS path, in addition to the lower "5V" voltage often found on USB ports. So for a fair test, measure the D1 Mini's "5V" pin while powering by USB. Then try powering by the 5V pin using the exact same voltage that was measured while using USB power. If the D1 Mini works doing this then perhaps the VReg IC is a bad performer.

Lastly, maybe the power supplies used during the "5V" powered tests were unlucky choices (the ESP8266 requires very clean/stable 5V power). Just to make sure all stones have been turned over I recommend trying a few other well regulated 5V/1A (or greater) supplies.

- Thomas

TD-er · #10 Post by **TD-er** » 01 Jul 2021, 12:26

Also, make sure to use wires which do have more copper than those cheap "breadboard wires".
I've made the same mistake only a few weeks ago, where I could not reach the 5V pins of a board I designed, so I added those breadboard wires (those with soldered on pins, not the crimped dupont plugs) of 10 cm and then tried to power it using proper thick copper from a beefy power supply.
But those relatively short breadboard wires were just too thin to keep the ESP board working reliable.
Even though that board does have large enough capacitors and an AMS1117 which can deliver upto 800 mA without problems.

Wookbert · #11 Post by **Wookbert** » 01 Jul 2021, 18:48

ThomasB wrote: ↑01 Jul 2021, 04:58 So I suggest checking the LDO VReg's exact part number and confirm its current rating is robust enough. I think >700mA is ideal, but maybe >500mA rating would be sufficient.

Took a close look. VReg marking say „S2RC“, which according to the web is a Nanjing Micro One ME6211. If that's indeed what it is and the datasheet is correct, then max. output current is 500 mA (page 1), while page 5 claims under Absolute Maximum Ratings 600 mA Output Current.

Hmm... What really confuses me, is that exact same setting with my PSU board (so top right in my initial posts picture) worked flawlessly for several months. Will do further tests.

@ThomasB Just came across this voltage regulator related thread. I'm not very good at SMD soldering and have little knowledge about electronics, but could try to replace the ME6211. Only question is: With what?

TI LM1117 as mentioned there!? (argh... No SOT23-5 package, not pin compatible).
TI TLV757P doesn't look too bad (according to my limited knowledge) ... 1Amp, and pin compatible (see datasheet)

: IMG_0878.jpeg (326.01 KiB) Viewed 37513 times

ThomasB · #12 Post by **ThomasB** » 01 Jul 2021, 19:44

VReg marking say „S2RC“, which according to the web is a Nanjing Micro One ME6211. If that's indeed what it is and the datasheet is correct, then max. output current is 500 mA (page 1), while page 5 claims under Absolute Maximum Ratings 600 mA Output Current.

I think a 500mA VReg will be OK if all you are powering is the Wemos D1 Mini and no other accessories. But to reduce the chance of power brownouts during random peak currents, I suggest adding a 330uF to 470uF 10V low ESR electrolytic directly on the D1 mini's 5V power pin. Short leads to 5V and Gnd (installed on D1 Mini not the PSU).

What really confuses me, is that exact same setting with my PSU board (so top right in my initial posts picture) worked flawlessly for several months.

ESP device current draw will vary, depending on the firmware release. I believe that the older releases used less power during WiFi calibration and connection negotiations.

BTW, your PSU shield looks interesting. Are you confident it can deliver clean/stable 5V under all conditions? Can you post a schematic? What are the part number and ratings written on the PSU's Hi-Link DC-DC power module?

Just to be clear, your issue might not be caused by the PSU. But given what you've reported, the 5V power would certainly be high on my troubleshooting radar.

- Thomas

Wookbert · #13 Post by **Wookbert** » 01 Jul 2021, 20:13

My PSU board
BOM

Hi-Link HLK-PM01 5VDC 0.6A
F1 – Eska 883110 Kleinstsicherung radial bedrahtet eckig 200 mA 250 V Träge -T-
F2 – Eska SW-102T Temperatursicherung 72 °C 10 A 250 V
MOV – TDK S14K300 Scheiben-Varistor 470 V
C1 – TANCAP CT4-0805B223K500C1 Keramik-Kondensator THT 22 nF 50 V 10 %
C2 – Würth Elektronik WCAP-ATG5 860020372001 Elektrolyt-Kondensator radial bedrahtet 2 mm 10 µF 16 V 20 % (Ø x H) 5 mm x 11

: Bildschirmfoto 2021-07-01 um 20.13.07.png (2.67 MiB) Viewed 37493 times

: Bildschirmfoto 2021-07-01 um 20.05.49.png (164.17 KiB) Viewed 37493 times

: Bildschirmfoto 2021-07-01 um 20.07.51.png (610.9 KiB) Viewed 37493 times

: D10EBBB9-C95E-4458-AA57-A44B7E5129B5_1_105_c.jpeg (245.67 KiB) Viewed 37493 times

: DC3B3949-8943-40DD-B415-DA73D4941552_1_105_c.jpeg (198.55 KiB) Viewed 37493 times

Wookbert · #14 Post by **Wookbert** » 01 Jul 2021, 20:21

Again: The D1 mini Pro board with that S2RC/ME6211 VReg did work with my PSU board (have to figure out which release). It DOES NOT work with the LED Blink sketch (which works – just like ESPEasy when powered via MicroUSB).

The following D1 mini (non-Pro) board works instantly on my PSU board, and/but has „4A2D“-marked VReg, which seems to be an actually much weaker 150 mA Torex XC6204A23ADR. See images below.

: 5B6C1CBD-B2F3-4ED0-9152-9466BD79F66D_1_105_c.jpeg (155.31 KiB) Viewed 37491 times

: 64DCB80C-918C-4D36-BB81-B8F34FA27B5E_1_105_c.jpeg (283.56 KiB) Viewed 37491 times

: 0197D235-62F9-4CD9-B171-D1783AF4F975_1_105_c.jpeg (251.37 KiB) Viewed 37491 times

ThomasB · #15 Post by **ThomasB** » 01 Jul 2021, 21:05

From reviewing the information it appears that your PSU board's DC-DC module is supplying 3.3V too. Or maybe not. Please ensure that the the PSU's 3.3V and the D1 Mini's 3.3V pad are not connected.
EDIT: Never mind, it appears that the 3.3V connection is optional (only used when you install a 3.3V DC-DC module).

The PSU's 10uF cap (C2) is a marginal value. I suggest a much larger value. I typically use 470uF or higher in this application, but if space is a problem then try 220uF. Also add the electrolytic to the D1 Mini as mentioned earlier.

The following D1 mini (non-Pro) board works instantly on my PSU board, and/but has „4A2D“-marked VReg, which seems to be an actually much weaker 150 mA Torex XC6204A23ADR.

The device with a 150mA rated VReg will not be a good choice for ESPEasy, even if you find that it is working as-is. The VReg should be upgraded.

Not promising that these suggestions will solve the problem. But they are highly recommended changes.

- Thomas

TD-er · #16 Post by **TD-er** » 01 Jul 2021, 23:47

The "600 mA" one you posted is in a SOT23 form factor and I really doubt a linear (!!) voltage regulator in that form factor is capable of handling over 200 mA.
Those small modules can't dissipate any heat, so they really can't handle much.

Not sure if the voltage regulator is the real problem here when trying to power it from the "5V" pin.
It could also be a problem where the USB to LAN chip is not initialized as it would be when connected to a PC.
To test this theory, you could also see issues when powering it via an USB cable which doesn't use the data pins but only the 5V and GND.

ThomasB · #17 Post by **ThomasB** » 02 Jul 2021, 02:39

The "600 mA" one you posted is in a SOT23 form factor and I really doubt a linear (!!) voltage regulator in that form factor is capable of handling over 200 mA.

I looked at the data sheet. It shows that the SOT23-5 package can support 250mW heat dissipation. So with 5.0V external power, and 3.3V out, the little fellow will only tolerate 150mA continuous, with short duration peak current of 500mA. So keep an eye on its operating temperature.

BTW, the USB power mode will tolerate higher continuous current because the "5V" voltage is much lower. Probably close to 4.0V due to the VBUS protection diode and typical USB voltages. So perhaps about 350mA continuous would be available from the little VReg, which I think is enough for ESPEasy. Keep in mind that 500mA peak current is available, it's the continuous currents that are the worry here.

Not sure if the voltage regulator is the real problem here when trying to power it from the "5V" pin.

You're probably right. But my Spidey senses continue to warn me that the power supply related bits need some attention.

- Thomas

TD-er · #18 Post by **TD-er** » 02 Jul 2021, 08:32

ThomasB wrote: ↑02 Jul 2021, 02:39 [...]
Not sure if the voltage regulator is the real problem here when trying to power it from the "5V" pin.
You're probably right. But my Spidey senses continue to warn me that the power supply related bits need some attention.

- Thomas

Yep, I'm also a bit worried about those tiny SOT23 voltage regulators.

Not sure where this protection diode is placed.
If it is placed between the USB port and the voltage regulator, then it might be the thermal protection kicks in which is prevented by the diode if it is powered via USB.

ThomasB · #19 Post by **ThomasB** » 02 Jul 2021, 20:10

If it is placed between the USB port and the voltage regulator, then it might be the thermal protection kicks in which is prevented by the diode if it is powered via USB.

Agreed. And if this is determined to be the winning difference then a possible solution is to add a silicon diode to the PSU shield. The diode will drop the DC-DC module's output voltage to about 4.5V, which helps reduce the VReg heating. Or possibly two series diodes will be needed. See schematic markup for details.
.

: wemos_diode.jpg (191.66 KiB) Viewed 37394 times

.
It's all conjecture at this point; Time to poke more sticks at the ESP device to see what is going on.

- Thomas

TD-er · #20 Post by **TD-er** » 02 Jul 2021, 21:31

Maybe better to poke with sticks at those people who decide to add even weaker voltage regulators to those boards.

Ath · #21 Post by **Ath** » 02 Jul 2021, 21:59

The only real 'poke' would be to withdraw our wallets, but as those Chinese manufacturers as the only ones that actually provide these ESP8266 units (AFAIK), we don't have much choice, I'm afraid

Wookbert · #22 Post by **Wookbert** » 03 Jul 2021, 03:20

OK, here’s to the experts, from the clueless guy... (Is that actually even proper english!?)

Did some more testing providing power to the 5V+GND pins of the suspicious D1 mini Pro.

Tried a 5V 2.4A genuine Apple iPad USB PSU (like those, good quality, high efficiency, low/no standby consumption) ... nothing.
Tried an adjustable Meanwell PSU (lowest possible voltage 4.6V, highest 6.0V) ... nothing.

But then I noticed something very interesting: Instead of a soldered connection, I used loose wires for quick board swapping. And if I kept disconnecting and reconnecting the 5V pin just often enough, that D1 mini Pro board became suddenly alive. Not one, but all of them. And regardless of the PSU model, even with my above shown HiLink-based PSU board.

I also tried to reproduce this with dis- and reconnecting the 230V plug, but this is nowhere near as „successful“ as fiddling around with the 5V connection.

Now, please explain this! And most important of all: What and where to tweak how, to make it work not only sometimes, but always? Can this be solved in the ESPEasy software somehow?

P.S.: Sorry, no oscilloscope here, Multimeter only. Any not too much clue about what I'm doing...

ThomasB · #23 Post by **ThomasB** » 03 Jul 2021, 03:50

Sounds like a reset issue, as suggested earlier by TD-er.

Try this.
1. Power up the device.
2. Confirm it isn't working (won't boot).
3. Briefly ground (pull logic low) the Wemos RST pin to force a hardware reset. Remove the ground.
EDIT: The Wemos D1 mini has a reset button. Briefly press it instead of grounding the RST pin.
4. Does it start running now?

- Thomas

Wookbert · #24 Post by **Wookbert** » 03 Jul 2021, 04:46

No, a short GND-> RST connection doesn't „turn on“ the D1 mini Pro clone. Unfortunately not.
EDIT: I pressed the reset button in the past, but that didn't help.

And what’s even stranger, is that while a was half an hour ago able to breath life into the every D1 mini Pro clone, within let’s say seconds of the 5V-disconnect-reconnect game, it now doesn't work.

PM your address and I'll mail you two of those D1 mini Pro clones plus my PSU board (the Hi-Link brick is 100-240V).

ThomasB · #25 Post by **ThomasB** » 03 Jul 2021, 04:51

And what’s even stranger, is that while a was half an hour ago able to breath life into the every D1 mini Pro clone, within let’s say seconds of the 5V-disconnect-reconnect game, it now doesn't work.

Hmm. Do they still run when you use USB power?

TD-er suggested that you power by USB but with the data wires removed from the USB cable (only 5V & GND connected). I think this experiment would be helpful in the diagnosis.

Also, change the electrolytic caps, per my previous suggestions.

PM your address and I'll mail you two of those D1 mini Pro clones plus my PSU board (the Hi-Link brick is 100-240V).

I'm in the USA. You're probably across the pond. So mailing stuff will delay a solution. Let's wait a couple days and see if anyone else has some ideas.

- Thomas

TD-er · #26 Post by **TD-er** » 03 Jul 2021, 09:37

I think the USB to serial chip is in an unknown state when powered via the 5V pin.
Since it can toggle the reset and GPIO-0 pin (which is needed to enter flash mode), it is possible either one is toggled.
You could try to add a stronger pull-up to GPIO-0 to see if it may improve success rate of booting.
Or simply power it via an USB power bank or something else that won't toggle the USB data pins.

Another option would be to add a capacitor + resistor to the EN pin, but I'm not sure if it is easy to access this pin on the D1 mini.
GND --- Capacitor --- EN pin --- resistor --- 3v3

ThomasB · #27 Post by **ThomasB** » 03 Jul 2021, 18:40

Since it can toggle the reset and GPIO-0 pin (which is needed to enter flash mode), it is possible either one is toggled.

Grab the voltmeter and measure these pins. That will confirm their power-on state.

I went through my ESP8266 collection and found the Wemos board like yours with the 4A2D VReg. It's the version that uses the shielded ESP8266MOD module. Unfortunately this is the board version that is working for you.

Another observation is that the non-working boards use the unshielded ESP8266 chip (discrete components instead of ESP8266MOD module). I vaguely recall a reported incident where an unshielded board was not working on some firmware versions. Maybe in your case, EMI/RFI from the adjacent DC-DC PSU is affecting the unshielded ESP8266 circuitry.

So another experiment would be to see what happens when you connect the Wemos D1 Mini to your PSU with ~15cm long wires. That is to say, get it further away from the DC-DC module.

- Thomas

TD-er · #28 Post by **TD-er** » 03 Jul 2021, 18:47

The issue with some unshielded boards is that it radiates high power WiFi signals directly into the ESP chip.
This caused all kinds of bit flips and thus the most difficult to reproduce crashes.
The fix there was to reduce TX power.
But this does not even seem to start if I understand correctly?

Wookbert · #29 Post by **Wookbert** » 04 Jul 2021, 17:15

ThomasB wrote: ↑03 Jul 2021, 18:40 I went through my ESP8266 collection and found the Wemos board like yours with the 4A2D VReg. It's the version that uses the shielded ESP8266MOD module. Unfortunately this is the board version that is working for you.

I identify this board version by the printed antenna and embedded/shielded ESP8266ex, although I'm sure that there are versions out there, which use yet another VReg than the 4A2D. On AliExpress it goes by the name „D1 mini“ (so without Pro).

The main reason I went for that other, problematic D1 mini Pro board (which has the S2RC/ME6211 VReg) is the ceramic antenna, which performs better than the printed one, according to my tests. Plus it has the IPX connector for an external antenna.

Did some further testing, all using my HiLink-based PSU board as is:

provided the power via 15 cm, twisted pair 0.4 mm² massive copper wire with the D1 mini Pro board distanced from the PSU = no change, still only randomly booting
tried grounding RST pin for a brief moment again = nothing
voltage on RST pin (climbs) instantly to 3.3V when connecting to 230V
voltage on D3 (GPIO-0) pin (climbs) instantly to 3.3V when connecting to 230V
when providing the HiLink power via 2-wires only to the MicroUSB receptacle, the D1 mini Pro always boots up

Also built a 3V3 version of my PSU board: It uses the HiLink PM-03 (3.3V 1A) and connects to the 3V3 pin, instead of the 5V pin of the D1 mini Pro:

measures 3.3V on both 3V3 and 5V pins
boots up just as randomly

Don't know what to do next. I mean it's cumbersome if we/I have/one has to specifically address each flavor of D1 mini (Pro) board. Perhaps some experts here should design the ultimate, flawless D1 mini board and we should kickstart it (assuming that the other ESP8266 firmware makers are having similar problems). Just an idea...

TD-er wrote: ↑03 Jul 2021, 09:37 I think the USB to serial chip is in an unknown state when powered via the 5V pin.
Since it can toggle the reset and GPIO-0 pin (which is needed to enter flash mode), it is possible either one is toggled.
1. You could try to add a stronger pull-up to GPIO-0 to see if it may improve success rate of booting.
2. Or simply power it via an USB power bank or something else that won't toggle the USB data pins.

3. Another option would be to add a capacitor + resistor to the EN pin, but I'm not sure if it is easy to access this pin on the D1 mini.
GND --- Capacitor --- EN pin --- resistor --- 3v3

I don't see how 2. makes a difference when connected to the D1 mini Pros’s 5V/GND pins, so I assume that’s meant to connect to the board’s MicroUSB receptacle. That would actually be a quick, simple fix: connect the 5V power provided by my PSU board not via the 5V/GND pins, but using a MicroUSB plug. Not very elegant though.

On 1.: Stronger Pull-Up. Like how and what? I assume that’s meant only brief to trigger booting. That would have to be done in software I guess. (I can't emphasize enough how clueless I am

)

Same applies for 3. Perhaps you can go into details.

I seriously wonder, what the manufacturers of this board think how this is going to work. Or are those 5V and 3V3 pins on the D1 mini (Pro) boards actually not meant to supply power to, but source power from (which is supplied via MicroUSB)??

ThomasB · #30 Post by **ThomasB** » 04 Jul 2021, 19:59

Your observation that older firmware works perfectly makes this a very interesting bug.

I have a hunch there is a hardware mistake in their clone's board layout. Or they installed the wrong value component in a place that matters.

I don't have any other troubleshooting ideas to toss at this. But someone may eventually stop by and provide a fix, so don't give up.

I agree that the simplest solution is to connect your 5V PSU to the USB+ power pin. That's not a horrible thing, but I empathize with the frustrations these type of detours create.

- Thomas

Wookbert · #31 Post by **Wookbert** » 04 Jul 2021, 20:17

ThomasB wrote: ↑04 Jul 2021, 19:59 Your observation that older firmware works perfectly makes this a very interesting bug.

Well, this has to be taken not with a grain, but rather a spoon of salt. I wrote this before I made the observation that the board does boot/work/ show signs of life randomly.

I mean I yet have to test all ESPEasy releases in the respective time frame, but the fact that the simple “blink on-board LED” sketch behaves no different, makes me believe, that it is not firmware release specific issue. Correct me if my assumption still might be wrong.

ThomasB · #32 Post by **ThomasB** » 04 Jul 2021, 20:22

Post a link to the item at the seller's online store. No promises, I might buy one and test it out.
- Thomas

Wookbert · #33 Post by **Wookbert** » 04 Jul 2021, 22:45

ThomasB, I’ve sent you a PM.

ThomasB · #34 Post by **ThomasB** » 05 Jul 2021, 01:13

@Wookbert, please see my PM reply.

I noticed that your posted board photos show that the "VPP rom programming" cap option is installed on CP2104 USB serial chip. After searching other clone makers of D1 Mini Pro board I notice that some omit this optional cap.

Compare the photo below (a random photo found on the interwebs) to your board. Please observe that the capacitor near the D4 silkscreen is not installed. However, your board has this component. Perhaps this optional cap is involved with your boot issue.
.

: VPP Programming Capacitory not installed.; WemosD1MiniProCap.jpg (197.1 KiB) Viewed 37117 times

.

Its purpose is to allow the board maker to program the CP2104's ROM (via the USB port) after board assembly.

As an experiment, remove the cap and see what happens (but save the cap so you can re-install it after the test). Maybe removing this component is the holy grail solution. Or maybe not.

- Thomas

Wookbert · #35 Post by **Wookbert** » 05 Jul 2021, 14:32

@ThomasB: Managed to carefully remove „your“ mentioned cap next to D4 using the Hot Air Gun on a fresh board (the plastic terminal would otherwise melt). Makes unfortunately no difference: no sign of life when powering through 5V pin. MicroUSB as always flawless.

(Do I have to solder that cap back into place? I guess not ... don't want to stress the board too much).

BTW: The D1 mini Pro boards we are discussing her were purchased on AliExpress from Hong Kong Feng Tai Co., Ltd. (Brand name „Eyewink“). There’s no guarantee however that all boards they ship are the same.

RobertM · #36 Post by **RobertM** » 05 Jul 2021, 18:47

This version is used for USB CP2104 which is also powered from USB 5V.
I assume that everything will work OK if you bridge the diode on which it marked "SL" under the 5V pin and power it with HLK-PM01 (5V)
It is possible that these outputs of this chip remain in some strange level that disrupts the ESP boot.

ThomasB · #37 Post by **ThomasB** » 05 Jul 2021, 19:30

Thanks for performing the test. Too bad it was a dead end idea. No need to reinstall the cap for now (but save it).

I assume that everything will work OK if you bridge the diode on which it marked "SL" under the 5V pin and power it with HLK-PM01 (5V)

I agree. There is a chance that the partially powered chip is triggering the ESP8266 to go into the wrong boot mode when power is applied.

The simplest workaround would be to bypass the protection diode. The only risk with this is that it would be unsafe to connect USB when the device is externally powered.
.

: VBUS Diode; WemosD1MiniProDiode.jpg (196.1 KiB) Viewed 37027 times

.

What is strange is that no one else has reported this problem. The official Wemos D1 Mini Pro board uses the CH340C, but there are a lot of clone makers selling a version with the CP2104.

- Thomas

ThomasB · #38 Post by **ThomasB** » 05 Jul 2021, 20:25

Follow up info:

Adafruit's Feather is an ESP8266 board with CP2104 USB interface. Their design took a different approach to the device's power bus. Instead of a power protection diode on the USB input, it is on the external 5V input. That makes a big difference. With this arrangement the CP2104 is powered with either voltage source. This could have been accomplished on the USB input too, but not in the way the D1 Mini Pro board implemented it.

So it appears the D1 Mini Pro clone board engineer has made a poor design choice. And yet another example of how we get what we pay for.

- Thomas

TD-er · #39 Post by **TD-er** » 05 Jul 2021, 23:04

Too bad it is hard to chose a "better" one, as those bad boards are also sold for way higher prices.

Wookbert · #40 Post by **Wookbert** » 06 Jul 2021, 00:53

TD-er wrote: ↑05 Jul 2021, 23:04 Too bad it is hard to chose a "better" one, as those bad boards are also sold for way higher prices.

Typo? Did you mean the good boards are sold for way higher prices?
Question is: Any recommendations for less flawed D1 mini Pro clones from AliExpress?

I hesitate to bypass the mentioned protection diode. Fear’s too big to inadvertently fry the USBus on my MacBook.

I did the inevitable and chose the jumper wire road

Foreseeing how I am I luckily added a „power stealing“ jumper to my PSU board layout, to which I now can directly solder the MicroUSB leads to.

: 204BCD1B-0ED0-4395-BF93-F6D17E74ED12_1_105_c.jpeg (313.55 KiB) Viewed 36985 times

ThomasB · #41 Post by **ThomasB** » 06 Jul 2021, 03:19

Question is: Any recommendations for less flawed D1 mini Pro clones from AliExpress?

I trust authentic Lolin branded boards if purchased directly from their official store. Admittedly I haven't tried the Lolin D1 Mini Pro V2, but for $5 USD ($8.5 shipped) it is a fair deal. It doesn't have a ceramic antenna (PCB and external only) and the VReg is regretfully the SOT23 size. But still worth trying out.
https://www.aliexpress.com/item/3272469 ... 9240.pic_3

- Thomas

TD-er · #42 Post by **TD-er** » 06 Jul 2021, 09:16

Wookbert wrote: ↑06 Jul 2021, 00:53
TD-er wrote: ↑05 Jul 2021, 23:04 Too bad it is hard to chose a "better" one, as those bad boards are also sold for way higher prices.
Typo? Did you mean the good boards are sold for way higher prices?
[...]

Nope, not a typo.
The same crappy boards are also sold for higher prices, so differentiating on price is no guarantee for better boards.

And if you need stable boards, and can design your own PCB's, you can also design your own board including an ESP.
JCLPCB now also can handle the ESP-07S or the ESP32 modules for SMT mounting and if you order > 10 or 30 boards the prices of those modules are quite OK.
I have made a number of boards which include those and they do a good job.
I do use the ESP modules which have an UFL connector for antenna and then use a small antenna which has an adhesive on the back so you can stick it to the enclosure where you like.

Just make sure to add the 2 transistors for flipping the GPIO-0 and RST in the right order to ease flashing.
I use those "Wemos CH340 breakout boards" like these: https://www.aliexpress.com/item/3289284 ... 9b7b004-24
Those have DTR and RTS/CTR pins in a row, which can be used by the flashing application to toggle the RST and GPIO-0 pin.

Just make sure to double check the default solder pad on the back of the board if it is set to 5V or 3v3.
I think it is best to leave it at 5V and include a 5V pin on your header + an AMS1117 on the board for powering the ESP. (these boards have the same under powered 3v3 regulator, older ones had the AMS1117)
You can look at the available schematics of the NodeMCU boards to see which pins need pull-up or -down resistors and how to wire the 2 transistors.
I advice to use 220 - 330uF to stabilize the 3v3 and place 100nF close to the power pins of any (!!) chip, like the ESP, but also sensors.

Wookbert · #43 Post by **Wookbert** » 06 Jul 2021, 15:50

@TD-er

My PSU board was actually by JLCPCB as well. Designing my ESP8266 board is an interesting idea, but way beyond my capabilities.

Perhaps we could turn this into an open-source community project in a not so distant feature, in order to create the ultimate, compact, modular and flawless ESP base unit. Think of a board with my shown form factor (to fit European wall boxes), with on-board 100-240V PSU (doesn't have to be Hi-Link, could get smaller; break-away PCB option in case you want to drive with an external PSU) and all sorts of matching shields.

I could manage the production, logistics and distribution side.

Opinion?

TD-er · #44 Post by **TD-er** » 06 Jul 2021, 16:04

Well it is a bit tricky as you make it for mains voltage.
If you design your own boards, that's not really an issue as you then also understand the risks (I hope

)
But manufacturing it for others to buy does add some risk to it since a lot of people don't understand the risks.

Also I would for sure use HiLink ones as that's an established brand, but it still needs some added protection like a thermal fuse and a normal fuse and also some cover to make sure the mains part cannot be touched by accident.
Such an enclosure also needs to be safe like able to handle higher temperature (thus PLA 3D printed parts are not an option) and may not shatter when hit with something.

In other words, I can for sure design a generic board with good parts etc. But I will not design one for mains power which can be sold to anyone.
I have made several designs which do work with mains power, but those are being used in bought enclosures for industrial use.
Even my own boards using mains power are not housed in 3D printed enclosures (at least when they leave my desk)

Wookbert · #45 Post by **Wookbert** » 06 Jul 2021, 16:22

First of all, this would have to be a non-commercial, open-source, hobby project. As such, I also wouldn't populate the mains power parts on the board. So design (and perhaps offer) the pre-populated board, but leaving the responsibility to solder the high voltage parts to the individual user, along with a clear warning.

One also doesn't have to necessarily take responsibility for manufacturing and distribution, but could simply provide the Gerber files and BOM through github, with instruction on how to order from JLCPCB. The project doesn't need to have any association with Let’s Control It/ESP Easy, if you don't want to. The whole idea is just to have something which works without pitfalls.

I'm fairly good in designing for 3d printing. Totally aware that PLA is no option for any cases which get warm. I either use ColorFabb nGen (PETG with glass transition @ 85°C) or Tritan (glass transition @ 100°C). If your PSU gets that hot, you have a problem anyway, but the thermal fuse in my shown PSU board blows instantly at 77°C (it blows so quick, I have to deep freeze it with ice spray in order to be able to solder it ... and even here I have to be super quick).

Got a lack of time right now, but would like to get back to you in the near future on this. I think the community would be happy to have a solution which doesn't have any issues, as shown in this thread.

TD-er · #46 Post by **TD-er** » 06 Jul 2021, 17:36

Well to be honest, I have been thinking about it too.
Especially for those that want to use ESP32 + ethernet, you either have to buy for quite a steep price (wESP32 board) or for somewhat larger boards like those from Olimex.
Those are both very good options by the way, but may be too big for use cases like you plan on using.
Also the Olimex boards do start from roughly 17 euro + VAT + shipping.

It is however possible to make ESP32 boards + LAN for a lot less, when using those LAN8720A boards from sites like Ali Express.

For my own test setups I would also prefer some board which is even easier to attach some sensors to it, so also for that purpose I have been thinking about designing some boards too.

However there is at this moment a very serious issue with availability of parts.
Last month alone, I have had to change designs 4 times already when re-ordering boards due to parts shortages.
Even the most simple transistors, which I picked before as they were in stock with over a few 100k pieces at JLCPCB, now may have a lead time of tens of weeks.
The LAN8720A chip now has a lead time of roughly 60 weeks at almost all sites like Mouser, Farnell, Digikey, etc.
The same for some very common mosfets to power simple motors for example.
So I can generate Gerber files, BOM & Pick-and-place files, but I'm sure those cannot be used in days or weeks from now.

Wookbert · #47 Post by **Wookbert** » 06 Jul 2021, 17:48

No need to hurry. I'm loaded with work anyway at this point. Will try to sketch my few ideas as soon as I find time, and approach you accordingly. No sooner than August I think. Hopefully the chip crisis relaxes a bit within the next 12 months.

ThomasB · #48 Post by **ThomasB** » 15 Jul 2021, 06:26

I evaluated a couple D1 mini pros (received from Wookbert). One randomly failed to boot (about once every ten tries).

When using external 5V the CP2104 USB chip is partially powered. So randomly it does not properly initialize the RTS and DTR pins, which control the ESP8266 Reset and Boot modes. Sometimes the ESP8266 is randomly put into boot mode at power up.

I devised a hardware mod that restores power to the CP2104 when powered by 5V. Only one cut and one jump. But it is difficult to do. Using the USB jack to power the D1 Mini Pro seems to be a better solution than applying the modification.

For reference, the mod removes VBUS power from U3 (CP2104) Pins 7-8 and C8 via a single copper trace cut. These three things remain connected together and are jumped over to +5V power at the cathode of D2. Please note that the anode of D2 remains connected to the VBUS (5V) pin on the USB jack.

Here is a revised schematic:
.

: Revised schematic; D1MiniProSchem_Mod_sm.jpg (317.22 KiB) Viewed 36535 times

.
This image shows the location of the cut and jump.
.

: Cut and jump.; D1MiniProDiodeMod.jpg (214.01 KiB) Viewed 36535 times

.
Here is the board I modified. It reliably boots from both 5V and USB power.
.

: Installed Mod.; d1minipro_mod.jpg (134.3 KiB) Viewed 36535 times

.

- Thomas

Wookbert · #49 Post by **Wookbert** » 15 Jul 2021, 14:14

Nice find. Thanks a lot. As I wrote above, that board had some sex appeal to me, because of the chip antenna and IPX connector, but obviously this clone (or all D1 mini Pros, even the genuine one?) has a flawed design.

Time to come up with something better, if not to say: the ultimate ESP boards

. Doing some concept work here...

P.S.: Why’s the „• Rainsun“ imprint on the chip antenna in the second picture missing? Did it come off in the IPA bath?

ThomasB · #50 Post by **ThomasB** » 15 Jul 2021, 18:32

Nice find. Thanks a lot.

You're welcome. Not a great outcome, but at least we confirmed that it was a hardware issue.

Why’s the „• Rainsun“ imprint on the chip antenna in the second picture missing? Did it come off in the IPA bath?

When I received the board the antenna was missing from the PCB. Found it loose in the static bag. In a hurry to diagnose the problem, I was sloppy and soldered it upside down.

EDIT: I unsoldered the SMT ceramic antenna to flip it over. Both sides look the same (no branding logo). However, there is a black mark printed on one end of this SMT antenna, so that was aligned topside, towards the ESP chip.

- Thomas

Who is online