After a really long time I received the boards. I finished the assembly for one and uploaded some code to it. So far it seems to work but I need to perform some more tests. I attached some pictures of the finished product. .

The silkscreen text is a little bit blurry and it was a little bit too small for the fab to print it so in some places is not right...anyways this is not important but the functionality is.
One note: the board round edges may be a little bit off the limits but that can be corrected by using some abrasive paper to remove the excess material(which I did in the above pictures and then the board fitted perfectly).

@Nca78

Elecrow is a good fab imho and yes they are very generous - I received too my boards today and instead of 5pcs I received 11pcs !!!. And this is not the first time they deliver more than requested. The price - 5$ for boards and 3.5$ shipping(standard mail) - 8.5$ in total. And yes they are the cheapest and pretty good quality pcbs - I'm very pleased about their services so far. They beat all the Chinese fabs including shipping so far(and the quality is pretty good - at least for what I need my pcbs for it's well beyond my expectations).

@NeverDie

Thanks. Well because it was cheaper/smaller than all of the Aliexpress modules found so far and it's shielded also(talking about the PTR5518 module here). So I found it to be perfect for this solution when this project was created(almost 2 months ago). Other than that nrf51 even if it's older or not so powerful than its nrf52 counterpart I found it to be more "mature" when it comes to the software part - just like atmega328p is for the Arduino world. If I remember well it was among the first one supported in sandeepmistry's nrf5 core: https://github.com/sandeepmistry/arduino-nRF5 or any other Arduino core for the nrf5xx modules so I assume its software support should be more richer than nrf52xxx.

@toyman

You can find it in my kicad repo: symbol and footprint

I didn't used it in any design so please double check.

@NeverDie

For me, elecrow is the best. But you can query this online tool which is very good: https://pcbshopper.com

OK...another remake for the classic Livolo 2 channels 1 Way EU switch and THE LAST ONE that I will make(well except for the 1 channel one which I need to work on it after this one) - @shabba I know that you waited for this. This is just the other Livolo hardware project of mine "ported" for the new hw revision that the Chinese manufacturer made. They made the 12V line available on the connector and moved it on the opposite side(the new connector is a 2mm pitch one and SMD - I choose to use the THT variant as it doesn't have any impact from a mechanical point of view and it's easier to solder for all of us the mortals around here).

The good news is that the power board is the same as the old one with some minor modifications(more or less):

1. the big MOSFET is in the same place but with a 45 degrees orientation
2. the power connector is a 2x7 pins one with the 12V line already supplied on one of the pins - less hacks for the board
3. the bistable relays are the same and drived by a ULN2003 IC(current buffer) in a TSSOP package

Thanks to @tonnerre33 for creating the new board outline which I borrowed from his project. The rest of the board is the same as the other one that I made with the dc-dc converter moved on the opposite side and with some rearrangements. Next I need to create the boards AGAIN using some external manufacturer(China of course) and then wait...and patience and... patience again and...wait.

And as always big thanks to the Mysensors community for their great effort and for maintaining this great project.

Oh...and for some reason the boards are not rendered correctly in the browser - I'm referring to the openhardware.io gerber viewer(both 2D and the 3D one) - maybe it has to do with the latest changes from KiCAD regarding the way the gerber are generated or maybe some bug in there? (I'm using the latest nightly version). The Linux gerbview utility displays them correctly.

@d00616

Hi,

This is a such great addition for this project. For long time I wanted something like this: a RF SOC with pretty good and usable software support and low power also(and a Cortex arch - pff..that's too much :simple_smile: ). I just wanted to congratulate you for your work and bringing this in to MySensors project - it's very neat and useful.

I tested it and it works just great so far. I would give you 100 likes on openhardware.io if it would be possible :simple_smile: . Thanks once again for your effort.

@NeverDie

For the swd connector - some simple wires soldered temporarily :simple_smile: and removed afterwards. The project doesn't support OTA - but can be added if desired as a DFU upload service - needs some extra work and research. But once the switches are flashed there's no need to reflash them as all the MySensors stuff is on the linux board node which can be reconfigured very easily through that web interface(the services behind it will restart automatically when the configuration file changes and the new settings are applied automatically also). I will prepare the required documentation about the whole system on my blog and there's a lot to write :simple_smile: .

Ok, so after posting some questions to CDSENET Aliexpress store(in regards to the E73-2G4M04S module):

1. Is the nrf52832 IC DEC1 pin decoupled with a 100nF capacitor as per Nordic datasheet recommendations?
2. Is the nrf52832 IC  DEC2 pin decoupled with a 100pF capacitor as per Nordic datasheet recommendations?
3. Are all the power pins of the nrf52832 IC decoupled as per  Nordic datasheet recommendations?
4. Are the external components already provided so that the internal DC-DC converter can be used?

I received this part of schematic:

The module that I bought is this one: CDSENET E73-2G4M04S

I hope that it helps demystify some things in regards to this module. Maybe @NeverDie can confirm this by checking his module connections as seen in the above schematic. I'm saying this because Chinese support is well..not so user friendly in my experience and I don't know if the above schematic is indeed what they used on their module or they just provided me an example schematic of how to use the NRF52832 IC in general instead.

nRF52832 2.4GHz Transceiver Wireless rf Module CDSENET E73-2G4M04S1B SMD 2.4 ghz Ble 5.0 Receiver transmitter Bluetooth Module

$2.95

@MiKa

I bought this switch(non-radio variant): https://www.aliexpress.com/item/Free-Shipping-White-Crystal-Glass-Switch-Panel-EU-Standard-VL-C702-11-Livolo-AC-110-250V/512770913.html?spm=2114.01010208.3.11.JqUULi&ws_ab_test=searchweb0_0,searchweb201602_4_10065_10068_10000009_10084_10083_10080_10082_10081_10060_10062_10056_10055_10037_10054_10033_10059_10032_10099_10078_10079_10077_10000012_10103_10073_10102_10000015_10096_10052_10053_10107_10050_10106_10051-10050,searchweb201603_9,afswitch_5,single_sort_0_total_tranpro_desc&btsid=518f103c-2c52-4311-9b72-2375b87824c5

The power/relays board looks like this:

@monte

Are you suggesting to implement something like esphome has? If that's the case then I would prefer yaml instead of json as it's easier to read and maintain imho.
The idea to have code generated from a configuration file is pretty neat and it simplifies tedious and repeating tasks a lot besides abstraction.
NodeManager should help as it seems to be pretty modular.

Started designing the enclosure for 3d printing ... I will use PET-G as it's pretty strong and not very flammable or so it seems after some tests. I don't know about electrical insulation properties yet ...This all very depends on what other compounds are used when it's manufactured.

As a general rule it seems that the input capacitor is choosed based on the needed output power which is obvious and the magic formula is (when using a bridge rectifier):

C (uF) = 2 x Pout (W)

So in my case one relay draws about 300mW when switching and we have two so let's take the worst case 600mW. Then the radio part and mcu should not take more than 30mW of power.
Summing up we have 630mW peak which translates to a 1.26uF capacitor and the nearest is 1uF.

I just tested with 1uF and it works just fine without a glitch.

So yeah learning new stuff as I go along. I wanted for a long time to learn how to design offline AC/DC converters because the power supply is the most important part of every circuit and so often overlooked.

As a side note: this doesn't mean that the input filtering capacitor is useless in the ac/dc power supplies as it also helps to store energy for the input stage which then gets transferred to the load by the pwm controller. Another useful thing is that it doesn't let the rectified wave voltage to drop too much which may not be very good for the input stage of the pwm controller making it to perform worse.

But in my case and particularly in this application it interferes when the switch "switches" between the two power supplies (well the other one is more of a "series regulator").

I will do some more tests. The most important thing for me right now is the power supply stability and then the thing I was waiting for the most: to create a 2 way setup using 2 switches to drive the same led bulb (for stairways).

Received my boards and started to perform some testing. It worked OK when powering front plates loaded with Bluetooth LE software which are drawing only ~500uA. But as soon as I tried to power a Bluetooth mesh node which draws constantly around 5-6mA then the stand-by supply (the one built around LNK364 IC) didn't acted fast enough when performing the switching causing the BLE mesh node to reset. In other words the power supply startup time was pretty slow for this kind of circuit.

After some serious investigations and headaches (around 3-4 days or so) I finally found why. It has to do with the AC-DC converter "capacitive loading". Initially I though that it was because of the output filtering capacitors and which are also used to store energy between cycling the bi stable relay on and off. As you all know capacitors are good for filtering the ripple and storing electric energy...BUT and very important for this project over here - if they're too big they make the power supply to react a little bit slow.

Why? The reason is very simple: if you use too large capacitors then it takes longer to charge (not to mention about high inrush currents) and because of that they slow down things affecting the AC-DC converter feedback loop.

Well, in a normal use of the power supply and when not cycling between on and off power states too fast this won't be noticed and doesn't matter too much. In my case it matters and guess what? - not the output capacitors were the culprit after tinkering a little bit with their values.

The input filtering capacitor was causing issues. Yes, I'm talking about the input filtering stage after the bridge rectifier. But in this case I started to "attack" the AC/DC power supply in a different way and that is: what happens at the input stage when the supply is switched on and off ?

Let's not forget that in this project the stand-by AC/DC converter is in series with the load (light bulb) and it "breathes" and draws power from the mains voltage and the leakage current from the light bulb which can be very small especially when using LED bulbs.

So what happens actually? As you all know the AC/DC converter won't work if the input AC voltage is less than 85V or so as most are rated to work between 85-265Vac right? But what this means after the rectifier/filtering stage where we have a continuous voltage (let's forget about ripple right now)?
Taking the minimum ac voltage value of 85V this translates to: 85 x 1.41 =~120Vdc. So untill the input filtering capacitor won't reach that dc voltage value the LNK364 chip won't work and here I'm referring to it's internal voltage regulator that powers the internal circuitry or it's "start up" circuit. (in reality it may be less than 120Vdc, but still...)

But then you may think: yes, yes but the input filtering capacitor is only 4.7uF - won't it charge fast enough? Well no, and not quite - let's not forget that it's an electrolytic capacitor with higher ESR and not only that - what matters here the most is the fact that the leakage current from the light bulb dictates it's charging rate an that current can be very small !!!

And now you have it - the input capacitor in this case slows down things and affects the power supply start up time. And it doesn't matter that much if you increase the output capacitors and it becomes even worse because then we end up affecting the feedback loop from the converter.

So what I did was to remove completely the input filtering capacitor and now it's damn fast - yes the power supply still works as expected but it's much faster between on and off cycles. And if you look at the original Livolo switch it doesn't have it either but now I understand why.

We have enough filtering at the output stage so it won't affect our powered circuits. It's a little bit noisy at the input but not that much I would say because we have a bridge rectifier. And the load doesn't draw too much current so it should keep up without the input capacitor which also stores energy for the primary inductor. The LNK364 internal MOSFET shouldn't be affected as it's rated to support the peak input values and even more (700Vdc) as the input capacitor was averaging a little bit the rectified voltage waveform (but not that much anyways because it's small value). I will try with a 1uF/2.2uF and see if that slow things again.

So yes..I was almost ready to give up on this project. Now I'm more happy than ever and also because I learned new stuff. And designing/playing with offline AC/DC converters is kinda new for me. There are still many things to learn of course.

I also need to thank @axillent for its "AC-DC at own" post from where I got inspired a little bit.

This board uses the same working principle behind the scenes as the Livolo or Sonoff no neutral switches (some use a triac based implementation). But again the idea is the same - we take advantage of the leakage current from the light bulb. So yes it still requires a capacitor across the bulb in some cases. This is because we power everything via the light bulb which is in series with our circuit and in the end it all comes down to physics law.

So no @arden , this one is not so different compared to the one that you already have. I did this because I'm a more of a diy person and I like to customize stuff and of course learn from it.
Another advantage is that you don't depend on external or 3rd party stuff and you can build/customize it whenever you want (of course this means that you know what you're doing considering the fact that you are dealing with the mains here).
As I live in EU I have only the live wire inside walls for light switches so this is another reason why. I know that there are another ways of accomplishing the same result but I don't want battery powered switches nor putting stuff on or inside my ceiling.

@AdamAnt Regarding the interference between MCU and the radio a pull-up resistor on the CSN line helps. By doing this you can program the MCU without any problems after soldering the radio module.

Now getting back to the main issue here:

1. Did you checked for shorts between pins (especially the ICSP part: mosi, miso, sck and also the reset pin to not be shorted to ground keeping the mcu in a reset state). Asking this because I see some pins shifted over there a little bit on the corresponding pads from the PCB.
2. Did you checked for continuity between the corresponding MCU pins and the ICSP header?
3. Can you post some pictures on how did you wired the USBASP programmer and the ICSP header from the PCB?

The schematic seems fine from my point of view. There may be some problems due to faulty PCBs also from the factory (it happened to me) - that's why I suggested doing the above tests with a multimeter (shorts/continuity test).

@alowhum

It's not portable but for the AVR architecture you can use:

void before() {
  wdt_disable();
  wdt_enable(WDTO_8S); // 8s timeout 
}

void loop() {
  wdt_reset();
}

Other watchdog options for the AVR architecture here.

@dzungpham0703

I replied in the nrf51822 thread. Yes I managed to make it work like the original with BLE stack. And using the wonderful Cypress Proc BLE solution including capsense and lots of goodies . Cypress really rocks! I love them.

The current consumption was lowered to 200uA without being connected and 700uA connected. Those numbers include LEDs current consumption and other blocks like touch sensing.

I'll get back with more updates when it's finished.