My poor effort - GitHub - borpin/esphome_components: ESPHome Components (but it did work and was robust)
Actually @awjlogan I remember now, this exists - GitHub - openenergymonitor/esphome_components: ESPHome Components @TrystanLea could add you to it.
Understood now, thanks. Potentially you could use the USB port on the Pi3/Tx++ as a host, particularly for serial port emulators (USB CDC class) like an MBUS dongle. That would be delightfully clean, and the USB stack I use does support host mode.
Just a configuration, for now. The Pi3/Tx++ can be configured to send the output as JSON and I think it will just need a lambda to decode it, or potentially thereās already a JSON decoder component. Truth be told, I havenāt looked too far into the software side yet, being focused on the Pi3 
I hadnāt quite realised just how cheap they were! That said, the ESP32-C3 is only about Ā£1.70 from a distributor - obviously I canāt match the assembled price of a Wemos, but itās really not expensive overall.
Have a look at that GitHub. That is what I did IIRC (I prompted the output in JSON to be available). It splits it into the parts for HA sensors and then reconstructs it as a new JSON to send. I couldnāt work out how to send the original JSON on directly IIRC.
Iām sure things are way too far down the line for the emonPi3, but for the emonPi4 this has to be an option
Hi Brian - do you mean including the M.2 connector on the emonPi[n] for an SSD? Only the Pi5 (and presumably future ones) bring out the PCIe lane on a header. My feeling is that the Pi5 is much too expensive to be a sensible choice for this application.
Using a Compute Module would be interesting - you can buy a range of SoMs with varying amounts of RAM, eMMC memory, and WiFi/BT connectivity depending on what you want so you only pay for the features you need. It also brings out all the high speed (USB, Gb Ethernet, PCIe, HDMI, MIPI) interfaces as well. Iāve designed a few products with the CM4 and itās nice to work with. The main difficulty is sourcing in small quantities, so either OEM would have to choose one to buy or you leave it to the end user to find and order the one they need.
Ok, the v0.2 boards have arrived and work nicely with the USB bootloader uploaded Three little pigs:
PCB assembly by Aisler - EU based and priced similar to the Chinese shops, and Iāve been very happy with the bare PCBs before. The assembly process was easy, and would definitely recommend using them in future.
Only one āmajorā mistake - the KiCAD footprint for the bicolour LED on the front is incorrect. Iād fixed this for another project, and clearly forgot to port it over to this one. Will open a PR with KiCAD as well, a bit annoying. Not a functional problem, but thereās no visual indication of activity. Good to catch at prototype stage.
Next step: finish off the firmware. Thanks for everyoneās input and comments so far, and always open to any other suggestions.
Looks great.
I still think a version without the RF would be worthwhile.
@awjlogan Thanks for the update. Looking good.
Hy everyone
Great job done so far. I would only suggest to drill ct clamp connectors pads to give them more strength. Correct me if Iām wrong. I was wondering if there is any chance to buy any spare board if someone have any assembled. Posted to Uk address
Kindness
Rob.
Hi Rob - thanks for the interest. The CTs attach through the 3.5 mm jacks, which have 3 large pads so thereās no need for additional support. Iāve revised the expander board to use a two row connector, which provides more mechanical stability along with the screws.
We are at the production candidate stage, looking for general availability mid to late 2025. The board and firmware are both open source, but itās not a āfunā build. OEM have the other 2 boards from the run for testing, and I have the 3rd for development
Feel free to ask/suggest anything in the meantime
Understood. I have couple of questions if I may
Regarding connectors. Push connectors from WAGO 250 SERIES family gives better look and quicker installation ( talking about green connectors ) but is is only my personal opinion. Iāve had also several issues with plug connectors on my board - sometimes under pressure (tightening)solder can break causing loose connection.
Are u considering adding Can Bus interface like mcp2562 just to be available on board to solder. I.C. for future development. ? This could bring more adoption to your board. Possible expansion where no network is available.
Last one. Could you please explain advantages of making this board 4 layers. Did this reduce space compared to price production increases. ?
For any thoughts I will be much appreciated.
Iām absolute lemon. So forgive me if Iām unclear. Happy coding.
Kindness
Rob
Hi Rob - thanks for the message. Just to go through the points below, hope this is helpful.
Two reasons we will stick with the existing ones, rather than something like the Wago 250. 1) Backwards compatibility with existing installations (e.g. people have their temperature sensors terminated with the correct pin), and 2) the casing dictates the choice of connector; the W 250s look like they need access above the socket to press the button.
This could be placed on the expander header. I donāt have space on the board to include this for CAN, MBUS, RS485 etc but can be put there for the people that need it. This would be useful for, e.g., heat pump monitoring. The only annoyance of using the expander header is that there isnāt a serial module available, so it would have to be bit banged. This may be addressed with another product in the futureā¦
Sure - the extra cost is very low these days, itās a 20% premium over the 2 layer board with the same tolerances and finish. This board has a lot of signals going into a 64 pin UFN package, and 4 layers makes the routing possible. All analog signals can be next to continuous ground and power can be put into planes. Additionally, the USB signals can be routed as a proper differential pair with reasonable widths (c.f. a 2 layer board!) as well as the (very short) 50R traces for the SPI and antenna signals.
Not at all, thanks for the queries and questioning the assumptions beneath them. Happy to answer anything else or include useful suggestions.
Thanks for update. Willing to order some boards from Jlcpcb. for testing based on your GitHub files. Is there any necessary changes I have to made. Perhaps you could kindly let me know. Please. So far I have noticed you have mentioned issue with on board led - so there is no indication from that point.
Thanks in advance.
Hi Rob - nice! Ok, a few things to note:
pi3-v1.0 branch. That corrects the footprint error for the LED, along with minor changes../generate.py and use the output in the folder ./output-48cee2e.J10 for programming.Send me a message if you need any other pointers. Itās only been me building it so far, and things should be documented but itās always easy to forget āassumedā knowledge. Are you ok building the firmware as well?