power meter IC selection

[gkasprow]

I did some survey around power metering ICs.
There are low cost power meter modules from China: PZEM-004 but their accuracy and safety is an issue. I decided to use very low cost chip from Microchip that has already support for many platforms.
There are also low cost modules available. Low cost means that it is far less expensive than Microchip devkit.
So I connected things together and started playing with software.
I didn't wanted to use software UART. Anyway, for some strange reason it didn't wanted to work.
So I modified the code to work with hardware serial port on Arduino Leonardo.
Another thing I discovered with this code is that it needs plenty of iterations to calibrate to board. First you have to flash CPU, connect load, ampmpeter, voltmeter, power meter, read values, enter into the code, flash CPU again, then read checksum that the ATM90E26 produces, put it into the code and flash again. Definitely not the way to go.

Current measurement is done using split-core transformer. I ordered different models from Aliexpress and tested one of them so far.
Voltage measurement was done via two 1M resistors connected between L and VIP, N and VIN. In this way I don't need an isolation and module can be on the safe side because leakage current is very low, far below 1mA.
I added Cayenne code and get some measurements via web application. So this approach seems to work well.

[gkasprow]

I did first attempt to build single channel of energy meter in a way that it is scalable and can be extended to any number of channels.
So I assumed that it must be a multi-drop connectivity and the protocol.
There are a few of them : I2C, CAN, RS485, 3-state LVCMOS
For connectivity within same module simple I2C or LVCMOS would do the job.
Protocol could be I2C or MODBUS.
The idea is to daisy-chain several modules.
But we want to have identical modules, with no address switches or jumpers.
The best would be geographical addressing.
And the number of connector pins need to be minimised - preferably 4 or so.
So the idea is to use autoconfiguration during boot time. Something that resembles USB re-numeration.
This could work in following way:

• the master talks to the first device on the bus (address 0). Other devices are disconnected.
• the unit 1 responds (with address 0), master changes its address to 1
• then the unit1 connects (extends) the bus to the unit 2 using analogue switch.
• the unit 1 talks to the second device on the bus (address 0). Units 3,4...n are disconnected
• then the unit 2 responds, master changes its address to 2
• the process continues until anything responds on address 0.
  During re-numeration process, the master reads ID table of device that contains descriptor of what the unit does.
  In this way after the process is done, we have a set of devices with unique addresses.

This approach works if each device is intelligent, it means that it contains some CPU.
There are very low cost, powerful Cortex M0 MCUs. Their price is below 1$
The MCU is powerful, i.e. has 4kB of SRAM, 16KB of FLASH and 2 serial interfaces that is sufficient for us.

I can imagine that such measurement channels could be placed side by side to form flat structure. This is critical in case of the fuse-boxes.
Below is the first estimate. The board size is 53.5 x 14mm


The board contains connectors for L, N and current transducer, energy metering IC, MCU, crystal, analogue switch and bus connectors.
One can imagine such board extending internal bus to external interface, i.e. MODBUS RTU