I’m new to this intetesting subject and have read alot on this site and forum. It seams to be much talk about accuracy and sampling regarding to arduinos limitations. Wouldn’t one be better of using a chip that was made for this? I have looked at th 78M6631 from Maxim and that seems to solve most problems. The arduino can take care of the calculated stuff that the chip spits out. What would be the draw backs of using this method? Sure its an extra cost of maybe 10 bucks but I think thats cheap to get 0.5 % accuracy on a 3 phase system.
The biggest issue with using a device lke the this is safety. The 78M6631 requires a direct connection to the mains.
For safety, you MUST isolate your low voltage circuit from the mains supply. If you don’t, it is possible with certain fault conditions, to have lethal voltage where you normally expect a safe 3.3 or 5 Volts.
I’ve no idea what [quote=“Malmen, post:1, topic:856”]
maybe 10 bucks
[/quote]
means, but the list price does not reflect the true cost of incorporating a device such as that. The application data shows galvanic connections to the power circuit. As you are no doubt aware, and as Bill points out, we rely on type-approved isolating transformers to assure the safety of our users. Were OEM to adopt the direct connection philosophy shown, it would immediately mean that the necessary isolation would be required after the device, and it would no longer be possible for users to modify or customise their devices. It would also mean that the device must be officially tested and certified before being put on sale, which adds further to the costs.
Also, you must not expect the overall accuracy to be 0.5% - that appears to be the accuracy of the Maxim device itself and does not take into account the accuracy of voltage and current transformers, which could add nearly an order of magnitude to that figure.
My aim was more at the enthusiasts that talks about optimizing algoritms and building their own HW. I dont think they have a problem solving the safety bit. You could for eg use transformers instead of resistors on this chip too if you like to separate the mains from logics even if I would think you get better readings without them. Im 100% with you that you need to develop safe stuff for the end customer so I do not criticize your excellent work! Love what you have done in this area 
Hi Maimen,
A year ago, more or less, I spend some time triying to design something tçcheap that allowed me to read 3 Phases with precision, since my EmonTX based solution wasn’t enough accurate. After some search and investigation, I endended buying Eastron SDM-630M series power analyzers. I couldn’t compete with something sub 80€ (vat excluded) which had CE marking, display and was 3 Phases, and measured the usual electrical parameters.
Now I’m triying to make something based on a Teensy 3.2 and RPi 3, all software based, and I aim to do the same as high end power analyzers.
I asked price for the PQube power analyzer, and they gave me the price of 2850 USD… and that’s why I’m trying to replicate it. But I think that there are cheap 3 phase simple power analizers around, for starting to do experimets with PCBs and mains in my humble opinion 
Have a nice day 
I saw your project thread and that looks intetesting! Especially using the Teensy for the work load. I dont know exactly what you want kind of data you are looking for but take a look at this and see what you think of the firmware of this analyzer
I’m not familiar with that Maxim device, but generally, as Maimen says, the same isolation techniques used in the OEM designs can also be used in the energy IC world. The ICs don’t really know or care how you get the low voltage I and V signals to them, so long as they’re a genuine reflection of the actual signals you’re trying to measure. They pretty much all offer very generous phase compensation adjustment, so they are expecting your front-end sensors to introduce phase errors.
The application notes for these devices are mostly directed at professional engineering teams working at meter manufacturers, so knowledge of isolation issues is assumed. I can point you to plenty of Atmel application notes that directly connect AVRs to the mains too, but we don’t exclude AVR usage on the basis that they can be connected to the mains.
If we take it as a given that the home hobbyist should always start with VTs and CTs as their front end sensors, then you can simply focus your design choice on where it makes the most sense to do all the power maths (8-bit AVR with a single 10-bit ADC Vs a dedicated IC).