Ive got a pretty full panel with an 8 circuit sub panel, but get a pretty good picture with 14 channels. 2 for mains. I run both legs of 240 circuits (US 120/240 single phase) through a single CT, so those are twofers. I have some circuits aggregated through single CTs. Finally, I subtract all of the measured circuits from main to produce a feed for “miscellaneous”. It’s amazing how steady that feed is as everything else changes. That indicates to me that the individual circuit measurements and mains are pretty accurate.
Bob,
I’m curious. How did you manage to get both legs in one CT?
(I’m thinking 2/0 copper or 4/0 Aluminum)
Bill,
Not talking about the mains, those are individually measured. That said, there are plenty of really big CTs out there, and if you were looking to do that, you can get some pretty good deals on them in quantity-one used on eBay.
I’m talking about my branch circuits, which are no larger than #10 copper. They fit easily through the SCT013s, and even the slightly smaller CR-Magnetics 3110 which I consider to be a far superior device yet available in modest quantities for about the same price. There are a bunch of Dent CTHMC-100’s available these days which have really low phase shift. They are voltage type and put out 333mv/100A and have a huge (25mm) round opening with tabs to wire tie to the conductor.
I’ve got three #12 circuits on one CT. To me, the larger issue with multiple circuits isn’t what will fit, it’s what is the voltage reference. Even with a best-practices installation, the neutral has some amount of resistance with respect to the ground on the center tap of the street transformer. As a result, unequal 120 loading on the two legs causes the voltage to sway from side to side by as much as a few volts. That is to say that the 240 stays pretty constant and when the voltage of one side goes down, the other side goes up. So if you are using one side as a voltage reference for power on the other, you will not be as accurate as if you had two voltage references. (IotaWatt can do that). The magnitude of the collective error is a function of the difference between the load on each side of the panel.
All that said, the goal isn’t revenue grade. 2% is more than adequate to inform the average user about the nature and extent of their consumption.
Gotcha.
You used the term “legs” which is the term usually used by electricians and others in the trade to refer to what you called mains, hence the question. (I’ve also heard electricians refer to them as “feeders”)
Here’s an update on the general ESP8266 stability issue. After a lot of burn in and monitoring I have some pretty solid data, but no absolute solutions:
-
The power issue is real. One particular nodeMCU that I had connected to a USB hub with a good 3A 5V supply would reliably reset whenever I woke up my desktop computer. There are a lot of other USB devices on the hub, but none that draw significant power. Nevertheless, I tried a 470uf capacitor on the voltage line which did nothing. More on this later.
-
When power is not an issue (symptom is reset for “external reset”) the other cause of reset is wdt reset. (Watchdog Timer Reset). I’ve researched this as much as possible but there doesn’t seem to be any definitive explanation of how the WDT timers work. Various docs speak of both a software wdt and a hardware wdt. I feed the dog very regularly (ESP.wdtFeed()) and I yield() to the higher power as often as possible. With the aid of a software trace in the RTC memory area, I’ve got it narrowed down to firing off when I re-enable interrupts after sampling a power cycle. So after interrupts have been disabled for 16-24ms. (That would be a bit longer in a 50Hz world). It occurs on average after about 12 hours of running.
-
Lastly, in my effort to go to a more modular commodity package, I have started using the Adafruit 8266 Huzzah Feather and all of the power problems on the USB hub have disappeared. Also, with limited experience to date, it appears the wdt issue may be better. The device has a different ESP, labeled ESP-12S, than the nodeMCUs. It bears the Ai logo of Ai-Thinker that is on my nodeMCU boards, but the ESP-12s designation is not on the others. I believe they are either 12E or 12F. The 12S is touted as having better wifi, but it is a new layout and maybe other problems were fixed. I’m optimistic.
All that said, one side benefit of the Adafruit Feather is that it uploads wicked fast at 921,600 baud. (8x over the nodeMCU 115,200).
At this point I think I can say that stability is no longer a big issue. Even if the device reboots a couple of times per day, it only takes about 15 seconds and with the real time clock, context is preserved and data integrity is sound. The data upload to internet service is streamed from the SDcard log, so even with WiFi interruption, the data is uploaded when service is restored.
1 Like
Thanks for the extensive update. My port of your project to Eagle stalled when they introduced the subscription model. Now I am redoing the port to KiCAD. To save boardspace and make soldering easier I am considering using RJ45’s to plug in multiple CT’s. What are your thoughts on this ? The Adafruit feather is where I went with my new designs as well, it uses a much more reliable ESP8266 module than random NodeMCU’s.
Here is my Adafruit feather based design: ATM90E26 and ESP8266 Energy Monitoring - Hackster.io . The other plus side is you can try out other Non-esp MCU’s as well.
Keep up the great work and let me know if you want some help or even funding. Happy to sponsor you a batch of PCB’s, I am looking to adapt your firmware for my ATM90 based projects.
I’ve done everything with the free version. I can see where a EE might need more, but as you can see, my stuff is simple.
As I’ve said, the stereo jacks are handy. That said, screw terminals are more economical and can save space. I haven’t crimped a cat5 cable to an RJ45 in a long time, but my recollection is that it’s a PIA. Last comment on this is that the difference in board space between any of these options is relatively small and doesn’t amount to any real savings, at least with the board house I’m using. How can you beat $1/board?
I have to say that I’ve got a lot of running hours with multiple devices and I really can’t see any difference between the two with respect to reliability. I’ve burned in at least a dozen nodeMCU 1.0 boards and they all worked great. Been using the feathers for the last month or so and same story. Only differences I can find are:
- feather costs more (quite a bit actually, especially when you factor in Adafruits shipping costs).
- Feather has a faster USB capability, but I can’t seem to reliably run them faster than 115,200 which is the same as the nodeMCUs
I had reliability issues for a long time, resolved before I started using the feathers. My problems were with wdt resets. I narrowed it down to what I am convinced is a problem in the interrupt handling of the Arduino/ESP8266 core. I stopped doing what I was doing to cause it and now the devices - either the nodeMCU or feather - run indefinitely without exceptions.
Well… that’s an ambitious project. Not familiar with that ATM90E26 chip, but did look at what I think is a similar intel chip briefly. My recollection is that it handled one channel, which didn’t fit my project. Also, as a hardware neophyte, that’s a long way past my capabilities at this point. So I’m very impressed with your project but have to confess that I don’t understand all of it.
Will keep you in mind as a technical resource. Thanks for the interest.
At 5x5cm $1 board there is no more gains by making it smaller. The only
gains you might get would be from going fancy to ENIG etc. while keeping
board space low. In my case trying to fit into the DIN rail M2 form factor.
Bob, it’s been interesting reading your journey to get to this point. The system looks great, although a lot of it is over my head. It might be something I’ll be able to use for stage two of a project I have just started (stage one is just reading the pulse outputs of the existing metering infrastructure).
One thing I can help with, although it’s not much is that Global Village were a big manufacturer of modems back in the 90’s, especially for Macs, and I’d guess that is where your plug pack came from.