I have used the N300 micro travel router from TP-Link in the past as a wireless client for an Ethernet device (to do the opposite of what you want). I used it at the goat house for many years and never had a issue with it it. It uses about a watt or two at the most. It is not POE, but uses USB 5V. I see they are still available, but almost double what I paid for them a decade ago (now about $30).
Some of the devices claim to run OpenWRT, which is very flexible, but you just need the very basic functionality of an access point (probably don’t want it in router mode). You will have two IP addresses to manage (one for AP and one for IotaWatt) but once set up, you should not need to touch the AP.
The ADC on the esp8266 is so bad that even if the one on the esp32 was 10X as good, it still would be mediocre at best. The ADC is critical to the accuracy of a measuring device. It pays to have a really good one if you care about accuracy at all.
I am not sure why people think using an RPI is a good thing for just about anything. They are decent for their original purpose (cheap replacement for a desktop computer, suitable for learning about computers without needing to spend hundreds of dollars).
To use an RPI as a generic AP seems like a waste of power. The RPI will use at least twice as much power and be harder to setup and need to be updated more frequently. If you have one in a box gathering dust, it might be cheaper in the short run, but not likely in the long run and certainly not if you include a value to your time setting it up. It also won’t do a better job (of just being an AP). Now, if you need it do more, perhaps. Even there, I don’t think it earns its keep based on power consumption. I use an old laptop as a virtualization host for VMs and/or containers. It does use more power (mine idles at 3-4W with the screen off, but can use almost 10X that if it needs to) than a single RPI, but not as much as the several it would take to host all the services I am running and I get much better performance and storage and reliability out of it.
If I may suggest adding two things for the ESP32 platform (I would like it on the actual but its already UL and some changes are forbidden by UL):
A header, port, connector to be able to connect a remote LED (Useful when doing an installation inside a box)
Using the ESP32 that has the IPEX antenna port. Also a convenience when the installation is made inside a metal box. I know having two versions would be painful in terms of inventory, but would be really useful for people who do not want to bother with an external antenna.
Using a dedicated 3.3V reference for the ADCs instead of the LDO. Although the LDO is probably good enough, using an VREF helps achieving the ADC specs, reducing noise, holding calibration for longer…
I am using an ESP32 with an IPEX connector, but it also has the internal antenna.
My current design does have a separate LDO off the 5V to provide the ADC Vref and Bias. I spent considerable time trying to make it lower noise. I’m not an EE and really don’t have the skill to make it optimal, but empirically got the noise down to the limits of my oscilloscope to measure.
The ESP8266 is way accurate enough without a separate Vref, but there are other advantages to doing this. With this design it’s possible to use any Vref voltage, so for instance I can use 1.5V and reduce the burden resistors to 10 which should make cheaper low capacity CTs like the SCT006 operate to their rated limits. Using 1V without burden resistors would allow use of 333mV CTs.
The ADC you are using is more capable than your oscilloscope if you use it in a fixed range. But if you got the bottom range of the scope then it should be good…
Usually one selects Vref so that the ADC has the best performance and we do signal conditioning to get whatever to the ADC range. But your setup is working ok so it is all fine I guess…
There are really good LDO’s in terms of low noise and OK in terms of accuracy. You could also use a Voltage Reference IC to go to lower voltage. Or you could use a reference with a pair of good resistors and an op-amp to buffer into Vref.
I have never used CTs in my designs, so I am a bit confused about them. I might experiment some with them… but I think my confusion comes from the fact that the datasheet specifies an output voltage of 333mV and then it talks about different burden resistor values depending on the current (I would guess to get the best performance)… but the CT itself states that it is a 50mA output CT… So the fact that I see 333mV at full scale, 50mA at full scale and different burden resistors, has me wondering…
One more note on the ADCs… In the datasheet, there is no mention for Vref != 5V or 2.5V… and all the typ graphs for linearity are when VDD = VREF = 5V or 2.5V… Usually not a problem and things behave just as you think they should behave… sometimes though…
@overeasy: I’ve read through this whole thread, and just wondering if you have an overall status update. It looks like you were initially focused on upgrading to ESP32 and adding built-in TLS support, and add-on unit, and were making good progress in that direction. Then there was a big long discussion on the viability of adding ethernet (and less likely PoE). Are you still exploring that direction, or are you back to your original design and just need to finish it. And then maybe maybe not get UL certification for it?
I’m personally most interested in the add-on, but I could make do with an additional base unit. But for work I’ve been evaluating IoTaWatt, and TLS (so we could upload data to a centralized service) could also be interesting. Ethernet and PoE would be less useful features for the majority of our installs.
That was a productive technical discussion that revealed a potential path. For a variety of reasons including cost, practicality and lack of significant demand I won’t be pursuing it. But as an open project, it is a useful discussion for others that may want to diverge from the version that I develop and support.
This effort has been in stasis as interest and demand for the V5 unit has completely consumed my efforts over 2021. There is a shopping list of remaining portions to be ported or redesigned for the ESP32. I think the most important is a redesign of the datalog to make it more general purpose and able to support other sensors and metrics that are possible with the ESP32.
Look for a restructuring of the IoTaWatt project over the next few months.
I would not expect Overeasy to add these features himself, it would be up to the userbase. Right now there is no room for additional features due to memory, pin and processor limitations.
Interesting… I just connect to my home network via VPN as I have many other devices, aside from my pair of IoTaWatt devices, that I like to be able to access when on the road.
I don’t see this as a requirement personally, although I can appreciate the desire for tighter security. Managing security certificates on a microcontroller does not appeal to me, though. Since these devices contain no information that I am personally worried about anyone on my home network seeing, this one is not much of a priority for me.
Care to expound on this a little more? What are the practical applications/benefits of Firebase?
Assuming this means additional CTs, I can definitely see the advantages of this. It would be nice to be able to monitor a few more of my circuits independently of one another.