Practical considerations for solar diverter

[dnb]

The subject of solar diverter design has come up a few times on here. I know I could buy one, but the ones that work properly cost rather too much for what they are, and the cheap ones don't do what I need... So I am making my own - it's a fun job that keeps my hand in with PCB design, layout, coding etc now that I don't get to do this any more at work. I have a design based on this one but with a "few" changes to add robustness, simplicity and a better display.

 

I have a few considerations to review, and this site is pretty good for finding these things out.

 

What do we think is the smallest input current we need to differentiate from zero? Is it smaller than 50mA, or is this sufficent resolution?

How accurate does the microcontroller clock need to be? At 1% it gives a 0.5Hz accuracy for mains frequency measurement, and I think this needs to be better - 50ppm seems more reasonable - this gets to near to a perfect mains frequency measurement.

Are 3 high current switched outputs enough? It's fine for my house plans now, but what else might I need?

 

[ProDave]

Interesting topic. I had not seen that design before.  I built my own completely from scratch using much simpler hardware.

 

There's a lot to take in to understand how that design works so I doubt I can usefully answer any questions about the operation of it.

 

My own design has 2 outputs, one for the SSR driving the immersion heater, and the other a wireless switched relay to turn on an electric convection heater.

 

 

[CC45]

blown my mind completely.  No idea how you know about all this..

 

Interesting though

[dnb]

Hopefully got my concerns resolved... Yes, resolution down to 100mA is useful, so it is included, and yes, accurate frequency measurement is needed.

 

So, can anyone tell me what this is going to be?

 

Board dimensions are 80mm x 95mm. I think it packaged up nicely.

Next steps are to double check the netlist and make sure the ground plane doesn't have too many holes poked in it. Then I can think about getting a prototype made assuming I can remember how to hand solder 0.8mm pitch TQFPs!

 

Typically for an engineer, spec creep happened. So in addition to the usual energy metering it has:

Current monitoring for all loads for the purposes of monitoring use and fault detection

RS485 interface for control and reporting

20x4 LCD character display

Lots of filtering and EMI suppression - sometimes I think half the components on a board don't do anything when all is working right...

Capacity for 5 control switches (may not need to use all of them in the code)

[ProDave]

You have put a lot of work into that. Clearly going for the 3 phase version.  I look forward so seeing the results.

[dnb]

Thanks. I used to this sort of thing to pay the bills so I hope I am not too rusty...

That one is single phase. It just has 3 current transformers for monitoring the loads. All a bit unnecessary for the case where everything works but it is useful for fault indication and various monitoring schemes I am planning. The main current clamp is off board.

[joth]

So I just managed to pickup a CARLO GAVAZZI RJ1P23I50E (data sheet) SSR for £25 on ebay so now idly thinking about how I'll use it.

My default thought is to lash something up on an ESP32 with ESPHome probably using a PZEM-004T V3 current clamp as the source, this way I automatically get monitoring/graphing in Home Assistant and it's easy to send on/off override commands to it from wherever.  My tank will have 4.5kW immersion, with 8kW PV, so slightly non standard hence the thought to DIY it. 

[MikeSharp01]

On 10/06/2020 at 13:35, dnb said:

I can remember how to hand solder 0.8mm pitch TQFPs!

I have to do it with a microscope these days.

 

Why didn't you choose an up spec micro controller module with all the interfaces and make the diverter controller portion shield pluggable? I guess processor speed might be an issue but there are some very fast boards out there. 

[dnb]

1 hour ago, MikeSharp01 said:

Why didn't you choose an up spec micro controller module with all the interfaces and make the diverter controller portion shield pluggable?

 

"All the interfaces" and "Shield pluggable" are the main reasons. I prefer, at least for simple projects to keep things properly simple. I don't want to build in too many complex toolchains, opportunites for obsolescence or have to worry about somebody else's PCB design and track routing. On the whole, a shield based design drives me towards a more convoluted power supply and doesn't save all that many components, although the easy USB and possibly TCP interfaces would be nice to have. Processor speed isn't an issue because my design needs nothing to happen particularly quickly by modern standards.

 

2 hours ago, MikeSharp01 said:

I have to do it with a microscope these days.

I can still get away with those optivisor things but it's getting harder.

[MikeSharp01]

9 hours ago, dnb said:

the easy USB and possibly TCP interfaces

I guess that is the bit I was wondering about. It feels much simpler to integrate with the house IoT if it is one of the things out of the box rather than via the house RS485 bus. I do agree on the tool chains issue, I now have tools, and special laptops, running tools going back to 1992 for some of the stuff here, EG the clock in our bell tower, no bell sadly, uses a PCB and micro controller board I designed in 1990 and only now am I thinking is it time to switch over to an ESP based module.

 

[dnb]

9 hours ago, MikeSharp01 said:

I now have tools, and special laptops, running tools going back to 1992 for some of the stuff here,

You and me both!

 

9 hours ago, MikeSharp01 said:

simpler to integrate with the house IoT

Yes it might be. I do have a few security concerns about everything sitting on a TCP based network, made up from random bits of I don't understand that won't get security updates... The well defined RS485 that has very little (but enough for the job) functionality appeals greatly!

[MikeSharp01]

13 hours ago, dnb said:

I do have a few security concerns about everything sitting on a TCP based network

I think we all do, but as it is the only current platform and more secure IoT systems will more than likely build off it one is left with the challenge of creating security systems. We have discussed this here before but never really got to the very bottom of it. I am something of a fan of secure mesh networks of IoT but the security stuff continues. Any system connected to the WWW is susceptible in the end and I am definitely not a fan of the house protocols around the smart speaker systems and all their gadgets which hoover data and have too many security loop holes. As a nation, probably the people of all nations, we need to get control of our data from these systems although big tech would argue that we have nothing to fear - but if that were true....

[joth]

Increasingly my attitude with ESPHome is to try and build autonomous devices that do not need the Home Assistant server running to operate (and ideally not even the WiFi local network) and just use the Home Assistant TCP smarts as a convenient development and monitoring platform. Being able to adjust code and reflash it insitu, and stream debug logs and monitoring graphs, all from a web-browser, suits me very well.

My ESPhome devices are all on my "IoS" VLAN with no access to the public internet and limited connectivity to my main VLAN (basically, just to the Home assistant server) so the attack surface is very reduced, if not removed all together.

 

For an example of autonomous behviour, I can have a single D1-mini both read the current export value from the main meter (modbus interface) and control the SSR to adjust the immersion load (simple analog output). This needs to Wifi or IP connection to work (unless I split the functions over 2 boards via UDP).  The IP network would just be used to control "boost" or "eco" mode on the immersion (i.e. override ON or OFF for a period of time).

[dnb]

11 hours ago, joth said:

My ESPhome devices are all on my "IoS" VLAN with no access to the public internet and limited connectivity to my main VLAN (basically, just to the Home assistant server) so the attack surface is very reduced, if not removed all together

This is why I'm looking at the RS485 network for the few things that need to talk to each other. (Most of my ideas don't require much comms - they have access to the data they need and should "just work" without too much intervention) I don't believe the IOT as it is currently being developed is particularly useful to consumers. It is however hugely useful to large corporations etc, and I see no reason to help them.

 

I was looking at a washing machine that could be configured to start when there was excess solar power. The machine seemed to think that making a connection to a server belonging to the manufacturer to supply serial numbers and usage stats was required for this function.

[MikeSharp01]

5 hours ago, dnb said:

The machine seemed to think that making a connection to a server belonging to the manufacturer to supply serial numbers and usage stats was required for this function.

This is where they make their mistake I feel. All these systems should be open and not demand a central server connection. I am happy to share my data but on my terms. Here the RS485 interface is indeed worthwhile but I have a concern that we won't stop this IoT bulldozer. However many industrial users of IoT have exactly the same concerns as us diminutive users so I have a small hope while anybody who thinks you can buy a car that does not tell the manufacturer exactly where you are and what you are doing has got another think coming.