Hello all.
I need a little advice. I am about to get a major service on my Samsung 6th Gen 8 kW heat pump that I’ve been monitoring using an EmonPi but without a heat meter. It’s taken a bit of fiddling with the “water law” but the pump is doing what it should but I suspect it could be optimised. For one thing, the water quality is appalling (it was retrofitted onto a filthy system) so there’s an opportunity to do things right.
My question: It’s been suggested that I use Homely to manage the pump so that I can take advantage of the Octopus Agile tariff (if Octobus ever come and put in a smart meter). Is that the best way to automate the Samsung? I’m modestly techie so need a fairly simple system, but I’m a little leery of shipping out my energy data to Homely or anyone else, if I can help it. Any thoughts?
The major service I’m about to do will include:
Draining
Checking the joints because I recently found a joint with no O-ring, ffs
Power flush with citrate
Refil with as soft or demin water as I can find.
Installation of a frost valve (so that I can increase heat capacity by getting rid of the glycol)
Putting in a heat meter (what should I buy? It’s not clear where it will sit in the system so I may have to leave it to the service people to get but I as thinking of the Axioma QP2.5)
Sorry that you’ve had no replies to your questions.
I’m not so sure that I am best placed to offer advice, as I don’t have experience of the monitoring products on sale through OEM, or of Homely, or even the various electricity tariffs. But FWIW I have found that:
The Samsung remote controller (the MIM-E03*N) displays – as opposed to monitors – all the info you need to optimise your setup (i.e. instantaneous heat generation and power consumption, thus CoP). Setting Water Law – FSV #2011/2 and #2021/2 (or #2031/2) – such that your house is just warm enough over a range of ambient temperatures, should get you 90% optimised without the need for a separate heat meter. I did install one (a Sontex because I have glycol, which cost me about £500 including installation cost/VAT), because I have a buffer tank and wanted to know what heat was actually entering the house rather than leaving the heat pump. But using it I found that 1) external heat losses were virtually negligible (about 150W on a really cold day), 2) the Samsung LWT and RWT temperature instruments and the circulating flow instrument give reasonably accurate results (compared with the heat meter) if averaged over time. and 3) the Samsung-published performance data for my model (HTQ, not Gen 6) were fairly consistent with my readings.
You don’t say whether you have a battery, but I suspect that you would need one to make best use of off-peak tariffs. I have no idea how Octopus Agile works in practice, but I have considered Octopus Cosy (3 low-tariff periods + 1 high-tariff period each day). By charging the battery (as well as running the heat pump flat out) during the low tariff periods you can ride through the high tariff period from the battery and making use of the house thermal inertia, even if you cook in the evening with electricity when the high-tariff period occurs. You wouldn’t need Homely for this – you could just set your roomstat programmer to switch off the heat pump during the high-tariff period. Also, a Homely for Samsung would cost about £360 (Midsummer Wholesale price, plus any installation cost) and in my case my electricity cost is so low (< £700 per year) that I doubt I’d get a financial payback, not to mention all the faff of setting the thing up. Long and short, so far I’ve decided to stick with an Octopus all-day tariff.
If you do want some heat pump monitoring, without risking putting your data onto the Internet (and thus potentially being vulnerable to hacking) you can do this at low cost (under £100 all-in) by installing an RS485-to-Ethernet adapter on your MIM, and recording the controller inputs/outputs (both Outdoor and Indoor Units) on your computer for later analysis. If you are interested in this I can give you details (recommended hardware, step-by-step installation instructions) on request. The Excel spreadsheet that the Samsung software generates is really useful to understand what is going on (cause-and-effect analysis), and allows you to change your FSVs from your computer without touching the remote controller.
Regarding glycol, have a look at Glycol: pros and cons. This offers some opinions (from experienced, if rather biased, community members). Personally, I’m sticking with glycol, mainly because if the frost valve operated (or malfunctioned), I’d lose my heating completely until I’d 1) figured out why the heat pump tripped on apparent low circulation flow, and 2) I’d recharged the circuit (with potential air-locking problems) or installed a replacement frost valve. (For me the big downside of glycol is that I can’t modify the circuit, such as a replacement radiator, without recovering and re-injecting the glycol, which is a pain. The changes in physical properties between water and glycol – density, viscosity, thermal conductivity – are fairly minor, and I doubt that I would spot the extra circulating pump power consumption or reduced heat pump condenser performance.)
This years Fernox water quality test revealed almost all the inhibitor had been used up, but the glycol concentration was still fine. A couple of bottles of inhibitor added and inhibition restored.
I have slightly modified my views on the type of glycol to use. At the moment I am using Alphi-11 which is propylene glycol based. Ethylene glycol has better thermal properties but is toxic. However, there are now products that use ethylene glycol and have an additive that circumvents the toxicity, Thermox DTX for instance, even slightly cheaper than Alphi-11. I will probably change to this when the Alphi-11 expires.
Hi @SellaTheChemist,
I had my Samsung Gen 7 R290 installed back in November with Homely. I wasn’t happy with the performance so ultimately turned it off in the end and haven’t gone back.
I did a write up on it here. If you have any further questions please feel free to ask.
I’m now switching between Agile and Go, depending on the price of Agile and what electrical devices I need to use that day. Agile has improve again now we’re heading into summer, so gets used more and more.
Homely is very much a fit and forget solution. It uses a Modbus interface to the outside unit (this is an extra box, I think you have to buy it separately). There is very little adjustment or information available to the user, very much of the “trust me, I am a doctor” school of thought. I guess it is intended for the less technically minded who want something that just works. Reviews in the interweb are mixed, some people love it, others have stopped using it.
Another option is as Sarah suggests. Here you are talking to the unit using the Samsung NASA protocol over RS485.
If you are technically minded, you could buy the Samsung Modbus interface that Homely uses and use Home Assistant to automate it.
Buy the Samsung Modbus interface and roll your own. This is what I have done (in Java, other languages are available). Schedules based on time and Octopus Agile prices, weather compensation, data logging, the normal stuff. So Homely look-alike without all the AI.
For the really ambitious you could use the RS485<->USB converter that Sarah mentioned and write your own NASA protocol driver. The NASA protocol has been pretty well reverse engineered on GitHub, so all the information is available.
I have been an Octopus customer from very early on in their history. This winter I moved from Agile to Cosy in February. I should have moved before, as Agile was not very good value for my usage pattern this winter. I am back on Agile now, it became good value again March/April time. I will probably spend winters on Cosy and summers on Agile in the future. So if Homely is restricted to the Agile tariff, it may not be the best value.
Octopus have an interesting feature when shifting tariffs. When you switch, the new tariff is active from previous midnight, not the next midnight as you might expect. So some people have developed scripts which they run at 11:00pm every day that calculate the days cost based on all the available tariffs. The cheapest is chosen and an automated tariff switch sent, so you always get the cheapest cost for the day.
Hi Andrea, i have a 16kw 6th gen Samsung and the best thing i have found is make sure your co trol is set so the compressor doesnt switch on and off, it wants to be set to room temp only on/off and make sure your all your rads trvs are fully open at all time. If you want to control a room just adjust the flow down a bit to that rad
Ensure your flow is as good as possible and leave on all the time during the cold months. Depending on building fabric if your like us and have solid walls half a metre thick then you can get away with going of at night but i dont bother i just have the controller set so i can plus or minus flow temp by 5 degrees amd put it at minus 5 overnight. Easy for me as controller is at the bottom of the stairs so its the firat thing i do either way.
I have no monotiring fitted but i do have an external electrical check meter on the heat pump and that tells me i use about 4200kwh per year to keep a 200 square metre 220 year old solid stone house at 20 degrees all year and hot water for 2 adults so thats all i need to know. Our epc says we require 25k kwh per year so must be doing something right
Your excellent listing of options highlighted for me the distinction between “automation” and “monitoring”.
The RS485-Ethernet adapter I mentioned above (costs between £25 and £50 depending on supplier), when used with the Samsung SNET-Pro2 software (free to download), essentially enables monitoring only. It displays - and optionally records in an Excel spreadsheet for subsequent analysis - controller inputs/outputs, such as refrigerant temperatures/pressures, outdoor temperature, compressor speed and current, EEV positions, circulating fluid temperatures and flow, DHW and room temperatures, and diverter valve positions. It is mainly Read Only (though you can Write FSV values manually). This option does require adapter configuration and virtual COM port creation (both are once-off procedures at initial set-up, using free-to-download software).
If this is all you want/need, you can also do this with an RS485-USB interface (costs from about £20, though the Samsung-branded version MIM-C02N - which does have additional functionality - costs nearly £200). This option requires you to run a 2-core cable between the controller and your computer (the above Ethernet option requires a cable - or a pair of power line adapters - to get the signal from the controller to your broadband router).
Neither of these options can be described as automation, which is what @SellaTheChemist was hinting at.
Maybe if you have time, Alun, you could flesh out the true automation options in your list, giving some pros and cons of each, along with details of the hardware/software required (maybe with rough purchase and any subscription costs) and the skills needed to implement them?
Well, I’ll try. The problem splits neatly into 2 parts.
Get the data into/out of the Samsung system into a PC/RaspberryPi/etc.
Use the data in the PC/etc. to take control of the Samsung system.
I will do this in a few posts otherwise it will just get humungous. Caveat is that I hate Microsoft and Windows, so the control part will be heavily biased towards Linux based solutions.
Part 1 - Get the data into/out of the Samsung system
There are once again 2 obvious ways to do this
Using the Samsung MIM-B19N Modbus interface. This is freely available in the UK from the likes of Midsummer/City Plumbing/etc. at 150GBP-ish. However, you can get it from Europe cheaper, for instance
If you look at the documentation that comes with this it looks like only a limited subset of registers are available over Modbus, but this (and many others) lists how to get more
The MIM-B19N connects to F1/F2 (which is the internal Samsung NASA bus). The installation instructions tell you to put it in the outside unit, but of course, it is an RS485 bus so you can equally connect it (more conveniently) at the indoor unit F1/F2 connections, doh!
Mine has been running reliably for years now. The only trouble it has given was locking up after a few months. I was polling Modbus too fast (10ms), reducing to 50ms and it has been reliable ever since.
So you now have a Modbus RTU interface (RS485). You now need a USB<->RS485 convertor to get it into the PC. You could use one of the cheap FTDI convertors
but I prefer to galvanically isolate systems from each other where possible, so this is a better solution
A bit of 2/3 core wiring, 120R Modbus termination resistor and you’re done.
Directly to the F1/F2 Samsung NASA bus. This is extensively discussed in
which had a description of the NASA protocol and all the Samsung registers, but the content seems to have disappeared, maybe Samsung have got annoyed that their proprietary information was being published, or maybe it is just temporarily broken, I don’t know. Anyway a trawl of Github should produce the same information.
So you can connect the USB<->RS485 convertors listed above (although in this case I would definitely advise the galvanically isolated one) and look at
to implement a protocol convertor.
RS485 is a physical layer, not a protocol. It does not define the structure of messages on the bus. Modbus RTU defines the structure of the messages, and NASA defines a different structure. Modbus is a master/slave protocol, there is one master that polls (requests data) and a slave responds (response data). So NASA could work in a similar way, or they could have defined some multi-master protocol to allow different masters to poll for data/send commands. If you just want to listen to messages on the bus, decode them and display, there is no problem. However, if you want to set register values, then you need to understand the multi-master protocol (if it exists) so that you can send write commands at the appropriate time. You could probably figure it out looking at a packet trace from S-NET …
This route is not for the faint hearted, if you have not got at least some software development experience, I would spend the 80EUR, it will be much less grief.
I took a copy of this page before the data disappeared. Here it is, converted to MS Word: Samsung NASA Codes.docx (77.1 KB)
Obviously this will slowly become out of date, but hopefully better than nothing…
Reliability. The system should run for at least a year without crashing, preferably not requiring updates in that time.
No cloud dependence. It should not rely on any cloud based services for any of the core functionality. It may use cloud based services for things like data display, tariff information, solar forecasting, etc., but should degrade gracefully if they fail, i.e. not lock-up or corrupt.
This also splits neatly into 2 parts
The infrastructure. The hardware/OS you are going to execute your control system on
The control system. The software you will run on your infrastructure to implement the control.
The infrastructure
A RaspberryPi/Pi-alike. Perfectly fine but I worry about SD and USB flash drive reliability. You could add external SSD drives but then you are getting towards a mini-pc.
A normal mini-PC. Perfectly good second hand PCs are available on Ebay (Dell Wyse, etc., 40->100GBP). You could load the Linux flavour of your choice, but I am going to suggest what I use, Proxmox (free)
This has several advantages
(local) web based interface. No horrible, arcane configuration files.
ZFS. You can choose to install ZFS during Proxmox installation so can improve your systems availability.
You can install VMs (Virtual Machines) and LXCs (Linux Containers, lightweight VMs). There are many pre-built VMs or scripts to install relevant LXCs, such as Mosquitto (MQTT), Node-Red, Home Assistant.
You can configure backup schedules for each VM or LXC. Never lose stuff again.
Now, you can do all this with any Linux distribution, but it involves installing and configuring individual tools, with Proxmox it is all in one place. Why bother with the complication of VMs and LXCs? You could just lob it all on a RaspberryPi using HA (Home Assistant). I think this is down to leaving your options open. If you decide one of the components is not for you, say HA, then you do not have to start from scratch, just replace that one component.
This should all be do-able by anyone with experience of using and installing software on a PC, no software skills required.
The Control System
Just lob HAOS on a RaspberryPi. This describes such an approach
You need to add the HA-Modbus integration
and configure it to the Samsung register map by writing a bunch of YAML. Surprisingly, I would have thought someone would have already published this in Github, but I can’t find anything. So a tedious but not complicated editing session required. Then write your control strategy in HA. There are other useful integrations you can add.
and Node-RED via HACS. Node-RED allows you to create control strategies visually by connecting together function blocks in a graphical UI. It is capable of generating much more sophisticated control strategies than HA (or maybe you prefer the UI compared to writing gibberish YAML).
Mosquito MQTT broker
An MQTT broker is like a real time database. You can “publish” a particular data “topic” to the broker (say Room Temperature) and subscribers can “subscribe” to the “topic” so that when it changes (is published to) the subscribers are sent a message with the new data. So you might want to send some of your HA generated data to a “proper” database for long term storage and to some sort of UI. The database and UI can subscribe to the topics of interest and automatically get updated when the data changes.
If it turns out you hate HA, then you have to start all over again. I use HA only for the UI bit of my system, so I am no expert. I have found it poorly documented ('though improving), and if there is no standard card/automation for what you want to do, you have to start descending into YAML. So you probably need at least a little software experience. There is an active forum to seek help on. I have found ChatGPT very useful, tell it what you want to do and it will spit out some YAML which gives you a good basis to start from.
The Proxmox approach.
Install a HAOS VM
Install Node-RED LXC
Install Mosquitto MQTT broker LXC
You can use the HA-Modbus integration as above, then publish any stuff for consumption elsewhere from HA to the MQTT broker. Or you could use Node-RED to talk to Modbus and publish to MQTT. Ideally there would be an existing simple executable that did Modbus2Mqtt. There seem to be some on Github, but that could well be a new rabbit hole to go down. Then write your control software in HA or Node-RED or a combination. To get this far you really haven’t written any (or much) software, just configured stuff.
If you decide that you hate HA and Node-RED, then write your own control software. Install another VM running the OS of your choice and away you go. I use Ubuntu with Netbeans IDE and write code in Java, the choice is huge.
Anybody can write code. When I started many decades ago there were ex-carpenters, history graduates, etc. It was rare to find a computer science graduate writing software. So what I should have said is “anybody can write bad code”. If this is your first foray into writing code, take a month to read all you can find on the interweb about software engineering and absorb it, then start. Otherwise, five years down the road you will be trying to debug an amorphous blob you no longer understand.
Set up a version control system. Git is the most popular, but unless it is a multi-programmer project I would recommend Subversion. You are much less likely to shoot yourself in the foot with Subversion. You only have to lose a weeks work once before you set one up anyway. And the ability to rewind to previous versions is invaluable when trying to find obscure bugs.
Write test harnesses for each module (you will be writing modular code, because you read about it in the software engineering month, right?). It will take you as long as writing the code, maybe longer. Test for expected outputs and response to invalid or unexpected data. If you find a bug later in your development that your test harness didn’t find, add a new test to check for the condition.
And make sure that if it all goes horribly wrong you can restore normality fairly quickly, either by graceful degradation or replacement, so that when you unexpectedly pop-your-clogs the poor person left is not having to deal with your departure as well as a house that no longer works. I have a switch that when thrown replaces my all-singing, all-dancing control system with an ordinary Honeywell system that anyone can understand.
If you don’t want to go down the coding route, then HavenWise sounds like a good alternative to Homely, but it does involve sharing your monitoring data. You will also need to install the WiFi kit, which you can do yourself as opposed to the Homely kit which costs around £500 to have installed.
You’ve missed a most important point just after here “[writing tests] will take you as long as writing the code, maybe longer.” This is “Now document what it was intended to do, what it does and what it doesn’t do, what it shouldn’t do, and anything that didn’t work and why. If you do this properly, it will almost certainly take longer than everything else, but save a huge amount of time when you need to revisit it.” And when you do revisit it and can’t work out how it ever worked, either you didn’t document it properly or it was too complicated to begin with.
That’s an interesting point, and provides a third way in addition to Modbus and NASA to control the heatpump. Looks like you have to install a Samsung MIM-H04EN (150GBP-ish) for a Gen 6 to get Wi-Fi access.
Then you can control it via Samsung Smartthings, which is how HavenWise accesses the heatpump. It installs in much the same way as the Modbus interface, F1/F2 and power.
I would imagine it is locked down pretty tightly from a security point of view so it would be difficult to use it for local control, and it suffers from the same drawbacks as Homely, cloud based and subject to arbitrary disconnection (Homely/HavenWise get sold to some big player, then change of focus by the big player, they decide to cease support. It has just happened, where Google bought Nest and have now announced they are dumping support for 1st and 2nd generation Nest Thermostats).
Still, an interesting option if you just want something that works without tinkering.
Let me know if I can help in any way. I think I went down the rabbit hole of the EHS F1/F2 or F3/F4 protocols
If you have specific insight request about particular intel you want from or to order to pass to the EHS, I’ll gladly help.
The only side I’ve not traced (just yet!) is the test mode to manipulate pumps and valves manually through the NASA protocol. Hopefully it’s not required for an almost perfect drive of the EHS through home assistant.
Cheers
