FGT-001 Fibaro Heat Controller, external sensor evaluation, part 1

[petergebruers 1,268]

I write this topic to help you decide if you want the "external sensor" (+ 15 EUR, +10 EUR bought as a bundle)... To avoid confusion: the FGT-001 does only displays temperature if you add the external sensor. Without the sensor, it displays "--". But that is not the main reason to buy one... You might have heard the TRVs cannot measure temperature accurately, because their sensors are "not in the right spot". But the question is... how bad is it? I encourage you to have a look at the setup, the graphs and, if possible, repeat the experiment. For me it is clear: I wanted the external sensor for the FGT even before I did the experiment, based on experience with older Danfoss valves. But this is not a general advise, it probably depends on a lot of factors.

 

Part 1 is this topic. I'll show you graphs of the temperature "in the room" and "near a valve". So you can decide for yourself. In "Part 1" I do not use my FGT yet! I compare the result of different temperature sensors in the room. I only want to know if they match.

 

Part 2 is not available yet, because the FGT has a bug that causes it to behave erratically. I'll publish results after running tests with the firmware release that will fix this. In "Part 2" I'll establish if the FGT + external sensor (a) does measure accurately (b) can control room temperature.

 

DISCLAIMER. "My room" is not "your room" and certainly insulation and heating system are different. If you want to make sure, please do your own measurements (and post them so we can all learn from your experiments).

 

Setup and measurements.

 

I have a DS18B20 sensor connected to a Z-Uno. It reports temperature every few minutes (after running it through a filter for smoothing). It is attached to my mechanical TRV, like this:

 

Please login or register to see this image.

/monthly_2017_12/ValveSensor.jpg.9dcac03e760143164475f0a902763e0d.jpg" />

 

As you can see, I really did "bugger all" to make this a professional solution, sorry, no time to waste... It is a temporary setup anyway. However, the measurements will be fine.

 

Now I need a reference device. I did slightly more than "bugger all" because this device has been my temperature reference since 2012. It is a DS18B20, probably from the same batch, but connected to a Fibaro UBS. I mounted it near "the sofa". Hence its name... "Sofa"...

 

Please login or register to see this attachment.

 

First, let us see if the sensors need calibration. So I mounted the valve sensor in the same spot as the sofa sensor.

 

Please login or register to see this attachment.

 

I would say they are identical. I am not a mathematician, I am an engineer, we have a different definitions of the concept: "the same".

 

So now let's put the valve sensor back as shown in the first picture.

 

Please login or register to see this attachment.

 

Let's see if I can explain what we see. The valve sensor = red almost always registers a temperature that is too high.

 

The left part of the graph is "the evening". If we ignore the dips on the graph, I'd say the sofa measures 1.5 - 2.0 °C lower. Is this plausible? Yes, when I used an LC13 on that radiator I had to set it about 1.5 higher.

 

The sudden drop and the right part of the graph are caused by two things: turning down the TRV and opening a door. The room next to this one is cooler (TRV is an LC12, controlled by the heating panel).

 

As you can see... For about one hour the TRV measures a temperature below the sofa sensor. I live in a could house, the TRV is close to a window. Temperature at the TRV drops quickly and the sensor does not have a lot of thermal mass (or digital filtering). The sofa sensor is probably not experiencing the same level of air flow (it is near a wooden panel).

 

At around 6:00 temperature increases in the other room, then I also increased the setting on the valve a bit (it is a mechanical valve). this explains the last part of the blue line.

 

Based on this graph I'd say the valve sensor is inaccurate, and I do not see a simple solution. Neither an offset, nor a gain correction can make the red an blue graph "fit".

 

Let's have a look at another time segment.

 

Please login or register to see this attachment.

 

Again, overall, the valve sensor reports higher temperatures.

Can you see I have trouble sleeping? Yes, at 4:00 I increased the temperature setting of the valve.

I cannot explain the complete graph, because I did not write down time and set-point of the valve. For comparison of the two sensors, this is no big deal.

 

Please note all graphs use the same temperature scale, because that is what we are trying to compare. But they do not use the same time scale.

 

Conclusion. In my house, in this room, I agree with the statement:TRVs cannot measure temperature accurately, because their sensors are "not in the right spot". So I understand why an FGT without external probe does not report temperature. It could do that, but then you would have to understand its limitations.

 

An FGT-001 with a sensor near the sofa should be able to control temperature more accurately. Whether this is important is up to you. I have learned to live with mechanical TRVs and LC12, LC13 without external sensor. I expect the FGT-001 plus external sensor to increase the comfort level. But right now, I really do not know. How people experience temperature is really complicated... Temperature control is only part of the experience...

 

Edit:

 

@Lode mentioned other sensors in the same room. Here is a graph of three sensors. The red and blue are the same (valve and sofa) and the third one is an FGMS-001 mounted roughly 0.5 m below the ceiling.

 

I think it is hard to say wether it better matches the sofa sensor. When you look at 1/3 of the graph and compare blue-black at 2/3 of the graph I'd say a simple offset will not do the trick. But as a rule of thumb, it measures about 1,5 °C higher because of the difference in height.

 

Please login or register to see this attachment.

 

@robmac made it clear to me... my topic title might be a bit misleading at the moment, because "Part 2" does not exist yet. So here's some more information.

 

12 hours ago, robmac said:

I replaced the Danfoss valves with three fibaro FGT-001 and one Fibaro external temp sensor. I am not particularly interested in accuracy but your charts are a bit concerning as they appear to show a lack of consistency. I am not sure I am seeing the same issues.

 

They should be concerning... But maybe there is a possibility of misinterpretation of my post and my graphs. I probably caused this myself, because the topic title is misleading. In hindsight. It says "FGT-001 Fibaro Heat Controller, external sensor evaluation, part 1". In fact, in part one, there is no FGT-001. And because of that, there is no external sensor either!

 

I compare three sensors in my room, while there is only one plain old TRV. And my point is exactly what you say: (a) Old valves do not perform well in my setup (b) it has a lot to do with the fact that an old valve has no other reference than the sensor inside its head. And indeed, may stuck-on "valve sensor" does not match either my FMGS nor my SOFA sensor. It is not an absolute temperature test. It is only a comparison of references. Because this is a mechanical valve, I could have noted each time I adjusted its setting. But I did not, because I was not really interested in the performance of this old, mechanical valve.

 

When Fibaro releases a firmware update, I'll redo the measurements, with external sensor. I hope to see the room temperature (aka SOFA) will be under control and will be stable and also match the set-point of the heating panel.

 

12 hours ago, robmac said:

My temperature sensor tracks the  Horstmann (Secure) thermostat temperature when placed beside it allowing for the different update interval and reporting sensitivity. I placed the sensor beside the Horstmann at 17:45 and by 18:00 it was tracking within 0.5 which is not bad.

 

I think as an absolute temperature sensor, the FGT-001 is quite accurate, the spec says +/- 0.5 °C and I have every reason to believe that. I could produce some graphs like the ones you posted. I'll think about that... I did have a quick look at the accuracy of the external sensor, but I did not save the data because I thought, it meets its specification, not much to tell about it. In part 2 I'll put the FGT external sensor next to my "SOFA" sensor and they should match. They also should match the programmed temperature.

 

12 hours ago, robmac said:

Is there any chance your issues are environmental.  Drafts would impact devices with different housings differently for example?

 

Absolutely... I have thought about explaining this. The short version: old house, not well insulated, two rooms 4x4 with a door that is mostly open, some chilly weather in Belgium (around 0 °C) and some western wind. Doors sometimes open, number of people in the room vary, ... So the absolute performance depends on so many factors! So Part 2 will be exiting, because then the FGT should try to keep the temperature near the sofa at the expected level! So when I write part two, I'll make notes of the exact circumstances.

 

Edited December 30, 2017 by petergebruers

[ggoetz56 9]

So once more

First of all Peter, thank you for the way you replied my post. I don't intend to say your tests are useless or even smart valves are useless.

But: If you read all the advertisement about energy saving and you like to buy a 20$-device to save 30% of your heating energy, these 20 bugs a wasted.

I think you agree.

 

Your question about external sensor Netatmo/MotionEye. To me it was not very helpfull. But every installation, even every room is different.

In this case the challange was/is (among others) 19° over the night, 24° at 6:30 and 19° after 8:00 and occasionally 24° for unplanned showers in a short time. No Apps should be used. It is about 25 qm bathroom with 1 small and 1 big radiator (150cmx80cm). I used the same valve as in your first picture. Since the former (before using smart valves) average bath temperature was 22° you need about 10 Minutes on "5" to heat up to 24°. A lot of energy was wasted, especially when you forgot to turn down the valves to "3" after the shower.

 

First I installed 2 LC-13 and create a heating-schedule. Sometimes it was a bit late 24° at 7:00. So I thought it depends on the room temperature and bought a motion eye. The reported temperature was no indicator for late heating. The overnight room temperature was about 20/21°. Not only for this I bought netatmo and try to correlate start-heating time with external temperature. I reset the LC to factory value each time i change the installation. That was no good idea for LC 13 because they lost any idea how to heat the room i believe. (Because they have a internal formula). Further more it was in April 2017 and outdoor temperature varies and perhaps not at least my LUA-code was not optimal. 

 

The other problem was the shower-situation. With LC-13 i had sometimes a delay of 1h when i pushed up temp on the valve to 28°. If i just change temp from 19° to 24° sometimes nothing happens. So you can imagine the family was not amused. So  I bought eurotronics spirit (with filrs) which has problems with manual changes because i jumps back to heating-panel in 15 seconds. You need to use a heating override VD for this. Imagine you stand in front of the valve look for your handy  start an app logg in ........ Before smart heating you turned the to the most left position - That's really smart. 

The solution now is burst mode. 2x Burstmode brings up a lot of hot water to the radiator and this enough for shower and clothing.

 

Having 2 systems, Danfoss (with a good algorithim for standing alone) and eurotronics (with FLIRS) ist no optimum, but this works best for me now. the danfoss is most of the time very late in closing the valve at 8:00. The temperature in the display is 19° (no z-wave issue) but it last 1 hour or more to close the valve. Sometimes it is late in the morning. (Evening and weekend would be separate discussions as well as living room etc).

 

So to me  a labor (test) situation is not very usefully and with the equipment that is now in the market it is too much effort to establish a good working environment

Just my opinion.

 

Regards

Gerhard

 

 

Edited January 23, 2018 by ggoetz56
grammer