Weird temperature readings

[cincura.net 13]

Hi, I have Window/Door Sensor 2 with this settings:

Please login or register to see this attachment.

 

Which for me is that the sensor will measure the temperature every 480 seconds and if it differs more than 0,1C the value will be sent to the controller (also it will be sent no matter what once every 3600 seconds). But when I look into history, I see this:

Please login or register to see this attachment.

 

Which to me doesn't make sense. In roughly 25 minutes the temperature went from 21,6 to 22,3, but why there isn't any temperature in between. It's surely more than 0,1 change in between and there's few 480 second intervals in 25 minutes where the measurement should/could happen.

 

Can it be explained?

[petergebruers 1,265]

2 hours ago, cincura.net said:

Can it be explained?

Yes, it is a bug and I have reported it to Fibaro but they have never investigated nor fixed this and they have investigated this (see answer by @m.roszak below)

 

One possible explanation is "self heating" of the chip... I am not 100% sure but it is an educated guess, if the device measures the temperature when the TX/RX and CPU were active then the temperature might actually be a bit higher than ambient. This seems to happen at random moments, but somewhat regularly so if you put that on a char eg 24h you will notice those "outliers" you have mentioned (there is no intermediate value, and a 1 degree "jump" is implausible). I did not investigate recently, from the top of my memory you'll see a 0.5 to 1.0 degree "jump" from time to time.

 

@m.roszak if your firmware engineers get bored, this is an interesting issue for them to investigate

Edited April 18, 2020 by petergebruers

[m.roszak 509]

We investigated this, I will check the status next week and get back to you

[petergebruers 1,265]

7 minutes ago, m.roszak said:

We investigated this

Thanks, I've edited my post, I wrongly assumed it was never investigated.

[m.roszak 509]

I am pretty sure that I have prepared test setup mysefl and found some inconsistencies... don't worry

 

I will let you know what is the status od this subject im next few days.

[cincura.net 13]

Thanks @m.roszak and @petergebruers, I'd really like to have this "stable". My plan is to use this door sensor as a temperature device and use this for managing heating, thus decent temperature readings are a must. If it happens to be not "fixable" easily now, I can probably use Motion Sensor, as I don't see this happening there (but might be that it didn't happen yet???).

[cincura.net 13]

Just for the good measure I removed the battery from the door sensor, waited few seconds, and put it back. Now it seems the issue is gone. That's good. But also little bit worrying, because I don't know what might trigger it in the future.

[m.roszak 509]

As this issue sometimes occurs it will be better for your use case to use another device.
I have checked the status and it is in our hardware engineer hands currently.  
 

[cincura.net 13]

1 minute ago, m.roszak said:

use another device

You mean different door sensor or different kind of device?

 

@petergebruers Have you seen this only with door sensors or also with other Fibaro devices?

[m.roszak 509]

Another device, this reproduce sometimes on D/W Sensors, never seen this on other modules. 

Also as device is "glued" to the surface it measures not only ambient temperature but some of surface temperature is transferred - so maybe it is not a good idea to use it for heating purpose  anyway

[cincura.net 13]

Thanks for the clarification.

 

11 minutes ago, m.roszak said:

Also as device is "glued" to the surface it measures not only ambient temperature but some of surface temperature is transferred - so maybe it is not a good idea to use it for heating purpose  anyway

In my case the plan was/is to just keep it lying on the surface, not gluing it.

[petergebruers 1,265]

23 minutes ago, cincura.net said:

@petergebruers Have you seen this only with door sensors or also with other Fibaro devices?

 

I have not seen it on the FGMS and also no issues with Flood Sensor. I can say this because those 2 sensors are in relatively controlled environments so I would notice if the temperature "jumped". Here is a graph of a motion sensor in our "TV Room". This room has West facing window and FGT-001 TRV. The temperature "jumps" of 0.5 °C you see here are nothing like the jumps you see with D/W Sensor 2... To be honest, to produce a really nice graph I would have to put this sensor in a room with slow changing temperature, eg in a cellar.

 

Please login or register to see this image.

/monthly_2020_04/2085630739_Screenshot2020-04-20at10_22_50.png.46a95d2562bdadd9dc1568e1f4cf9610.png" />

 

I do have a Flood sensor in the cellar, look how stable the measurement is... I forced Grafana to use the same Y axis scale... 

 

 

18 minutes ago, cincura.net said:

In my case the plan was/is to just keep it lying on the surface, not gluing it.

 

I was indeed going to say that a Door Sensor would not be in the right place to measure temperature, and neither would the Motion Sensor or Flood Sensor work in its intended position. If you look at my FGMS graph you would think the room is 1 to 1.5 degrees hotter than "head height" and it would also register airflows in a different way, compared to a sensor on eg e cupboard.

 

Of course, I do not know how much experience you have with temperature measurement and control so you may already know about thermal mass and so on.

[cincura.net 13]

4 minutes ago, petergebruers said:

I have not seen it on the FGMS and also no issues with Flood Sensor.

Thanks.

[petergebruers 1,265]

I also own a bunch of DS18B20 sensors, connected to the older FGK-101 Door Window sensor and the Unversal Binary Sensor. Those are good too but beware of low quality clones...

 

I like the UBS because I can "poll" it frequently if I want to (eg every 2 minutes) and the DS18B20 has low thermal mass so produces fast and accurate results

 

 

Here is a graph of the same room with the FGMS, yellow line is FGMS, green line is measured by DS18B20 + UBS, sampled every 2 minutes

 

 

Please login or register to see this attachment.

[cincura.net 13]

I have few FGK-101 sensors too, mostly because it runs on battery and I can connect the DS18B20 on a cable (waterproof) and i.e. put it into the pool. I haven't played with UBS yet, because it needs power and I don't have usage for that kind of setup yet.

[cincura.net 13]

Funny or weirdly enough, I'm looking now at the temperature in Door Sensor "1" using DS18B20 and I see the jumps too. The probe is in over 20000 liters of water, so I'm really sure it's not heating/cooling that fast.

Please login or register to see this image.

/monthly_2020_05/img.png.e2c89f1fffd61930e1ecfe0b6afc83ac.png" />

The 27,12 is the weird one. Other values seem, more or less, plausible-ish (although having 0,2 difference in 4 minutes (01:19 vs 01:15) isn't probably realistic either, it's good enough for me in this case). Looking at the times, it looks like it's the one that's sent as a periodic report (not the threshold one).

 

In this case we can exclude the chip/board heating affecting the temperature readings. Obviously it might be bad quality DS18B20, but it is suspicious.

 

/cc @petergebruers

 

[petergebruers 1,265]

14 hours ago, cincura.net said:

The 27,12 is the weird one. Other values seem, more or less, plausible-ish (although having 0,2 difference in 4 minutes (01:19 vs 01:15) isn't probably realistic either, it's good enough for me in this case). Looking at the times, it looks like it's the one that's sent as a periodic report (not the threshold one).

Did you buy that DS18B20 form a reputable source? Do you have an Arduino and does the test script say your sensor is genuine? I am asking because some of the "clones" or "fakes" or whatever you want to call them have higher conversion noise than the original Dallas chip! See previous post...

[petergebruers 1,265]

14 hours ago, cincura.net said:

The probe is in over 20000 liters of water, so I'm really sure it's not heating/cooling that fast.

 

I had to think long and hard and came up with more questions than answers

 

Agreed, that amount of water cannot change temperature quickly... But the body of water does not necessarily have a uniform temperature and currents can exists, and I guess sometimes water enters and leaves the tank which would not change the average temperature, but maybe does so locally. I don't have such a tank or anything I could use for that to experiment.

 

What you could do to get an idea of what is going on is increase the reporting frequency of that FGK.

 

Or put that probe in a bucket of water and put that in cellar

 

Just thinking out loud. It is an intriguing topic

 

If you want me to do some experiments, I do have some spare sensors and a spare UBS. Not sure if I have a FGK-101 though. I also could add a few to my Z-Uno (fake and genuine DS18B20)

 

Please login or register to see this link.

[cincura.net 13]

8 hours ago, petergebruers said:

Did you buy that DS18B20 form a reputable source? Do you have an Arduino and does the test script say your sensor is genuine? I am asking because some of the "clones" or "fakes" or whatever you want to call them have higher conversion noise than the original Dallas chip! See previous post...

That's something I need to check. At the moment I don't have any available Arduino at hand, maybe some ESP, I have to check my box of "stuff" .

 

8 hours ago, petergebruers said:

Agreed, that amount of water cannot change temperature quickly... But the body of water does not necessarily have a uniform temperature and currents can exists, and I guess sometimes water enters and leaves the tank which would not change the average temperature, but maybe does so locally. I don't have such a tank or anything I could use for that to experiment.

This is tank with no input/output of water and the water is there still (especially during night). And although I'm not a hydro-thermo-something engineer, I think 0,2C in these conditions is really an error - but not sure on which side.

[petergebruers 1,265]

Okay, intriguing question. I've "thrown" a FGK-101 + DS18B20 in my cellar and I'll let you know how it goes once I've got enough data points 

[petergebruers 1,265]

I have 12 hours worth of data... I see no "blips" on the temperature readings of my FGK-101 + DS18B20 (genuine Dallas sensor).

 

Here's the test setup:

 

Please login or register to see this image.

/monthly_2020_05/IMG_20200503_225501.jpg.29ee0730f17e3f84d170bec66d9921ea.jpg" />

 

So that's the floor in my cellar and the "Flood Sensor" is the one I've talked about before...

 

Here's the effect of putting the sensor in that environment, it gives an idea of the thermal constant of the DS18B20 sensor (plus a bit of plastic, it is mounted in the housing of the FGK)

 

 

Le't's zon in on the art after about 23:30 up to now. Pay attention to the vertical scale, it is 0.05 °C per division!

 

 

On 5/2/2020 at 8:01 PM, cincura.net said:

although having 0,2 difference in 4 minutes (01:19 vs 01:15) isn't probably realistic either, it's good enough for me in this case

Based on this graph and my previous (anecdotal) experience with the DS18B20 I would say noise is indeed lower than 0.2 °C... It is probably a lot lower than that.