Back to Smartcitizen.me

Accurate Measurements for Temperature and Noise

Sensors
#1

Hello, I have owned the SCK for two weeks and did several tests to find the best setup.
I am still investigating complex measurements like TVOC or PM, I started with simpler ones like temperature, humidity, and noise.

I am convinced that the design choice of putting the sensor board directly over the data board is the root cause of most problems. A cable connection, even short, would have been much more sensible.

Said that, the kit in practice needs to be used “open air” and in vertical position. Due to self heating of data board, and limited air circulation, horizontal position leads to errors of 2-3 degrees, in excess, on temperature. As the temperature and humidity values are used to correct other measurements, the impact of these errors is even greater.

Not all the vertical positions are equivalent, also. The best orientation is with the temperature sensor on the bottom, collecting air directly from below. In all the other positions, the air has to pass along to the data board before being measured, and this is enough to cause a 0.5 - 1 degree error.

I was able to compare SCK with a professional thermometer, 0.1 degree accurate, and with the setup suggested above, I found an agreement within 0.2 degrees. For same reasons, humidity measurement is a bit less accurate, I find a systematic error of 5 %, in defect.

About noise, basically the adopted sensor is not well suited for indoor measurements, having a minimum value of 45 dBa. In fact, I never measured less than 48 dBa, even in a dead silent room, because the power management chip, which is very close to the noise sensor, emits a multitone whistle in operation. I reduced the problem putting a felt disc above the “sounding” chip. This gave me additional 3 dBa of measurement range, reaching the 45 dBa limit.

As stated above, any enclosure worsens sensor performance. I tried some, but the sensor board needs to be in free air, both surfaces, so the only “enclosure” I am using is a sheet of plastic material, in vertical position, with the components fixed on the surface. For indoor use, this is acceptable.

#2

Hi @nicola.pomaro

Thanks again for all your tests. This kind of posts is beneficial to move forward on improving the sensors collectively. Some comments I could share:

  1. We had to do a lot of trading while designing the board. It was impossible to create a board that had the perfect conditions for all the sensors while yet being cheap and compact.

  2. There’s a static offset for temperature on the firmware that tries to cope with the battery/charger temperature issues by compensating the temperature readings on each situation. However, @oscgonfer and @victor started to work on a much-improved algorithm that includes self-calibration. Feel free to contribute any updates here:

  1. I will be fantastic if you could share with the forum any recommendations on enclosure design to help improve the temperature readings. We think the idea of custom enclosures that benefit a specific kind of text, temperature, noise, particulates, indoor, outdoor is the best option right now.

  2. If you want to separate the two boards, you could try to use standard jumper cables. That’s why we use such a standard connector to join the two boards instead of a fancier connector. However, we can guarantee this works perfectly as the I2S signal from the microphone is quite fast and quite sensible to medium changes.

1 Like
#3

Hello.
So at present temperature and humidity offsets are written in the code, and it would require to edit and recompile it to change them.

In my opinion, the average user can have access to a good thermometer, and could calibrate the measurement for the particular situation of placement and enclosure, but editing the firmware could be a more complicated and risky task. It would be better if offsets or calibration factors could be inserted in a special page during setup procedure, so to change them one only has to repeat the setup procedure.

Yes I2S signal is quite critical, I would avoid to add more than 5-10 cm of cable, but this should allow, for example, to put the sensor board outside the enclosure and the rest inside, so reducing heat and noise issues.

#4

I have encountered similar issue in building my own env. mon. system. I have my temperature/humidity/pressure sensor mounted adjacent to an inlet air hole. A fan at the distant end of the enclosure draws air out of the enclosure and in the process forces air to flow past the device. The fan has to be selected carefully to be compatible, I found a 30 mm fan that runs on 3-5 volts using ~100 MA
With all that I am not yet convinced temperature readings are accurate since they disagree with my only other reference … A red spirit type column thermometer. There is a 2 degree discrepancy. But the device measurements agree with the weather bureau.