Hello Team Pimoroni,
I already have some Pimoroni sensors such as the BME680. Of all the ones that were tested on the market, the Pimoronis were the most valuable.
i hope the Sensor-boards from Adafruit are near same same quality.
However, since the accuracy (from the factory delivery) of the typical temperature and humidity sensors is not nearly enough for me (+/-2 C° and maybe more than 5% for humidity), I am now looking for a really good and very precise sensor.
Here I came across two items. The TMP117 and the SHT45.
Question: how accurate are these sensors when delivered?
have you checked them?
have they already been calibrated?.
Is there important information about the sensors that you should definitely pay attention to?
I think the descriptions on the websites describe how great and accurate they are, but I think there is still very important information that you should know when handling or using it for the first time.
Or are they already ready for immediate use and you don’t have to be very careful about cleanliness and contamination or even the possible soldering of connections (key word heat) (if any not use the Stemma connectors) etc.?
I would be very happy to get some important information about the sensors here so that I can really enjoy the parts for a long time
To my knowledge Adafruit doesn’t do in house calibration testing, but both those sensors are models that have factory certified calibration that are very stable at common values so highly unlikely to be out of spec.
My personal experience with a variety of sensors from AF has been that they are spot on for their rated accuracy.
most important information to know is that if it has what looks like a plastic seal over the sensor (without a color pull to remove tab) DON’T puncture or try to remove it… it’s an integral part of the sensor package.
AF is really good about compatibility on breakouts, so they’re 3.5v/5.0v compatible, with a usual +/-0.2v safety margin. Any other special features specific to them will be in the datasheet
Thanks for the information.
I could already see that AF sensors still work best.
But I was very disappointed about the 1-2 C° deviation of the old BME680 or 280 sensors. For a weather station or a temperature measurement in the house, a deviation of 1-2C° is extremely high. After all, one would like to build better measuring stations than those usually known from China.
During my research at the time, I found out that some Chinese manufacturers probably also used the same BME680 sensors, but that this was probably damaged when soldering and mounting on PCB boards due to incorrect production.
Yes, I have read the data sheet for the TMP117 sensors and that they are calibrated. However, I could not read anything about the fact that these were tested again after the assembly on a PC board (even if only in bullet points) and that worried me a bit.
I just see a very high-precision sensor that only has a deviation of 0.1 C° so that I can use it to calibrate my 1-2C° wrong sensors.
I hope this works with the TMP117 and SHT45 sensors.
I have now bought these two sensors from Pimoroni and will then see whether they show the same temperature values.
Normally they should. If they deviate by more than 0.5 or even 1 C° I wouldn’t be exactly happy.
That’s why I asked about the handling.
Thanks for the info that you shouldn’t put on the plastic wrap
Thank you and best regards
Achim
ps.
3.3v -5V voltage :
the AF and Pimoroni boards always had a voltage regulator on them. Since I found out that the sensors react very sensitively if they can tolerate a higher voltage than (in the factory delivery) only max. 3.6V, I only use ESP boards when I build sensor stauions.
I’ve been pretty disappointed in the past with the accuracy of temperature sensors. I use an old, certified, mercury thermometer (left over from my chemical colour photography days) to check against. I keep the sensor on short wires, 10-15 cm from the mc board, and leave the glass thermometer and the sensor on the desk top to settle before taking readings. I try to reduce drafts. The heat capacity of the sensor board is also a problem.
When you read the datasheets, you will always get data related to the sensor. Not the data of the breakout. Or the setup (MCU, power-supply, wire-distance and so on). If you want more precision than 1-2 °C, you need expensive scientific equipment and need to calibrate your measurements.
So the best thing to do is to take all your sensors, try to set them up in a “standard” way for you, and do a calibration (this wont be linear). You will never measure the “real” temperature, but that is anyway nonexistent. In normal rooms, a temperature gradient of 1-2 °C is absolutely normal. I once read that a human has an output of about 100W, so alone this will change the “real” temperature.
be aware the TMP117 has a built in temp offset, which should be already set correctly, but ymmv. if readings are consistently biased, setting that will correct it.
just a small side note, to my knowledge none of the commodity breakout mfrs do post assembly calibration on anything more than a spot check basis (a few units per run) if they even go that far. The only ones that are likely to are made to order shops and integrated final product mfrs
For the price we pay for these boards we cannot expect the makers to do individual checks and callibration on each one. The ‘blurb’, lifted from the chip makers literature, is a bit of an ‘oversell’ and needs to be taken with a small pinch of salt.
Sorry but I see things differently…
if a sensor is already 1-2 c° wrong from the manufacturer, and the customer then also recognizes this, then every measurement would always be wrong, no matter where.
If, on the other hand, the accuracy of 0.1 C° specified by the manufacturer is correct, it is entirely up to the user to add the well-mixed room air to the sensor.
Example. If I use a sensor to measure the temperature in the obscure corner of the room and then in the bottom corner, I’m only ever interested in the temperature of the place where I’m measuring and that’s where I want it to be exactly. If I measure the temp at body height, for example 1.2 meters, I am interested in the value there at this point.
But what I’m definitely not interested in is a temperature that fluctuates by 1-2 degrees.
I don’t have to build a complex weather station or room sensors, I can use ready-made Chinese radio detectors for $5.
The point is this:
It is true that exact and calibrated sensors with a deviation of 0.1 °C become expensive if they are laboriously calibrated again by the end manufacturer before the measuring device is delivered. This makes the product expensive !!!
As a customer of Adafruit or Pimoroni, I therefore buy products here because I trust that measures are taken here during production that the manufacturer of the actual sensor also prescribes. That has to do with the max temperature of the soldering of the chips. In china, this is often NOT taken into account, which is why we get so much scrap there.
If manufacturers like Adafruit and Pinoroni would at least do random tests here, then they would not have to refer to the manufacturers’ data sheets, but could also tell the customers that each batch of boards is tested 2 or 3 times and the deviation is then 0.1 to 0 ,x °C.
that would be great.
But for me it only remains a trust and hope that the sensors also deliver what is in the data sheet, although what is there is very much dependent on the production of the end manufacturer.
And then the sensor can cost $10, $20, or $50 and is exactly as good as a $250 calibrated device
i don’t can check that with a 100% calibrated Temp-Sensor… but i checked with a TESTO Temp-Tool what have a accuracy 0,5 C°.
Im sure this TMP117 and SHT45 are now the sensors Nr1 if any use uC.
I had a play with multiple temperature sensors in the past comparing them against my Jacob Jensen room thermometer. That has no particular claims of being calibrated but it was within 0.1 of some industrial thermometers used to check air conditioning, all details and charts in Adafruit Forums: External temperature sensors vs CLUE’s onboard
Extracting some of the interesting points from that:
The sensors on a hobbyist microcontroller board tend to suffer a lot from local heating effects and there’s often no or iffy mitigation in the design.
I can echo bablokb’s tip on room gradients. In winter it’s probably 22 (celsius) on one side of my room and 18 or less near windows and with curtains open that will cause all sorts of airflows even in a still room. My desktop PC and screens create a local microclimate. An open window is clearly even worse! In offices there’s all the air conditioning moving warmed air around.
Analogue sensors require a lot more diligence. My (experimental breadboard) setup was borked because I had some dodgy crimping on cheap cables. In the digital world you’ll probably get away with those high resistance points. ADC issues also apply, for example nRF52840 is a bit noisy for ADC.
Thermistors respond to changes incredibly quickly but this means you’ll continually get different readings so you need a physical or sampling design to deal with this.
I’ve not done much with humidity sensors but I have noticed some can pick up each human breath nearby suggesting some caution is required there with sampling/integration.