cdenney
Posts: 24
Joined: Sat Jul 28, 2018 2:35 am

High reliability, high accuracy temperature/humidity sensor

Mon Feb 11, 2019 5:40 pm

I need a temperature/humidity sensor that will have high reliability and high accuracy for continuous use of weeks at a time. I have a test build right now that uses the AM2302 sensor from Adafruit, but the sensor failed the other day during testing (at least, as far as I can tell it failed. Any script that I run that uses the DHT read_retry command will never return a value, and it won't crash, it seemingly just keeps retrying over and over again). Additionally, before it failed, when I compared the temperature results it was relaying compared to a high accuracy probe thermometer from Thermoworks (calibrated at 0 C and 100 C), it was usually off by several degrees. I'm not sure if this was because of lag time in the readings or just plain inaccuracy.

I'm looking for recommendations for sensors that are high accuracy, fast response (preferably on the order of seconds), and will be reliable over the course of weeks when used in high humidity (>50%) conditions. The accuracy and response rate of the temperature sensor is more important the humidity sensor. The humidty sensor can be anything +/-5% and a response time on the order of minutes would be ok, as long as it is similarly reliable.

Thanks

JohnsUPS
Posts: 74
Joined: Fri Jul 06, 2018 2:13 am
Location: USA

Re: High reliability, high accuracy temperature/humidity sensor

Wed Feb 13, 2019 12:19 am

So long as the temperature range you wish to measure is between -55 degC to +125 degC, I would recommend the Dallas 1-wire temperature sensors. They're accurate (as long as they're sourced from a reputable dealer/not fakes) and have a programmable resolution.
The data conversion time is approximately 750ms for 12-bit resolution, faster for fewer bits.
Temperature response time is fairly fast, as the raw sensors (I use the TO-92 packages) do not have much mass. Some of the ones you can buy are encapsulated, so expect those to not settle as fast.

Perhaps encapsulated sensors are in order here, as they would be sealed and resistant to the high humidity levels you described.

Currently, I have a data center temperature monitoring system installed utilizing five sensors spread along 369 feet of wire, and it has been working fine for months now.

pfletch101
Posts: 383
Joined: Sat Feb 24, 2018 4:09 am

Re: High reliability, high accuracy temperature/humidity sensor

Wed Feb 13, 2019 3:45 pm

cdenney wrote:
Mon Feb 11, 2019 5:40 pm
I'm looking for recommendations for sensors that are high accuracy, fast response (preferably on the order of seconds), and will be reliable over the course of weeks when used in high humidity (>50%) conditions. The accuracy and response rate of the temperature sensor is more important the humidity sensor. The humidty sensor can be anything +/-5% and a response time on the order of minutes would be ok, as long as it is similarly reliable.

Thanks
For those of us who use Temperature/Humidity sensors outside during the winter in non-desert environments, a high relative humidity is >90%; 50% is pleasantly dry! It matters, because it is stability in the >90% range (and, though to a somewhat lesser extent, below 10%) that 'separates the men from the boys' in the humidity sensor stakes. Do you need stability in near-condensing environments?

User avatar
Joel_Mckay
Posts: 266
Joined: Mon Nov 12, 2012 10:22 pm
Contact: Website

Re: High reliability, high accuracy temperature/humidity sensor

Fri Feb 15, 2019 6:08 am

I like to use the bme280 or a mcp9808 ( free samples https://www.microchip.com/wwwproducts/en/mcp9808 ).
Being i2c, they are low-power, clock-skew tolerant digital interfaces, and are comparatively reliable/repeatable... Just keep an eye on the 7-bit bus address allocation needs before you order chips, and future you will thank yourself. ;-)

SPI is also good as it usually has a hardware protocol handler, but should be reserved for high data rate applications given the limited 2-device count.

We found devices with 1-wire are often terrible on anything with dynamic CPU clocking like the Pi. I am constantly surprised how many people don't understand why a polling busy-loop for 1-wire i/o is so wasteful. Such operations can even pin the CPU every-time a DHT style sensor needs power-cycled to recover from hanging every few thousand reads... failures/corruption seems rather random, so people always seem to think... "aha, I solved it this time"... but no... give it a few weeks and you will see the issues again. ;-)

Good luck,
J

User avatar
rpdom
Posts: 13743
Joined: Sun May 06, 2012 5:17 am
Location: Chelmsford, Essex, UK

Re: High reliability, high accuracy temperature/humidity sensor

Fri Feb 15, 2019 7:21 am

Joel_Mckay wrote:
Fri Feb 15, 2019 6:08 am
We found devices with 1-wire are often terrible on anything with dynamic CPU clocking like the Pi. I am constantly surprised how many people don't understand why a polling busy-loop for 1-wire i/o is so wasteful. Such operations can even pin the CPU every-time a DHT style sensor needs power-cycled to recover from hanging every few thousand reads... failures/corruption seems rather random, so people always seem to think... "aha, I solved it this time"... but no... give it a few weeks and you will see the issues again. ;-)
True 1-wire is pretty reliable on the Pi.

The DHT sensors use a different single wire interface. If you read the datasheets you should notice that they do not mention 1-wire anywhere.
Signature is on holiday.

User avatar
Joel_Mckay
Posts: 266
Joined: Mon Nov 12, 2012 10:22 pm
Contact: Website

Re: High reliability, high accuracy temperature/humidity sensor

Fri Feb 15, 2019 11:41 am

rpdom wrote:
Fri Feb 15, 2019 7:21 am
True 1-wire is pretty reliable on the Pi.
The DHT sensors use a different single wire interface. If you read the datasheets you should notice that they do not mention 1-wire anywhere.
True, the example DHT polling libs are usually more likely to have timing issues being invoked from user-space... but its still a hard pass for bit-banging GPIO4 with the kernel 1-wire module (please correct my assumption if something was improved recently)...
If you have to go 1-wire, than at least have a look at a protocol converter like the $1.20 DS2482 bridge ( https://www.maximintegrated.com/en/prod ... 2-100.html )... matching 2 successive reads within given precision bounds is also a good idea. ;-)

Cheers,
J

Idahowalker
Posts: 445
Joined: Wed Jan 03, 2018 5:43 pm

Re: High reliability, high accuracy temperature/humidity sensor

Fri Feb 15, 2019 12:27 pm

A + on the bme280 SPI connected sensors. Mine has been running in a project for 8+ months. My program, on an ESP32, does a read of the bme280 every 1 second.
Without knowing why you are deleting my postings, I will not know how...

User avatar
rpdom
Posts: 13743
Joined: Sun May 06, 2012 5:17 am
Location: Chelmsford, Essex, UK

Re: High reliability, high accuracy temperature/humidity sensor

Fri Feb 15, 2019 9:09 pm

Joel_Mckay wrote:
Fri Feb 15, 2019 11:41 am
If you have to go 1-wire, than at least have a look at a protocol converter like the $1.20 DS2482 bridge ( https://www.maximintegrated.com/en/prod ... 2-100.html )... matching 2 successive reads within given precision bounds is also a good idea. ;-)
Nope, sorry. Been running a number of DS18B20s for years on my Pis now. Occasionally get a bad reading and have code to allow for that and retry. No need for overcomplicating things with extra circuitry.
Signature is on holiday.

JohnsUPS
Posts: 74
Joined: Fri Jul 06, 2018 2:13 am
Location: USA

Re: High reliability, high accuracy temperature/humidity sensor

Sun Feb 17, 2019 3:17 am

I have also found the DS18B20's reliable. I'm also using the OWFS file system and a DS2482 (for driving long distance and to protect the Pi) for reading the sensors, and it works great. No overhead problems. For short runs I'd probably use a GPIO pin (buffered, of course).

The DHT sensor data sheets don't mention the 1-wire protocol because it is proprietary to Maxim/Dallas.

User avatar
joan
Posts: 13805
Joined: Thu Jul 05, 2012 5:09 pm
Location: UK

Re: High reliability, high accuracy temperature/humidity sensor

Sun Feb 17, 2019 9:26 am

JohnsUPS wrote:
Sun Feb 17, 2019 3:17 am
I have also found the DS18B20's reliable. I'm also using the OWFS file system and a DS2482 (for driving long distance and to protect the Pi) for reading the sensors, and it works great. No overhead problems. For short runs I'd probably use a GPIO pin (buffered, of course).

The DHT sensor data sheets don't mention the 1-wire protocol because it is proprietary to Maxim/Dallas.
The DHTxx don't mention the 1-wire bus protocol because they don't use it. They use a so called 1 wire interface (ground, power, data line) but it is not a bus based interface and is in no way compatible with the Dallas 1-wire bus.

User avatar
joan
Posts: 13805
Joined: Thu Jul 05, 2012 5:09 pm
Location: UK

Re: High reliability, high accuracy temperature/humidity sensor

Sun Feb 17, 2019 9:31 am

Another device to consider is the I2C based Si7021 temperature and humidity sensor.

JohnsUPS
Posts: 74
Joined: Fri Jul 06, 2018 2:13 am
Location: USA

Re: High reliability, high accuracy temperature/humidity sensor

Mon Feb 18, 2019 11:57 am

joan wrote:
Sun Feb 17, 2019 9:26 am
JohnsUPS wrote:
Sun Feb 17, 2019 3:17 am
I have also found the DS18B20's reliable. I'm also using the OWFS file system and a DS2482 (for driving long distance and to protect the Pi) for reading the sensors, and it works great. No overhead problems. For short runs I'd probably use a GPIO pin (buffered, of course).

The DHT sensor data sheets don't mention the 1-wire protocol because it is proprietary to Maxim/Dallas.
The DHTxx don't mention the 1-wire bus protocol because they don't use it. They use a so called 1 wire interface (ground, power, data line) but it is not a bus based interface and is in no way compatible with the Dallas 1-wire bus.
Correct. That is why I just stated protocol and didn't use the word interface. The DHT's do seem to be popular though.

Return to “Automation, sensing and robotics”