Precision thermometer for data logging

[feedback.loop]

Silicon Labs Si7051 was used to build a 0.1 degree C accurate thermometer with a serial port over USB. A small 0.96" OLED display with SPI interface was also used, as well as Atmel ATMEGA328P micro-controller. The software was written using Arduino development environment and Adafruit_SSD1306 library for OLED display.

https://youtu.be/pj1HOM35vH4

Si705x data sheet: http://www.silabs.com/Support%20Documents/TechnicalDocs/Si7050-1-3-4-5-A20.pdf

OLED display library: https://github.com/adafruit/Adafruit_SSD1306/blob/master/Adafruit_SSD1306.h

Arduino sketch: https://github.com/sandy-s/Si705x/blob/master/m328p-spi-oled.ino

[TiN]

Sweet, I have same display unused I wish to have time to do similar..

[feedback.loop]

I was not sure where to put this video. Let's consider it the first practical application of the thermometer in question.

https://youtu.be/80ypF-Ix2p8

[TiN]

feedback.loop
Congratulations becoming baby-level volt-nut. Get your wallet ready. 

If you want to chat about this addiction in russian, feel free to reach out, I'd be glad to help you dispose extra cash on LTZ's. 

[feedback.loop]

feedback.loop
Congratulations becoming baby-level volt-nut. Get your wallet ready. 

If you want to chat about this addiction in russian, feel free to reach out, I'd be glad to help you dispose extra cash on LTZ's. 

No doubt, you know how to dispose of non-trivial amounts of cash. 
I am quite alarmed watching you acquiring increasingly expensive gear in a hurry.
Let's see if I can practice some self-control.
By the way, thanks for the KX reference project and all your excellent tear-downs, repairs and such. 

[TiN]

There is less competition up there, as most of volt-nuts already gave up long ago  .

[branadic]

Nice to see your solution.
I build myself a setup using SHT25 (humidity + temperature), MS5611 (pressure + temperature) and TMP112 (temperature as variable element to measure at certain points within a reference voltage circuit). A STM32F405 µC box reads the sensor data and is connected via USB.
Andreas has expanded his readout software for 3458A via GPIB-USB adapter, so that ambient data are captured together with voltage readings. This works fine.

I have now finished a new temperature sensor that is based on 2x PT1000 and a TDC-based (time-to-digital-converter) readout circuit, just to see what is possible with low amount of work. Need to show this peace of kit within the next days.
I will measure it against our PT100 (<0,02K accuracy) temperature system in a calibrated climate chamber in near future in a limited temperature range.

-branadic-

[Dr. Frank]

Hello feedback.loop,
welcome in the volt-nuts community.
I really enjoy your slowly-narrated videos, like this one.

Your standards deliver less noise, more stability and better uncertainty, if you also use them at full scale, like your DMMs.
That means, output these 10V in the 10V range, instead of 1/10 F.S. in the 100V range.

So dial in 10.00000 on your Fluke 341A, same on the EDC.

I'm looking forward, what your further encounters will be.
Next time, it will be probably the assembling and launch of your LTZ1000.
That smells like 45°C oven temperature, i.e. 12k/1k oven setting?

So you'll also need some colder place, 30°C is quite warm for a lab.

Frank

 

[feedback.loop]

Your standards deliver less noise, more stability and better uncertainty, if you also use them at full scale, like your DMMs.
That means, output these 10V in the 10V range, instead of 1/10 F.S. in the 100V range.

So dial in 10.00000 on your Fluke 341A, same on the EDC.

I'm looking forward, what your further encounters will be.
Next time, it will be probably the assembling and launch of your LTZ1000.
That smells like 45°C oven temperature, i.e. 12k/1k oven setting?

So you'll also need some colder place, 30°C is quite warm for a lab.

Frank

Good point about using full scale. Will do this next time.
Where do I get a colder place?  I think I am stuck with what I have for now. I don't really want (yet) to install a conditioner and live with its noise and extra power consumption.

[CalMachine]

Very nice!     Once again, I enjoy watching your videos.  We seem to be at a similar level of joining the volt-nuttery world.  My wallet has already taken a severe hit.

I'm in similar process getting my own thermometer circuit built for the internal temp of my oven that I am getting around.  I was using the BME280 inside the oven for preliminary testing, but it's not accurate enough.  This Si7051 looks like a good candidate for my needs.

[Dr. Frank]

Good point about using full scale. Will do this next time.
Where do I get a colder place?  I think I am stuck with what I have for now. I don't really want (yet) to install a conditioner and live with its noise and extra power consumption.

Well, you have called your video 'Low Budget Metrology', so that's the right way to go with temperature, and also with any struggle to mitigate temperature variations.

Maybe you have a basement in your house?
That's the perfect place to set up an analogue calibration lab - even recommended by NIST (or where did I find this hint?).

This location is cold, even in hot summers, i.e. temperature is stable to +/- 2°C. It's a bit EMC - shielded, due to missing windows, and all the iron mesh in the concrete.. and it's absolutely free.

So, there's also not that big pressure, to have extremely low T.C.s in your references.

Frank

[Echo88]

Maybe the ADT7320 also fits your needs CalMachine. It has higher resolution than the Si7051.

http://www.analog.com/media/en/technical-documentation/data-sheets/ADT7320.pdf

[Andreas]

Silicon Labs Si7051 was used to build a 0.1 degree C accurate thermometer

Hello,

thanks for sharing.

One thing: if you read all the footnotes in the data sheet the 0.1 deg C
accuracy is only for a small 5 deg C range around 37 deg C.

What would be interesting for me:
how much self heating of the sensor is produced by reading it all 100 ms.
With branadics solution and some sensors (break out boards from ebay) I found out that
the self heating is up to 1 deg C depending on what is additionally on the board.
(voltage regulator, I2C pull up resistors, voltage regulator, LED).
You can find this out with a small fan switching on/off.

But anyway room temperature measurement is not very exact.
When I open the door in my "lab" I have a 0.2 deg C jump on the temperature sensors.
So even my uncalibrated NTCs together with a 24 bit ADC give a good result.

with best regards

Andreas

[feedback.loop]

Silicon Labs Si7051 was used to build a 0.1 degree C accurate thermometer

Hello,

thanks for sharing.

One thing: if you read all the footnotes in the data sheet the 0.1 deg C
accuracy is only for a small 5 deg C range around 37 deg C.

What would be interesting for me:
how much self heating of the sensor is produced by reading it all 100 ms.
With branadics solution and some sensors (break out boards from ebay) I found out that
the self heating is up to 1 deg C depending on what is additionally on the board.
(voltage regulator, I2C pull up resistors, voltage regulator, LED).
You can find this out with a small fan switching on/off.

But anyway room temperature measurement is not very exact.
When I open the door in my "lab" I have a 0.2 deg C jump on the temperature sensors.
So even my uncalibrated NTCs together with a 24 bit ADC give a good result.

with best regards

Andreas

Yes, +-0.1C from 35.8C to 41C, and +-0.13C from 20C to 70C. So let's call it 0.13C accuracy, which is not bad at all.
Regarding self heating, they say that with 1Hz sampling the average current is 195nA. You want 10Hz sampling, for which they don't give direct numbers, but give conversion time and current consumption during conversion. The chip works down to 1.9V, so you may want to run it together with the micro-controller from 2V to reduce self heating.

[Echo88]

Why would anyone use 10Hz sample rate for a room-temperature thermometer, especially if it degrades the performance through selfheating? After all the measurement/reference-stuff sits in high thermal capacity enclosures.

[branadic]

Selfheating should be no problem using the system I set up with PT1000.
Why? Well, a C0G capacitor 10nF - 33nF is charged to 3,3V and discharged through the PT1000 down to a threshold voltage, discharge time is measured with ~22ps resolution. Thus only a few nC are flowing through the resistor while discharge and selfheating is extremly low.

[Andreas]

Why would anyone use 10Hz sample rate for a room-temperature thermometer, especially if it degrades the performance through selfheating? After all the measurement/reference-stuff sits in high thermal capacity enclosures.

I only asked this because it is used in his arduino sketch.

Regarding self heating, they say that with 1Hz sampling the average current is 195nA.

The data sheet is very fuzzy with all values. So it would be better to have some "real life" values.
If I calculate back this is only true for 11 bit resolution and typical conversion times and currents.

And what is about I2C-communication with 4 mA?
Is this only valid for special commands or also for reading out the results?
What will happen with I2C power consumption of the sensor if there is also the display on the same line?

I know I´m a little bugger. (asking always that, what is not mentioned in the data sheet).

With best regards

Andreas

[feedback.loop]

Why would anyone use 10Hz sample rate for a room-temperature thermometer, especially if it degrades the performance through selfheating? After all the measurement/reference-stuff sits in high thermal capacity enclosures.

I only asked this because it is used in his arduino sketch.

Regarding self heating, they say that with 1Hz sampling the average current is 195nA.

The data sheet is very fuzzy with all values. So it would be better to have some "real life" values.
If I calculate back this is only true for 11 bit resolution and typical conversion times and currents.

And what is about I2C-communication with 4 mA?
Is this only valid for special commands or also for reading out the results?
What will happen with I2C power consumption of the sensor if there is also the display on the same line?

I know I´m a little bugger. (asking always that, what is not mentioned in the data sheet).

With best regards

Andreas

The reason I chose to sample every 100ms is that I wanted fast response to requests over the serial port, but didn't want to complicate this first version with interrupts. You are quite right, this is not optimal at all. Perhaps I should work on the next version with 1Hz or less sampling for displaying locally, and with interrupts from the serial port. This way it would respond to requests even faster, and avoid the cost of sampling too often.
And you are right that the micro-controller, display and other parts can heat up and distort the measurements.
There is no limit to perfection.

[branadic]

Here is a picture of my thermometer, one with SMD PT1000 and one with THT PT1000. There are two temperature sensors, one on top and one on the bottom of each board.
This setup is revision 1, just to see how it works and how they compare to each other.

-branadic-

[feedback.loop]

More data (about 230 hours) from my experiment measuring 10V from Fluke 341A (still 1/10 of the 100V range - silly me). This is the continuation of the same experiment I have started for the video. The weather was cooler since then, so the ambient temperature has never hit 35C again. Still about the same temperature coefficients as shown in the video: about -1.5ppm/C as measured by Datron 1071 and about -2ppm/C as measured by HP 34401A.

I think that despite of unknown tempco of Fluke 341A calibrator I can at least conclude that Datron has about 0.5ppm/C better tempco than HP.

I stopped this experiment and started another one, but this time using Fluke 341A 10V range full scale. Let's see how much better this will be.

The parts for LTZ1000 reference have arrived. I hope to build it soon.

[feedback.loop]

I collected about 36 hours of measurements of 10V, full scale this time, from Fluke 341A. Unfortunately, I started experiencing some problems with GPIB, so the experiment was interrupted. The results are clearly better than 10V as 1/10 of 100V range. This time I see about 3ppm change over about 4 degrees C, which is less than 1ppm/C. And Datron 1071 still looks a bit more stable than HP34401A.

[feedback.loop]

More data (about 230 hours) from my experiment measuring 10V from Fluke 341A (still 1/10 of the 100V range - silly me). This is the continuation of the same experiment I have started for the video. The weather was cooler since then, so the ambient temperature has never hit 35C again. Still about the same temperature coefficients as shown in the video: about -1.5ppm/C as measured by Datron 1071 and about -2ppm/C as measured by HP 34401A.

I think that despite of unknown tempco of Fluke 341A calibrator I can at least conclude that Datron has about 0.5ppm/C better tempco than HP.

I stopped this experiment and started another one, but this time using Fluke 341A 10V range full scale. Let's see how much better this will be.

The parts for LTZ1000 reference have arrived. I hope to build it soon.

After playing with LTZ1000 it seems to me that my previous conclusion that Datron 1071 is more stable than HP34401A was wrong. It must have been an illusion. It seems now that my Datron 1071 has about +0.8 ppm/C temco, which combined with about -2 ppm/C tempco of Fluke 341A so the result looked better than more stable HP34401A, which, let's say, just showed the naked -2 ppm/C of the Fluke 341A. I am a bit disappointed by Datron 1071, and quite pleased with HP 34401A.