Radiometric vs Non-Radiometric Leptons

Discussion related to "under the hood" OpenMV topics.
GL-ITC
Posts: 86
Joined: Tue Feb 25, 2020 7:09 am

Re: Radiometric vs Non-Radiometric Leptons

Postby GL-ITC » Tue Oct 20, 2020 12:14 pm

Yes I think I see where your coming from now...the 20 degree blackbody may need to be cooled as opposed to heated if the ambient temperature is too high. Another thing to think about!

Thanks,
G
GL-ITC
Posts: 86
Joined: Tue Feb 25, 2020 7:09 am

Re: Radiometric vs Non-Radiometric Leptons

Postby GL-ITC » Wed Oct 21, 2020 5:48 am

Hi Stub,

Apologies for the endless stream of questions, can I ask what sort of a heater element are you using? Is it a cartridge heater or some form of a ptc heating element which seem to be the more power efficient option.

I had initially thought of using a cartridge heater but a ptc heating element may be the better option.

Thanks,
G
Stub
Posts: 25
Joined: Mon Dec 23, 2019 8:30 am

Re: Radiometric vs Non-Radiometric Leptons

Postby Stub » Wed Oct 21, 2020 5:54 am

Hi,

I used 3 of these at 6V they are fine (with the FET). https://www.robotshop.com/uk/5v-heating ... 11289.html Other distributors are available, its a sparkfun part.
GL-ITC
Posts: 86
Joined: Tue Feb 25, 2020 7:09 am

Re: Radiometric vs Non-Radiometric Leptons

Postby GL-ITC » Wed Oct 21, 2020 5:57 am

Out of interest have you checked to see how uniform the heat distribution is on the plate? Or do the three of those heating pads completely cover the back of the plate to mitigate that issue?
Stub
Posts: 25
Joined: Mon Dec 23, 2019 8:30 am

Re: Radiometric vs Non-Radiometric Leptons

Postby Stub » Wed Oct 21, 2020 6:10 am

Yes (fluke spot meter) and yes - it's fine for my purposes as the camera is looking for the average in the blob.
GL-ITC
Posts: 86
Joined: Tue Feb 25, 2020 7:09 am

Re: Radiometric vs Non-Radiometric Leptons

Postby GL-ITC » Wed Oct 21, 2020 6:10 am

Thats great, thanks for the help.

G
GL-ITC
Posts: 86
Joined: Tue Feb 25, 2020 7:09 am

Re: Radiometric vs Non-Radiometric Leptons

Postby GL-ITC » Wed Nov 04, 2020 10:37 am

Hi Stub,

In terms of your blackbody heating algorithm did you take the approach of providing current to the heating pad if the temperature is not at the set point and then turn off when the temperature is at the set point or did you go much more robust in your approach. Again I was thinking of applying the Arduino as a PID controller and taking an approach similar to http://electronoobs.com/eng_arduino_tut24_2.php. Im just wondering if you tried this straightforward on/off approach. I only ask as I'm still waiting for parts to be delivered so I cannot try myself yet.

Kind regards,
G
Stub
Posts: 25
Joined: Mon Dec 23, 2019 8:30 am

Re: Radiometric vs Non-Radiometric Leptons

Postby Stub » Wed Nov 04, 2020 11:14 am

Hi,

Yes I tried the on/off approach and it was not good enough. I didn't need a formal PID controller though. Four set points around the control point and PWM 90%, down to 40% controlled any overshoot nicely. Not sure if it would be robust enough as it stands for other set points though.
GL-ITC
Posts: 86
Joined: Tue Feb 25, 2020 7:09 am

Re: Radiometric vs Non-Radiometric Leptons

Postby GL-ITC » Fri Nov 06, 2020 3:03 am

Hi Stub,

I've another quick question if thats alright, how did you mount the heating pad to the sheet of steel, did you glue it to it or something as basic as that?

Thanks,
G
Stub
Posts: 25
Joined: Mon Dec 23, 2019 8:30 am

Re: Radiometric vs Non-Radiometric Leptons

Postby Stub » Fri Nov 06, 2020 4:52 am

As it is a prototype I used electrical tape just in case I needed to disassemble it to make any improvements. But glue was the intention.
GL-ITC
Posts: 86
Joined: Tue Feb 25, 2020 7:09 am

Re: Radiometric vs Non-Radiometric Leptons

Postby GL-ITC » Thu Nov 12, 2020 10:04 am

Hi Stub,

I got my blackbodies working eventually. I initially tried implementing a pid controller approach which sort of worked but not well enough. So I basically done what you said with the different pwm and set points etc.

I currently power the the Arduino with 9v through the power jack. I'm powering an LCD display and DS18B20 temperature sensor from the 5v pin but I'm powering the two heating pads from the VIN pin so they should be getting ~9v before the onboard regulator if my understanding is correct. My question for you is, are you using the VIN pin to get the 6v power to the heating pads? I'd ideally like to supply them with less power if it meant more longevity.

Thanks,
G
Stub
Posts: 25
Joined: Mon Dec 23, 2019 8:30 am

Re: Radiometric vs Non-Radiometric Leptons

Postby Stub » Thu Nov 12, 2020 11:11 am

I have not done any work on making the power supply elegant. I simply have two. One for the arduino and another set to 6V for the heaters...
GL-ITC
Posts: 86
Joined: Tue Feb 25, 2020 7:09 am

Re: Radiometric vs Non-Radiometric Leptons

Postby GL-ITC » Fri Nov 13, 2020 5:27 am

Ok. I'm only using two heating pads on some 2mm steel thats 100mm x 100mm in size. Im using this power supply https://thepihut.com/products/arduino-u ... pply-9v-2a to power everything which seems to work quite well. I may need to add a 5v pwm fan to the 20 degree blackbody for its eventual end system.

I'm hoping to do a test to see what improvement using two blackbodies in the FOV will make when the camera is recording the temperature of a pot of water cooling from ~50 degrees. I'll record the actual temperature of the water with a DS18B20 temperature sensor, the default output the camera gives me for a roi in the pot of water and the adjusted output for the same roi using the blackbodies to scale the values. I'll let you know how it goes.
Stub
Posts: 25
Joined: Mon Dec 23, 2019 8:30 am

Re: Radiometric vs Non-Radiometric Leptons

Postby Stub » Fri Nov 13, 2020 5:30 am

Hi,

Thanks - that will be interesting...

Cheers
GL-ITC
Posts: 86
Joined: Tue Feb 25, 2020 7:09 am

Re: Radiometric vs Non-Radiometric Leptons

Postby GL-ITC » Wed Nov 18, 2020 3:00 pm

Hi Stub (and anyone else interested in improving radiometric output),

I have some preliminary results using the two blackbodies in the field of view to improve temperature readings in the 20 to 40 degree temperature range. I have performed various experiments to get to this stage. The test I used to assess the accuracy was to use a container of water that was cooling from 40 degrees to 20 degrees. The temperature of the water was monitored using a DS18B20 temperature sensor submersed in the water. I used water because it was something that would change temperature naturally as it was cooling and because its emissivity is good at ~.95.

Initially results were poor. The primary reason for this is was I failed to account for the fact that the sensor would be influenced by the ambient temperature, not just the surface of the blackbody that it was touching. This meant my readings were 4 degrees Celsius out. The simple fix for this was to insulate the temperature probe from the surrounding environment. In my case I used a special insulation used in furnaces etc due to its really high insulating value. This alone improved accuracy dramatically with the temperature readings now out by 1.8 degrees Celsius on average.

The temperature sensor I use to monitor the temperature of the blackbody is also a DS18B20 temperature probe. Due to its cylindrical shape the thermal contact between the probe and plate is quite small meaning there is an element of inaccuracy introduced here too. There are two solutions that need to be implemented here. Firstly I need to change from the probe version of the DS18B20 to the 3 pin version as it has a flat edge meaning I will be able to get better thermal contact between the blackbody and the sensor. I have yet to implement this step as I'm waiting for the sensors to be delivered. The other thing I could and did do was add heat transfer paste between the probe and the plate. This was done to remove any air gaps between the probe and the plate which would impact temperature readings and also as the name suggests to improve heat transfer between the plate and the probe. This resulted in the temperature readings of the water being out by 1.16 degrees on average with a maximum of 1.96 degrees compared to an average of 3.9 degrees with a maximum of 7.1 degrees when using the default output of the camera.

There are numerous ways to improve the system even further. Firstly improving the thermal contact between the sensor and steel plate will probably be of benefit. In my scenario I allow my blackbodies fluctuate anywhere from 19.75 degrees Celsius to 20.25 degrees Celsius on the 20 degree blackbody and 39.75 degrees Celsius to 40.25 degrees Celsius on the 40 degree blackbody. This is obviously a definite source of inaccuracy as I assume they are at 20 and 40 degrees Celsius respectively which they are not. Attached is a basic graph of the results from the latest test. The red line represents the Flir 3.5 default output when I set the min and max temperatures to 10 and 50 degrees on the openmv h7. The green line is the temperature of the water reported by the DS18B20 probe submersed in it. The purple line represents the "calibrated" temperature of the water based on the two blackbodies. The purple line is a little bit bouncy due to the minor temperature ranges in the blackbodies which I listed above. The more accurate the blackbodies are the more accurate the camera output will be.

Screenshot 2020-11-18 at 19.17.20.png

The thermal image shows the regions of interest I am using for my calculations

boundary.png
boundary.png (7.77 KiB) Viewed 45 times

I suspect adding heat transfer paste between the heating pads and the piece of steel would probably make the system more efficient also but this is only a small thing.

Thanks,
G

Return to “Technical Discussion”

Who is online

Users browsing this forum: No registered users and 2 guests