A new affordable thermal imager

[jeremy]

Regarding the window: At the moment the camera lens is recessed by about 1 mm but we have no final decision (yet) about using a window or not - those tiny little things are quite expensive.

Are you looking at germanium, chalcogenide glass or ZnSe? or something else? Do you plan on doing your own calibration of the units, or are you relying on the calibration of the lepton?

[Andreas_Thermal]

Excellent project -- I'd be really really interested to use this! Looking forward to the kickstarter.

If your looking for more LUT's to play with I posted a file with many in it here https://www.eevblog.com/forum/thermal-imaging/flir-e4-thermal-imaging-camera-teardown/msg713511/#msg713511. (not all of them are good but with 8gb of space it sounds like it would be a minimal impact)

Thank you - nice! Yesterday we did the first take of our crowdfunding campaign video - I had to use your black-green-white gradient for the night vision scene, it looks so nice  .

Are you looking at germanium, chalcogenide glass or ZnSe? or something else? Do you plan on doing your own calibration of the units, or are you relying on the calibration of the lepton?

I was looking for ZnSe windows because they are cheaper than the ones made of germanium but I didn`t do to much researching about windows up to now, so I don`t have an idea about the price of chalcogenide glass (i can imagine that this could be also a cheap option for us). At the moment I calibrated each of our prototypes manually, according to this experience, we definitely have to do our own calibration.

[Fraser]

For info ZnSe is a relatively soft material. It scratches easily. This is why you do not see it used as lens protectors on commercial cameras. IIRC ZnSe also needs a materials safety warning for the user as it can be toxic.

Be very careful how you attach ZnSe, it does not like some adhesives and reactsvwith them.

When considering options, it is often a good idea to research how the bigger companies approach the issue of lens protection. An obvious case being fire fighter cameras. They have robust Germanium windows to protect the lens..... Very expensive. The FLIR ONE G2 uses a Chalcogenide glass such as GASIR as it is a good balance of performance and cost. Ask yourself if a lens protector is truly essential with such a small lens ? PCB CCTV board cameras do not bother with such due to small recessed lenses.

If needed when working with fluids, to protect against splashes, a thin thermally transparent film may be used as a lens protector. That is what FLIR provides for its industrial users. Just mount the film on a ring that attaches to the camera. The Chinese make auxilliay lenses for mobile phones that attach with magnets and look very neat.

Fraser

[Andreas_Thermal]

Thank you for the information about ZnSe! I never got in touch with a ZnSe lens, and I thought they will survive harder conditions (as far as I know they are mostly used in Co2 laser applications).

Ask yourself if a lens protector is truly essential with such a small lens ?

That`s why I didn`t do further research up to now - at the moment our focus is to lower the price of our camera. On thing I can say about the missing window up to now is that I have one of our prototypes in my pocket all day (without any cover) and I never had problems with the lens.

I`ll take a look at the thin film you suggested - this sounds promising to me

Andreas

[Fraser]

Anreas,

The film that is used on commercial thermal camera lens protectors is POLYOLEFIN Shrink Wrap, as used in teh food and cosmetics industry. It is also to be found around some new Audio CD cases.

I buy mine from Bee Beautiful ( www.beebeautiful.org.uk ). It is very cheap stuff.



Common thicknesses are 12 and 15 Micron. For thermal cameras, the 12 Micron is best. The effect on the image is more than acceptable.

I recommend that the film is attached to some form of ring or frame with glue or other securing mechanism. It is then gently heated with a hair dryer and it shrinks drum tight  No wrinkles to cause image distortions.

FLIR sell disposable lens protectors that just clip onto the front of the lens. For a simple POLYOLEFIN item they are very expensive though.

The film is easy to find on ebay with a search fro Polyolefin.

http://www.ebay.co.uk/itm/20mtrs-Polyolefin-Shrink-Wrap-Film-25-micron-450mm-/351366578838?hash=item51cf148296:g:1GAAAMXQVT9SvXDC

http://www.ebay.co.uk/itm/20m-Polyolefin-Heat-Shrink-Wrap-Film-250mm-500mm-Centrefolded-/272104814150?hash=item3f5ab5f646:g:xEQAAOSwd4tULs~S

http://www.ebay.co.uk/itm/5m-or-10m-Polyolefin-Heat-Shrinkwrap-Film-15mu-500-1000mm-Centrefolded-/182055746914?var=&hash=item2a635e3d62:m:mr2uAuQtLWMdJXJmD7EKmxw

http://www.ebay.co.uk/itm/13-25-Micron-250-700-mm-Polyolefin-Heat-Shrink-Wrapping-Film-Centre-Folded-/161819677503?var=&hash=item25ad34533f:m:maNrLAg6RzmvbcXPsQGwAKg

Fraser

[OrBy]

Any chance that this will be up gradable to a 3rd gen Lepton when they start becoming available to the public?
(It may even be a suggestion to wait till they are as the higher res would be awesome)

[Andreas_Thermal]

The film that is used on commercial thermal camera lens protectors is POLYOLEFIN Shrink Wrap, as used in teh food and cosmetics industry. It is also to be found around some new Audio CD cases.

Thank you Fraser for sharing your resources!

Any chance that this will be up gradable to a 3rd gen Lepton when they start becoming available to the public?
(It may even be a suggestion to wait till they are as the higher res would be awesome)

Technically it would be no problem - I also thought about an upgrade solution to the newer core, the only issue for me is that customers have to open the case and from there it can get complicated in terms of warranty handling. Af few months ago as I started the project it was planned to be open source - maybe I`ll get back to this idea when FLIR releases their latest Lepton for OEM use.

Cheers,
Andreas

[Andreas_Thermal]

Just wanted to give you short update: we have integrated the Lepton 3 core in C-THERMAL, works great!

Sorry for the C-THERMAL advertisings in the video, but I also use it for promotional purposes .

Checkout the short comparison of Lepton 2 and Lepton 3 image quality: https://youtu.be/HPWR_mpzY9I

[OrBy]

Wow - seeing them side by side like that really highlights how much better the gen 3 core looks!

[Trax]

Shut Up and take my money!
Where/When can I buy it?
Around 300$ for a unit that does not need a cellphone its perfect for me.

[Fraser]

I suspect the problem at the moment is sourcing a Lepton 3 core. FLIR does not appear to be making them available to OEM's at the moment.

The current cheapest way to obtain a Lepton 3 is to buy a FLIR One Gen 2 camera ! ...... when there is stock available that is.

Fraser

[Andreas_Thermal]

Shut Up and take my money!
Where/When can I buy it?
Around 300$ for a unit that does not need a cellphone its perfect for me.

Hey Trax, we are going to start a crowdfunding campaign on Kickstarter soon - you can register for a notification on www.c-thermal.com.
You can also visit me in Leibnitz or Graz, there I could show you our working prototypes 

I suspect the problem at the moment is sourcing a Lepton 3 core. FLIR does not appear to be making them available to OEM's at the moment.
The current cheapest way to obtain a Lepton 3 is to buy a FLIR One Gen 2 camera ! ...... when there is stock available that is.

According to my information, we don't have to wait very long until the Lepton 3 gets available to OEM's 

[Fraser]

@Andreas Thermal,

Thanks for the info on the Lepton 3. That is great news 

Fraser

[tomas123]

@ Andreas_Thermal

How do you solve the problem of missing radiometric mode of lepton 1 and 3 for temperature measurements?
http://www.flir.com/cvs/cores/knowledgebase/index.cfm?view=67336

Yes, I know that the Flir One is radiometric...

see data sheet "8.4.2 Radiometry Enabled"
https://www.eevblog.com/forum/thermal-imaging/flir-one-thermal-imaging-camera-teardown-and-hacks/?action=dlattach;attach=206832

this  "14 Bit output values" are linear to Radiance [W/m²] (with an offset)
...and not linear to Temperature [K]"

I wrote a excel sheet for the planck law at 10600 nm wave length over the temperature range
There you can see, that the spectral radiance in [W / m^2] (per nm wave length) is not exact linear

[Trax]

Shut Up and take my money!
Where/When can I buy it?
Around 300$ for a unit that does not need a cellphone its perfect for me.

Hey Trax, we are going to start a crowdfunding campaign on Kickstarter soon - you can register for a notification on www.c-thermal.com.

Done

I was waiting for a Device like that since mikes "prototype".
The issue for me with the existing solutions was that they all would require me to buy a new smartphone, USB OTG needed. What I don't want to do as the current one still fulfills all my needs, especially that recently there even was an android 6.0 custom rom for it released.

You can also visit me in Leibnitz or Graz, there I could show you our working prototypes 

I'm located in Vienna so to Graz it till be a bit of a long drive, but may be I can find time for that...
Or will you be visiting Vienna in near future, for some reasons of Austrian bureaucracy or so?

Cheers
Trax

[Bill W]

Quote from: tomas123 on October 18, 2015, 04:52:44 AM

    I wrote a excel sheet for the planck law at 10600 nm wave length over the temperature range
    There you can see, that the spectral radiance in [W / m^2] (per nm wave length) is not exact linear

I would recommend extending that idea by adding the radiances calculated for 7500, 8500, 9500, 10500, 11500, 12500, 13500nm as you can also factor in lens transmission and sensor sensitivity.

It may change your curve slightly, but you should find that it gives a very good fit 0 - 1000°C

Bill

[Trax]

I'm wondering if the range of this lepton sensors can be extended using a filter?
Sure you would loose on resolution but if you want look to 300°C i guess you don't need sub 1°C resolution.

[encryptededdy]

I'm wondering if the range of this lepton sensors can be extended using a filter?
Sure you would loose on resolution but if you want look to 300°C i guess you don't need sub 1°C resolution.

If you use a ND filter you shouldn't lose temperature resolution / NETD, you will only lose the ability to make lower temperature measurements (ie. anything too cold will clip the bottom end of the sensor).

You will only lose NETD if the sensor is put into some kind of a low gain mode.

[tomas123]

I would recommend extending that idea by adding the radiances calculated for 7500, 8500, 9500, 10500, 11500, 12500, 13500nm as you can also factor in lens transmission and sensor sensitivity.


Bill

Edit field F29 of the Excel sheet for display another wave lenghts.

[Ben321]

I just had to try out one of your gradients which looks very night vision like  - As we have 8 Gb of internal memory, your suggestion is really no big issue to implement for us   

Regarding the window: At the moment the camera lens is recessed by about 1 mm but we have no final decision (yet) about using a window or not - those tiny little things are quite expensive.

Use a silicon window like the FLIR One has. FLIR managed to keep that pretty cheap. Alternatively, purchase a used Germanium window (better LWIR transparency than silicon) from eBay. Bet you could find one on there for about $100 or less if you are lucky. If you are trying to keep your price down, you should depend on eBay for your parts. You can get some pretty good deals on there. There may be a scratch or two on it, but as long as it isn't too bad (is near the edge of the window, rather than in the middle of it), and as long as your product isn't used for anything mission critical (like medical, industrial, or military usage) it should be fine. Ordinary consumer products like "point and shoot" digital cameras almost always have some small defects in them (both in the optics, and in the sensor, that's why they are SO CHEAP compared to professional devices like thermal imagers which are NOT aimed at the consumer market, as FLIR products are 100% WITHOUT DEFECT, making them HUGELY expensive). Consumer usage is just more tolerant of those defects than professional usage.

[Bill W]

If you use a ND filter you shouldn't lose temperature resolution / NETD, you will only lose the ability to make lower temperature measurements (ie. anything too cold will clip the bottom end of the sensor).

You will only lose NETD if the sensor is put into some kind of a low gain mode.

er no.
The ND filter will reduce the radiance from all temperatures just like closing an iris does.  With constant system noise your end image MDTD will rise. 
That is the problem with NeTD, it is a sensor parameter not a camera parameter.  It is also abused by many sensor manufacturers, eg FLIR quote NeTD at f/1 and then put a f/1.4 lens on the camera.

You could however win a bit with a 10um cut filter (but see Tomas's spreadsheet to sort out radiance to temperature scales)

Bill

[Bill W]

Posted by: tomas123

Edit field F29 of the Excel sheet for display another wave lenghts.

That is not what I meant.  I am suggesting copying the sheet 6 times, set each sheet to the different wavelengths and then sum the radiance result from all the sheets with appropriate transmission / sensitivity factors on a new master sheet.

R (total) = a*R(T)(7500) + b*R(T)(8500) + c*R(T)(9500) + d*R(T)(10500) + e*R(T)(11500) + f*R(T)(12500) + g*R(T)(13500)

where:
R(T)(7500) is the radiance of a black body at temperature T for 7500nm,  and assumed representative of the radiance from 7000 to 8000 nm.
a - g are the optic transmissions / sensor absorptions for each 1000nm block
I assume zero transmission below 7000 and above 14000

Bill

[tomas123]

sorry, but I think, your approach is unnecessary complex

https://en.wikipedia.org/wiki/Planck%27s_law

The sensor manufacturers give us a "Peak Spectral Sensitivity" and calibration values are conversion factors for temperature=function(radiance)
You don't need special investigations of band with.

[Bill W]

Sensors are calibrated viewing (typically) 20°C and 40 °C images.

Stored factory gain / offset tables convert sensor ADC output to an arbitrary linear-in-radiance scale based on the radiances of the 20° and 40 °C images.  This process probably invisible to the user.
The 'actual-temperature' images are then calculated using either a measured or calculated conversion curve much as you have plotted.  Around ambient a curve for 10.6um will be fine.

However when dealing with say 1000°C, so much of the radiant energy is in the lower wavelength areas that you do need to deal with what is essentially  'colour correction'.  The FLIR response curve you posted would otherwise result in undereading temperatures for very hot objects.
The 7.5 - 8.5um band is 15% of an ambient scenes' radiance (between 7.5 - 14.5um), 18% of a 100°C scene but is 32% of the radiance from a 1000°C scene.
A consequence of https://en.wikipedia.org/wiki/Wien's_displacement_law

regards
Bill

www.fire-tics.co.uk

[tomas123]

A great answer 

My first thought was, that a necessary ND filter for 1000°C will also move the peak of radiance.
But a ND filter is not a wavelength converter 


With a ND filter, the camera needs certainly new calibration values for the function Temperature=f(RAW).
And Andreas_Thermal hasn't answered to my question, how he calculates this function.