Yet another cheap thermal imager incoming.. Seek Thermal

[mikeselectricstuff]

Did anyone found any kind of temperature sensor on PCB in this Seek Thermal design?

There is a diode on the rear which I'm pretty sure is used as a temp sensor.

[eneuro]

After dinner realized that metalic lens holder will tend to more quicly uniform any local heating up, so it must be die enclosure much worse thermal conductivity lead to this gradient effect like heating glass block lying on the board with top cornets exposed to heat from hot air around lens holder with hole and air gaps and additionaly one side corners not masked by lens holder.


While at the corners relative area/per volume is higher it will got hotter additionally side conners from shutter side are exposed to heat mixed and pumped by shutter in calibration events,
so while lens holder is cooler heat from central top sensor will be cooled and heat up at lower rate than snensor enclusure block cornets
and due to bad thermal conductivity of senesor enclusure gradients in its block must exists like in glass heated from top and side corner.

In Flir E4 design sensor is completely hidden inside huge thermal capacity metal cave and even looks like additionaly thermal connection to sensor bottom PCB is made with temp sensor coupled with it to measure this ambient metal cave temperature  in which sensor operates-they really nice and hard try to make this temp uniform around sensor inside, so had to put inside such nice shutter shown before but worth to see it again, to avoid any interference from any side of die enclosure which could create gradients in sensor block 

[efahrenholz]

@eneuro

If any heat was getting soaked into the die, a shutter event would include this during subtraction of the flat field. If you look at miguelvp's photo of the hot ring, you see the outside ring is warmer than the data in the center. This shows that the lens isnt aligned properly, but most importantly, that the bottom flat(what the lens sits on) of inside the lens housing is reflecting the sensors heat back at itself. Also, I bet a good deal of money on the image circle is barely large enough to cover the sensor. Its made to a low cost, so its clearance will be tight.

 A few people are tied up on this being a convection problem. A metal housing will heat fairly evenly, but it's not even directly sitting on a good heat conductor, so it isn't really able to get PCB heat. And its too far away from the sensor so there's a gap where glue takes up the space. This isn't a convection problem, it's a radiation problem. The shutter isn't responsible for this either. It covers the sensor entirely during a calibration. That's all it needs to do. Then it moves completely out of the way. I tested this as well as mike.

Look at the photo of the lens housing inside, and then look at the photo miguelvp made. Taken into consideration the problem gets worse with time (heat buildup), and you have your culprit. The sensor isn't visible to itself until it's warm enough that it radiates significant heat back off the lens housing onto itself. This throws the flatness off because it is adding to the data where it shouldnt be there at all. Use the high-low mode and look at ice, the hi is almost 20 degrees from low. Look at a hot object, the hi and low are close together. The radiation will always add to the scene, but the biggest difference will be on cold scenes. Hot scenes only have a minor impact. Scenes with wide temperature ranges won't show the gradient because the gradient only shows itself when the palette is squished to fill a small range.


The lens housing is too compact and as a result sits close to the sensor die. The die is small, so the focal point is closer to the lens, look at any phone camera. The smaller the die and lens ratio, the closer they must be together to achieve focus. This is one of the issues of a 12um sensor of this resolution. Its a compromise.

[Fraser]

@efahrenholz,

Very nice analysis. Thank you.

That diaphragm in front of the microbolometer does seem quite invasive to the signal path and it seems very reasonable to believe that it could be causing radiation reflection issues for the microbolometer. I was not expecting narcissus effect from the lens tube but you may well be on to something here.

I am still considering options to replace the lens tube with something less enclosed, and now with less of a diaphragm, if any.

Thanks again.

Aurora

[eneuro]

Scenes with wide temperature ranges won't show the gradient because the gradient only shows itself when the palette is squished to fill a small range.

Nope, you call gradient something that in fact is thermal image destruction in the corners and is more visible in one of them for reasons I wrote and modeled with glass block.
Wide temperatures are even on this router thermal image and  image destruction easy to see in the corners even from raw sensor data calibration frames provided there earlier-a few averaging makes this cleary visible and after this small mechanical investigation I've made now fits into this what is seen there in one of these flat surface frames.


We'll se how it will look like on latest raw sensor data provided by @miguelvp - those output images by @miguelvp are not compenated and horizontal lines still visible-more advanced image processing is neded there.

In this @Mike photo from its Seak teardown there is no such small range but temperature difference close to 5*C on flat surface and it is not acceptable for anything more serious but the toy.


Another review - now we know where to look for image destruction and catch this worst case scenario corner where most of the heat comes inside sensor cage   


Probably there below i sno reason for such warm corners while it is pointed on the table-my guess is gradients in the corners visible and there is nothing to compare to higher price Flir equipment-from its mechanical design now we know Seek Thermal dongle for the moment is a thermal toy 


Thx, @mikeselectricstuff for this thread and teardowns - I've saved $200   

Update:
Another example in dark night... complete destruction of thermal image at the corners with one much stronger than rest of the other corners,
but ok depending on wind conditions and ambient temperature there maybe air heated by this animal created this aurora and it is not close to animal shape, but you never know while this deveice is not reliable...


OK, lets wait for hardware update of Seek Thermal and than maybe will give them the second chance?

[efahrenholz]

@mike,

If your lens isn't glued down, perhaps you could take a white sheet of paper, measure the opening on the inside of the lens, cut that out of the paper and slowly lower it closer to the camera after different lengths of time. I believe you will see a bright ring around the image after two minutes. Also I believe the ring will disappear if you move the paper closer, like the same as side blinder for a horse.

[efahrenholz]

@efahrenholz,

Very nice analysis. Thank you.

That diaphragm in front of the microbolometer does seem quite invasive to the signal path and it seems very reasonable to believe that it could be causing radiation reflection issues for the microbolometer. I was not expecting narcissus effect from the lens tube but you may well be on to something here.

I am still considering options to replace the lens tube with something less enclosed, and now with less of a diaphragm, if any.

Thanks again.

Aurora

I don't believe you will be able to remove the lens housing from the design. It was intentional and I'll tell you why. Seek wanted this camera module on cell phones. Devices that blast all sorts of radiation which would easily be picked up by the small bond wires. They are short, but they are not so short that they can't receive interference. So the lens holder was probably designed as a Faraday cage of sorts. It is sitting close to the bond wires, and likely acts as a small ground plane. Granted the whole module sits in a magnesium case, but its thin. If you look at other thermal cameras, most don't have Bluetooth and wifi or a cellular signal being blasted from the board. They needed a way to kill the noise getting into the bond wires without a radical design change.

This is my theory at least.

[efahrenholz]

I did come up with an idea that maybe we could remove the lens from the lens housing and just glue it to the rubber grommet, and shove that into the module case like it was designed to. This eliminates the housing, leaving an unobstructed view, but the case will warm up, which means radiation from the case will shine down onto the sensor. My module does get a little warm, so it would likely be more beneficial to leave the housing in place, to reduce noise to the bond wires, and to prevent internal case radiation from blinding the sensor. The only solution I can foresee is external through the lens calibration. 

[mikeselectricstuff]

I don't believe you will be able to remove the lens housing from the design. It was intentional and I'll tell you why. Seek wanted this camera module on cell phones. Devices that blast all sorts of radiation which would easily be picked up by the small bond wires. They are short, but they are not so short that they can't receive interference. So the lens holder was probably designed as a Faraday cage of sorts.

The lens holder is not electrically connected, so would not be very useful as a shield.

I did come up with an idea that maybe we could remove the lens from the lens housing and just glue it to the rubber grommet, and shove that into the module case like it was designed to

Wouldn't be accurate enough.

[Fraser]

I have just seen my first SEEK iOS ebay auction:

http://www.ebay.co.uk/itm/Seek-Thermal-Camera-LLW-AA-iOS-/371186239600?&_trksid=p2056016.l4276

Either they are selling something that they do not have yet or the iOS version is shipping.

Aurora

[efahrenholz]

@mike,

Fair enough. Let's just call the lens housing a thermal barrier.

[eneuro]

The only solution I can foresee is external through the lens calibration.

Before saw Flir's E4 approach where this cool shooter is put between sensor and lenses BUT inside metal cage with accurate ADT7031 temp sensor thermally coupled I was thinking that putting shutter in front before lens could do the job, but... no way 
During calibration we need to know reference ambient temperature and while sensor and shutter is in one closed cage it can be kept in uniform distribution and quite easy to controll in changing environment, while when outside such cage we do not know which is its temperature and it is difficult to controll it.
Of course I maybe wrong but quick review of few teardown and quite easy to see that it have to be consistent design form mechanical point of view and thermal conductivity else difficult to get any accuracy and deal with external heat.

Probably Seek Thermal wanted make this dongle as small as possible for mobile gadets people, so while this nice integrated Flir E4 shutter is wider has bigger footprint, they Seek device had to have very different shape to fit everything together-so they did it how we seen and run into problems or it was calculated risk with estimated expected returns from not happy customers vs much more volume of unaware ones which are not able to notice something which is such invisible like temperature when haven't got another equipment....

[efahrenholz]

@miguelvp,

Any more progress on seeker?

[Fry-kun]

Here's linux/python code: https://github.com/Fry-kun/pyseek
It's based on cynfab's code from https://www.eevblog.com/forum/testgear/yet-another-cheap-thermal-imager-incoming/msg544080/#msg544080 -- which pretty much mirrors sgstair's Seek interface from https://github.com/sgstair/

Code is really simple right now, only displays live capture on screen.

To start:

Code: [Select]

$ sudo python seek.py

To quit: press Escape

Didn't have much time to play with it yet, but seems like a lot of frames look completely black for no reason... possibly contrast issue?)

[miguelvp]

@miguelvp,

Any more progress on seeker?

Still working on it, but it's the week so I only have a couple of hours to put on it, but I did find that I can use the first frame to compensate with the shutter without an external reference, will add progress later on.

Adding the sliders made it unstable again, so I'm also addressing that.

[efahrenholz]

You mean you found a way to remove the need to take an external image through the lens?

[miguelvp]

Yes but it leave a veil noise residue, btw I'm not using pixel 207 to get rid of the lines, frame one on boot has enough data to clear it, but I need to scale based on temperature because it shows up on hot spots



Compared to not using frame1:

[miguelvp]

But it facing the camera down shows up with a slight gradient,



If I do take frame 1 and change the contrast it seems it was generated by the actual sensor so it might be stored in the device somewhere during factory calibration, but the ring doesn't correspond to the placement of my lens, since it looks way more centered than mine.



Or maybe it's the same, not sure:


Edit: It does fit but it doesn't match the gradient, it's like if they calibrated them somewhat different than when they are running, maybe they were calibrated at boot boot time.



This is frame 1 overlaid when the shutter was colder, seems like a better match:

[marshallh]

Someone get this man a nobel peace prize

[Fraser]

+1 

I hope SEEK Thermal are suitably impressed with the excellent work that is being done on thsi forum. They could possibly achieve great things if they were to co-operate with certain members of this thread and provide assistance with non-commercially sensitive design details.

Aurora

[eneuro]

I hope SEEK Thermal are suitably impressed with the excellent work that is being done on thsi forum.

Yep, probably now they now how to NOT design thermal dongles, but do not care too much about this forum, because their target customers are not electronic guys who can hack their circuits and writ eown better software, but gadget fans, who will be able now make thermal blob of their cats, dogs, etc and will have another good reasons to chat on facebook etc. about this 

BTW: Gadget fans have no idea what gradient means at all and it doesn't matter that for the moment difficult to find thermal image from this Sick Thermal cam without destruction in the corners-they are happy thay can in restaurant now check which caffe cup is hot and go at night throutgh park with this dongle in hand saying and showing that now someone behind the trees can not only steel crappy phone worth $200 but additionally Sick Thermal dogle as a bonus  which is in their hands

[Yushir0]

I've noticed that the seek won't produce any video files on my phone running cyanogenmod and I suspect that it is due to missing a codec.  Can someone generate a short video and run the following command from a Linux command line so that I can figure out what codec they are using?

mplayer -vo null -ao null -frames 0 -identify <filename>.mp4 | grep "ID_VIDEO_FORMAT"

[Andrew Seltzman]

I think this thermal gradient issue(brightening around the corners) may be an issue with the lens, not the sensor itself. I made a FLIR photon 320 do the exact same thing when I tried to substitute a non-athermalized Thorlabs germanium lens for the factory lens. Has anyone tried substituting the lens from a high end thermal imager onto the seek?

See pictures near the bottom of
http://www.rtftechnologies.org/general/thermal-imager-flir-photon-320.html

Factory lens:


non-athermalized Thorlabs lens:

[efahrenholz]

I believe its a calibration for the lens housing. Stored right in the flash. But it appears they are applying it assuming alignment from assembly hasn't altered much. However as we have seen, many units have a gradient showing up in different places.

So seeks engineers knew about the problem, and provided a general solution, but seems this is ineffective.

[miguelvp]

It might be as simple as stretching the frame to compensate from where it heats more. So giving more weight to the right bottom of frame 1 might solve the issue.

It's not like they could have 2 minutes worth of powered up devices queued up on their production line, and that will still cause a problem when the device first powers up.

I did order a ZnSe lens 101.6 mm (4") focus length and it probably will take another week or two to get here.

Edmunds Optics lists ZnSe as toxic, maybe because they offer uncoated lenses, mine comes from China:
http://www.ebay.com/itm/20mm-CO2-ZnSe-Laser-Collimating-Lens-Laser-Engraver-Cutter-Focus-101mm-4-/261541512155

So other than not exposing it to acid is it safe to handle like I've seen others do, or is there enough acidity in your hands to make it react (probably not visibly but releasing gases to your skin)?

I do have disposable gloves that my wife buys by the box so I'll use them to handle the lens just in case, or am I being paranoid? Thinking about it, I might just do it anyways to keep oils out of the lens.