Yet another cheap thermal imager incoming.. Seek Thermal

[Fraser]

Mike has a FLIR E4(E8) with close up lens attachment. A better choice than the Fire FLIR 

He is a very busy chap at the moment (talk of 6000 LED's). We need to be patient.

I must say I am already enjoying the dissection of the SEEK and all of its foibles.... this will be a very interesting on-going thread much like the E4 Teardown 

[cynfab]

Here is one from mine, pointed at the floor, radiant floor heat turned off, so it has a large thermal mass.

   ...ken...

This was taken right after powering up the camera.

[ricksastro]

Here's one from mine,   Looks this way either pointed at the floor or camera face down on the desk.   Turned on the min/max temps to give an idea of the bad magnitude of the problem.

Starts out looking OK for the first few seconds then gets worse as time goes by.

[waldo]

I get exactly the same temperature distribution as rickastro after it has settled for about 2 minutes.  Whatever the cause, seems like it might be something that can be corrected for in software if it is always that repeatable.

[miguelvp]

Here is mine pointing at the wooden floor on a spot I didn't step in a while.

Since the Shield doesn't quite work for displaying I can't activate the min/max (well I can activate it but it doesn't show)

Edit: also taken right after powering up the camera

I'll turn it on again and let it settle for a couple of minutes and take another one.

[Sensor Geek]

2-minute MP4 attached showing face down from cold startup. After one minute the gradient is obvious and continues to expand.

http://myfilehost.nfshost.com/20141025_seek_gradient.mp4

Edit: Moving lines in the video are an interference pattern caused by my camera. Not visible to the eye.

Edit 2: Seek just released an update which was applied prior to that video and after I first noted the issue a few comments back. Currently running:

Hardware: 0.0.0.5
Camera s/w: 1.3.0.0
App: 1.3.0.1

[miguelvp]

I get the a similar result after a while, first attached image

But one interesting thing is that my Nvidia Shield has a magnet near the clam and I noticed that it holds the shutter closed if I'm close enough to it. (Of course don't do this for a prolonged time or you might burn the electromagnet that opens the shutter if it's trying too hard to open the shutter).

Anyways, I took a shot when it was closed, 2nd attached image

Software doing the temperature drift? because with the shutter closed the sensor is fine.

Edit: looks like a bug to me, where their correction image keeps adding an error overtime. Edit 2: on second thought it would affect the sensor with the shutter closed as well, not sure why it's happening, the lens casing?

[miguelvp]

Seems like the spot is common to all cameras.

I think this is thermal interference from the lens assembly or the lens itself warming up in some spots.

Since the calibration is done without the lens then it doesn't compensate

[Fraser]

All very interesting data guys, please keep experimenting 

The fact that the shutter may be operated by an external magnet is a useful discovery  No harm should occur as the camera does not have shutter position monitoring so will not increase current to the solenoid coil 

From what miguelvp has seen, the shutter being activated manually removes the issue from the displayed image....as he states, that firmly places the later optical path components in the frame for causing the issue as the shutter only blocks the signal path from the lens. Interesting but I currently have no suggestions as to what is happening here

Mike with his 'brain the size of a planet' may well have some luck working out what is happening on his stripped unit, but that will need to wait until he has more time for 'toys' 

[Fraser]

If SEEK engineers are reading this forum thread, feel free to comment either privately to me, or as an anonymous forum member, on this issue. We may be able to assist you in correcting this anomaly in your camera    The EEVBLOG forum contains some very knowledgeable techs who may be a good UAT group for you 

Aurora

[miguelvp]

I don't think there is much they can do. I did another experiment and sure enough, even after 10 minutes if I force the shutter close there is no drift but the lens or light path shows the drift. I will suspect the lens itself.

Only obvious solution is to somehow make sure the lens retains a flat thermal signature. Being that small it probably wouldn't be hard to preheat it at certain temp and keep it at that temp, but that will add into cost to them. Or maybe the assembly could have a heat sink around the lens.

Edit: or redesign it so that the shutter is over the lens instead of under the sensor.

Edit: forgot to attach the images, both are after 10 minutes, the last one forcing the shutter closed with a small magnet behind the camera sensor.

[Fraser]

Something to be considered as well is the autospan function of the camera.

When pointed at a uniform temperature surface the camera should automatically decrease its span and so effectively increases its sensitivity to temperature gradients in the optical path.

Hmmmm so many questions and no answers yet. At least if it is a physical issue it may be able to be easily identified via a series of tests.

The presence of the temperature gradient 'plume' at different temperature spans would be an indicator of whether we are seeing a genuine serious issue or one that only occurs when the unit is running at its most sensitive. I can even see some uniformity issues in my other cameras when they are running at their smallest temperature spans.

[miguelvp]

True, they could offer a manual calibration by software that compensates for the lens when pointing at a uniform room temperature surface, giving a mask that they can multiply the image with to compensate for the lens assembly thermal issues.

[ricksastro]

Actually, the fact that the shutter closed images look good point exactly to the shutter itself as the source of the gradient if the calibration is working correctly.    The calibration is neutralizing the shutter gradient, and thus we get negative shutter gradient overlay on the real image.

[Fraser]

Excellent .... this is what I like about brain storming a problem as a group  ....indeed the shutter could still be the issue. Experimentation with it by only partially moving it across the signal path should show any temp gradient across it as it moves into view by causing non uniformity when the lens is covered by a uniform temperature material.

Partially moving the shutter without physical access to it may be a challenge though ?

I am also wondering what could cause temperature change across the shutter. A warm component beneath it or thermal currents passing over it and out of the solenoid port ? Sadly the lens holder obscures a decent view of the components around the microbolometer FPA.

Its getting late in the UK now so I will have to sleep on this. It would be great to read more thoughts and test results from my fellow forum members across the pond in the USA. All good fun    If I had a SEEK in my lab right now I would be examining the lens assembly and shutter flag with my FLIR E4, fitted with its best close-up lens. With the right lens you can actually look into the SEEKS lens and see the shutter temperature gradient, if present.

Aurora

[miguelvp]

Actually, the fact that the shutter closed images look good point exactly to the shutter itself as the source of the gradient if the calibration is working correctly.    The calibration is neutralizing the shutter gradient, and thus we get negative shutter gradient overlay on the real image.

Could be that as well, I don't know if I can use the magnet to force it open and test it by placing it face down on a uniform temperature surface.

Edit: ok I can force the shutter open externally with a group of neodymium magnets (probably overkill) under the Seek Logo. I'll let it warm up and post the results

I hope I don't permanently magnetize something

[miguelvp]

So ricksastor was right. I did let the camera warm up for a while and let it keep on clicking for 5 minutes.

Then I placed the magnet to force it open waited 20 seconds and took a picture of the wooden floor. Then I pointed it at the door (using a usb micro b extension cord so I don't know which orientation it was taking the picture) and took another one to prove that the shutter was open. No clicks happened between both snapshots and the magnets held the shutter open until I was done.

It has different patterns but I wasn't too methodical, however it does show the thermal signature problem is on the shutter as ricksastor predicted.

[Fraser]

Excellent work 

Now to work out what causes the temperature gradient on the FFC shutter flag piece.

Correcting the issue may involve no more than adding a small thermal shield between the heat source and the shutter.

Aurora

[miguelvp]

To recap on the external magnet operation.

Magnet behind the sensor opposite to the lens, forces the shutter close.
Magnet under the Seek thermal logo opposite to the USB connector, forces the shutter open.

Edit: use at your own risk, I don't think it will do permanent damage but I might be wrong.

[miguelvp]

So, this explains a lot.

Why does the shutter is constantly clicking away? Well it calibrates the sensor with the flag that is not thermally uniform, so after opening the shutter it will get out of calibration fast, so it needs to calibrate again.

I'm going to do one last test, let it warm up. Force the shutter open and take a snapshot, then as quick as I can force the shutter close and take another one. it should give us a negative image of the drift if I'm not mistaken.

[Fraser]

A quick look at the PCB design and shutter position leads me to the following comments:

1. The likely greatest sources of heat in the SEEK are the switching regulator transistor and the microprocessor. If we treat them as warm air generators within the SEEK's sealed case and consider convection air movement we may find our problem.
2. From my brief view of the PCB the shutter operates from the top of the SEEK module where hot air will collect over time. This could warm up quite quickly as the micro will be a decent source of heat.
3. There is a large through-hole under part of the shutter arm and flag. The hole could potentially provide a focussed warm air path from the rear of the PCB to the front. The micro is on the rear so it is not unreasonable to think that the rear of the SEEK module will be warmer than the front in terms of air flow.
If I am correct, the air around the micro will warm up quicker than that around the lens assembly and this air will rise to the top of the module where it will find a path to the lens assembly via the through hole. Such a warmer air current is not great for the FFC shutter temperature gradient across the flags surface as it is in the path of said air current..
4. Proof of the theory would involve opening the camera shell and blocking the through hole in the PCB adjacent to the lens.

Additional thought....

Operating the SEEK without its rear shell in place would also stop the warm air build up that drives the air current through the hole and onto the shutter. This would prove the issue id thermal convection heating of the shutter flag rather than direct radiant heat from another heat source.

All this is coming from a pretty tired brain in the UK so forgive me if I have rambled on a bit !

EDIT,

I have added a picture taken by Marshall to help show the issue with the through hole above the lens assembly. (Hope this is OK with you Marshall)


Aurora

[mikeselectricstuff]

A quick look at the PCB design and shutter position leads me to the following comments:

1. The likely greatest sources of heat in the SEEK are the switching regulator transistor and the microprocessor. If we treat them as warm air generators within the SEEK's sealed case and consider convection air movement we may find our problem

the switching reg is at the opposite end, and is unlikely to get hot, if you assume it handles all the input current ( which it doesn't but I don't have a figure for the seperate regs), and it's 75% effifcient, dissipation will be about 60mW

However the linear reg for the sensor is a lot closer, and probably gets a lot warmer - I'll measure it at some point. I can measure current via the resistor array on the output.

[miguelvp]

Can the source of the heat be the actual sensor?

This is what I did, let it warm up then force the shutter open for a while and took a picture.
Then I immediately forced the shutter close and took the 2nd snapshot, The flag seems to be fine.
Held the shutter close for a while removed the magnet and took the 3rd snapshot.

I think the sensor is heating up the flag.

[Fraser]

The microbolometer runs at around 30C in most non TEC stabilised cameras. If you look at yourself in a mirror using the SEEK you will see a warm glow from the lens. This is the narcissus effect of the camera seeing its own microbolometer heat through the lens structure.

If the shutter is actually touching the microbolometer it could suffer conducted heat contamination quite quickly. If the contamination is radiated heat onto the shutter we have a serious problem and I cannot think of a quality countermeasure to such.

See here for narcissus effect examples relating to TIC's:

http://books.google.co.uk/books?id=b-MqbyPwAuoC&pg=PA125&lpg=PA125&dq=narcissus+thermal+camera&source=bl&ots=3bIumYKQNw&sig=5Qsi5GIISpN8p3FDYF8F-V152xs&hl=en&sa=X&ei=_UJMVJqtBKzV7Ab4koG4AQ&ved=0CCMQ6AEwAA#v=onepage&q=narcissus%20thermal%20camera&f=false


My brain is fading fast so time for bed. Thanks for all the good work on this folks.

TTFN

Aurora

[miguelvp]

The only device I have that can use the Seek camera is my Nvidia Shield so I can't see it updating live, only after snapshots. My wife is out of town so I can't use her phone to test this further (or easier).

I don't think the shutter is touching the microbolometer but I do think the temperature inside the lens enclosure is hotter than outside, so when the flag is inside its getting baked a little and since they keep calibrating often it doesn't have time to cool down and it gets worse overtime.

I'm going to do one more experiment. Now that everything is back to room temperature I'm going to start it and force the shutter closed for 30 seconds and see if that accelerates the problem.

Edit: inconclusive, I will have to wait until my wife gets home tomorrow to set with her galaxy s4 so I can see what is happening real time.